本篇博文主要内容为 2026-09-02 从Arxiv.org论文网站获取的最新论文列表，自动更新，按照NLP、CV、ML、AI、IR、MA六个大方向区分。
说明：每日论文数据从Arxiv.org获取，每天早上12:30左右定时自动更新。
提示: 当天未及时更新，有可能是Arxiv当日未有新的论文发布，也有可能是脚本出错。尽可能会在当天修复。
目录
概览 (2026-09-02)
今日共更新867篇论文,其中:
- 自然语言处理共186篇(Computation and Language (cs.CL))
- 人工智能共312篇(Artificial Intelligence (cs.AI))
- 计算机视觉共152篇(Computer Vision and Pattern Recognition (cs.CV))
- 机器学习共205篇(Machine Learning (cs.LG))
- 多智能体系统共10篇(Multiagent Systems (cs.MA))
- 信息检索共21篇(Information Retrieval (cs.IR))
- 人机交互共33篇(Human-Computer Interaction (cs.HC))
多智能体系统
[MA-0] GlossoGen: Emergent Language in Complex Multi-Agent LLM Interactions
【速读】：该论文旨在解决多大语言模型（LLM）智能体在交互过程中语言演化的机制与影响问题，尤其关注其对安全性、可监控性以及对LLM语言表征的理论理解所带来的挑战。其核心解决方案是提出GlossoGen平台，用于在复杂情境下研究多智能体语言演化过程。关键发现包括：在压力环境下，具有部分信息的智能体间会自发演化出具有组合性与形态生成能力的语言，且这些语言显著偏离人类可理解的英语语义结构；语言演化依赖于效率驱动、智能体背后模型的能力强度，以及“尸检”阶段（postmortem stage）以达成语言约定的机制。此外，研究揭示了语言传播的新规律：新语言可通过使用经验习得，智能体在学习中扮演主动角色；虽然新型语言的产生需强模型支持，但弱模型可在语言形成后有效学习并应用。综上，该研究表明当前LLM具备累积性文化演化潜力——这一特性此前仅被证实存在于人类之中，且异构智能体群体能发展出超越其最低能力水平的协同能力。
链接: https://arxiv.org/abs/2609.01491
作者: Elias Stengel-Eskin,Newton Sander,Carlos Bonetti,Sasha Boguraev,James Bowler,Hale Sirin,Simon Kirby
机构: University of Texas at Austin (德克萨斯大学奥斯汀分校); AE Studio; Schmidt Sciences; University of Edinburgh (爱丁堡大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注: GlossoGen code: this https URL Paper code: this https URL
Abstract:The growing rate at which LLM agents interact with one another raises key questions about language evolution in multi-LLM-agent settings, with implications for safety and monitorability as well as for linguistic accounts of LLMs. To address these questions, we introduce GlossoGen, a novel platform for studying multi-agent language evolution in complex scenarios. Within GlossoGen, we build the SaveVeyru scenario, which requires agents with partial information to communicate under pressure. We find that language evolution does occur between LLM agents, that the resulting languages are compositional and morphologically productive, and that they deviate from the LLMs’ English prior in ways that render them incomprehensible to humans. Moreover, we identify several qualities essential to this evolution: pressure towards efficiency; the strength of the models backing the agents; and access to a “postmortem” stage in which agents can agree on linguistic conventions. Importantly, we observe that different conditions govern the transmission of language to new agents. Specifically, we find that agents learn new languages from usage alone, take an active role in this learning, and that while stronger models are required for novel language emergence, weaker models can learn an existing language once it has emerged. Taken together, our results indicate that current LLMs have the potential for cumulative cultural evolution – previously attested only in humans – with mixed populations of agents developing capacities that go beyond their lowest common denominator.
[MA-1] Classic AI Scaffolding for LLM Social Agents
【速读】：该论文旨在解决大语言模型（Large Language Models, LLMs）在社会模拟中仅能生成局部流畅的对话回合，却难以实现连贯、有结构的社会行为序列化问题。其核心挑战在于，真实人类互动（如餐厅午餐、酒店入住）本质上是具有角色分工、脚本流程、物质状态、义务承诺、时间约束及收尾条件的有限社会事件（bounded social episodes）。为应对这一问题，论文提出EpisodeSim——一种混合式LLM-代理架构，其关键创新在于将经典人工智能（Classic AI）中的结构化控制机制以自然语言形式表示，并通过调用大语言模型进行解释与执行，从而实现对社会情境的动态建模。其中，“世界主控”（World Master）作为核心组件，负责维护共享现实、构建场景、裁定行动提议、追踪行为后果与义务状态，并控制事件的终止条件。实验通过在两个独立场景下的小规模定性消融分析验证了设计主张：尽管大语言模型提供了局部语义流畅性，但真正的社会模拟一致性依赖于持续存在的、类经典人工智能式的结构支撑，以组织跨时间的行为逻辑。
链接: https://arxiv.org/abs/2609.01167
作者: Anatole Gershman
机构: Carnegie Mellon University (卡内基梅隆大学)
类目: Multiagent Systems (cs.MA)
备注:
Abstract:Large language models can produce locally plausible social turns, but fluent next-turn generation is not enough for social simulation. Human encounters such as restaurant lunches and hotel check-ins are bounded social episodes with roles, scripts, material state, obligations, commitments, timing, and closure conditions. We present EpisodeSim, a hybrid LLM-agent architecture that represents classic-AI structures as natural-language control state interpreted by LLM calls. A World Master maintains shared reality, constructs scenes, adjudicates proposed actions, tracks effects and obligations, and controls closure. Experiments with small qualitative ablations on two held-out settings support a design claim: LLM fluency supplies local texture, but coherent social simulation benefits from persistent classic-AI-style scaffolding that organizes behavior over time.
[MA-2] Update for Decisions Not Freshness: Goal-Oriented Status Updating and Selective Offloading at the Network Edge
【速读】：该论文旨在解决边缘-云协同计算环境中，边缘节点（EN）在部分可观测条件下对用户任务进行本地执行、远程卸载或拒绝决策的优化问题。由于边缘节点仅能通过间歇性更新的缓存获取远端服务节点（SN）的状态信息，导致状态更新与任务调度形成异步闭环，传统基于信息年龄（Age of Information, AoI）的鲜度驱动策略无法有效反映状态更新对后续任务决策的实际影响。为此，论文提出了一种协同式事件驱动强化学习框架CoSMO（Co-design of Semantic-state Management and Offloading），其核心在于通过可实现的任务效用（realized task utility）协同优化语义状态管理与选择性卸载。CoSMO在SN端采用基于循环半马尔可夫双深度Q网络（Double DQN）的代理，联合决策“发送/不发送”及下一决策周期；在EN端则采用任务终止型离策略值学习代理，基于局部观测与过时的远程语义信息做出分层的“门控-路由”决策。两个代理虽保持独立的观测与价值目标，但共享同一实现实用性信号，无需集中式控制。实验表明，在多种工作负载下，CoSMO相较于最优对比方法，在按时完成率上的相对提升平均达18.6%–21.2%，在三种严格过载点下的容量感知决策准确率提升平均为17.6%–17.9%，验证了其在复杂动态环境中的有效性。
链接: https://arxiv.org/abs/2609.01082
作者: Jianpeng Qi,Qiyang Zhang,Chao Liu,Jing Sun,Yimei Liu,Yanwei Yu,Yingjie Wang,Wei Ni
机构: 未知
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Multiagent Systems (cs.MA); Networking and Internet Architecture (cs.NI)
备注: 16 pages, 13 figures
Abstract:In an edge–cloud collaborative edge-computing environment, an edge node (EN) must decide whether each user task should be executed locally, forwarded to a remote service (or cloud) node (SN), or rejected. The EN observes its local state directly but receives the SN state only through an intermittently refreshed cache. Status updating and task control therefore form an asynchronous closed loop under partial observability. Freshness-driven schemes, including those based on Age of Information (AoI), do not directly value an update by its effect on subsequent task decisions. We propose CoSMO (Co-design of Semantic-state Management and Offloading), a cooperative event-driven reinforcement learning (RL) framework that coordinates semantic status management and selective offloading through realized task utility. CoSMO learns a compact representation of the heterogeneous SN service state. At the SN, a recurrent semi-Markov double deep Q-network (Double DQN) agent jointly selects send/no-send and the next decision interval. At the EN, a task-terminal off-policy value-learning agent makes hierarchical gate–route decisions from local observations and stale remote semantics. The agents maintain separate observations and value targets but share the same realized task-utility stream, without centralized execution. Across the evaluated workload families, CoSMO’s reported relative improvement in on-time completion rate over the best-performing competing method averages 18.6%–21.2%. For capacity-aware decision accuracy across the three strict-overload points, the corresponding reported gains average 17.6%-- 17.9%.
[MA-3] ChatDev 2.0: A No-Code Multi-Agent Platform for Developing Everything EMNLP2026
【速读】：该论文旨在解决大语言模型（Large Language Model, LLM）驱动的多智能体系统（Multi-Agent System, MAS）在开发过程中面临的表达能力与易用性之间的权衡问题。现有代码框架虽具备高表达性但工程复杂，而无代码构建工具虽简化了开发流程，却限制了智能体间的交互模式为预定义的工作流。为此，论文提出ChatDev 2.0: DevAll（简称DevAll），一个无代码平台，实现了高度表达性与易用性的统一。其核心解决方案在于：一方面，通过声明式可执行图抽象与循环感知的执行引擎相结合，支持异构智能体及动态、循环交互的建模与执行；另一方面，集成可视化界面，使用户可在无需编写代码的情况下完成多智能体系统的构建、运行、监控与调试，包括人机协同步骤。实验表明，DevAll在三个代表性任务上复现了当前最先进的多智能体系统性能，且无需针对特定任务编写编排代码，验证了其作为通用型LLM-MAS平台的有效性。
链接: https://arxiv.org/abs/2609.00714
作者: Yufan Dang,Shu Yao,Bowen Lai,Chenting Xu,Ruijie Shi,Wai-Shing Leung,Huatao Li,Chen Qian,Zhiyuan Liu
机构: Tsinghua University(清华大学); Shanghai Jiao Tong University(上海交通大学); Shanghai Innovation Institute; Fudan University(复旦大学); Peking University(北京大学); Massachusetts Institute of Technology(麻省理工学院)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Multiagent Systems (cs.MA)
备注: Accepted at EMNLP 2026 Demo Track
Abstract:Large language model (LLM)-based multi-agent systems (MAS) have shown strong potential for solving complex tasks, yet their development forces a tradeoff: code frameworks are expressive but engineering-intensive, while no-code builders simplify authoring but constrain agent interactions to author-defined workflows. We present ChatDev 2.0: DevAll (hereafter DevAll), a no-code platform for building, executing, and inspecting heterogeneous MAS that delivers both high expressiveness and ease of use. In terms of expressiveness, DevAll pairs a declarative executable graph abstraction with a cycle-aware execution engine, so that heterogeneous agents and dynamic and cyclic interactions can be represented and executed within a single framework. For ease of use, an integrated visual interface lets users author, run, monitor, and inspect MAS, including human-in-the-loop steps, entirely without writing code. Experiments demonstrate that DevAll reproduces state-of-the-art MAS across three representative tasks at competitive performance and without task-specific orchestration code, highlighting its effectiveness as a general-purpose platform for LLM-based MAS. DevAll is available at this https URL.
[MA-4] Control-Data Flow Separation: Stable Prompt Optimization in Multi-Agent LLM s
【速读】：该论文旨在解决多智能体大语言模型（multi-agent LLM）系统中提示词（prompt）优化时出现的协议破坏问题。其核心挑战在于，传统提示词往往同时承担生成任务相关文本内容与定义执行关键协议（如消息路由、输出格式化、终止信号等）的双重角色，二者相互纠缠，导致对提示词的优化可能无意中破坏执行协议，进而引发整个智能体流水线的失效。解决方案的关键在于提出“控制-数据流分离”（control-data flow separation）机制：将执行关键的控制逻辑以类型化、可验证的程序对象形式表示，而将任务相关的自然语言内容作为可优化的数据流进行处理。该设计使优化器能够专注于提升任务性能，同时避免因提示词漂移（prompt drift）影响消息路由或格式化等底层接口的稳定性。在合成推理、协作式内容生成及保险评级工作流等多个场景中，该框架实现了100%的最终协议有效性，并持续提升了任务表现。
链接: https://arxiv.org/abs/2609.00621
作者: Wentao Zhang,Syed Shariyar Murtaza,Junaid Ahmad Bhatti,Utkarsh Soni,Yifan Nie,Eugene Wen,Yuntian Deng
机构: University of Waterloo(滑铁卢大学); Manulife(宏利金融)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Multiagent Systems (cs.MA)
备注:
Abstract:Prompt optimization can improve multi-agent LLM systems, but the prompts being optimized often serve two entangled roles: generating task-relevant content and specifying execution-critical protocols, such as message routing, output formatting, and termination signals, on which the underlying code relies. As a result, a prompt edit intended to improve content generation can inadvertently corrupt the protocol and cause the entire agent pipeline to fail. Our key observation is that these two roles have different representations: execution protocols are typically structured, while task-relevant content is usually expressed in unstructured language. Based on this, we propose control-data flow separation, where execution-critical control is represented as typed, validated program objects, while task-relevant language remains the optimizable data flow for agent communication. This design allows optimizers to improve multi-agent behavior without exposing the routing or formatting interface to prompt drift. Across synthetic reasoning, collaborative review generation, and insurance rating workflows, our framework empirically achieves 100% eventual protocol validity while consistently improving task performance.
[MA-5] CUDA-Harness: Harnessing Agent ic CUDA Kernel Generation and Optimization from Natural Language
【速读】：该论文旨在解决从自然语言直接生成高性能CUDA内核（Text2CUDA）所面临的挑战，核心问题在于如何在缺乏领域专业知识的情况下，实现从高阶语义理解到低阶内核实现与验证的无缝衔接。现有基于大语言模型（LLM）的方法多集中于从PyTorch等高层框架向CUDA的转换（Torch2CUDA），而忽视了对自然语言输入的深层语义解析及底层内核正确性保障。此外，这些方法因依赖预设测试用例，易受奖励欺骗（reward hacking）影响。本文提出CUDA-Harness框架，其关键解决方案包括：引入中间结构化生成（Intermediate-Structured Generation），以桥接高层语义理解与底层内核生成；构建基于合成的验证机制（Synthesis-Based Verification），通过隔离测试数据和渐进式验证降低奖励欺骗风险；提出反馈自适应演化策略（Feedback-Adaptive Evolution），在优先保证正确性的前提下优化性能。实验表明，CUDA-Harness在多个维度上均展现出优越性能，并具备跨大模型、跨硬件平台以及支持C-to-CUDA转换的泛化能力。
链接: https://arxiv.org/abs/2609.00058
作者: Qi Fan,An Zou,Yehan Ma
机构: Shanghai Jiao Tong University(上海交通大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA); Programming Languages (cs.PL); Software Engineering (cs.SE)
备注:
Abstract:Developing high-performance CUDA kernels demands specialized knowledge in algorithm implementation, correctness validation, and hardware-aware parallel optimization, creating a substantial expertise barrier and making generating CUDA kernels directly from natural language (Text2CUDA) essential. Meanwhile, the general-purpose code generation capability of Large Language Models (LLMs) prompts a series of works exploring LLM-based CUDA kernel generation. They mainly focus on transpilation from high-level frameworks such as PyTorch to CUDA (Torch2CUDA) rather than Text2CUDA, where models must understand the high-level input semantics and handle low-level kernel implementation and validation. Additionally, these methods are vulnerable to reward hacking due to reliance on predefined test inputs. In this paper, we propose CUDA-Harness, a framework for harnessing agentic CUDA kernel generation and optimization from natural language. Specifically, we introduce Intermediate-Structured Generation to connect high-level semantic understanding with low-level kernel generation. To dilute reward hacking in Text2CUDA, we construct Synthesis-Based Verification to provide isolated test data and progressive validation. Furthermore, we propose Feedback-Adaptive Evolution, a kernel evolution strategy that prioritizes correctness while optimizing performance. Finally, through extensive experiments, we demonstrate the effectiveness of CUDA-Harness, with further evaluations illustrating generalization across LLMs, hardware platforms, and to C-to-CUDA transpilation.
[MA-6] RAPIDMap: Rapid Multi-Agent Pipeline for Interpretable Disaster Mapping from Satellite and Street-view Imagery
【速读】：该论文旨在解决灾后快速、可靠制图中面临的三大核心挑战：现有基于人工智能的方法依赖大量人工标注、缺乏跨灾害类型的泛化能力，以及仅依赖单一模态数据（如仅卫星影像）导致信息不全面。其解决方案的关键在于提出一种名为RAPIDMap的零样本可解释多智能体快速制图框架，通过集成四个协同智能体——灾害感知代理（Disaster Perception Agent, DPA）、图像修复代理（Image Restoration Agent, IRA）、损毁识别代理（Damage Recognition Agent, DRA）和灾害制图代理（Disaster Mapping Agent, DMA），实现对卫星与街景影像的多源异构数据融合。该框架无需人工微调，具备跨灾害类型（如地震、洪水、飓风等）的零样本泛化能力，并生成结构化、可直接用于应急响应的地图级灾情情报，同时提供恢复建议，显著提升灾后评估的效率与准确性。
链接: https://arxiv.org/abs/2609.00046
作者: Yifan Yang,Lei Zou
机构: 未知
类目: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI); Computers and Society (cs.CY)
备注: 4 pages, 7 figures, CaGIS Conference 2026
Abstract:Rapid and reliable disaster mapping of impacted areas, damaged infrastructure, and affected populations is essential for emergency response and recovery. However, existing AI-based approaches often require extensive manual annotation, lack cross-hazard generalization, and rely on single-modal observations. To address these challenges, this paper proposes RAPIDMap, a rapid multi-agent pipeline for zero-shot interpretable disaster mapping from satellite and street-view imagery. The framework integrates four intelligent agents: Disaster Perception Agent (DPA), Image Restoration Agent (IRA), Damage Recognition Agent (DRA), and Disaster Mapping Agent (DMA). By combining remote sensing and street-view data, RAPIDMap eliminates the need for manual fine-tuning, generalizes across multiple disaster categories, and generates structured, map-ready disaster intelligence with recovery recommendations.
[MA-7] EULER: Exploring Underused Links with Evidence-Checked Return for Multi-Agent Mathematical Discovery DATE
【速读】：该论文旨在解决数学领域中跨不同数学结构、不变量与工具之间进行问题迁移的高成本与低可行性问题，即在不同数学子领域间构建有效“桥梁”以实现定理证明或反例构造的自动化。其核心挑战在于如何高效识别并验证能够连接源域与目标域的可执行操作路径，同时确保推理链的逻辑正确性与结论回溯性。解决方案的关键在于提出EULER——一个基于多智能体系统的框架，将“桥接”（bridge）作为搜索的基本单元，通过竞争性地探索直接路径、邻近域路径与远距离域路径，动态评估每条桥接的有效性：仅当桥接引入了源表示无法执行的操作，且目标侧证据经由可验证的蕴含关系成功返回至原始猜想时，该桥接才被保留并积累预算。为提升效率与可靠性，EULER引入六组有序压力测试，在正式搜索前剔除无效桥接；实验表明，桥接特异性压力测试使错误结论从9个降至3个，而桥接材料与目标本域操作的协同作用带来+4.2个额外可解决任务，显著优于单一机制。研究还发现，领域间距离并非成功预测因子，而“可执行操作增益”与“有效返回能力”才是关键判据。
链接: https://arxiv.org/abs/2609.00032
作者: Ren Zhenzhuo
机构: EULER Team(欧拉团队)
类目: Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注: 55 pages, 10 figures, 29 tables; includes a 13-page companion candidate-proof manuscript as Appendix P
Abstract:Mathematical communities work with different objects, invariants, and tools, so transferring a problem across them is expensive and often skipped. We present EULER, a multi-agent system that takes such a transfer–a bridge–as its unit of search. Around a fixed conjecture, EULER runs direct, adjacent-domain, and distant-domain routes in competition; a bridge keeps its budget only if it supplies an operation the source representation cannot execute and its target-side evidence returns to the original statement along a checked implication. Six ordered stress tests reject invalid bridges before expensive search begins. We evaluate EULER on 120 recent conjectures. The conjectures were frozen before search and screened for contamination, and are drawn from public papers by authors who had recently published in the Journal of Combinatorial Theory, Series A, a leading journal in combinatorics. EULER produced 10 proofs and 3 refutations, plus 45 scoped partial results. Two mechanisms held up under ablation: bridge-specific stress tests cut incorrect conclusions from 9 to 3, and bridge material combined with a target-native operation yielded a positive interaction of +4.2 resolved tasks that neither factor produced alone. Domain distance did not reliably predict success; executable operation gain and valid return did. Comments: 55 pages, 10 figures, 29 tables; includes a 13-page companion candidate-proof manuscript as Appendix P Subjects: Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA) Cite as: arXiv:2609.00032 [cs.AI] (or arXiv:2609.00032v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2609.00032 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[MA-8] Long-Horizon State Tracking in LLM s: Executing MD5 through a Deep Sequence of Dependent Tool Calls
【速读】：该论文旨在解决大语言模型（LLM）在长时序任务中因误差累积导致的端到端失败问题，尤其关注模型在多步依赖任务中精确维护中间状态的能力。现有代理评估基准常因混淆状态追踪与指令理解、缺乏对照组及存在幻觉式最终答案等捷径，无法准确揭示长序列失败的根本原因。为此，作者设计了一个严格控制的实验：要求模型逐步计算MD5哈希值，共执行196次依赖性工具调用，跨越64轮，需在每一步中精确传递四个32位状态变量（a,b,c,d）。由于算法从零实现（RFC 1321），每一步均可与真实轨迹对齐并逐比特校验结果，从而确保任何失败均由状态传递错误引起。实验结果显示，gpt-oss-120b（每标记仅激活约5.5B参数）在零温度、固定提示下成功完成多数运行并输出正确哈希。最严苛设置下，所有基础工具均由另一LLM替代，形成“驱动-工作”双模型架构，进一步排除外部精确算术参照的影响。关键成功因素为两点：一是在每轮中保留模型自身推理过程于上下文；二是通过启用思维能力的工作模型进行投票以纠正模运算错误。通过故障溯源分析，可明确区分状态保持、算术计算与服务调用三类失败来源。
链接: https://arxiv.org/abs/2609.00012
作者: Dheeraj Mohandas Pai,Lu Xian
机构: 未知
类目: Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR); Multiagent Systems (cs.MA)
备注: 6 Pages, 2 figures
Abstract:Long-horizon tasks remain uncommon in large language model (LLM) evaluation, and for a reason: when each step depends on the last, per-step accuracy that looks excellent in isolation decays catastrophically, as errors cascade and the end-to-end failure probability grows sharply with length. Existing agentic benchmarks report end-to-end success but confound this state-tracking difficulty with instruction interpretation, give no control group that isolates it, and are vulnerable to shortcuts such as a hallucinated final answer, so they cannot say why a long run fails. Whether an LLM can carry exact intermediate state across many tool calls at all is itself not well established. We test this cleanly by having the model compute a cryptographic hash, MD5, step by step: a sequence of 196 dependent tool calls over 64 rounds while it carries four 32 -bit words (a,b,c,d) in its own context from one call to the next. Interpretation is trivial and, because we implement MD5 from scratch (RFC~1321), we align every call to the ground-truth trace and check the digest to the bit, so any failure is pure bookkeeping. gpt-oss-120b, a mixture-of-experts model with only \sim 5.5B active parameters per token, at temperature 0 with a short fixed prompt, carries the full state across all 196 calls and returns the correct digest on a majority of completed runs. In the strongest setting we replace every primitive tool with a second LLM, so a driver and a worker compute the whole hash from scratch with no exact-arithmetic oracle in the loop. Two ingredients decide success and neither changes the weights: keeping the model’s own reasoning in its context each turn, and voting over a thinking-enabled worker to remove its modular-arithmetic slips. We localize the residual failures by origin, separating state-carrying from arithmetic and from serving.
[MA-9] Harness Engineering: Anatomy Architecture and Evolution of Coding Agents – A Source-Code Study of Eleven Systems
【速读】：该论文旨在解决生成式AI（Generative AI）时代中智能体（Agent）运行时环境——即“Harness”——缺乏系统性设计规范与实证基础的问题。当前，尽管大量生产级编码智能体已部署，但其底层运行时架构仍处于高度定制化、非标准化的阶段，导致可复用性差、安全风险高且难以规模化演进。本文通过分析11个生产级编码智能体运行时（包括Claude Code、Codex CLI、Gemini CLI等）及首个元运行时（Omnigent），首次为“运行时工程”（Harness Engineering）这一新兴学科提供了最全面的实证基础。其核心解决方案在于：明确界定“Harness”的定义，将其分解为七个标准子系统，并在此框架下对所有系统进行结构化剖析，揭示出29种反复出现的设计模式与13项跨系统观察。研究发现，在约四百万行Python、TypeScript和Rust代码中，无一运行时引入通用型智能体框架，亦无采用向量嵌入检索代码，表明行业普遍依赖手写异步循环与确定性检索机制。此外，通过对比同一套系统在一季度内的源码演化，揭示了从工具到平台的范式转变——行为策略由提示词描述转向配置驱动，而“运行时托管”（Harness Hosting）作为第三角色被引入。最终，论文提出18项设计建议与一个90行最小可行运行时模板，推动智能体运行时从临时拼凑走向可复用、可扩展的平台化基础设施。
链接: https://arxiv.org/abs/2609.00006
作者: Paul Barbaste,Tristan Darrigol,Germain Vu,Tom Wiltberger
机构: Wavestone AI Lab(瓦斯特恩人工智能实验室)
类目: oftware Engineering (cs.SE); Multiagent Systems (cs.MA)
备注: 83 pages, 7 figures, 18 tables. Second, substantially expanded edition of an April 2026 study: corpus grown from eight to eleven systems plus a meta-harness contrast point; the retained April snapshots provide a controlled longitudinal comparison
Abstract:An agent is a model plus a harness – the runtime that couples an LLM to the world through a loop, tools, context management, safety controls, orchestration, and extension surfaces. Harness engineering, named as a discipline in early 2026, is the design and evolution of that runtime. This paper gives the young discipline its most comprehensive empirical foundation to date: a source-code anatomy of eleven production coding harnesses (Claude Code, Codex CLI, Gemini CLI, Mistral Vibe, OpenHands, Aider, Mini-SWE-Agent, Hermes, Pi, OpenCode, OpenClaw), plus Omnigent, the first meta-harness, analyzed as a contrast point. We define what a harness is, map its seven canonical subsystems with the minimal and maximal implementation of each, and dissect all eleven systems along those subsystems. The audit yields 13 cross-cutting observations and a catalog of 29 recurring design patterns. Two absences survive a threefold corpus expansion: across roughly four million lines of Python, TypeScript, and Rust, no agent runtime imports a general-purpose agentic framework, and none retrieves code with vector embeddings; the field runs on hand-rolled async loops and deterministic retrieval. this http URL skills lead MCP in adoption (9/11 vs. 8/11), and ACP ships in six systems with a new third role: harness hosting. Because the original eight systems were re-pinned rather than replaced, the study also contains a controlled longitudinal sample – the same harnesses source-diffed across one quarter – showing convergence becoming imitation and behavioral policy migrating from prompt prose to configuration. These threads converge on the paper’s thesis: in the first half of 2026 the coding harness completed a turn from tool to platform. The paper closes with 18 design recommendations and a 90-line minimum-viable-harness scaffold.
自然语言处理
[NLP-0] Beyond Scores: Understanding LLM -as-a-Judge Mechanisms in Summarization Evaluation EMNLP2026
【速读】： 该论文旨在解决大语言模型（LLM）作为自然语言生成（NLG）质量评估工具时，其内部评分机制不透明的问题。尽管生成式AI（Generative AI）驱动的评估器被广泛用于自动评分与训练信号生成，但其决策过程仍缺乏可解释性。为此，研究提出一种基于八类扰动攻击的分析框架，覆盖可读性（Readability）与充分性（Adequacy）两大维度，构建了包含成对干净与污染摘要、可控错误强度及显式词元级修改映射的生成流水线，并结合因果追踪、对数透镜词汇投影与注意力头剔除等四组实验，对Themis（Llama-3-8B）和Prometheus（Mistral-7B）两个评估模型进行系统剖析。关键发现表明，两类评估器均采用分阶段结构化评估流程：在第15层以下，注意力机制执行局部错误对比并将其路由至最终输入位置；在第15层以上，多层感知机（MLP）级联整合信号，并在残差流中于晚期层（Themis为第26层，Prometheus为第25层）实现决策“结晶”。通过同规模基础模型（Llama-3-8B）对照实验进一步揭示，微调过程并未从零构建评估管道，而是对预存架构进行精炼——具体表现为抑制第15层以下MLP在末位置的贡献，并将决策结晶深度提前两层，说明微调本质上是对已有底层结构的优化而非重建。该研究为理解生成式评估器的内部运作提供了可解释性依据，并公开了全部代码与数据以支持复现。
链接: https://arxiv.org/abs/2609.01604
作者: Himil Vasava,Ming Jiang
机构: University of Wisconsin-Madison(威斯康星大学麦迪逊分校)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: Accepted at EMNLP 2026 Main Conference
Abstract:LLM-based evaluators of natural language generation (NLG) quality are widely deployed as scoring tools and as automated training signals, yet the internal procedure by which they assign a rating remains poorly understood. We investigate this procedure mechanistically through an eight-attack perturbation taxonomy across the Readability and Adequacy dimensions of NLG quality, a generation pipeline that produces paired clean and corrupt summaries with controlled error intensity and explicit token-level modification maps, and a four-experiment battery of causal tracing, logit-lens vocabulary projection, and attention-head knockout applied to Themis (Llama-3-8B) and Prometheus (Mistral-7B). Both evaluators implement a structured, coherent evaluation pipeline operating in two stages: below layer 15, attention performs local error comparison and routes the result to the final input position; above it, the MLP cascade integrates the signal and writes the rating, with the decision crystallizing in the residual stream at a sharp late layer (L = 26 on Themis, L = 25 on Prometheus). Furthermore, a base-model control at the same scale (Llama-3-8B) reproduces the routing architecture and crystallization but not the stage separation, isolating the two mechanisms that fine-tuning specifically installs, suppression of below-L15 MLP contribution at the last position and a two-layer advance of the crystallization depth, indicating that fine-tuning sculpts an existing substrate rather than building the pipeline from scratch. We release the source code and data at this https URL
[NLP-1] Efficient SWE Agent Benchmarking via Trajectory-Aware Evaluation
【速读】： 该论文旨在解决软件工程智能体（Software Engineering Agents）在真实基准测试中评估成本高昂的问题，其核心挑战在于：每个任务通常需要多步的代码探索、修改与测试执行，导致全量评估耗时且资源消耗大。现有高效评估方法虽通过选取代表性子集来估算完整基准的表现，但其主要依赖结果层面的信息（如通过/失败记录或静态任务语义），忽视了智能体求解过程中的动态行为信息。为克服这一局限，本文提出一种融合过程与结果信号的新型框架——特权轨迹感知项目反应理论（PTA-IRT）。该方法的关键创新在于引入历史执行轨迹作为“特权信息”（Privileged Information），包括智能体探索的上下文、尝试的代码修改及求解路径等过程级证据，用于指导子集选择与能力估计。相比传统仅基于结果的项目反应理论（IRT）基线，PTA-IRT在低校准预算条件下，在四个SWE基准上均显著提升了评分预测精度与排名恢复能力，展现出更强的评估效率与准确性。
链接: https://arxiv.org/abs/2609.01603
作者: Kefeng Duan,Dewu Zheng,Yanlin Wang,Xiwen Wang,Ensheng Shi,Xilin Liu,Yuchi Ma,Jiachi Chen,Mingwei Liu,Zibin Zheng
机构: 未知
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: Under review
Abstract:Evaluating software engineering agents on realistic benchmarks is costly, since each task may require multi-step code exploration, modification, and test execution. Existing efficient evaluation methods select representative subsets to estimate full-benchmark performance, but are largely result-only: they fit historical pass/fail response matrices or static task semantics, discarding how agents solve problems. We propose PTA-IRT, a Privileged Trajectory-Aware Item Response Theory framework that fuses process and outcome signals. Historical execution trajectories supply process-level evidence beyond pass/fail, such as explored context, attempted edits, and solving paths, which PTA-IRT uses as privileged information for calibration subset selection and ability estimation. Under low calibration budgets, PTA-IRT consistently outperforms prior IRT baselines on score and ranking recovery across four SWE benchmarks. Code and data are publicly available at this https URL.
[NLP-2] Adaptive Critical Token-Aware Retrieval for Repository-Level Code Generation
【速读】： 该论文旨在解决仓库级代码生成任务中因大模型（LLM）输入长度限制导致的上下文不足问题，尤其关注在生成过程中对关键代码位置（即“关键令牌”）缺乏细粒度仓库上下文支持的问题。现有方法虽采用检索增强生成（RAG）策略提升仓库上下文召回效果，但通常仅提供任务层面的通用上下文，未能精准识别并为生成过程中的决定性位置提供针对性的上下文支持。由于自回归生成过程中错误常集中于少数关键令牌位置，一旦生成错误，将导致后续代码语义偏离并引发功能失败。为此，本文提出ACToR——一种自适应的关键令牌感知检索框架，其核心在于在生成过程中动态识别关键令牌，并按需触发目标检索以在这些决定性位置注入精确的仓库上下文；同时设计了一种位置感知加权机制，优化稠密检索器对生成更具信息量的上下文的优先排序。实验在RepoExec和CoderEval两个代表性基准上验证了ACToR的有效性，分别实现了8.4%和15.4%的相对性能提升。此外，研究系统量化了关键令牌的影响，揭示其在生成失败中的核心作用，进一步证明了靶向检索策略的必要性。
链接: https://arxiv.org/abs/2609.01601
作者: Kefeng Duan,Dewu Zheng,Yanlin Wang,Terry Yue Zhuo,Mingwei Liu,Jianxing Yu,Jiachi Chen,Ensheng Shi,Xilin Liu,Yuchi Ma,Zibin Zheng
机构: Sun Yat-sen University (中山大学); School of Artificial Intelligence, Sun Yat-sen University (中山大学人工智能学院); Monash University (莫纳什大学); CSIRO’s Data61 (澳大利亚联邦科学与工业研究组织数据61实验室); Zhejiang University (浙江大学); Huawei Cloud Computing Technologies Co., Ltd. (华为云计算技术有限公司)
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: Under review
Abstract:The repository-level code generation task requires synthesizing code that satisfies task requirements while remaining consistent with the target repository context. Since real-world repositories often exceed the input length limits of LLMs, existing approaches commonly adopt retrieval-augmented generation (RAG) to provide repository-specific context. Despite improving repository-context retrieval, existing methods typically provide context as task-level support, without explicitly identifying the critical tokens that require fine-grained repository context during generation. During the autoregressive generation process of LLMs, errors often concentrate at a small number of decisive positions: once such tokens are generated incorrectly, subsequent code may follow an incorrect semantic path and eventually lead to functional failure. We refer to these positions as “critical tokens”. In this paper, we propose ACToR, an adaptive critical token-aware retrieval framework for repository-level code generation. ACToR identifies critical tokens during generation and triggers targeted retrieval on demand to provide repository context at these decisive positions. In addition, we design a position-aware weighting method for dense retrievers to prioritize context that is more informative for generation. We evaluate ACToR on two representative repository-level benchmarks, RepoExec and CoderEval. Experimental results show that ACToR consistently outperforms state-of-the-art methods, achieving relative improvements of 8.4% on RepoExec and 15.4% on CoderEval. Beyond performance gains, we systematically quantify the impact of critical tokens, revealing their central role in major generation failures and highlighting the necessity of targeted retrieval strategies. We provide the code and data at this https URL.
[NLP-3] CordisBench: Can Language Models Reason About Component Lifecycles in Dynamic Agent Harnesses?
【速读】： 该论文旨在解决生成式智能体（Generative AI）在动态组件系统中因插件变更引发的依赖传播与清理（cleanup）问题，核心挑战在于模型需具备对系统生命周期进行推理的能力，包括识别受影响组件、预测特定卸载顺序下的系统状态、判断条件在所有或部分顺序下是否成立，以及选择可成功执行的重配置方案。其解决方案的关键在于提出CordisBench——一个包含1,200个问题的基准测试集，结合受控形式化环境与Cordis运行时系统（负责管理组件依赖与清理），通过确定性任务特定评分机制评估模型在低推理成本（2至32次相关交互）下的表现。实验表明，尽管当前效率导向模型在小规模系统中表现良好，但随着交互数量增加，尤其在预测最终状态及跨卸载顺序推理时可靠性显著下降；额外推理资源虽能提升部分模型性能，但代价高昂（如GPT-5.6 Luna在中等推理强度下每题需近3,000个推理标记）。研究进一步发现，在这些受控实例中，该成本可被规避：一个独立的有限参考语义（finite reference semantics）在所有528个可执行问题上均与Cordis执行结果一致，为高效且准确的推理提供了理论基础。
链接: https://arxiv.org/abs/2609.01600
作者: Damien Sileo,Dimitri Kachler
机构: Univ. Lille (里尔大学); Inria (法国国家信息与自动化研究所); CNRS (法国国家科学研究中心); Centrale Lille (里尔中央理工学院); UMR 9189 - CRIStAL (CRIStAL 9189联合研究实验室)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 13 pages, 6 figures, 5 tables. Code: this https URL ; Data: this https URL
Abstract:Dynamic agent harnesses let language models change the software that shapes their own execution. This flexibility brings a new reasoning burden: a local plugin change can propagate through dependencies and cleanup. We introduce CordisBench, a 1,200-question benchmark of this lifecycle reasoning. It combines a controlled formal setting with programs executed against Cordis, a runtime that manages component dependencies and cleanup, and asks models to identify affected components, predict state after a specified teardown order, determine which conditions hold under all or some orders, and choose reconfigurations that succeed when executed. Across these tasks, we evaluate three efficiency-oriented models at low reasoning effort with 2, 4, 8, 16, 24, or 32 relevant interactions, using deterministic task-specific scoring. Models usually handle small systems well but grow less reliable as more interactions become relevant, especially when predicting final state and when reasoning across teardown orders. Additional inference effort recovers marked gains for some models. The cost is nontrivial: on our 16-interaction subset, GPT-5.6 Luna uses nearly 3,000 reasoning tokens per question at medium effort. For these controlled instances, that cost is avoidable: an independent finite reference semantics agrees with Cordis execution on every observation and action outcome used for scoring across all 528 executable questions.
[NLP-4] he Rise of Verbal Reinforcement Learning
【速读】： 该论文旨在解决语言智能体（language agents）在发展过程中如何有效利用自然语言反馈以提升性能与对齐性的问题。当前，自然语言正逐渐成为优化语言智能体的核心反馈渠道，能够以人类可理解且模型可解析的形式传递意图、偏好及因果结构。为此，论文提出了“言语强化学习”（Verbal Reinforcement Learning, VRL）这一统一范式，并围绕两个核心维度——言语反馈在智能体生命周期中的作用时机及其所影响的机制——构建了系统性分类框架。其解决方案的关键在于提出三大支柱：（1）语言作为基础信号（Language as Grounding Signal），即通过语言定义任务目标、状态空间和奖励结构，实现任务的语义化建模；（2）语言作为反思性反馈（Language as Deliberative Feedback），即在推理阶段利用自然语言指导智能体的决策过程，无需更新模型参数；（3）语言作为学习信号（Language as Learning Signal），即通过语言形式的反馈直接调整模型参数，实现训练过程中的持续优化。该分类体系不仅厘清了语言在不同阶段对智能体行为塑造的独特作用，也为构建更强大、更对齐的语言智能体指明了关键挑战与未来方向。
链接: https://arxiv.org/abs/2609.01597
作者: Kshitij Tayal,Arun Sharma,Genta Indra Winata,Anirban Das,Sambit Sahu
机构: Capital One(资本一号); University of Minnesota (明尼苏达大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Natural language is emerging as a primary feedback channel for improving language agents, capable of conveying intent, preferences, and causal structure in forms interpretable by both humans and modern language models. We call this paradigm Verbal Reinforcement Learning (VRL) and offer the first unified account of it. We organize the field around a single axis, \textitwhen verbal feedback takes effect in an agent’s lifecycle and \textitwhat it modifies, yielding three pillars: (1) \textbfLanguage as Grounding Signal, where language defines the task itself by specifying goals, states, and reward structures; (2) \textbfLanguage as Deliberative Feedback, where natural language guides reasoning at test time without the need to update model parameters; (3) \textbfLanguage as Learning Signal, where language-based feedback shapes model parameters through training. Within each pillar, we synthesize representative work, distinguish key subcategories of approaches, and outline the distinct role language plays in shaping agent behavior. Together, this taxonomy shows how verbal reinforcement is reshaping agent development, while also defining the challenges and opportunities for building more capable and aligned agents.
[NLP-5] StudentSim: Training LLM -based Student Simulators
【速读】： 该论文旨在解决生成式 AI 教学助手（AI tutor）在个性化辅导中面临的根本性挑战：缺乏针对不同学生个体特征（如优势、薄弱点及偏好指导方式）的有效反馈信号，而真实学习者数据的收集成本高且效率低。现有学生模拟方法存在明显局限——状态追踪模型难以处理解释与纠正等复杂交互，而基于大语言模型（LLM）的角色扮演虽能流畅响应指导，却无法可靠匹配被模仿学生的实际认知水平。为此，论文提出 StudentSim 框架，其核心创新在于采用“联合训练+个体化微调”的两阶段策略，将稀疏的单个学生数据通过群体训练提炼通用表征，并在此基础上实现对每位学生的专属建模。该框架生成的学生模拟器不仅能准确复现学生的真实行为（行为保真度，F），还能在导师引导下动态调整自身反应（指导响应性，R）。为系统评估，作者构建了 StudentSimEval 标准化评测协议，覆盖国际象棋、第二语言英语写作和数学三个领域，使用去标识化的公开学习者数据集进行统一测试。实验表明，StudentSim 在所有任务中均显著优于 GPT-5.4 和 Maia2 模型，在国际象棋任务中分别达到 F=0.51、R=0.91，远超 GPT-5.4 的 0.23 和 0.72。进一步验证显示，以 StudentSim 作为奖励模型用于教学策略强化学习，可生成更精准、更具针对性且更受专家认可的个性化辅导系统，证明了其在构建高效自适应教学系统中的关键价值。
链接: https://arxiv.org/abs/2609.01591
作者: Ke Yang,Chenglong Wang,Michel Galley,Chandan Singh,Jeevana Priya Inala,ChengXiang Zhai,Jianfeng Gao
机构: Microsoft Research; University of Illinois Urbana-Champaign(伊利诺伊大学厄本那-香槟分校)
类目: Computation and Language (cs.CL)
备注:
Abstract:AI tutors are most useful when they adapt to each student’s strengths, weaknesses, and preferred guidance, but evidence about which guidance works for which student is sparse, slow, and costly to collect from real learners. Student simulators can provide this signal as a proxy, yet existing approaches are limited: state-tracking models fit student behavior but struggle to process explanations or corrections, while LLM role-play follows guidance fluently but does not reliably match the competence of the student being imitated. We present StudentSim, a training framework that turns sparse per-student data into individualized simulators through pooled training followed by per-student specialization. The resulting simulators both mirror a student’s own responses and update them under tutor guidance. We also introduce StudentSimEval, a standardized protocol covering 60 students across chess, second-language English writing, and mathematics, using public learner datasets with de-identified records shared for research. StudentSimEval measures behavioral fidelity (F), or how well a simulator matches a student’s responses, and guidance responsiveness ®, or how readily it updates under tutor guidance, with all methods fit and evaluated on the same records. Across all three domains, StudentSim outperforms GPT-5.4 on both metrics. In chess, StudentSim reaches F=0.51 and R=0.91, compared with 0.23 and 0.72 for GPT-5.4 and 0.45 and 0.27 for Maia2. As a proof of concept, using StudentSim as a reward model for tutor reinforcement learning produces a chess tutor that expert humans rate as more accurate, better-guided, and more personalized than a no-RL baseline and a tutor trained against a GPT-5.4 simulator reward. Code is available at this https URL.
[NLP-6] he Structure of Quantization Damage in LLM s: Why the Next Bit Should Be Spent Globally NEURIPS2026
【速读】： 该论文旨在解决大语言模型（LLM）在后训练量化（Post-training Quantization, PTQ）过程中精度损失分布不均且难以通用优化的问题，尤其关注如何高效分配有限的额外精度预算以最小化性能下降。其核心挑战在于：尽管量化可显著降低推理成本，但不同模型层对量化敏感度差异大，而现有方法缺乏普适性指导来确定哪些层最需恢复精度。论文的关键解决方案是通过因果混合精度干预（causal mixed-precision intervention）作为基准——即逐层将各层提升至8比特并测量其对准确率的恢复效果，在9个开源模型、4种架构族中系统评估了量化损伤的位置。研究发现，量化损伤并非集中于特定任务回路（task circuits）、计算位置或权重统计特征，而是呈现扩散性分布；除Qwen3-8B外，其余8个模型中约半数层即可恢复75%的精度损失。更重要的是，在相同精度预算下，全局性地采用更细粒度的量化策略（如组块128兼容的粒度）优于局部修复最具恢复潜力的单一层，性能提升达21–52点，包括对损伤高度集中的Qwen3-8B也表现更优。此外，研究还揭示残差连接部分在8比特下接近无损（近似无损），且峰值恢复位置与同一架构家族内的结构特性相关，但跨家族不具可比性。因此，论文得出结论：依赖廉价信号预测量化损伤位置并不可靠，必须通过因果干预验证；在当前预算设定下，全局细化量化粒度应作为默认策略，而非选择性保护关键层。
链接: https://arxiv.org/abs/2609.01587
作者: Jundong Hu,Shekar Ramachandran
机构: PayPal AI(贝宝人工智能)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: Preprint. Under review at a NeurIPS 2026 workshop. 11 pages, 4 figures, 8 tables
Abstract:Post-training quantization (PTQ) is widely used to reduce the cost of serving large language models (LLMs), but its accuracy cost is uneven and is often tuned per model. We study where quantization damage occurs and how to allocate a small additional precision budget. Using causal mixed-precision intervention as ground truth (raise each layer to 8-bit in turn and measure the accuracy it recovers) across 9 open-weight models in 4 architecture families, we test 3 intuitive hypotheses: that quantization damage lives in task circuits, where the model computes, or in weight statistics. None of them predicts which layers benefit from restored precision. Recovery is instead diffuse: for 8 of 9 models, recovering 75% of the gap takes roughly half the layers; the lone exception, Qwen3-8B, is sharply concentrated. At a matched precision budget, spending it globally on finer quantization granularity beats locally repairing the most recoverable layers for all 8 group-128-compatible models (all but OpenLLaMA, whose width rules out group-128), by 21-52 points, including the concentrated Qwen3-8B. We report 2 secondary findings: the residual is budget-limited (8-bit is near-lossless in our evaluation across RTN, GPTQ, and AWQ), and the location of peak recovery correlates with architecture within a family, though not across families. Within this budget setting, global granularity is a better default than selectively protecting critical layers. More broadly, cheap signals that correlate with quantization damage do not necessarily identify where restoring precision improves accuracy; this must be tested with causal intervention.
[NLP-7] Closing Cost-Quality Gap in Document VLMs: Difficulty-Aware Data Curation and Quality-Adjusted Deployment Economics
【速读】： 该论文旨在解决在受监管行业中从数亿份文档中高效提取结构化字段的高成本问题。现有方案如定制化光学字符识别（OCR）流水线覆盖范围有限，隐私法规禁止使用外部模型，而满足质量标准的开源视觉语言模型（VLM）部署成本高于人工标注。其解决方案的关键在于构建一个基于混合专家（Mixture-of-Experts, MoE）架构的VLM（总参数量350亿，活跃参数30亿），并在内部生产数据与由难度感知管道筛选的开放域文档（兼顾版式多样性、事实可提取性及跨模型一致性）上进行微调。该模型仅需单张H100 GPU即可部署，并通过提示工程支持异构工作流，性能超越所有可部署的非推理类基线模型，且规模大一个数量级。基于生产遥测数据校准的质控与修正成本的质量调整成本分析表明，该模型相较人工基准降低预期成本超80%，相比最优开源模型亦降低超过50%，而更大规模基线模型仍不具备经济可行性。
链接: https://arxiv.org/abs/2609.01575
作者: Maksim Evdokimov,Matvey Ivanov,Dmitrii Tsiupin,Olga Tsymboi,Anatolii Potapov,Aleksandr Ivanov
机构: T-Tech
类目: Computation and Language (cs.CL)
备注:
Abstract:Extracting structured fields from hundreds of millions of documents annually remains costly in regulated industries: bespoke OCR cascades cover only a fraction of workflows, privacy rules preclude external models, and existing open-source VLMs that clear quality thresholds cost more to serve than human annotation. We present a deployed document-understanding system built on a Mixture-of-Experts VLM (35B total, 3B active), fine-tuned on in-house production data mixed with open-domain documents curated by a Difficulty-Aware pipeline for layout diversity, fact-extractability, and cross-model consistency. Fitting on a single H100 and serving heterogeneous workflows via prompting, the model leads all deployable (non-reasoning) baselines up to an order of magnitude larger. A quality-adjusted cost analysis, with confirmation and correction costs calibrated from production telemetry, shows it reduces expected costs by over 80% against the human baseline and by more than 50% against the best competing open-source model, while larger baselines remain economically unviable.
[NLP-8] Scaling Near-Optimal SFT-RL Annotation Budget Allocation from Small to Large LLM s EMNLP2026
【速读】： 该论文旨在解决大语言模型（LLM）后训练过程中，如何在监督微调（SFT）与强化学习（RL）之间合理分配固定标注预算这一开放性问题。现有研究仅描述了粗略趋势（如在低数据场景下SFT占主导），缺乏系统性的资源配置框架，且未考察最优比例在不同模型规模间的可迁移性。本文提出以“近优区域”（near-optimal region）为核心概念，即在性能接近峰值的一定容忍度内（如2%-10%）的所有SFT-RL资源配置组合，而非寻找单一最优比例。实证结果表明，该近优区域在小容忍度下仍具有较宽范围，且随模型规模增大而进一步扩展，并能可靠地从小型代理模型迁移至大型目标模型。这一发现揭示了一种实用策略：仅通过小型代理模型的实验即可识别出可迁移的近优区域，从而避免在大规模模型上进行耗时耗力的穷举式搜索。该结论在多种任务、模型族以及基于偏好（preference-based off-policy）和奖励监督（reward-supervision on-policy）的不同强化学习方法中均保持一致。此外，研究还分析了SFT与RL数据标注成本不对称性对近优区域位置的影响，进一步增强了方案的实际指导意义。
链接: https://arxiv.org/abs/2609.01573
作者: Jingtan Wang,Arun Verma,Xiaoqiang Lin,Zhengyuan Liu,Nancy F. Chen,Daniela Rus,Bryan Kian Hsiang Low
机构: National University of Singapore(新加坡国立大学); Agency for Science, Technology, and Research (A*STAR)(新加坡科技研究局); Singapore-MIT Alliance for Research and Technology Centre(新加坡-麻省理工学院科研中心); CSAIL, Massachusetts Institute of Technology(麻省理工学院计算机科学与人工智能实验室)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Accepted at EMNLP 2026
Abstract:How to divide a fixed annotation budget between supervised fine-tuning (SFT) and reinforcement learning (RL) during LLM post-training remains an open problem. Existing work characterizes only broad trends (e.g., SFT dominates in low-data regimes), lacks a principled allocation framework, and does not examine whether the optimal ratio transfers across model sizes. We frame this problem in terms of near-optimality: rather than seeking a single optimal SFT-RL ratio, we characterize the near-optimal region, the set of allocations within a specified tolerance of peak performance. Empirically, this region is wide even for small tolerances (2-10%), widens with model scale, and transfers reliably from small proxy models to large target models. This yields a practical strategy: small proxy-model experiments suffice to identify a transferable near-optimal region, eliminating the need for exhaustive large-scale search. Our results hold consistently across tasks, model families, and both preference-based off-policy and reward-supervision on-policy RL methods. We further analyze how the asymmetry in annotation costs between SFT and RL data shifts the near-optimal region.
[NLP-9] From Production Traffic to Post-Training: Building a Self-Hosted LLM That Covers the Corporate Request Mix
【速读】： 该论文旨在解决企业在数据驻留（data-residency）约束下自托管大语言模型（LLM）所面临的资源碎片化问题：随着新模型不断迭代上线而旧模型难以及时退役，导致服务集群规模持续扩张，进而过度消耗有限的GPU资源。其核心解决方案是通过系统性识别并弥合生产环境中三大关键质量维度——指令遵循、函数调用与内部任务分布——的性能差距，将超过200个内部应用的流量统一收敛至单一模型。为避免多目标联合优化带来的跨领域奖励干扰，研究提出分轴训练独立的广义近端策略优化（GRPO）专家，并采用两阶段球面线性插值（SLERP）进行融合。每个专家专门针对特定失效模式（如语义坍塌、过度调用、冗余生成等）进行优化，实现领域专用修复。实验表明，在非推理模式下，该方案在自研基准测试中以约1/7的参数量超越更大规模基线模型，在指令遵循（69.6→65.8）、函数调用（0.79→0.77）和内部任务分布（0.85→0.83）等指标上表现更优，同时提升通用对话性能；该模型已承载平台50%的流量（每月1.16亿请求），且服务成本显著降低。
链接: https://arxiv.org/abs/2609.01572
作者: Olga Tsymboi,Dmitrii Stoianov,Ramil Latypov,Danil Taranets,Daniil Dryabin,Mikhail Gashkov,Viktor Zelenkovskiy,Aleksandr Fida,Gleb Alektorov,Nikita Gulyakov,Arthur Babkin,Aleksandr Medvedev,Pavel Gein,Anatolii Potapov
机构: 未知
类目: Computation and Language (cs.CL)
备注:
Abstract:Data-residency constraints force enterprises to self-host LLMs, but continuous adoption of newer models without decommissioning their predecessors expands the serving fleet, fragmenting a finite GPU pool. We consolidate traffic from over 200 internal applications onto a single model by closing quality gaps identified through production error analysis along three axes: instruction following, function-calling, and internal task distribution. Quality is tracked by offline benchmarks stratified to production traffic and scored by deterministic verifiers or calibrated LLM judges. Rather than optimising all objectives jointly, which introduces cross-domain reward interference, we train a separate GRPO expert per axis and merge them via two-stage SLERP. Each expert’s reward exposes a distinct failure mode, namely semantic collapse, over-calling, and verbosity hacking, each requiring a domain-specific fix. In non-reasoning mode the recipe surpasses a \sim7\times larger by total parameters baseline on the in-house Arena with 69.6 to 65.8, instruction following with 0.85 to 0.83, and function-calling with 0.79 to 0.77, while lifting general dialogue benchmarks. The model absorbs 50% of platform traffic, 116M requests per month, at a fraction of the serving cost.
[NLP-10] Selective Agent Guidance via Entropy: Learning Autonomous Policies from Imperfect VLM Teachers
【速读】： 该论文旨在解决如何高效利用昂贵且不完美的视觉-语言模型（Vision-Language Model, VLM）作为教师指导来训练一个轻量级、自主的强化学习（Reinforcement Learning, RL）策略，以克服直接将VLM用作策略时存在的高计算开销、缺乏环境交互优化能力以及重复系统性错误等问题。其解决方案的关键在于提出SAGE（Selective Agent Guidance via Entropy）框架：该框架仅在学习者处于不确定性状态时才调用VLM获取建议，通过执行教师建议进行训练，并将指导信息蒸馏至轻量级的RL策略中；同时，SAGE引入基于环境反馈的优势度量对教师动作进行加权，从而动态评估并筛选有价值的指导，避免盲目采纳所有建议。实验表明，在稀疏奖励的视觉推理与导航任务中，SAGE能够实现无需依赖VLM的自主决策，且在多个环境中超越无指导的RL基线，甚至在某些场景下优于其教师模型。此外，该方法显著减少了对VLM的调用频率（仅在部分训练步骤中触发），并在部署阶段完全无需VLM调用。研究结果表明，VLM无需作为固定策略使用，而可作为临时、不完美但具有价值的指导源，其有效性通过与环境的交互被验证和内化。
链接: https://arxiv.org/abs/2609.01567
作者: Matteo Merler,Giovanni Bonetta,Davide Zago,Rossella Cancelliere,Bernardo Magnini
机构: Fondazione Bruno Kessler(布鲁诺·凯斯勒基金会); University of Torino(都灵大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 9 pages, 3 figures, 4 tables in the main text, 27 pages, 4 figures, 9 tables including Appendix
Abstract:Vision-Language Models (VLMs) provide useful priors for interactive decision-making, but using them directly as policies is expensive and brittle: they must be queried at every step, do not improve from environment interaction, and can repeat systematic errors. We study how to learn a cheap autonomous policy from an online, expensive, and imperfect but informative VLM teacher. We propose SAGE (Selective Agent Guidance via Entropy), a framework that queries a VLM only when the learner is uncertain, executes the suggested action during training, and distills guidance into a lightweight Reinforcement Learning (RL) policy. Because VLM advice is not always reliable, SAGE can weight teacher-action distillation using environment-derived advantages rather than treating all suggestions as equally useful. Across sparse-reward visual reasoning and navigation tasks, SAGE learns policies that act without VLM guidance at evaluation time and improves over unguided RL in several environments, including settings where the learned policy exceeds its VLM teacher. The results show that selective guidance is most beneficial when the VLM can help the agent discover high-reward trajectories, and less useful when unguided exploration already succeeds or teacher actions do not lead to informative experience. SAGE also reduces VLM usage by prompting the teacher only on a fraction of training steps and requiring no VLM calls at deployment. Overall, our results suggest that VLMs don’t need to be used as fixed policies to be useful; they can instead act as temporary, imperfect sources of guidance whose value is tested and internalized through interaction.
[NLP-11] From Confusion to Clarity: Confusion-Aware Retrieval and Knowledge Injection for Text Classification EMNLP2026
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在处理具有大量语义相似标签的分类任务时，因预训练阶段未捕捉领域特定差异而导致的分类性能下降问题。其核心挑战在于：当候选标签间语义相近时，仅通过嵌入相似度检索前K个候选标签的方法无法有效区分这些高度相似的标签，导致模型缺乏足够的判别信号。为此，作者提出一种无需微调的框架，其关键在于：（1）识别模型难以区分的标签对；（2）将易混淆标签纳入候选集以增强上下文可区分性；（3）生成针对性的判别规则以明确区分相似候选标签。该框架所生成的规则具备良好的跨模型迁移能力，可应用于更小、成本更低的模型中，显著提升分类性能。在WOS、Flipkart和LEDGAR三个基准数据集上的实验表明，该方法相较传统检索基线在宏平均F1得分上最高提升10.0个百分点，且小型模型（2B–20B参数量）通过跨模型迁移实现最高达11.5个百分点的性能增益。
链接: https://arxiv.org/abs/2609.01564
作者: Manish Gupta,Chaitanya Giri,Jayasimha Talur
机构: Amazon(亚马逊)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: EMNLP 2026 (Industry Track)
Abstract:Large language models (LLMs) struggle to classify text into taxonomies with many semantically similar labels, as the distinctions are domain-specific and not captured by pre-training. To handle large label spaces, a common approach retrieves top- K candidate labels by embedding similarity and prompt the LLM to choose among them. However, top- K retrieval reduces the number of candidates but does not help the model tell similar ones apart. When two similar labels both appear as candidates, the model lacks the signal to choose correctly between them. We propose a framework that (1) identifies which label pairs the model struggles to distinguish, (2) expands the candidate set to include confusable labels, and (3) generates targeted rules to differentiate between similar candidates. The framework requires no fine-tuning, and the generated rules transfer to smaller, cheaper models. On three benchmarks (WOS, Flipkart, LEDGAR), our approach improves Macro F1 by up to 10.0pp over retrieval baselines, with smaller models (2B–20B) gaining up to 11.5pp via cross-model transfer.
[NLP-12] A systematic Approach to constructing a Chance-and-Risk Matrix for Semiconductor Supply Chains
【速读】： 该论文旨在解决半导体供应链在地缘政治紧张、地理集中度高及技术快速迭代背景下，缺乏可扩展的系统持续从公开企业披露信息中提取、结构化并优先排序风险情报的问题。其解决方案的关键在于构建一个端到端的自动化管道：首先通过大规模检索获取半导体企业公开文档，继而利用大语言模型（Large Language Models, LLMs）识别其中描述的风险与机遇；随后将这些要素组织成知识图谱，实现与类别、来源及关联事件的精准链接；再通过三层次融合机制——算法公式、基于LLM的语义相关性修正以及专家验证——完成去重与优先级排序。该方法在五家产业链不同环节企业上的应用生成了76,207个经评分的风险与机遇条目，独立验证显示92.6%有效；自动化排名与专家判断的Spearman相关系数平均达0.55（风险）和0.72（机遇），揭示贸易限制为跨企业主导风险。
链接: https://arxiv.org/abs/2609.01563
作者: Ema Salkić,Alexander Fichtl,Philipp Ulrich,Hans Ehm,Marta Bonik,Georg Groh
机构: Infineon Technologies AG (英飞凌科技公司); Technical University of Munich (慕尼黑工业大学)
类目: Computation and Language (cs.CL)
备注: Copyright 2026 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works
Abstract:Semiconductor supply chains face escalating risks from geopolitical tensions, geographic concentration, and rapid technological shifts, yet no scalable system continuously extracts, structures, and prioritizes risk intelligence from public corporate disclosures. We present an end-to-end pipeline that retrieves corporate documents for semiconductor companies and uses large language models (LLMs) to extract the risks and opportunities they describe. It organizes these into a knowledge graph linking each item to its category, sources, and related events, then merges duplicates and ranks them with a three-layer mechanism combining an algorithmic formula, an LLM relevance adjustment, and expert validation. Applied to five companies across the value chain, the pipeline produces 76,207 scored items, of which an independent check finds 92.6% valid. The automated rankings match expert judgment at an average Spearman correlation of 0.55 for risks and 0.72 for opportunities, and the resulting matrices identify trade restrictions as the dominant cross-company risk.
[NLP-13] SDARE-Bench: Evaluating Large Language Models on Conversational Stigma Detection and Response in Dyadic and Group Dialogue EMNLP2026
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在涉及社会判断的咨询与决策场景中对污名化（stigma）识别与回应能力不足的问题。现有评估基准多依赖静态提示和固定格式任务，忽视了日常交流中的对话情境与受众影响，导致对模型在真实社交语境下表现的评估存在显著缺陷。为此，研究提出SDARE-Bench——首个基于情景的基准测试框架，用于评估LLMs在污名检测与开放式回应生成两方面的表现，包含1,138组双人对话与1,388组群体对话样本。实证结果表明，8个主流LLM在识别污名要素方面表现普遍不佳，尤其在群体对话中更为严重；在开放式回应生成中，群体场景下的污名表达显著高于双人场景，且模型对污名的抵抗能力较弱、建议内容更不切实际。通过基于1,392条人工标注回应训练的分类器进行评估，在人为构建的群体压力情境下，污名表达率高达平均97.5%。研究揭示了污名响应是大语言模型在复杂社交对话环境中反复出现的安全漏洞，凸显了在动态人际互动中提升模型社会伦理敏感性的迫切需求。
链接: https://arxiv.org/abs/2609.01548
作者: Stephanie Fong,Yiwen Jiang,Zimu Wang,Hongxi Yang,Yaling Shen,Hiu Weh Naomi Chow,Heung Ying Lai,Xiangyu Zhao,Qingyang Xu,Zhongxing Xu,Jiahe Liu,Guilherme C. Oliveira,Vincent Lee,Zongyuan Ge,Dominic Dwyer
机构: Monash University(蒙纳士大学); University of Liverpool(利物浦大学); University of Edinburgh(爱丁堡大学); Federation University(联邦大学); Orygen, The University of Melbourne(奥里根，墨尔本大学)
类目: Computation and Language (cs.CL)
备注: Paper accepted at EMNLP 2026
Abstract:Large Language Models (LLMs) are increasingly used in advice seeking and decision making that may affect social judgements. Despite stigma’s profound effects on people and communities, benchmarks remain scarce. Existing general-domain evaluations typically rely on static prompts and fixed-format tasks, overlooking conversational contexts and audience effects in everyday communication. To address these gaps, we introduce SDARE-Bench, the first scenario-based benchmark evaluating both stigma detection and open-ended response generation in LLMs, comprising 1,138 dyadic queries and 1,388 group dialogue. Empirical results across 8 LLMs consistently demonstrate poor identification of stigma components, especially in group dialogues. In open-ended response generation, stigma expression was substantially higher in group settings than in dyadic, with weaker resistance to stigma and more unrealistic advice. Responses were evaluated using a classifier trained on 1,392 human annotated responses. In constructed group pressure settings, stigma expression rates further increased to a striking average of 97.5%. Our findings identify stigma response as a recurring LLM safety vulnerability, especially in socially complex conversational contexts.
[NLP-14] Knowledge Distillation During Mid-Training Favors Reasoning over Factual Recall KR
【速读】： 该论文旨在解决生成式语言模型在不同训练阶段中，基于逻辑值的知识蒸馏（Logit-based Knowledge Distillation, KD）效果不一致的问题，尤其关注中段训练阶段（mid-training）——即在高质量语料上进行自监督学习的中间阶段——时，标准前向KL蒸馏方法导致事实记忆能力下降的现象。其核心问题是：尽管在预训练阶段，前向KL蒸馏能同时提升推理与事实记忆表现，但在中段训练阶段，该方法虽仍促进推理能力发展，却显著抑制了事实知识的获取。解决方案的关键在于识别出这一现象的根本原因——教师模型在程序性知识数据上比知识密集型数据更具置信度，而学生模型早期便已习得低熵的事实性知识，从而造成师生间在知识分布上的不对称性。为此，作者提出切换蒸馏（Switch Distillation），通过引入教师预测熵作为轻量级路由信号，仅在教师对特定词元具有高置信度时执行蒸馏，否则回退至标准交叉熵损失。该方法有效缓解了知识分布不平衡问题，在多种教师规模下均优于现有蒸馏方案，并在后训练阶段持续保持更高的推理与常识理解性能，同时几乎完全保留事实记忆能力（96.7–96.8%），显著缩小了事实召回差距。
链接: https://arxiv.org/abs/2609.01532
作者: Jacqueline He,Howard Yen,Shuyue Stella Li,Margaret Li,Hanqing Zeng,Yinglong Xia,Benyu Zhang,Zhuokai Zhao,Qiang Zhang,Pang Wei Koh,Luke Zettlemoyer,Wen-tau Yih
机构: Meta AI(Meta AI); University of Washington(华盛顿大学); Princeton University(普林斯顿大学)
类目: Computation and Language (cs.CL)
备注: 33 pages, 13 figures, 9 tables. Code is publicly available at this https URL
Abstract:Logit-based knowledge distillation (KD) is used to train smaller language models (LMs) via supervision from stronger teachers, but whether its benefits are consistent across training stages remains unclear. Through controlled experiments, we find that forward Kullback-Leibler (KL) distillation–the standard KD formulation–with post-trained teachers behaves fundamentally differently during mid-training, an intermediate phase of self-supervised learning on curated corpora. Surprisingly, while forward KD simultaneously improves reasoning and factual recall during pre-training relative to standard next-token prediction (NTP), it instead slows factual recall acquisition during mid-training despite continued reasoning gains. We trace this stage dependence to an asymmetry in teacher confidence across data domains and the student’s evolving knowledge state: teachers are more confident on procedural than knowledge-intensive data, while students acquire low-entropy factual knowledge earlier in training. To mitigate this imbalance, we propose Switch Distillation, a simple mid-training objective that distills on tokens where the teacher is confident, using teacher predictive entropy as a lightweight routing signal, and otherwise falls back to cross-entropy. Switch Distillation consistently outperforms existing distillation objectives across teacher sizes. Relative to standard NTP, it achieves 1.61-1.71x the reasoning performance and 1.13-1.19x the knowledge and commonsense performance while preserving 96.7-96.8% of factual recall. Crucially, these benefits persist after post-training: Switch Distillation closes the factual recall gap while maintaining 1.25-1.32x and 1.13-1.20x gains in reasoning and knowledge and commonsense, respectively.
[NLP-15] HarnessDev: Can LLM s Create and Evolve Their Own Agent Harness?
【速读】： 该论文旨在解决当前生成式AI代理（Generative AI Agent）在实际部署中依赖外部执行基础设施（即“代理吊架”，agent harness）所带来的性能不确定性问题。现有评估方法通常仅报告特定吊架下的下游任务表现，忽略了模型自身构建和优化吊架的能力，导致对代理自主性与适应性的评估不足。为此，论文提出HarnessDev基准测试框架，将评估单位从任务输出转向可运行的执行基础设施本身，涵盖“创建”（Creation）与“演化”（Evolution）两个阶段：在创建阶段，代理从最小种子和少量案例出发，自主构建完整的执行系统；在演化阶段，代理基于下游执行反馈迭代优化自身构建的吊架以提升性能。评估指标包括能力（在保留测试集上的任务成功率）与效率（执行令牌成本）。实验结果表明，尽管部分生成吊架在写作与机器学习实验任务上达到或超越人工设计参考方案，但在代码生成、搜索与研究任务上仍显著落后，且执行成本差异巨大；演化虽带来一定性能提升，但效果不稳定，且难以泛化至未见任务，进一步实验显示性能增益高度依赖于执行吊架的模型本身，揭示了跨模型迁移能力有限的问题。因此，解决方案的关键在于建立以“可运行基础设施”为核心的新评估范式，并揭示当前模型自主构建与优化执行环境能力的局限性。
链接: https://arxiv.org/abs/2609.01437
作者: Yuhao Wu,Jingyuan Zhang,Jiajun Shi,Xinping Lei,Qingshui Gu,Yuxuan Zhang,Zexuan Wang,Chen He,Chen Huang,Maojia Song,Zhiyuan Zeng,Shaowen Wang,Jinkai Liu,Yunfeng Shi,Jiaheng Liu,Shen Yan,Wenhao Huang,Ge Zhang,Wenxuan Zhang
机构: 未知
类目: oftware Engineering (cs.SE); Computation and Language (cs.CL)
备注: Project page: this https URL
Abstract:As agents move from research prototypes to deployed tools, their capability increasingly depends on model-external execution infrastructure, commonly termed the agent harness. Changing this harness while holding model weights fixed can substantially alter task performance. Current agent evaluations typically report downstream performance under a chosen harness, leaving a model’s ability to develop the harness itself comparatively underexplored. We introduce HarnessDev, a benchmark that shifts the unit of evaluation from task outputs to runnable infrastructure. HarnessDev covers two stages. In Creation, the agent starts from a minimal seed and a small number of cases, then builds a complete execution system. In Evolution, it starts from its own created harness and iteratively revises it using downstream execution feedback, with the goal of improving benchmark performance. We then evaluate each constructed harness on capability (task success on held-out benchmarks) and efficiency (execution-token cost). The reported Creation results cover six creator LLMs, four domains, and five downstream benchmarks totaling 2,207 unique downstream instances, with hidden evaluation tasks withheld from development. We find that generated harnesses remain substantially behind mature human-engineered references on code and on search and research, while matching or exceeding the selected references on writing and machine-learning experimentation, with large variation in execution cost. Evolution produces some performance gains, but they are unstable and transfer only partially to held-out tasks. Experiments with a fixed runtime model further show that the gains depend strongly on the model executing the harness, indicating limited transfer across models.
[NLP-16] Citing Less Critically: LLM s Reshape the Rhetoric and Reach of Scientific Citation EMNLP2026
【速读】： 该论文旨在解决生成式人工智能（Generative AI）在科学写作中引用文献时是否能准确再现人类作者的修辞意图这一关键问题。其核心挑战在于，现有大语言模型（LLM）在生成引用时可能偏离人类引文行为的语用特征，从而影响学术交流的批判性与社会网络关联性。解决方案的关键是提出一种“掩蔽引用任务”（masked-citation task），通过让LLM在给定引用上下文中生成替代性引用句，构建一个可与人类引文直接对比的反事实语料库。研究采用大语言模型作为评判者（LLM-as-a-judge）对引用意图进行分类，并利用包含2000万条边的共作者网络量化被引作者之间的社会距离。分析揭示三个主要模式：（1）相较于人类，LLM引用表现出显著更低的批判性；（2）LLM倾向于过度引用热门且陈旧的文献，尤其在对比性引用中更明显，而人类则更常引用近期、小众的研究；（3）人类在支持性引用中倾向于引用其紧密社交网络内的作者，而LLM则更频繁引用社会距离较远的作者。这些差异具有双重效应：一方面拓展了引用的广度，突破了学者的社交圈层；另一方面削弱了引用的批判性并加剧了可见性偏差，重塑了科学引文的修辞格局与传播范围。
链接: https://arxiv.org/abs/2609.01432
作者: Yixuan Liu,Lin Chen,Zhuoqi Liu,Jianglin Lu,Dakota Murray
机构: Northeastern University(东北大学); University at Albany, State University of New York(纽约州立大学阿尔巴尼分校)
类目: Digital Libraries (cs.DL); Computation and Language (cs.CL); Computers and Society (cs.CY); Social and Information Networks (cs.SI)
备注: Accepted at the EMNLP 2026 main conference
Abstract:Scientific citations carry rhetorical intent. Scholars may cite prior work positively (supporting), negatively (contrasting), or neutrally (mentioning). As large language models (LLMs) increasingly assist scientific writing, whether they reproduce citations with the same rhetorical intent as humans remains unclear. We introduce a masked-citation task to compare human and LLM-generated citation behavior. For each citation context, an LLM generates a replacement citation sentence, producing a counterfactual corpus directly comparable to human citation. We analyze what, whom, and how models cite, using an LLM-as-a-judge to classify citation intent and a 20-million-edge coauthorship network to measure social distance between cited authors. Across six popular LLMs and 1,746 top NLP conference papers (63k+ contexts, 132k+ citations), three patterns emerge: (1) Compared with human citation, LLMs cite significantly less critically; (2) LLMs over-cite popular and older papers, a tendency amplified for contrasting citations where human writing more often draws on recent, niche work; (3) Whereas humans often cite within their close social network, especially for supporting citations, LLMs tend to draw on more socially distant authors. Together, these differences are double-edged: LLM citation reaches beyond a scholar’s close collaborators while being less critical and amplifying visibility bias, reshaping the rhetoric and reach of scientific citation.
[NLP-17] From Rollouts to Recipes: Self-Contained Post-Training for LLM s EMNLP2026
【速读】： 该论文旨在解决大语言模型后训练（post-training）过程中采用统一训练策略处理所有样本的局限性，尽管模型自身的推理轨迹（rollout）已揭示不同样本存在异质性的学习状态。其核心问题是：如何根据每个样本的实际学习表现动态调整优化策略，以提升训练效率与效果。解决方案的关键在于提出一种行为条件驱动的自适应框架——Self-Routing，该框架基于样本的推理正确性（rollout correctness）与置信度（confidence）判断其行为状态，并据此将样本路由至不同的优化路径：GRPO（Generalized Reward Policy Optimization）、在线策略自蒸馏（on-policy self-distillation）、正则化或跳过更新。该方法无需外部教师模型、额外标注或复杂采样机制，实现了训练过程的自我调节。实验在Qwen3与Qwen3.5骨干模型上的数学推理任务中验证了其有效性，结果表明Self-Routing显著优于均匀的GRPO、OPS（on-policy self-distillation）以及固定混合策略等基线方法；进一步分析显示，随着训练进程，路由分布动态变化，有效减少了对低信号或已稳定样本的冗余更新，提升了训练资源的利用效率。
链接: https://arxiv.org/abs/2609.01422
作者: Yifei Li,Lingling Zhang,Muye Huang,Zihan Ma,Jiashuai Liu,Jun Liu
机构: Xi’an Jiaotong University (西安交通大学); Shaanxi Province Key Laboratory of Big Data Knowledge Engineering (陕西省大数据知识工程重点实验室); Zhongguancun Academy (中关村学院); National Engineering Research Center for Visual Information and Applications (国家视觉信息与应用工程研究中心)
类目: Computation and Language (cs.CL)
备注: 14 pages, 5 figures. Accepted at EMNLP 2026
Abstract:Post-training large language models usually applies a single training recipe to all samples, even though the model’s own rollouts reveal different sample-level learning states. We propose Self-Routing, a behavior-conditioned post-training framework that uses rollout correctness and confidence to decide how each sample should be optimized. Depending on its behavior state, a sample is routed to GRPO, on-policy self-distillation, regularization, or skipping, allowing training to adapt without external teachers, extra annotations, or additional sampling. Experiments on mathematical reasoning across Qwen3 and Qwen3.5 backbones show that Self-Routing consistently improves over uniform GRPO, uniform OPSD, fixed mixtures, and simpler routing baselines. Further analyses show that the routing distribution changes over training and reduces unnecessary updates on low-signal or already stable samples.
[NLP-18] EdiTikZ: Scientific Figure Editing from Revision Trajectories
【速读】： 该论文旨在解决科学图表生成中从文本或图像生成可直接发表的高质量图表所需进行迭代优化这一关键挑战，而现有方法普遍依赖昂贵的专有代理系统、侧重于评估而非实际编辑能力，或通过合成编辑构建训练监督信号。其核心解决方案在于利用自然发生的科研文献修订与演进轨迹作为可扩展的监督来源，首次构建了大规模的基于修订数据的科学图表编辑数据集DaEdiTikZ——通过挖掘arXiv、GitHub和TeX SE上的39.1万对合理的TikZ编辑对，并借助视觉-语言模型（VLM）结合渲染后的图像与TikZ代码，推断出78.1万条有向编辑指令。为验证模型性能，研究进一步提出了人工精修的基准测试集DaEdiTikZ-Bench（包含790个实例），并训练了两个基于Qwen3.5架构的轻量级EdiTikZ模型（4B和9B），采用联合重建与编辑学习策略，并通过强化学习（RL）引入互补奖励机制以优化渲染保真度与编辑执行准确性。实验表明，所提出的9B模型在自动评估中优于所有对比基线，在人类评估中表现超越GPT-5.6-Sol，且与Gemini-3.1-Pro相当；即使在严重分布外（out-of-distribution）的场景下，其性能仍保持竞争力，接近GPT-5.6-Sol在2K上下文长度限制下的表现。研究成果包括模型与数据集的开源发布。
链接: https://arxiv.org/abs/2609.01409
作者: Christian Greisinger,Zhixue Zhao,Steffen Eger
机构: University of Technology Nuremberg(纽伦堡应用技术大学); University of Sheffield(谢菲尔德大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Vision-language models (VLMs) have shown strong performance in generating scientific figures from text or images. However, producing publication-ready figures requires iterative refinement, making scientific figure editing an important yet largely unexplored task. Existing approaches rely on costly proprietary agentic systems, focus primarily on evaluation, or construct training supervision from synthetically generated edits. Instead, we leverage naturally occurring scientific revision and development trajectories as a scalable source of supervision. To this end, we introduce DaEdiTikZ, the first large-scale dataset of revision-derived scientific figure edits, constructed by mining 391K plausible TikZ edit pairs from arXiv, GitHub, and TeX SE and inferring 781K directed edit instructions with a VLM conditioned on rendered figures and TikZ code. We further introduce DaEdiTikZ-Bench, a human-refined benchmark with 790 instances, and train two compact Qwen3.5-based EdiTikZ models (4B and 9B) by jointly learning reconstruction and editing, followed by reinforcement learning (RL) with complementary rewards for rendered fidelity and edit application. Automatic evaluation places our 9B model above all tested baselines, while human evaluation with 9 annotators and 4,320 ratings places it above GPT-5.6-Sol and on par with Gemini-3.1-Pro. Under severe out-of-distribution shifts, it remains competitive with GPT-5.6-Sol near its 2K training sequence-length regime. Models and datasets will be released.
[NLP-19] When Tokenization is Secretly Output Supervision EMNLP2026
【速读】： 该论文旨在解决语言模型中分词（tokenization）机制对模型学习过程与内部表征影响的深层问题，尤其关注在自回归模型中，输出端分词粒度如何作为监督信号的一部分，决定模型单次前向传播需处理的任务复杂度。传统观点将分词视为输入预处理步骤，但本文指出这一框架不完整：输出分词方式直接决定了模型所接收的监督信号，进而影响学习难度、训练动态及模型内部表示的形成。其核心解决方案在于提出一种新颖的输入与输出分词解耦实验设计，通过控制变量验证了输出分词对任务性能、训练过程和模型内态的显著影响，而输入分词的影响则相对较小且不具决定性。研究结果表明，不同分词策略不仅改变输入表示，更实质上改变了模型所“学习的任务”，因此跨模型比较时，差异可能源于任务定义而非模型能力本身。实证调查进一步揭示，当前120篇自然语言处理领域权威论文中，仅有约10%报告模型的数值分词方式，而69%在未说明分词设置的情况下进行跨分词策略比较，凸显了该问题的普遍忽视。本文通过将分词重新定位为输出监督（output supervision）的组成部分，为长期以来观察到的分词对性能的系统性影响提供了理论解释框架。
链接: https://arxiv.org/abs/2609.01386
作者: Tanja Baeumel,Josef van Genabith,Simon Ostermann
机构: German Research Center for AI (DFKI); Saarland University; Center for European Research in Trusted AI (CERTAIN)
类目: Computation and Language (cs.CL)
备注: Accepted to EMNLP 2026 Main Conference
Abstract:Tokenization in language models is treated by default as an input preprocessing decision. We argue that this framing is incomplete: in autoregressive models, tokenizer granularity determines what the model must resolve in a single forward pass, and therefore the supervision signal it receives. This affects both the difficulty of the learning problem and the representations that emerge inside the model. We test this in a controlled experiment on numeric reasoning with a novel decoupling of input and output tokenization. As the output supervision view predicts, differences in task performance, training dynamics, and model internals are induced by output tokenization and largely invariant to input tokenization. This may matter in practice, because models with different tokenization strategies differ not only in input representation but in the task they were trained on. Comparisons between models may thus partly reflect task definition rather than ability. A survey of 120 recent *CL papers on numeric reasoning confirms that this is rarely acknowledged: only about 10% report the numeric tokenization of the models they evaluate, while 69% compare across tokenization, and thus supervision, regimes without reporting it. While prior work documents that tokenization consistently affects model performance, there is no principled account of why. We argue that framing tokenization as output supervision provides that account.
[NLP-20] Polish ModernBERT: The Long and Short of Polish Language Understanding
【速读】： 该论文旨在解决波兰语（Polish）领域中高效、长上下文编码器缺乏的问题，尤其针对现有基于BERT/RoBERTa架构的波兰语编码器在处理长文本任务时性能不足与参数效率低下的局限性。其核心挑战在于如何在保持高精度的同时提升模型对长序列（如8K token）的建模能力，并实现更优的计算效率。解决方案的关键在于提出Polish ModernBERT系列模型，采用改进的ModernBERT预训练范式，通过分阶段选择实验优化训练策略，并构建首个覆盖法律主题分类、意识形态倾向预测、文学情节事实一致性评估及人权侵犯判定等任务的长上下文基准。该模型在多个任务上表现出色，在30项评测中整体性能领先，尤其是8K上下文版本在长文本任务中显著超越同类波兰语RoBERTa-8K基线，同时以更少参数（如Base-8K仅149M vs. 190M）实现更低的峰值内存占用和推理延迟，展现出卓越的效率优势，且在低于300M参数的编码器中于波兰语检索任务上取得最佳表现。
链接: https://arxiv.org/abs/2609.01379
作者: Michał Perełkiewicz,Sławomir Dadas,Rafał Poświata,Małgorzata Grębowiec
机构: National Information Processing Institute (国家信息处理研究所), Warsaw (华沙), Poland (波兰)
类目: Computation and Language (cs.CL)
备注:
Abstract:Encoder-only Transformers remain effective for discriminative and representation-learning tasks, yet Polish encoders still largely rely on BERT/RoBERTa-style architectures. We introduce \textbfPolish ModernBERT, a family of four Polish encoders available at Base and Large scales, each with 512-token and 8K context variants. We adapt the ModernBERT pretraining recipe through staged selection experiments and release a long-context benchmark covering legal topic classification, ideological decision-direction prediction, factual-consistency assessment over literary plot summaries, and human-rights violation assessment. Across 30 tasks, Polish ModernBERT achieves the best overall performance among the evaluated Polish encoders, reaching 83.99 and 85.11 for the Base-8K and Large-8K models, respectively. On long-context tasks, the 8K variants improve over matched Polish RoBERTa-8K baselines from 67.47 to 77.15 and from 75.88 to 78.49 at the Base and Large scales, respectively. The Base-8K model achieves this gain with 22% fewer parameters (149M vs.\ 190M). Efficiency measurements in representative inference setups show lower peak memory usage and latency than matched Polish RoBERTa baselines in both 512-token and 8K settings. Polish ModernBERT-8K-Base additionally achieves the best result on a Polish retrieval benchmark among the evaluated encoders below 300M parameters.
[NLP-21] IntroConformal: Conformal Factuality Guarantees for Large Vision-Language Models via Introspective Signals EMNLP2026
【速读】： 该论文旨在解决大视觉语言模型（Large Vision-Language Models, LVLMs）在生成内容时难以保证事实正确性的问题，尤其针对现有方法依赖外部验证器或生成时置信度信号所引入的额外依赖关系及对高置信度错误输出失效的局限性。其核心解决方案是提出一种无需训练的自洽风险控制（Conformal Risk Control, CRC）框架——IntroConformal，通过模型内部的内省信号实现有限样本、分布无关的事实性保障。该框架的关键在于利用模型自身产生的两种内省性指标：一是基于隐藏层表示的逐层语义稳定性（layer-wise semantic stability），作为符合性得分；二是提出的验证概率（verification probability），能够捕捉模型对命题事实性的自我判断。实验表明，IntroConformal在多种LVLM架构上均满足符合性风险保证，同时显著降低拒绝率，并在命题级事实性判别能力上达到或优于依赖外部验证器的基线方法。
链接: https://arxiv.org/abs/2609.01375
作者: Md. Atabuzzaman,Christian Alexander,Chris Thomas
机构: Virginia Tech(弗吉尼亚理工学院)
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL)
备注: EMNLP 2026 main conference
Abstract:Large Vision-Language Models (LVLMs) have achieved strong multimodal performance, yet ensuring the factual correctness of generated content remains challenging. Existing methods that provide statistical guarantees on factuality typically rely on external verifiers or generation-time confidence signals, which introduce auxiliary dependencies or often fail for confident but incorrect outputs. We argue that reliable factuality control can instead be achieved through introspective signals derived from the model itself. We introduce IntroConformal, a training-free Conformal Risk Control (CRC) framework that provides finite-sample, distribution-free factuality guarantees. We first instantiate it with layer-wise semantic stability, a conformity score derived from hidden-state representations, and then propose verification probability, a stronger score capturing the model’s self-administered judgment on claim factuality. Across multiple LVLM architectures, IntroConformal satisfies the conformal risk guarantee while substantially reducing abstention and achieving competitive or superior claim-level discrimination relative to external verifier-based baselines.
[NLP-22] Behaviorally Effective LoRA Writes Are Sparse and Structured
【速读】： 该论文旨在解决低秩适配（LoRA）中一个关键问题：尽管LoRA通过固定更新秩来提升效率，但其原始参数化形式无法揭示哪些具体参数部分真正承载了模型的行为变化。为此，作者提出了一种名为**学习基LoRA（Learned-Basis LoRA）**的解决方案，其核心在于通过一种“学习基延续”（learned-basis continuation）的训练策略，将未经约束的适配器所学得的写入（write）列转换为模块级正交基，并冻结该基底后在约束参数空间中继续训练。实验表明，在14个从无约束到有约束形式的精确切换中，转换时刻的保留精度不变，重构写矩阵的相对弗罗贝尼乌斯误差不超过0.25%。进一步的同状态延续测试验证了不同写入子空间会导致相同检查点产生显著不同的行为演化，从而确立了写入几何结构作为因果状态变量。此外，无需重训练的投影测试显示，有效写入信号集中于学习得到的写入空间，而在随机或冻结激活的PCA控制组中几乎消失。研究发现，这种集中性在局部与全局尺度上均显著存在：在GSM8K、MathQA和AQuA任务中，每模块前k个组件的连续优化在k=2、4时即达到最优，且全局排名测试表明，学习所得的前16和前32个组件优于随机匹配子集，尤其在GSM8K/Qwen和MathQA/Qwen上表现更优。单方向消融分析进一步揭示，仅有少数晚期的q_proj、o_proj和down_proj组件具有显著的行为影响，证明行为有效的LoRA写入是稀疏且结构化的。
链接: https://arxiv.org/abs/2609.01374
作者: Haruto Sato,Yuki Tanaka,Ren Nakamura,Aoi Kobayashi,Mei Ito
机构: 未知
类目: Computation and Language (cs.CL)
备注:
Abstract:Low-rank adaptation fixes the rank of the update, but it does not identify which parts of a trained write actually carry behavior. We study that question directly and show that behaviorally effective LoRA writes are sparse, structured, and far more concentrated than the raw low-rank parameterization suggests. We use Learned-Basis LoRA, a learned-basis continuation recipe, to expose that structure. The recipe warms up an unconstrained adapter, converts its learned write columns into a module-wise orthonormal basis, freezes that basis, and continues training inside the constrained parameterization. Across 14 exact switches from unconstrained to constrained form, held-out accuracy is unchanged at the conversion step and reconstructed write matrices differ by at most 0.25% relative Frobenius error. Same-state continuation then shows that the same trained checkpoint develops differently under different write subspaces, establishing write geometry as a causal state variable. A no-retraining projection test shows that useful write signal stays inside the learned write space and largely disappears from random or frozen-activation PCA controls. The concentration pattern is strong at both local and global scales. Across GSM8K, MathQA, and AQuA, per-module top-k continuation reaches its optimum at k in 2, 4 in all twelve seed-level cases we test. A stricter global ranking test shows that learned top-16 and top-32 subsets outperform matched random subsets, especially on GSM8K/Qwen and MathQA/Qwen. Single-direction ablations further reveal a sparse set of late q_proj, o_proj, and down_proj components with outsized behavioral impact. Subjects: Computation and Language (cs.CL) Cite as: arXiv:2609.01374 [cs.CL] (or arXiv:2609.01374v1 [cs.CL] for this version) https://doi.org/10.48550/arXiv.2609.01374 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Haruto Sato [view email] [v1] Tue, 1 Sep 2026 15:09:43 UTC (1,385 KB) Full-text links: Access Paper: View a PDF of the paper titled Behaviorally Effective LoRA Writes Are Sparse and Structured, by Haruto Sato and 4 other authorsView PDFHTML (experimental)TeX Source view license Current browse context: cs.CL prev | next new | recent | 2026-09 Change to browse by: cs References Citations NASA ADSGoogle Scholar Semantic Scholar export BibTeX citation Loading… BibTeX formatted citation loading… Data provided by: Bookmark checked="checked"class=“labs-tab-input”> Bibliographic Tools Bibliographic and Citation Tools Bibliographic Explorer Toggle Bibliographic Explorer (What is the Explorer?) Connected Papers Toggle Connected Papers (What is Connected Papers?) Litmaps Toggle Litmaps (What is Litmaps?) scite.ai Toggle scite Smart Citations (What are Smart Citations?) Code, Data, Media Code, Data and Media Associated with this Article alphaXiv Toggle alphaXiv (What is alphaXiv?) Links to Code Toggle CatalyzeX Code Finder for Papers (What is CatalyzeX?) DagsHub Toggle DagsHub (What is DagsHub?) GotitPub Toggle Gotit.pub (What is GotitPub?) Huggingface Toggle Hugging Face (What is Huggingface?) ScienceCast Toggle ScienceCast (What is ScienceCast?) Demos Demos Replicate Toggle Replicate (What is Replicate?) Spaces Toggle Hugging Face Spaces (What is Spaces?) Spaces Toggle TXYZ.AI (What is TXYZ.AI?) Related Papers Recommenders and Search Tools Link to Influence Flower Influence Flower (What are Influence Flowers?) Core recommender toggle CORE Recommender (What is CORE?) Author Venue Institution Topic About arXivLabs arXivLabs: experimental projects with community collaborators arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv’s community? Learn more about arXivLabs. Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?) mathjaxToggle(); We gratefully acknowledge support from our major funders, member institutions, , and all contributors. About Help Contact Subscribe Copyright Privacy Accessibility Operational Status (opens in new tab) Major funding support from
[NLP-23] How Correct Is Your Answer? A Semantic Correctness Framework for Open QA Evaluation
【速读】： 该论文旨在解决开放域问答任务中答案正确性评估的可靠性问题，尤其针对现代大语言模型（LLM）在自由回答形式下的评估瓶颈。由于开放式答案存在多种可能的正确表达形式，且错误类型多样（如不完整、自相矛盾、过度生成、支持错误前提等），现有基于人工判断或相似度计算的评估指标往往无法区分这些质性差异，导致评估结果失真。其解决方案的关键在于提出一个可复用的语义正确性分类体系（semantic correctness taxonomy），将开放式回答划分为八个有序类别，有效区分冗长但正确的回答与包含幻觉内容的回答；同时构建了两个大规模基准数据集——CAP-Correctness（8.8k样本，覆盖多个主流问答数据集）和CAP-Statements（11k样本，用于将问答对转换为可用于自然语言推理（NLI）训练与基于陈述的评估的命题式语句）；并提出一种上下文感知精确度（Context-Aware Precision, CAP）指标，通过双向自然语言推理（bidirectional NLI）对条件化陈述进行评分，在单调性协议下验证其能有效尊重分类体系的预期排序，显著优于现有基线方法。
链接: https://arxiv.org/abs/2609.01369
作者: Elitsa Yotkova,Violeta Kastreva,Petar Velkov,Hristo Boyanov,Dimitar Dimitrov,Ivan Koychev,Preslav Nakov
机构: 未知
类目: Computation and Language (cs.CL)
备注:
Abstract:Reliable evaluation of open-ended question answering remains a bottleneck for measuring answer correctness of modern LLMs. Unlike multiple-choice tasks, free-form answers may be correct in many surface forms and may fail in qualitatively different ways, including incompleteness, contradiction, overgeneration, and endorsement of false premises. Existing judgment-based and similarity-based metrics often collapse these distinctions. We address this gap with three reusable contributions. First, we introduce a semantic correctness taxonomy that assigns open-ended answers to eight ordered classes, separating verbose-but-correct answers from those contaminated by hallucinated content. Second, we release CAP-Correctness, an 8.8k-example benchmark spanning widely used QA datasets, and CAP-Statements, an 11k-example dataset for converting question-answer pairs into declarative statements for natural language inference (NLI) training and statement-based evaluation. Third, we introduce CAP (Context-Aware Precision), a reference-based metric that scores question-conditioned statements using bidirectional NLI. Under a monotonicity protocol testing whether metrics respect the taxonomy’s intended ordering, CAP outperforms established baselines.
[NLP-24] Investigating Linear Probe Robustness to Linguistic Register Medical Specialty and Corpus Shifts in Medical QA EMNLP2026
【速读】： 该论文旨在解决大语言模型（LLM）隐状态中是否存在可泛化的“真实性方向”（truth direction）这一问题，即线性探测器能否在不依赖特定数据集的情况下，通过单次前向传播识别事实性错误。其核心挑战在于先前研究对真实性方向在输入分布变化下的泛化能力存在分歧，而这种分歧难以澄清，因跨数据集的探测迁移实验往往同时混杂多种输入变化因素。为此，作者在医学问答（QA）场景中系统分离出三个关键变量：写作风格（register）、领域（medical specialty）和语料库（corpus），构建了一个包含500个MedQA条目的基准数据集，每条数据被重写为四种风格（教材体、患者体、临床笔记体、口语体），并标注临床专科信息，同时与另外两个考试语料库（MedMCQA 和 MMLU-medical）进行跨数据集评估。通过对四个开源权重的LLM（2–8B参数）进行探测，研究发现：真实性方向对写作风格（均值Δ_register ≈ 0.10 AUROC）和医学专科（Δ_specialty ≈ 0.03）具有较强鲁棒性，但在不同语料库间表现不一致——在MMLU-medical上下降0.12 AUROC，而在MedMCQA上下降0.21 AUROC，后者约为风格差异的两倍。该结果在第二代生成器及真实人类撰写的患者问题中均得到复现。因此，该研究的关键结论是：尽管真实性方向在医学领域内部相对稳定，但其泛化能力会因语料库结构差异而显著退化，且这种退化无法由问题格式解释，表明线性探测所捕捉的信号部分依赖于特定数据集的结构特征，而非纯粹的医学知识本身。
链接: https://arxiv.org/abs/2609.01361
作者: Nishant Mishra,Ameen Abu-Hanna,Iacer Calixto
机构: Amsterdam UMC, University of Amsterdam (阿姆斯特丹大学医学中心，阿姆斯特丹大学); Amsterdam Public Health, Methodology (阿姆斯特丹公共卫生方法学)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: Accepted to EMNLP 2026 (Main Conference). 9 pages, 4 figures (main text); 28 pages total, including appendix. Code and data: this https URL
Abstract:Linear classifiers trained on hidden states of a large language model (LLM), linear probes, can flag factual errors from a single forward pass. Geometrically, that implies that true and false statements separate along a stable direction in hidden state space, i.e., the truth direction. Prior work disagrees on whether this generalises across input shifts, but the disagreement is hard to interpret because cross-dataset probe transfer experiments confound several kinds of input change at once. We isolate three such variables in medical question-answering (QA): writing style (register), domain (medical specialty), and corpus (dataset). We build a benchmark using 500 MedQA entries, each rewritten into four styles (textbook, patient, clinical note, colloquial), annotated with clinical specialty, and grouped with two other exam corpora, MedMCQA and MMLU-medical, for cross-dataset evaluation. Probing four open-weight LLMs (2–8B), we find that the truth direction is largely robust to writing style (mean \Delta_\textregister \approx 0.10 AUROC on held-out facts) and to medical specialty ( \Delta_\textspecialty \approx 0.03 ), but degrades unevenly across corpora: by 0.12 AUROC on MMLU-medical and by 0.21 on MedMCQA, roughly twice the register gap. The register result replicates with a second generator and carries over to human-written patient questions. The truth direction is therefore largely stable within the medical domain but breaks under some corpus shifts, and question format does not explain the break, which suggests that the signal a linear probe recovers is partly bound to dataset structure rather than to medical knowledge alone.
[NLP-25] Separating Syntax from Language: A Mechanistic Account of Translation in Multilingual LLM s EMNLP
【速读】： 该论文旨在解决多语言大语言模型（mLLM）在机器翻译过程中，跨语言表征转换机制不清晰的问题。现有研究认为翻译可分解为概念内容的独立表征与语言特异性形式生成两个阶段，但其内部细节仍不明确。本文的关键突破在于揭示翻译过程具有更高的模块化结构：目标语言的生成可进一步分离为句法结构构建与表面语言形式实现两个独立阶段。通过构建控制变量的多语言数据集，结合因果干预与探针分析，研究发现模型在生成目标语言时，优先构建目标语的词序结构，随后才实现具体语言的表面形态表达。研究还识别出部分注意力头对句法变换具有选择性敏感性，而对语言身份保持不变，表明句法结构的形成是翻译中一个独立的功能阶段。这一发现拓展了以往对翻译过程的分解认知，揭示了多语言模型中功能分化组件如何协同实现翻译任务。
链接: https://arxiv.org/abs/2609.01356
作者: Mikhail Sonkin,Tanja Baeumel,Daniil Gurgurov,Josef van Genabith,Simon Ostermann
机构: Saarland University (萨尔兰大学); German Research Center for Artificial Intelligence (DFKI)(德国人工智能研究中心); Centre for European Research in Trusted AI (CERTAIN)(欧洲可信人工智能研究中心); University of Göttingen (哥廷根大学)
类目: Computation and Language (cs.CL)
备注: Accepted to EMNLP Findings 2026
Abstract:Multilingual large language models (mLLMs) achieve strong performance in machine translation, yet our understanding of the mechanisms by which they transform representations from one language to another remains incomplete. Prior work suggests that translation decomposes into separable processes within an mLLM, where conceptual content is first represented independently, followed by a production into language-specific form. In this work, we show that translation is even more modular than previously assumed and that the output language production in translation processes is actually further separable into a syntax and a surface language process. We construct controlled multilingual datasets that isolate cross-linguistic differences in word-order and use causal interventions and probing to track how representations are transformed during translation. We find that models first construct target-side word-order before realizing the target language surface form. We identify individual attention heads that are selectively sensitive to syntactic transformations while remaining largely invariant to language identity. These results establish the commitment to a syntactic structure as an independent stage in translation, extending prior decompositions and showing how translation is implemented by functionally different components within mLLMs.
[NLP-26] Where the Verifier Fails: A Category-Level Audit of Reward Signals in RLVR
【速读】： 该论文旨在解决生成式 AI（Generative AI）在自动评估中因验证器（verifier）对答案格式敏感而导致的不可靠性问题，尤其关注验证器在处理自由文本答案时的误拒率（false negative）及其根本原因。其解决方案的关键在于采用变异测试（metamorphic testing）对验证器本身进行系统性检验，通过构造数学意义保持不变的等价答案变体（即形式重写），从而确保任何拒绝结果均为可证明的错误，无需人工介入判断。研究通过对四个主流验证器在307,420个评估实例上的分析发现：（1）同一验证器内部自验证率差异显著，跨度达41.3个百分点，同一库的不同配置在49.9%的输入对上产生分歧；（2）错误主要集中在空格与标点符号处理，占默认LaTeX配置下不一致失败的93.0%，尤其是尾随句号或换行符占据主要误差预算；（3）分离拒绝与执行失败后揭示，相似总体错误率的验证器其失败原因相反，且参考数值级联验证器因相对容差的尺度不变性，对“数量级”错误呈现阶跃式接受行为——低于10⁴时接受率为0%，达到或超过10⁴时则为100%。这表明验证器设计中的格式敏感性与容差机制是导致评估失真核心因素。
链接: https://arxiv.org/abs/2609.01354
作者: Esther Xin
机构: Independent Researcher(独立研究员); Google(谷歌)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 8 pages, 3 figures, 8 tables. Code, transform suite, contract matrix, and per-sample verdict records at this https URL
Abstract:Reinforcement learning with verifiable rewards (RLVR) and standard benchmark evaluation both rely on an automatic verifier that turns a free text answer into a binary reward. Prior work reports that one evaluation harness accepts only about 94% of its own ground truth answers, blaming LaTeX parsing. That is an aggregate: it does not say which answer forms consume the error budget. We supply the decomposition. We apply metamorphic testing to the verifier rather than the model, generating certified equivalent answer variants, that is, rewrites that preserve mathematical meaning by construction, so that any rejection is a provable false negative needing no human adjudication. We then measure rejection per answer category across four widely used verifiers over 307,420 verdicts. We find three things. (1) Self validation ranges from 53.8% to 95.2% on identical inputs, a spread of 41.3 points. The published figure describes one implementation, not the task; two configurations of the same library disagree on 49.9% of pairs. (2) The residual is not spread across parsing categories but concentrated in whitespace and punctuation, which account for 93.0% of in contract failures for the default LaTeX configuration. A trailing period or newline dominates the budget. (3) Separating rejection from execution failure shows that verifiers with similar aggregate error fail for opposite reasons, and that a reference numeric cascade accepts off by one wrong answers as a step function of magnitude, from 0% below 10^4 to 100% at or above, because its relative tolerance is scale invariant.
[NLP-27] CHARM: Character Hallucination for Multicultural Role Play Benchmark EMNLP2026
【速读】： 该论文旨在解决生成式角色扮演大语言模型（Generative AI）在模拟角色时存在的知识边界误判问题，尤其关注模型在面对超出角色认知范围的提问时，既未能准确识别边界（Boundary-Awareness），又未能遵守边界而擅自生成内容（Boundary-Compliance）。其核心挑战在于现有评估方法难以区分角色幻觉（character hallucination）是源于对知识边界的认知缺失，还是虽已识别边界却仍选择违规回答。为此，研究提出CHARM——一个涵盖40位来自五大文化语言区域的真实与虚构角色的多文化基准测试集，并通过母语评审者验证其有效性。该基准采用支持“回避”（abstention）的多项选择题，分别探测时间边界（Temporal，如历史与现代之间的差异）与跨宇宙边界（Cross-Universe，即角色叙事或历史宇宙之外的实体）。研究设计了两阶段评估框架，将边界感知能力与边界合规性分离评估。实验结果表明，模型的幻觉主要由合规性失败驱动：尽管多数模型能明确意识到问题超出了角色的知识范畴，但仍倾向于生成看似合理但不符合角色设定的答案。进一步分析发现，此类错误中大量为参数化覆盖（parametric override）现象——即模型内部存储了相关事实，但在推理过程中未能抑制输出。此外，研究还揭示了不同文化背景角色在边界处理上的系统性差异，反映出模型知识中各文化代表性不均的问题。
链接: https://arxiv.org/abs/2609.01352
作者: Sunkyung Han,Nahyeon Park,Gaeun Seo,Seunghyun Yoon,JinYeong Bak
机构: Sungkyunkwan University (成均馆大学); Adobe Research (Adobe 研究院)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 16 pages, 1 figure. Accepted to Findings of EMNLP 2026
Abstract:Role-playing large language models (LLMs) are expected to adopt a character’s style while also respecting that character’s knowledge boundaries. Prior evaluations detect character hallucination but rarely distinguish whether errors arise from failure to recognize a boundary or from failure to comply despite recognition. We introduce CHARM, a multicultural benchmark of 40 real and fictional characters drawn from five cultural-linguistic regions, and validated by native reviewers. It probes two boundary types, Temporal (historical vs. modern) and Cross-Universe (entities outside a character’s narrative or historical universe), using abstention-enabled multiple-choice questions. We propose a two-stage evaluation that separates Boundary-Awareness (explicit recognition that a query is out of scope) from Boundary-Compliance (abstention when answering concrete questions). Evaluations across six LLMs show that hallucination is driven predominantly by compliance failures. Models frequently acknowledge that a query lies outside the character’s knowledge yet still provide factual, out-of-character answers. By re-posing the same questions to the target character, we confirm that a large fraction of these cases are verified parametric overrides; the model stores the relevant fact but fails to suppress it. We also observe systematic cultural variation in these failures, consistent with imbalances in how characters from different regions are represented in model knowledge.
[NLP-28] Probing Factual Knowledge Transfer with Training Data Interventions EMNLP2026
【速读】： 该论文旨在解决多语言语言模型在持续预训练过程中，是否能够将源语言（如英语）中习得的事实知识有效迁移至目标语言（如波斯语）的问题。其核心关切在于区分知识迁移与直接从目标语言数据中学习的差异。解决方案的关键在于提出一种基于干预的评估框架：以英语预训练模型为起点，对波斯语数据进行持续预训练，并系统性地在不同粒度上移除特定事实。为此，研究构建了SIFT数据集，包含500个三元组，覆盖20种关系，按事实主体的文化归属划分为全球普遍性实体与波斯相关实体，并配备原生波斯语填空模板，实现训练数据中事实的可控移除与效果评估。实验结果表明，事实迁移极为有限——在最严格的移除条件下，绝大多数英语习得的事实未能成功迁移到波斯语。此外，研究发现句子级共现关系的移除不足以消除事实信号；而使用较易的负样本集会因奖励浅层关联启发式策略而夸大迁移表现，相比之下，在更难的负样本集上性能显著下降，揭示出当前评估方法中的偏差。最后，研究证实源语言实体频率对迁移能力具有决定性影响，波斯相关事实在英语语料中本就稀少，导致其几乎无法迁移。
链接: https://arxiv.org/abs/2609.01341
作者: Romina Oji,Marc Braun,Marcel Bollmann,Marco Kuhlmann,Jenny Kunz
机构: Linköping University(林雪平大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Accepted at EMNLP 2026 Main Conference
Abstract:Do multilingual language models transfer factual knowledge across languages during continued pretraining, or do they mostly recall facts learned directly from the target-language data? To answer this question more reliably, we propose an intervention-based framework: starting from an English-pretrained model, we continue pretraining on Persian data from which specific facts have been systematically removed at varying levels of granularity. We construct SIFT, a resource of 500 triples across 20 relations, stratified by the cultural origin of each fact’s subject into general (globally prominent) and Persian-related entities, designed for both systematic fact removal from training data and evaluation, with natively written Persian cloze templates. Our results show that fact transfer is very limited: under the strictest removal condition, a large majority of English-acquired facts fail to transfer into Persian. We further show that sentence-level co-occurrence removal is insufficient to eliminate fact signal, and that easier (randomly selected) negative candidate sets substantially inflate apparent transfer by rewarding shallow associative heuristics, while performance on a harder candidate set that allows for less reliance on heuristics is much lower. Finally, we show that source-language entity frequency has a large influence, with Persian-related facts, which are orders of magnitude rarer in the English corpus, hardly transferring.
[NLP-29] Exploring Sparse Autoencoders in Text-Based Causal Confounding Adjustment EMNLP2026
【速读】： 该论文旨在解决在基于文本数据的因果推断中，如何有效调整文本内隐藏混杂因素（confounding information）的问题。核心挑战在于文本表示的构造需在保持足够丰富性以捕捉必要混杂变量（从而实现无偏效应估计）与保证表示足够稀疏以满足有限样本重叠性（finite-sample overlap）和降低估计方差之间取得平衡。为应对这一权衡，作者提出一种基于稀疏自编码器（Sparse Autoencoders, SAE）的新型因果调整流程，通过条件独立性检验迭代筛选出最小且最优的SAE特征子集，实现高效、可解释的混杂调整。实验结果表明，在包含二元混杂因子的标准半合成评估中，SAE表示相较于其他文本表示方法具有更低偏差和更高覆盖率；其良好的可解释性还支持进一步的反事实检验（falsification）。此外，作者引入了更贴近真实场景的半合成评估框架，采用多标签数据作为未观测混杂因子，发现现有现成调整方法在此类复杂情境下表现不足，亟需进一步研究。
链接: https://arxiv.org/abs/2609.01322
作者: Mian Zhong,Katherine A. Keith,Anjalie Field
机构: Johns Hopkins University (约翰霍普金斯大学); Williams College; Cohere
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: Long paper accepted at EMNLP 2026 main conference, 25 pages, 16 figures
Abstract:In many settings, studying causal questions based on text data requires adjusting for confounding information within texts. Yet there is a tradeoff in constructing text representations for adjustment: they must be sufficiently large and/or dense to preserve the confounding variables necessary for unbiased effect estimation, but sufficiently small and/or sparse to satisfy finite-sample overlap and yield low-variance estimates. To address this tradeoff, we turn to sparse autoencoders (SAEs), and propose a novel causal adjustment pipeline that iteratively selects a minimal set of SAE features via conditional independence tests. We find that SAE representations achieve better adjustments (lower bias and and higher coverage) than alternative representations in standard semi-synthetic evaluations with binary confounders, and their interpretability offers opportunities for falsification. We also introduce a more realistic semi-synthetic evaluation that uses multi-label data as the unobserved confounders and find off-the-shelf adjustment methods require increased investigation for these more complex settings. Code: this https URL
[NLP-30] Reliability Challenges in Diffusion Vision-Language Models EMNLP2026
【速读】： 该论文旨在解决生成式视觉-语言模型（Generative Vision-Language Models, LVLMs）中扩散模型（Diffusion-based LVLMs, dLVLMs）在幻觉（hallucination）与偏见（bias）等可靠性问题上的系统性评估缺失问题。尽管扩散模型相较于自回归（Autoregressive, AR）模型在并行解码、双向上下文建模和可控生成方面具有优势，但其在实际应用中的可靠性特性尚未得到充分研究。本文提出首个针对dLVLMs的系统性可靠性评估框架，对比六种扩散模型与竞争性自回归基线，在四个维度上进行评测。其解决方案的关键在于揭示扩散生成机制所引发的独特可靠性模式：首先，dLVLMs在二元视觉问答任务中逆转了AR模型的“是”偏向；其次，虽幻觉率与AR模型相当，但语言质量显著下降；再次，对代表性不足的种族群体表现严重准确率崩溃，且呈现与性别极性相反的偏见；最后，当正确选项长度短于干扰项时，模型准确率出现坍塌，这与扩散过程首步去噪阶段产生的长度先验（length prior）密切相关。此外，后期去噪步骤中低置信度提交的词元与幻觉内容高度相关，揭示了一种仅存在于扩散生成过程中的可解释性信号。这些现象因模型家族而异，表明可靠性不仅受训练数据影响，更由生成范式本身决定。
链接: https://arxiv.org/abs/2609.01318
作者: Md. Atabuzzaman,Chris Thomas
机构: Virginia Tech(弗吉尼亚理工学院)
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL)
备注: EMNLP 2026 main conference
Abstract:Diffusion-based Large Vision-Language Models (dLVLMs) have recently emerged as a compelling alternative to autoregressive (AR) LVLMs, offering advantages in parallel decoding, bidirectional context, and controllable generation. Despite rapid progress, their reliability properties remain largely uncharacterized. We present the first systematic reliability evaluation of hallucination and bias in dLVLMs, benchmarking six diffusion models against competitive AR baselines across four dimensions. Our key findings are: (1) dLVLMs reverse the yes-bias of AR models in binary visual queries; (2) they achieve competitive hallucination rates yet exhibit degraded linguistic quality; (3) they collapse to near-zero accuracy on underrepresented racial groups with opposite-polarity gender bias; and (4) they exhibit accuracy collapse in multiple-choice settings when the correct option is shorter than its distractors, associated with a length prior that emerges at the first denoising step. Tokens committed at late denoising steps with low confidence further correlate with hallucinated content, pointing to a mechanistic signal unique to diffusion generation. These patterns vary across model families, suggesting reliability is shaped by the generative paradigm together with training data.
[NLP-31] Explore Before Committing: Hypothesis-Guided Search for Deep Research Agents
【速读】： 该论文旨在解决深度研究型智能体（deep-research agents）在处理复杂问题时因采用单一演化轨迹而导致的决策偏差问题。其核心挑战在于：当智能体在早期搜索阶段面临多个合理方向时，若过早选定某一路径而缺乏充分的对比证据，后续的工具调用将不断强化该错误路径，从而增加最终失败的概率。针对这一问题，论文提出的关键解决方案是HypoSearch，其核心机制在于通过生成轻量级假设作为软搜索提示（soft search hints），在有限范围内并行探索多个独立分支，并在各分支间比较证据后再做出最终决策。该方法有效实现了对模糊探索的具象化引导以及在当前路径薄弱时及时转向，显著提升了搜索的鲁棒性与准确性。实验表明，HypoSearch在四个深度研究基准上均优于单轨迹搜索和标准并行基线，且在降低工具调用次数的同时实现性能提升，例如使Qwen3.5-122B在BC-small上的表现从46.7提升至60.0；此外，初步的监督微调研究也验证了这些行为信号可有效构建紧凑的训练轨迹，缓解未过滤数据带来的性能退化。
链接: https://arxiv.org/abs/2609.01294
作者: Ruochen Zhou,Zhengyu Chen,Luan Zhang,Siyang Gao,Yee Whye Teh,Shiqi Chen
机构: City University of Hong Kong(香港城市大学); Meituan(美团); Independent(独立研究者); University of Oxford(牛津大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Deep-research agents answer complex questions by interacting with search and browsing tools, yet they often search along a single evolving trajectory. Our trajectory-level analysis reveals a common failure mode in which the agent may encounter an early search state with several plausible directions, but follow one direction before collecting enough comparative evidence. Once this happens, subsequent tool calls tend to reinforce the same path, increasing the chance of failure when the initial direction is misleading. We further find that successful trajectories reduce this risk through two behaviors: grounding vague exploration in concrete candidates and shifting directions when the current path is weak or incomplete. Based on these findings, we propose HypoSearch, which generates lightweight hypotheses as soft search hints, explores them through bounded independent branches, and compares branch-level evidence before commitment. Across four deep-research benchmarks and three backbone models, HypoSearch consistently outperforms single-trajectory search and standard parallel baselines, improving Qwen3.5-122B from 46.7 to 60.0 on BC-small while using fewer tool calls than five independent trajectories. A pilot supervised fine-tuning study further shows that these behavioral signals can curate compact training trajectories and reduce degradation from unfiltered data.
[NLP-32] Some Emotions Run Deeper: Layer-wise Probing and Causal Intervention in Large Language Models EMNLP2026
【速读】： 该论文旨在解决大语言模型（LLM）中情感表征的深度分布问题，即情感信息在模型层级中的可访问性是否仅由模型架构决定，还是也受文本来源语境特征的影响。现有研究多基于单一语料库进行逐层分析，难以区分模型固有特性与数据特异性对情感表示位置的影响。本文通过跨三个不同情感表达显性程度的语料库（推特帖子、Reddit评论、自传体叙述）和八款1B–9B参数量的开源大模型（Llama、Qwen、Granite系列），系统考察了情感在模型内部的表征层级。其解决方案的关键在于：结合层间探针（layer-wise probing）、离线特征归一化、在线前向干预、迁移性分析及早退出分类器等多种方法，揭示出情感表征的最佳探针层在不同语料中系统性地从输入邻近层延伸至超过模型深度一半的位置，且该趋势在控制文本长度分布后依然成立；进一步发现，针对探针选定区域的前向干预导致测试准确率下降5–6个百分点，显著优于随机选取等宽区间（p < 0.01）；同时，所选特征带在不同数据集与情感类别间具有迁移能力，表明情感信息存在部分共享的表征结构而非严格对应特定情绪的独立子结构；最后，基于探针选择的早退出表示在平均性能上比全深度输出提升6.9个百分点，验证了高效情感表征的早期可获取性。该研究为理解情感在大模型中的内在组织机制提供了实证依据，并提出了一种基于动态探针引导的轻量化情感推理范式。
链接: https://arxiv.org/abs/2609.01279
作者: Tian Fang,Gaël Guibon,Davide Buscaldi
机构: Université Sorbonne Paris Nord, CNRS, Laboratoire d’Informatique de Paris Nord, LIPN; LORIA, CNRS, Université de Lorraine
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Accepted at EMNLP 2026 Findings
Abstract:Emotion is expressed in text along a wide spectrum, from surface lexical cues to inferences entangled with content. Most layer-wise analyses of emotion in LLMs use a single corpus, leaving open whether the depth at which emotion becomes accessible is a property of the model or also of the text source. We investigate this across three datasets spanning different degrees of explicitness and contextualization in emotion expression (Twitter posts, Reddit comments, and autobiographical narratives) and eight 1B–9B open-weight LLMs from the Llama, Qwen, and Granite families. We combine layer-wise probing with offline feature scaling and online forward interventions, transfer analyses, and an early-exit classifier. We find that (i) the best probing layer shifts systematically across corpora, from input-adjacent layers to over half model depth, and this ordering persists after matching label-by-length-bin distributions; (ii) across the evaluated settings, forward-pass interventions on probe-selected bands reduce test accuracy by 5–6 points more than same-width random bands ( q 0.01 ); (iii) selected bands transfer across datasets and emotion categories, suggesting partially shared affective information rather than strictly per-emotion substrates; and (iv) probe-selected early-exit representations outperform full-depth exits by 6.9 percentage points on average.
[NLP-33] From Base Rollouts to RL Reasoning : A Budgeted Search Perspective EMNLP2026
【速读】： 该论文旨在解决强化学习增强语言模型推理能力（Reinforcement Learning with Verifiable Rewards, RLVR）中一个核心问题：RL带来的性能提升究竟是源于生成了基线模型原本无法达到的新型推理路径，还是仅仅通过优化采样效率，使模型更有效地探索其已有能力范围内但低概率被采样的推理轨迹。其解决方案的关键在于提出统一解码框架（Unified Decoding Framework, UDF），将不同解码策略（如逐标记采样、束搜索、树搜索及序列级重采样）统一建模为在共享预算空间内可执行的策略，并通过pass@k、自洽性、best-of-N和首次完成成功率等指标进行后验评估。研究发现，在多个基准测试（Math500、AIME、GPQA、IFEval）上，基于基线模型的解码表现可通过一个“预算化操作点转移规则”（Budgeted Operating-Point Transition Rule, BOPTR）近似描述，即 $ N_\mathrm{Base} \approx \alpha N_\mathrm{RL}^\beta $，且指数参数受具体任务条件调节。实验表明，该规则在十种跨模型家族、四类未参与拟合的基准上均保持较高预测精度，即使在无RL检查点或完全无强化学习监督的情况下仍具鲁棒性。这些结果支持一种“内部化搜索”的解释：在当前实验设置下，大部分RL增益可归因于对基线模型已有推理能力的采样效率优化，而非引入全新的推理模式。作者强调，所观察到的缩放规律仅是对此特定方法与模型群体的行为描述，而非参数层面等价性的证据，并将UDF与BOPTR作为行为诊断工具使用。
链接: https://arxiv.org/abs/2609.01274
作者: Wenhe Sun,Cunxiang Wang,Zijun Yao,Yixin Cao
机构: 复旦大学( Fudan University); 中国科学院自动化研究所( Institute of Automation, Chinese Academy of Sciences); 腾讯(Tencent); 香港科技大学( The Hong Kong University of Science and Technology)
类目: Computation and Language (cs.CL)
备注: Accepted to Findings of EMNLP 2026
Abstract:Reinforcement learning with verifiable rewards (RLVR) improves language-model reasoning, but how these gains relate to inference-time decoding and search remains unclear. Does RL create reasoning the base model lacks, or shift the rollout distribution toward trajectories it can already reach but rarely samples? We study this behaviorally with a Unified Decoding Framework (UDF), which expresses token-level sampling, beam-like search, tree search, and sequence-level resampling as executable policies over a shared budgeted operating space, scored post hoc with pass@ k , self-consistency, best-of- N , and first-finish success. Using paired Base/RL checkpoints from SimpleRL-Zoo, we ask whether an RL default-policy curve can be approximated by a structured path of Base operating points. On Math500, AIME, GPQA, and IFEval, the pass@ k recovery path follows a Budgeted Operating-Point Transition Rule (BOPTR), N_\mathrmBase \approx \alpha N_\mathrmRL^\beta , with benchmark-conditioned exponents. On Qwen2.5-7B, BOPTR gives the lowest transfer error among the non-oracle rules we test, 3.41 pp (95% CI [2.32, 5.53]); a three-seed replication gives 3.07 \pm 0.39 pp. The rule extends to ten models across four families (3.28 to 4.87 pp on checkpoints added after fitting), to four benchmarks it was never fitted on (5.03 pp vs. 4.44 pp in fit), and holds without an RL checkpoint for the target model (4.19 pp) or without RL supervision of any kind (5.08 pp). These results support a qualified internalized-search reading: under the recipe we test, much of the measured RL gain corresponds to a change in sampling efficiency toward operating points the base model can already reach under search. We treat the scaling patterns as descriptive of this recipe and cohort, report where they break down, and use UDF and BOPTR as behavioral diagnostics rather than evidence of parameter-level equivalence.
[NLP-34] What Does an Agent ic Software Engineering Benchmark Measure? Profiling Task Demands and Agent Behaviour Beyond What Category Labels Reveal
【速读】： 该论文旨在解决当前生成式软件工程基准测试（benchmark）中标签语义模糊、无法准确反映任务实际工程需求的问题。现有基准多以“缺陷修复”或“功能实现”等名义类别标签进行归纳，但相同标签下的任务因数据构建流程（curation pipeline）差异显著，导致其真实任务需求存在本质区别。为此，作者提出一种基于实证软件工程研究的三轴分析框架——传播性-新颖性-中心性（Spread–Novelty–Centrality, SNC）剖面，用于量化代码库层级任务的实际工程要求。其核心解决方案在于：通过SNC剖面揭示不同基准任务在代码传播范围、创新程度及核心模块依赖性上的真实差异，并发现标签作为任务需求的代理指标不可靠；同时，模型行为（agent behavior）能暴露人类标注答案所未体现的任务隐含要求，例如任务表述方式直接影响生成代码规模；此外，任务需求与成功概率呈一致相关性，高成功率运行集中于低SNC区域，但成功的行为特征具有模型家族特异性——Claude通过精准匹配黄金标准范围实现最优表现，而Qwen则需超越黄金标准范围才可成功，且编辑不足是两类模型共同失败的关键信号。
链接: https://arxiv.org/abs/2609.01271
作者: Radin Shayanfar,Keheliya Gallaba,Ahmed E. Hassan
机构: 未知
类目: oftware Engineering (cs.SE); Computation and Language (cs.CL)
备注:
Abstract:Agentic software engineering benchmarks are typically summarized by nominal category labels such as “bug fix” or “feature implementation,” yet benchmarks carrying the same label are built through very different curation pipelines. A label thus reveals little about the engineering work a benchmark demands. We introduce the Spread–Novelty–Centrality (SNC) profile, a three-axis characterization of the demands of repository-level coding tasks, grounded in empirical software engineering research. We apply the profile to five widely used benchmarks and 14,922 trajectories of two model families at three scales, and report three findings. (1) A label is an unreliable proxy for task demands, as every pair of benchmarks is statistically separated on at least two SNC axes, and the separations trace back to specific curation decisions. (2) Agent behaviour reveals demands that the human-written gold solution cannot. Agents produce larger solutions than the gold where problem statements withhold hints and smaller ones where curation inflates the gold. How a task is phrased shapes what an agent produces. (3) Task demands correlate with success uniformly, with resolved runs concentrating in the low-SNC region for every family and scale, whereas the behavioural signatures of success are family-specific. Claude succeeds by matching the scope of the gold solution, and its parity share on files rises from 0.17 at the smallest scale to 0.54 at the largest. Qwen succeeds by exceeding the gold scope at every scale, and editing too little marks failure for both families.
[NLP-35] Ready to Speak: Aligning LLM s for TTS-Friendly Text Generation EMNLP2026
【速读】： 该论文旨在解决当前大型语言模型（LLM）生成的文本虽在语法和语义上表现良好，但缺乏适合语音合成（Text-to-Speech, TTS）系统自然朗读的问题。传统方法依赖下游重写模块对输出进行适配，存在额外开销与失真风险。本文提出将生成TTS友好文本视为一种偏好对齐（preference alignment）问题，直接通过训练使LLM原生生成更适合语音播放的表达形式。其核心解决方案是引入两个跨领域的偏好数据集——CORA与Recipe，分别涵盖对话与食谱场景中成对的TTS友好与非友好响应，并构建包含基于模式的启发式指标、TTS→ASR评估流程以及MUSHRA听觉评测的人类判别体系，以多维度评估生成质量。实验表明，采用可解释特征驱动的特征感知采样与调优（FaST）框架，在不使用黑箱奖励模型的前提下，显著优于多种对齐基线方法，实现了在TTS友好性与有用性之间的最优权衡。此外，研究发现不同评估指标间存在强相关性，验证了高效启发式方法在可靠评估TTS友好性方面的可行性。
链接: https://arxiv.org/abs/2609.01246
作者: Thibaut Thonet,Jos Rozen,Laurent Besacier
机构: 未知
类目: Computation and Language (cs.CL)
备注: EMNLP 2026 - Main Conference
Abstract:Current Large Language Models (LLMs) are primarily optimized for written text, often producing outputs that are grammatically correct and helpful yet poorly suited for spoken delivery via Text-to-Speech (TTS). In this work, we study how to make LLMs natively generate TTS-friendly text, which we frame as a preference alignment problem: instead of relying on downstream rewriting modules, we directly align LLMs to generate text optimized for spoken delivery. We introduce two preference datasets spanning different target domains, CORA and Recipe, which contain paired TTS-friendly and TTS-unfriendly responses. We further propose an evaluation suite combining a pattern-based heuristic metric, a TTS \to ASR evaluation pipeline, and a MUSHRA listening study with human judges. Our experiments compare the recently proposed Feature-aware Sampling and Tuning (FaST) framework – leveraging interpretable features instead of a black-box reward model – against an array of alignment baselines on the TTS-friendly generation task. Notably, we found that FaST achieves the best overall tradeoff between TTS-friendliness and helpfulness across various settings. We also identified a strong correlation between our different metrics, highlighting the ability to reliably assess TTS-friendliness via an efficient heuristic.
[NLP-36] Post-Training Science for Supervised Fine-Tuning
【速读】： 该论文旨在解决监督微调（Supervised Fine-Tuning, SFT）过程中超参数与训练策略选择高度依赖经验、缺乏系统性指导的问题。当前实践中，学习率、批量大小、LoRA与全量微调的选择、训练轮次、优化器类型及数据配置等关键决策通常需针对每个新模型和数据集重新摸索，导致效率低下且可复现性差。其核心解决方案是通过一个统一的实验框架——单变量扫描（sweep that varies one lever at a time），在多个维度上系统评估这些因素的影响：涵盖两种模型家族（Qwen3 与 Llama）、密集模型与混合专家（Mixture-of-Experts, MoE）架构、四组真实客户SFT数据集，并覆盖LoRA与全量微调两种范式。该实验设计以客户定义的评估标准作为训练目标，确保训练数据与评价指标内在一致，从而提供可控、可验证的基准测试环境。研究重点揭示了最优学习率与批量大小随模型规模、模型家族及数据特性变化的规律，检验了超参数选择规则的跨域迁移能力；量化分析了LoRA相较于全量微调的权衡关系及其适配器秩（rank）与缩放因子（alpha）对学习能力的影响；评估了验证损失及其他指标（如损失曲面平坦度）对下游性能的预测可靠性；探究了后训练收益随模型规模与数据量的增长规律，扩展至2350亿参数的MoE模型；明确了指令遵循能力退化的训练轮次阈值，并对比了几何感知优化器与AdamW的性能差异。所有结论均附带不确定性度量，为实际应用提供可信赖的决策依据。
链接: https://arxiv.org/abs/2609.01244
作者: Charles O’Neill,Mudith Jayasekara,Harry Partridge
机构: 未知
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注:
Abstract:Every supervised fine-tuning run forces the same chain of decisions, such as learning rate, batch size, LoRA or full fine-tuning, how many epochs, which optimiser, and what data to feed the model. Each of these is typically rediscovered from scratch for every new model and dataset. Here we measure them under one instrument: a sweep that varies one lever at a time, and spans dense and mixture-of-experts models in two families (Qwen3 and Llama), on four real-world customer SFT datasets, for both LoRA and full fine-tuning. These datasets give a controlled testbed: each task carries an evaluation built with the customer, and its training data is produced by iterative supervised fine-tuning that refines model outputs until they pass that evaluation, so the supervised target is internally consistent and the task judge we report against is the criterion the data was built to satisfy. We ask how the optimal learning rate and batch size move with model scale, family, and data, and whether one selection rule transfers across them; what LoRA trades against full fine-tuning, and how its rank and alpha set what the adapter can learn; whether validation loss (or other metrics, such as loss landscape flatness) faithfully ranks downstream quality; whether post-training gains scale with model size and data volume, on a model ladder extended through mixtures-of-experts to 235B parameters; how many epochs to train before general instruction-following erodes; and whether a geometry-aware optimiser improves on AdamW. Each recommendation is paired with a measure of its uncertainty.
[NLP-37] owards AI-Assisted Clinical Trial Matching: Practical Considerations Multicenter Evaluation and Real-World Deployment
【速读】： 该论文旨在解决癌症临床试验中因患者入组不足而导致的高失败率问题，核心挑战在于现有AI系统仅聚焦于患者的资格筛选，缺乏在真实临床工作流中的有效性验证。其解决方案的关键在于提出TrialGPT 2.0——一个面向真实世界部署的AI辅助临床试验推荐系统，不仅评估患者是否符合某项试验的准入标准，更进一步结合患者的当前临床需求与本地工作流程优先级，智能识别值得深入考虑的试验，并提供结构化、可审查的解释以支持专家决策。研究通过回顾性与前瞻性多中心评估（涵盖政府、学术型肿瘤中心、患者倡导组织及NIH转诊流程），证明该系统在288例病例中使91%的案例在前10项推荐中包含临床医生推荐的试验，同时将临床医生的筛选时间减少55.0%；在为期六个月的精准肿瘤学多学科会诊（tumor board）前瞻性应用中，成功发现常规流程遗漏的试验机会，使患者参与临床试验的可能性提升90.9%。为保障科学可复现性，研究还发布了由临床医生构建的NIH-TrialBench数据集，包含126个多样化合成患者案例及其对应场景。上述结果表明，生成式AI（Generative AI）可通过提升临床效率并揭示被忽视的试验机会，有效助力癌症临床试验的加速入组与扩展。
链接: https://arxiv.org/abs/2609.01202
作者: Yin Fang,Qiao Jin,Shubo Tian,Lauren He,Maya Geer,Noor Naffakh,Ryan Huu-Tuan Nguyen,Zifeng Wang,Jimeng Sun,Charalampos S. Floudas,James L. Gulley,Kamilia Moalem,Catarina Martins Maia,Amanda Nottke,Juan W. Valle,Melinda Bachini,Lourdes Rocha-Nussbaum,Kari Ramage,Nikita Curry,Megan Barnes,Mandy Mansaray,Darlene Gabeau,Craig E. Grossman,Heath Skinner,Michael Burczynski,NIH-TrialBench Consortium,Zhiyong Lu
机构: National Library of Medicine, National Institutes of Health (美国国立卫生研究院国家医学图书馆); University of Illinois Chicago (伊利诺伊大学芝加哥分校); University of Illinois Urbana-Champaign (伊利诺伊大学厄本那-香槟分校); National Cancer Institute, National Institutes of Health (美国国立卫生研究院国家癌症研究所); The Cholangiocarcinoma Foundation (胆管癌基金会); Division of Cancer Sciences, University of Manchester (曼彻斯特大学癌症科学系); Office of Patient Recruitment, NIH Clinical Center (美国国立卫生研究院临床中心患者招募办公室); UPMC Hillman Cancer Center, University of Pittsburgh (匹兹堡大学UPMC希尔曼癌症中心); NIH-TrialBench Consortium (NIH试验基准联盟)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Computers and Society (cs.CY)
备注: 43 pages, 12 figures
Abstract:Clinical trials are essential for advancing cancer care and drug development, but many fail because of insufficient patient enrollment. While there is growing interest in using AI to support patient recruitment, existing systems largely perform eligibility assessment alone and have rarely been evaluated in real-world oncology workflows. Here we present TrialGPT 2.0, an AI-assisted clinical trial recommendation system designed for real-world deployment. Rather than asking only whether a patient may qualify, the system also assesses which trials warrant further consideration given the patient’s current clinical needs and local workflow priorities, and provides structured, inspectable explanations for expert review. Importantly, we evaluated TrialGPT 2.0 retrospectively and prospectively across multiple oncology-focused settings, spanning government, academic cancer-center, patient-advocacy, and NIH referral workflows. In retrospective multicenter cohorts comprising 288 cases, TrialGPT 2.0 retrieved at least one clinician-recommended trial in its top 10 recommendations for approximately 91% of cases while reducing clinician screening time by 55.0%. In a six-month prospective evaluation embedded in an active precision oncology tumor board, TrialGPT 2.0 contributed additional trial opportunities missed by the routine workflow, expanding patient access to clinical trial participation by 90.9%. To support scientific reproducibility, we also introduce NIH-TrialBench, a clinician-authored dataset comprising 126 diverse synthetic patient vignettes and matching scenarios from 11 NIH Institutes and Centers. Together, these results support the value of AI to assist clinical trial matching by improving clinician efficiency and identifying frequently overlooked trial opportunities, ultimately helping to expand and accelerate accrual to cancer trials.
[NLP-38] FinLifeBench: Exhaustive Life-Event History and Financial-State Reconstruction from Longitudinal Banking Dialogue
【速读】： 该论文旨在解决生成式人工智能（Generative AI）在长期银行交互场景中难以维持完整、实时且可追溯的客户生命周期记录的问题。现有基准测试多聚焦于问答任务、有限对话轮次或特定信息召回，而忽视了对跨时间跨度的全周期事件重建与金融状态演进的系统评估。其解决方案的关键在于提出FinLifeBench这一新型基准，通过同一累积对话流中的两个核心任务：一是重构每个生活事件实例及其首次建立会话的锚点，二是重建连续时间点上的34条金融状态路径。该基准包含6,000个八轮韩语银行会话，源自20个独立合成轨迹，提供确定性且全面的黄金标准（gold standard），涵盖24类生活事件和34条状态路径，并采用共识质量保障机制。实验表明，在全上下文条件下，11个大语言模型（LLM）在事件锚定召回率上从15个会话时的0.591下降至300个会话时的0.445，主要错误源于事件遗漏而非锚点定位偏差；同时，金融状态重建常将过时或已被覆盖的信息误判为当前状态，最佳模型在15个会话时的全局一致性准确率（GCA@15）仅为0.470。两项重建任务之间的性能关联性较弱，揭示出模型虽具备识别证据的能力，却仍无法有效维护完整且时间一致的纵向记录，凸显了长时序记忆与状态演化建模的深层挑战。
链接: https://arxiv.org/abs/2609.01198
作者: Hangyeul Lee,Juyoung Oh,Jaeyong Ko,Sunmin Kim,Jaeik Park,Hyunkyu Kim,Jungmin Son,Pilsung Kang
机构: Seoul National University (首尔国立大学); KakaoBank, Financial Tech Lab (Kakao银行，金融科技实验室)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 9 pages, 3 figures, 3 tables
Abstract:Repeated banking interactions require assistants to maintain complete, current, and traceable customer records as life changes emerge incidentally in routine requests. Existing benchmarks emphasize question answering, bounded episodes, or targeted recall rather than exhaustive longitudinal reconstruction. We introduce FinLifeBench, which evaluates two tasks over the same cumulative dialogue: reconstructing every life-event instance with its first-establishing session and reconstructing a complete 34-path financial state at consecutive checkpoints. The benchmark contains 6,000 eight-turn Korean banking sessions from 20 independent synthetic trajectories, with deterministic, exhaustive gold for 24 event types and 34 state paths and consensus quality assurance. Across eleven LLMs under a full-context condition, event-anchor recall falls from 0.591 at 15 sessions to 0.445 at 300. Errors are driven primarily by omitted events rather than poor anchor localization, while financial-state reconstruction frequently treats superseded or potentially outdated information as current; the best GCA@15 reaches 0.470. Performance on the two reconstruction tasks is only weakly associated. These results show that models can localize evidence for recovered events while still failing to maintain complete and temporally valid longitudinal records.
[NLP-39] CaRL-EM: Cost-Aware Reinforcement Learning for Entity Matching with LLM s ACL2026
【速读】： 该论文旨在解决生成式 AI（Generative AI）在实体匹配（Entity Matching, EM）任务中面临的两大核心问题：一是现有方法在多候选场景下缺乏灵活性，通常采用独立的成对决策或依赖人工设计的复合流水线，难以适应复杂的真实场景；二是忽略了大规模推理时的计算成本，导致效率低下。其解决方案的关键在于将基于大语言模型（LLM）的实体匹配建模为一个成本感知的序列决策问题，并提出一种基于强化学习的控制器 CaRL-EM。该控制器能够根据锚记录的状态、候选集及当前成本动态选择操作符（如匹配、比较、选择、决策）和模型能力，以最大化质量与成本之间的权衡目标。通过与抽象操作符交互，CaRL-EM 实现了无需重训练即可适配不同底层 LLM 后端的可复用性，显著提升了系统在多样化数据集和领域上的零样本迁移能力，并在保持或提升匹配质量的同时，显著降低了推理开销，展现出更优的质量-成本平衡。
链接: https://arxiv.org/abs/2609.01195
作者: Chaohui Guo,Michel Klein,Zhisheng Huang
机构: Vrije Universiteit Amsterdam (阿姆斯特丹自由大学)
类目: Computation and Language (cs.CL)
备注: Accepted to ACL 2026 Main Conference
Abstract:Entity matching (EM) requires fine-grained contextual understanding and domain knowledge. Recent work shows that large language models (LLMs) can serve as strong matchers across domains, but most methods either make independent pairwise decisions or rely on manually designed composite pipelines, thus lacking flexibility in realistic multi-candidate settings. At the same time, they typically ignore inference cost at scale. We formulate LLM-based EM with candidates as a cost-aware sequential decision problem and propose CaRL-EM, a reinforcement learning controller that manages LLM operations. Given the state of an anchor record, its candidate set, and the cost, CaRL-EM adaptively chooses among different operators (Match/Compare/Select/Decide) and model capacities to maximize a quality-cost objective. The policy interacts with abstract operators, allowing the same controller to be reused with different underlying LLM backends at inference time without retraining. Experiments on 7 benchmarks show that CaRL-EM (i) learns to dynamically plan the usage of inexpensive and expensive operators based on task complexity, (ii) achieves robust zero-shot transfer across diverse datasets and domains, and (iii) consistently achieves a better quality-cost trade-off than strong LLM-based baselines and manually designed pipelines, yielding a lower inference cost at comparable or higher quality.
[NLP-40] PersuaRL: Reinforcement Learning-Driven Multi-Expert Selection for Persuasive Dialogue Generation in Insurance EMNLP
【速读】： 该论文旨在解决生成式对话系统在保险领域中缺乏有效说服力的问题，尤其是在需要建立信任与清晰沟通的场景下，现有大语言模型（Large Language Models, LLMs）虽能完成事实性对话，却难以实现情境敏感且具有说服力的交互。其核心解决方案是提出PersuaRL框架，该框架基于强化学习（Reinforcement Learning），使LLM驱动的对话代理能够根据动态对话上下文，自适应地探索、选择并协调多个专家模块中的策略，从而提升对话的说服效果。为支持该方法，研究还构建了InsureDial——一个聚焦于汽车保险领域的说服性对话数据集，以捕捉该领域特有的沟通细微差别。实验结果表明，PersuaRL在多个基准说服性对话数据集（包括InsureDial）上的自动评估与人工评估中均显著优于基线模型，生成了更具情境适配性和说服力的回应。
链接: https://arxiv.org/abs/2609.01188
作者: Rohan Kirti,Akash Ghosh,Aryan Vats,Niladri Ghosh,Shipra Shriparn,Roshni Ramnani,Anutosh Maitra,Sriparna Saha
机构: Indian Institute of Technology Patna(印度理工学院帕特纳分校); Ramakrishna Mission Vivekananda Educational and Research Institute(拉玛克里希纳使命维韦卡南达教育与研究学院); Accenture Labs(埃森哲实验室)
类目: Computation and Language (cs.CL)
备注: EMNLP Findings 2026
Abstract:Large Language Models (LLMs) are revolutionizing digital communication by powering conversational agents deployed across domains such as customer service, digital sales, and insurance. These agents, built on LLMs, can understand user input, retrieve relevant information, and generate coherent responses. However, while they excel at factual communication, they often lack the ability to engage in truly persuasive, context-sensitive dialogue, especially in domains like insurance, where trust and clarity are critical. Building on this need within the insurance domain, our work focuses on improving the persuasiveness of digital agents, aka LLMs. To support this, we introduce InsureDial, a Persuasive Insurance Dialogue dataset, designed to capture the nuances of persuasive communication specific to motor insurance interactions. We introduce PersuaRL, a reinforcement learning-based framework that equips LLM-driven dialogue agents with the ability to adaptively explore, select, and coordinate strategies across multiple expert modules, guided by the evolving dialogue context, to achieve more effective persuasion. We conduct extensive automatic human and qualitative evaluations on two benchmark persuasion dialogue datasets, including our InsureDial. Our evaluations consistently demonstrate that PersuaRL outperforms baseline, generating contextually appropriate and highly persuasive responses.
[NLP-41] LLM PEDIA: Browsing Verifying and Comparing the Parametric Encyclopedic Knowledge of LLM s
【速读】： 该论文旨在解决大模型在标准评测基准（如MMLU）上表现趋近饱和所隐含的“可及性偏差”（availability bias）问题，即评测仅覆盖实验者预先设定的问题集，无法全面反映模型真实知识能力。其核心解决方案是构建LLMPEDIA——一个基于生成式AI（Generative AI）参数化记忆的动态、可浏览的知识库，通过递归生成约130万条来自GPT-5-mini、DeepSeek-V3.2和Llama-3.3-70B三类模型的非检索式文章内容，并对分层抽样的原子性主张进行系统性审计，依据维基百科与精选网络资源判定其支持、反驳或证据不足状态。结果显示，在随机样本中真实率为68.4%，显著低于MMLU的90%以上水平，且30.5%的主张因缺乏足够证据而无法判断，揭示了长尾知识缺失与潜在幻觉的存在，扩展了此前GPTKB在三元组层面发现的覆盖缺口至自由文本领域。该系统以五个一键式视图（链路遍历探索、主张级事实性评估、跨模型与政治人格对比、主题引导钻取）实现对每一条主张及其结论的透明化呈现，所有页面、主张与裁决均具有稳定URL，形成一个实时开放的百科全书，使用户能够逐条检视大模型知识边界。
链接: https://arxiv.org/abs/2609.01182
作者: Muhammed Saeed,Simon Razniewski
机构: ScaDS.AI Dresden/Leipzig; TU Dresden (德累斯顿工业大学), Germany
类目: Computation and Language (cs.CL)
备注:
Abstract:Flagship language models appear saturated on benchmarks like MMLU (Hendrycks et al., 2021), scoring above 90% - yet benchmarks test only what the experimenter thought to ask, the availability bias of fixed question sets. LLMPEDIA makes this bias measurable and browsable. We recursively materialized ~1.3M articles from three model families’ parametric memory (GPT-5-mini, DeepSeek-V3.2, Llama-3.3-70B) without retrieval, then audited a stratified sample of atomic claims against Wikipedia and a curated web stack, coloring every claim supported, refuted, or insufficient (Saeed and Razniewski, 2026). On a uniform random sample the true rate is 68.4% - more than 21 pp below MMLU - with 30.5% of claims insufficient: assertions no benchmark probes and the world’s largest encyclopedia cannot adjudicate - long-tail knowledge or plausible hallucination, the evidence cannot tell - extending to free text the coverage gap GPTKB established for triples (Hu et al., 2025). The resulting live, open encyclopedia lets visitors inspect this frontier one claim at a time through five one-click views - link-traversal exploration, claim-level factuality, cross-model and political-persona comparison, and a guided topic drill-down - each page, claim, and verdict at a stable URL. LLMPEDIA is live at this https URL
[NLP-42] Subword Segmental BabyLMs: Learning to Tokenise for Sample-Efficient Pretraining
【速读】： 该论文旨在解决标准语言模型（Language Model, LM）训练流程中依赖预定义子词分词（subword tokenisation）所带来的局限性。传统方法使用固定的分词策略（如Byte Pair Encoding, BPE），可能无法充分适应特定语言或任务的语义结构，导致模型在低资源场景下的样本效率低下。为此，本文提出一种可学习的子词分段建模（subword segmental language modelling）范式，将分词过程作为训练的一部分进行端到端优化，使模型能够自主发现更优的子词单元以提升训练目标性能。其核心解决方案在于设计两种新型可学习分词的语言模型：SubSegGPT（基于解码器的自回归模型）与SubSegDeBERTa（基于编码器的掩码生成与分词联合模型），二者均在预训练阶段同时学习文本表示与子词划分策略。实验表明，在2026年BabyLM挑战赛的Strict和Strict-small赛道上，所提出的模型显著提升了零样本评估表现及样本效率；进一步分析揭示，模型在训练过程中逐渐收敛至兼顾形态学对齐与细粒度分割的子词单元。这证明了可学习子词分词机制在降低对预设分词依赖、增强模型泛化能力方面的有效性。
链接: https://arxiv.org/abs/2609.01151
作者: Francois Meyer
机构: University of Cape Town (开普敦大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:In the standard LM training pipeline, subword tokenisation is applied as a preprocessing step. Subword segmental language modelling is an alternative paradigm in which tokenisation is learned during training, allowing the model to discover subword units that optimise its training objective. In this paper, we present our submission to the 2026 BabyLM Challenge, for which we develop two new subword segmental LMs: SubSegGPT and SubSegDeBERTa. SubSegGPT is a decoder-only model that learns tokenisation during autoregressive pretraining. SubSegDeBERTa is an encoder-based model that jointly learns to generate and tokenise masked words. We train both for the Strict and Strict-small tracks. Our top submission to Strict is SubSegDeBERTa, which achieves notable gains in zero-shot evaluation. Our top submission to Strict-small is SubSegGPT, which outperforms tokenisation-based baselines. Our results show that learnable subword tokenisation can improve sample-efficiency for BabyLM pretraining. We analyse the subword learning dynamics of our models and find that tokenisation gradually converges on subword units that balance morphological alignment and fine-grained segmentation.
[NLP-43] On the Design Fundamentals of Pixel Text Representation Learning EMNLP2026
【速读】： 该论文旨在解决现有像素-文本编码器在高分辨率文档泛化、视觉-文本对齐、多语言理解以及避免视觉捷径学习等方面的局限性问题。其核心挑战在于如何在像素空间中实现鲁棒的视觉文本表征学习，尤其是在面对复杂布局、多语言内容和低资源压缩场景时的表现不足。解决方案的关键在于提出并验证四个关键设计原则：（1）采用可变图像分辨率与渲染字体大小作为高分辨率文档泛化的空间代理；（2）依赖自然图像-文本对进行语义锚定，防止仅文本坍缩；（3）引入布局感知的文本渲染机制以抑制像素级捷径学习；（4）通过两阶段多语言课程学习实现有效的跨语言对齐。基于这些原则，研究构建了可扩展的训练范式，训练出Pixel Linguist II——一个原生分辨率视觉编码器，支持实时渲染、统一对比式文本接地及大规模多语言训练（共2.8亿样本）。该模型在英文、跨语言和多语言视觉语义相似度（Visual STS）及ViDoRe基准上均达到新最优性能，并显著提升多模态大模型（MLLM）下游任务表现；尤其值得注意的是，其在80%视觉令牌压缩下仍保持鲁棒性，展现出卓越的光学上下文压缩潜力。
链接: https://arxiv.org/abs/2609.01147
作者: Chaohao Yuan,Ruifeng Yuan,Zhuoxu Huang,Yu Rong,Hong Cheng,Hou Pong Chan,Chenghao Xiao
机构: The Chinese University of Hong Kong; DAMO Academy, Alibaba Group; Fudan University; Aberystwyth University; Hupan Lab; University of Macau; Shanghai University of Finance and Economics
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL)
备注: EMNLP 2026
Abstract:Text-rich visual inputs require models that can read, retrieve, and compress language directly in pixel space, yet existing pixel-text encoders struggle with fixed resolution pretraining, visual shortcut learning, weak visual grounding, and multilingual visual text understanding. In this work, we investigate the fundamental design principles required for robust visual text representation learning. Through systematic controlled ablations, we identify four critical components: variable image resolutions and rendered font sizes provide spatial proxies for high-resolution document generalization; natural image-text pairs are indispensable for grounding and prevent text-only collapse; layout-aware rendering helps prevent pixel-level shortcuts; and a two-stage multilingual curriculum enables effective cross-lingual alignment. By integrating these principles into a scalable training recipe, we train Pixel Linguist II, a native-resolution vision encoder trained with on-the-fly rendering, unified contrastive grounding, and a multilingual curriculum over 280M training examples. Pixel Linguist II sets new state-of-the-art results on English, cross-lingual, and multilingual Visual STS and ViDoRe, while also enabling better MLLM downstream evaluation. Notably, Pixel Linguist II remains robust under 80% visual token compression, showing great promise for optical context compression. Our code and resources are available at this https URL.
[NLP-44] Does task decomposition improve automatic NLG evaluation? EMNLP2026
【速读】： 该论文旨在解决生成式 AI（Generative AI）在自然语言生成（Natural Language Generation, NLG）评估中缺乏低成本、可复现且无需参考文本的评价方法的问题。当前主流的“大语言模型作为裁判”（LLM-as-a-judge, LLMaJ）框架虽具潜力，但其性能提升常依赖于将复杂评估任务分解为多个子任务（task decomposition），而这一策略的有效性尚未得到充分验证。本文的关键发现是：在多个NLG数据集上的系统性对比表明，采用任务分解的LLMaJ方法并未显著优于不使用分解的公平基线；此前报道的性能提升实际上源于使用人工标注作为训练数据，而非任务分解本身所带来的增益。此外，当存在人工标注时，无需任务分解的LLMaJ方法可达到与人类标注者相当的评估性能。因此，该研究的核心结论是——任务分解并非提升LLMaJ性能的关键因素，而高质量的人工标签才是决定评估准确性的关键。
链接: https://arxiv.org/abs/2609.01139
作者: Sebastian Steindl,Nikos Voskarides,Alberto Gasparin,Diego Marcheggiani
机构: Amazon(亚马逊); Amazon(亚马逊); Amazon(亚马逊); Amazon(亚马逊)
类目: Computation and Language (cs.CL)
备注: Accepted to EMNLP 2026
Abstract:The LLM-as-a-judge (LLMaJ) framework has emerged as a promising solution for cheap, reproducible, reference-free Natural Language Generation (NLG) evaluation. Prior work seeks to improve LLMaJ by decomposing evaluation tasks into simpler sub-tasks. In this work, we systematically compare LLMaJ methods with and without decomposition on multiple NLG datasets. We find no evidence that LLMaJ with task decomposition leads to performance gains over a fair baseline that does not use decomposition. Instead, we find that previously reported performance gains in decomposition-based LLMaJ stem from using human labels as training data, and not task decomposition itself. Also, we find that, when human labels are available, LLMaJ without using task decomposition can perform comparably to human annotators.
[NLP-45] Overfitting Mitigation via Singular Value Decomposition in Minimum Bayes Risk Decoding EMNLP2026
【速读】： 该论文旨在解决最小贝叶斯风险（Minimum Bayes Risk, MBR）解码在文本生成过程中易受评估指标过拟合的问题。具体而言，传统MBR解码通过优化单一评估指标来选择最优假设，但这一过程往往导致其他未优化的评估指标性能下降，即存在“指标过拟合”现象。为缓解此问题，论文提出SVD-MBR方法，其核心在于将成对的效用矩阵视为含噪信息信号，并利用奇异值分解（Singular Value Decomposition, SVD）进行低秩近似，仅保留前k个主成分，从而有效分离出真实共识信号与评估指标噪声。实验表明，SVD-MBR能够显著提升多维度通用评估指标的表现，实现更稳健的解码。此外，研究发现该去噪效果具有指标依赖性：神经类评估指标（如基于深度语义的指标）具备较强的低秩一致性，适合通过SVD进行信号提取；而表面特征类指标则难以区分信号与噪声，去噪效果有限。
链接: https://arxiv.org/abs/2609.01135
作者: Riza Setiawan Soetedjo,Yusuke Sakai,Hidetaka Kamigaito,Katsuhiko Hayashi,Taro Watanabe
机构: Nara Institute of Science and Technology (奈良尖端科学技术大学院大学); The University of Tokyo (东京大学)
类目: Computation and Language (cs.CL)
备注: Accepted to EMNLP 2026 Main
Abstract:Minimum Bayes Risk (MBR) decoding enables high-quality text generation by selecting the hypothesis that maximizes a utility metric over sampled pseudo-references. However, it is highly susceptible to metric overfitting: it can irregularly inflate the chosen utility metric at the direct expense of other unoptimized evaluation metrics. To mitigate this, we introduce SVD-MBR, which frames the pairwise utility matrix as a noisy information signal. By computing a low-rank approximation via Singular Value Decomposition (SVD) and retaining only the top- k components, we effectively decouple true consensus from metric noise. Experiments demonstrate that SVD-MBR successfully regularizes decoding, yielding substantial gains across a range of generalized metrics. Furthermore, we reveal that this denoising is metric-dependent: neural metrics encode a robust low-rank consensus ideal for SVD, whereas surface-level metrics struggle to separate signal from metric noise.
[NLP-46] Latent Recurrent Thoughts: Recurrent Refinement of Proposed Latents for Reasoning with Frozen LLM s
【速读】： 该论文旨在解决传统链式思维（Chain-of-thought, CoT）推理在离散标记空间中固有的局限性：每一步推理以文本形式固化，错误易传播，且高质量推理轨迹的生成依赖于可模仿的示范轨迹。为此，论文提出一种在模型连续表示空间中进行推理的新范式——通过在连续潜在空间中执行迭代推理，避免了离散文本生成带来的误差累积与轨迹依赖问题。其解决方案的关键在于双轴设计：一是保持大型语言模型（LLM）冻结，仅利用其在序列建模与解码方面的优势；二是引入一个小型循环推理网络（recurrent reasoner），通过多步有界残差修正对连续潜在状态进行迭代优化，从而将计算深度与模型规模解耦，使潜在状态成为多次迭代处理的结果而非单次前向传播的产物。该方法被具体实现为潜变量循环思维（Latent Recurrent Thoughts, LRT），由任务专用提议器生成初始潜在状态，循环推理器逐步优化，最终由冻结的LLM解码答案。实验表明，在无推理轨迹监督但有答案监督的任务（如Countdown-4、Sudoku）以及自然语言推理任务（HumanEval、MBPP、StrategyQA）上，LRT在相同解码器、提示、数据和训练预算下显著优于现有冻结解码器的连续空间推理方法，并在相同骨干模型上以极小的推理计算开销超越非思维模式的链式思维提示。
链接: https://arxiv.org/abs/2609.01117
作者: Zhaoliang Chen,Jie Fu
机构: Emory University(埃默里大学); IQuest Research
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Chain-of-thought reasoning unfolds in discrete token space: each step is committed as text, errors propagate, and eliciting good traces presupposes traces to imitate. Reasoning instead in a model’s continuous representation space - where intermediate states are vectors rather than words - sidesteps these constraints, but leaves open how those latent states should be computed. We approach this along two axes. First, we keep a large language model (LLM) frozen and use it for what it is already good at - modeling and decoding sequences - while a small auxiliary network supplies continuous latent thoughts as input. Second, we produce those latents by recurrence: a tiny recurrent reasoner refines them over many steps, decoupling the depth of computation from the size of the model, so that the latents are a product of iterative processing rather than a single forward pass. We instantiate this as Latent Recurrent Thoughts (LRT): a task-dedicated proposer supplies base latents, a recurrent reasoner refines them through bounded residual corrections, and the frozen LLM decodes the answer. On symbolic reasoning with answer supervision but no reasoning traces (Countdown-4, Sudoku) and on natural-language reasoning (HumanEval, MBPP, StrategyQA), LRT substantially outperforms prior frozen-decoder continuous-space reasoning methods under an identical decoder, prompt, data, and training budget, and outperforms non-thinking-mode chain-of-thought prompting on the same backbone at a small fraction of its inference compute.
[NLP-47] EDRAC: Benchmarking Arabic Dialect Reading Comprehension
【速读】： 该论文旨在解决阿拉伯语方言（Dialectal Arabic, DA）在机器阅读理解（MRC）与生成式问答（QA）任务中严重缺乏高质量、大规模标注数据的问题。相较于现代标准阿拉伯语（Modern Standard Arabic, MSA），现有阿拉伯语问答基准大多聚焦于正式书面体的MSA或选择题形式，对自然口语化方言的覆盖极为有限。为此，论文提出首个大规模方言阿拉伯语机器阅读理解与生成式问答基准——EDRAC，涵盖埃及、摩洛哥、阿联酋、叙利亚和沙特五种主要方言。EDRAC包含499段源自真实口语交互的文本片段及4,977个通过人机协作流水线生成的问答对，该流程结合了迭代生成、大语言模型（LLM）作为裁判评估与人工验证，确保数据质量。研究在EDRAC上对阿拉伯语专用及多语言大模型进行评估，采用词法与语义指标，结果揭示了语义答案质量与方言忠实度之间存在显著差距，凸显了现有评价指标在方言生成任务中的局限性。因此，该研究的关键在于构建一个真实且具有挑战性的方言阿拉伯语自然语言处理基准，推动未来相关领域的发展。
链接: https://arxiv.org/abs/2609.01113
作者: Noor Abo Mokh,Kirill Chirkunov,Teresa Lynn,Nizar Habash,Reham Marzouk,Malik H. Altakrori,Younes Samih,Muhammed Abu Odeh,Nour Rabih,Rahaf Alshahrani,Hamad Alshehhi,Hamdan Al-Ali,Muhra Almahri,Besher Hassan,Mohamed Anwar,Abed Alhakim Freihat,Preslav Nakov,Alham Fikri Aji
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Dialectal Arabic (DA) remains under-resourced compared to Modern Standard Arabic (MSA), particularly for machine reading comprehension (MRC) and question answering (QA). Existing Arabic QA benchmarks primarily focus on formal written MSA or multiple-choice QA, with limited coverage of naturally spoken dialects. Here, we aim to bridge this gap. We introduce EDRAC, the first large-scale benchmark for dialectal Arabic machine reading comprehension (MRC) and generative QA, covering five major dialects: Egyptian, Moroccan, Emirati, Syrian, and Saudi Arabic. EDRAC contains 499 passages derived from naturally occurring spoken interactions and 4,977 corresponding QA pairs generated through a human–LLM collaborative pipeline combining iterative generation, LLM-as-a-judge evaluation, and human verification. We benchmark Arabic-centric and multilingual LLMs on EDRAC using lexical and semantic metrics. Our results reveal substantial gaps between semantic answer quality and dialectal fidelity, highlighting the limitations of existing evaluation metrics for dialectal Arabic generation. EDRAC provides a realistic and challenging MRC benchmark for future research on dialectal Arabic NLP.
[NLP-48] ClinTraceBench: Source-Verifiable Longitudinal Clinical Reasoning over EHR-Derived Dialogues EMNLP2026
【速读】： 该论文旨在解决临床大语言模型（LLM）助手在处理多访次患者病程时，其采用的紧凑历史表示方法（如检索、结构化时间线、LLM摘要、代理记忆等）是否能够有效保留临床推理所需的纵向信息信号这一关键问题。现有方法虽提升了可扩展性，但其对长期病程关系的表征能力尚未被系统评估。论文提出ClinTraceBench基准，包含385个基于MIMIC-IV数据集构建的经验证对话及事件ID溯源信息，涵盖九项任务分类（T1–T9）与从L0到L5的多层次验证体系（人类审计一致性达98.92%）。通过在6,271个问题上对八种历史表示策略（包括无上下文基线、仅末次就诊、全上下文、BGE-M3稠密检索、两种压缩方案及两种代理记忆系统Mem0、A-Mem）在四种骨干模型（DeepSeek-V3、GPT-4o-mini、Haiku 4.5、Sonnet 4.6）上的综合评估，揭示四项核心发现：（SP4）通过受控的T3注入探针证明，压缩过程导致关系信息丢失——即使在构造前已提供关键语句，Mem0、A-Mem与LLM摘要仍仅恢复0–5.3%的注入阳性样本；（SP1）压缩策略在多访次趋势与跨患者比较任务中存在聚合代价；（SP2）“盲视全上下文”差距高达+29.8个百分点（GPT-4o-mini）至+62.7个百分点（Haiku）；（SP3）拒绝回答行为随上下文长度非单调增长。在帕累托前沿上，Haiku在全上下文条件下超越Sonnet（25.76 vs. 106.21），颠覆了“模型越大越优”的常规认知。解决方案的关键在于建立可验证的纵向推理基准，量化不同历史表示方式对临床推理能力的影响，从而推动更具可解释性与可靠性的人工智能辅助临床决策系统的发展。
链接: https://arxiv.org/abs/2609.01111
作者: Huimin Wang,Zhengyi Zhao,Yutian Zhao
机构: Shenzhen University (深圳大学); The Chinese University of Hong Kong (香港中文大学); Dealism
类目: Computation and Language (cs.CL)
备注: Findings of EMNLP 2026
Abstract:Clinical LLM assistants must reason over multi-visit patient trajectories, yet whether the compact history representations used to scale them—retrieval, structured timelines, LLM summaries, agentic memory—preserve the longitudinal signal clinical reasoning needs has not been measured. We introduce ClinTraceBench: 385 MIMIC-IV-derived verified dialogues with event-ID provenance, a nine-task taxonomy (T1–T9), and L0–L4 deterministic + L5 human-audit validation (98.92% agreement). We evaluate eight history representation strategies—a no-context floor, \textitlast-visit-only, \textitfull-context, BGE-M3 \textitdense-retrieval, two compression schemes, and two agentic-memory systems (\textitMem0, \textitA-Mem)—across four backbones (DeepSeek-V3, GPT-4o-mini, Haiku~4.5, Sonnet~4.6) on 6,271 questions: 32 cells, 200,672 predictions. Four findings: (SP4) a controlled T3 injection probe isolates compression-induced \textitrelation loss—with the attribution sentence present \textitbefore construction, \textitMem0, \textitA-Mem and \textitllm-summary still recover only 0–5.3% of the injected positives; (SP1) compressed strategies pay an aggregation tax on multi-visit trends and cross-patient comparisons; (SP2) the blind-to-full gap spans +29.8 ~pp (GPT-4o-mini) to +62.7 ~pp (Haiku); (SP3) abstention scales non-monotonically with context length. On the Pareto frontier Haiku dominates Sonnet under \textitfull-context (\ 25.76 vs.\ \ 106.21), inverting the ``biggest backbone wins’’ heuristic.
[NLP-49] Hints Help But Do They Teach? Evaluating Skills Transfer in Code Generation
【速读】： 该论文旨在解决生成式 AI（Generative AI）在代码生成任务中，当提示（hint）能够将一个失败的生成程序修复为通过测试的程序时，这种提示究竟提供了原本缺失的信息，还是仅起到引导模型朝其已有能力范围内的解法方向进行调整的作用。研究通过可执行评估方法，在HumanEval+和MBPP+基准上对Qwen2.5-3B-Instruct与Phi-3.5-mini模型进行了验证。结果显示，相关提示虽能挽救部分失败案例（如Qwen中36/79个、Phi-3.5-mini中42/101个），但大量被提示挽救的样本实际上也可通过普通采样方式获得，且未使用提示的样本已成功解决46个（Qwen）或57个（Phi-3.5-mini）问题，其中分别包含31个和36个本可通过相关提示挽救的案例。进一步机制分析表明，相关与无关提示均激活了模型内部一个稳定的隐状态方向，持续施加该方向虽带来14次成功修复和18次退化，但整体准确率无显著提升；基于低秩干预的学习策略也仅表现出正向但不精确的效果。此外，完整文本规范可解决22/24个上下文定义问题，远优于所测试的虚拟键值（virtual-KV）前缀（5–11个）。后生成阶段的隐藏状态探测器在跨基准迁移中表现良好（合并AUROC达0.806和0.780），但其最优选择优势相对于词元置信度尚未达到统计显著性。综上所述，尽管相关提示具备一定的故障修复能力，但大多数被挽救的解法本质上仍可通过常规采样实现，而当前测试的内部干预手段未能证明其具备通用任务迁移能力。
链接: https://arxiv.org/abs/2609.01106
作者: Will Badr
机构: University of Leeds (利兹大学)
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:When a hint turns a failing generated program into a passing one, does it provide missing information or merely steer the model toward a solution it could already produce? We test these hypotheses on HumanEval+ and MBPP+ using executable evaluation. For Qwen2.5-3B-Instruct, adaptive relevant hints rescue 36 of 79 selected failures; an unrelated hint rescues 19, while eight unhinted samples solve 46 and recover 31 of the 36 relevant-hint rescues. Phi-3.5-mini shows the same pattern: relevant hints rescue 42 of 101 failures, an unrelated hint rescues 17, and unhinted sampling solves 57, including 36 of the 42 relevant-hint rescues. Because the hint conditions use different attempt budgets, these comparisons do not isolate a purely semantic effect. Mechanistic tests on Qwen identify a stable activation direction shared by relevant and unrelated hints. Persistently adding this direction yields 14 rescues and 18 regressions, with no detectable net accuracy gain; learned low-rank interventions have a positive but imprecise estimated effect. Full textual specifications solve 22 of 24 context-defined problems, versus 5-11 for tested virtual-KV prefixes. Post-generation hidden-state probes transfer across benchmarks, with pooled AUROC 0.806 and 0.780, but their top-one selection advantage over token confidence is statistically unresolved. Overall, relevant hints can rescue failures, but most rescued solutions are already reachable through ordinary sampling, and the internal interventions tested here do not establish task-general capability transfer.
[NLP-50] When Modality Gap Reduction Fails: Prediction-Level Hubness in CLIP EMNLP2026
【速读】： 该论文旨在解决预训练模型CLIP中图像与文本表示之间的模态差距（modality gap）在零样本分类任务中虽被减小，却未能稳定提升下游性能的问题。其核心问题在于：尽管平均层面的图像-文本对齐度降低，但零样本分类准确率并未随之一致提升，这表明仅优化平均对齐不足以保证性能增益。解决方案的关键在于揭示了这种不一致性背后的深层原因——即决策结构（decision structure）的变化。研究发现，零样本分类的准确率不仅依赖于平均对齐程度，更受类别间决策边界（class-wise decision margins）的影响。通过线性校正（Linear correction）这一可解析分析的案例，作者发现模态差距校正会改变不同类别之间的相对相似性关系，导致预测结果过度集中于少数几个类别，形成“预测层级聚心”（prediction-level hubness）现象。实验进一步验证，无论采用线性还是基于学习的校正方法，准确率下降均与预测集中度上升显著相关。因此，该研究提出应从下游决策结构的角度评估模态差距校正的效果，而不仅局限于平均对齐指标。
链接: https://arxiv.org/abs/2609.01103
作者: Shota Sato,Hajime Kiyama,Tosho Hirasawa,Mamoru Komachi
机构: Hitotsubashi University (一桥大学)
类目: Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV)
备注: Accpeted to EMNLP 2026 Main Conference
Abstract:Reducing the modality gap between image and text representations in CLIP is widely expected to improve cross-modal alignment and downstream performance. However, a smaller average image-text gap does not necessarily lead to consistent accuracy gains. We analyze this mismatch from the perspective of the decision structure in zero-shot classification, i.e. selecting the most similar class-text prototype for an input image. Zero-shot accuracy depends not only on average image–text alignment, but also on class-wise decision margins. Using Linear correction as an analytically tractable case, we show that modality gap correction can alter the relative decision structure among classes and cause predictions to concentrate on a small subset of classes. We refer to this output-space failure mode as prediction-level hubness. Furthermore, experiments across multiple datasets show that accuracy degradation under gap correction is consistently associated with increased prediction concentration, both for Linear correction and for learning-based correction methods. This provides a systematic explanation of why modality gap reduction does not consistently improve CLIP zero-shot accuracy from the perspective of downstream decision structure. Our results suggest that gap correction should be evaluated not only by average alignment, but also by its impact on downstream prediction structure.
[NLP-51] Beyond Magnitude: Contrastive Routing for Modular Mixture-of-Experts EMNLP2026
【速读】： 该论文旨在解决当前混合专家（Mixture-of-Experts, MoE）架构中路由机制依赖于跨所有标记共享的结构化表示，导致专家专业化能力受限的问题。其核心挑战在于现有路由策略基于绝对激活值，难以捕捉细粒度的语义差异，从而削弱了专家对特定任务或语言结构的区分能力。解决方案的关键在于提出对比路由机制（Contrastive Routing Mechanism, CoRM），通过将每个输入标记与层隐藏状态的指数移动平均（Exponential Moving Average, EMA）作为参考状态进行对比，而非直接依据绝对激活强度进行路由。CoRM利用每个专家独立的投影空间计算输入标记与参考状态之间的亲和力差距，将路由信号聚焦于低维、高度可分的子空间。这一设计使专家的路由边界更符合语言学结构，显著提升了专家的专业化程度。实验表明，CoRM在九个零样本推理基准上，相对于标准Top-k MoE基线，将平均零样本准确率提升0.67至1.69个百分点（Top-1）及1.38至1.77个百分点（Top-2），仅增加2.9%参数量和2.6%每标记浮点运算量，实现了性能与效率的高效平衡。
链接: https://arxiv.org/abs/2609.01100
作者: Nikolaos Xiros,Dimitrios Damianos,Maria-Eleni Zoumpoulidi,Leon Voukoutis,Vassilis Katsouros,Georgios Paraskevopoulos
机构: Institute for Language and Speech Processing, Athena Research Center(雅典研究中心语言与语音处理研究所)
类目: Computation and Language (cs.CL)
备注: Accepted to EMNLP 2026. 14 pages, 7 figures
Abstract:In current Mixture-of-Experts architectures, routing is performed based on representations dominated by structure shared across all tokens, limiting expert specialization. We show that contrasting each token against an Exponential Moving Average of the layer’s hidden states, rather than routing on absolute magnitude, concentrates the routing signal onto a low-dimensional, highly separable subspace. Building on this, we propose the Contrastive Routing Mechanism (CoRM), which scores each expert by the gap between its affinity for the incoming token and its affinity for this shared reference state, interpreted through a distinct per-expert projection. The resulting experts have routing boundaries that align with linguistic structure significantly more than the Top-k baseline. Our experiments show that CoRM improves average zero-shot accuracy by +0.67 to +1.69 points (Top-1) and +1.38 to +1.77 points (Top-2) over standard Top-k MoE baselines on nine zero-shot reasoning benchmarks, at the minimal cost of 2.9% added parameters and 2.6% added FLOPs per token.
[NLP-52] StateSwap: Probing Support-Elimination Hidden States in Multiple-Choice Questions EMNLP2026
【速读】： 该论文旨在解决大语言模型在面对同一多选题时，因题目呈现方式（支持导向与排除导向）不同而产生不一致回答的问题。其核心问题是：这两种框架是否诱发了模型内部表征的差异，从而导致推理行为不一致。解决方案的关键在于提出一种双框架协议，通过仅微调提示语中的表述方式（支持或排除导向），保持评估目标不变，以隔离框架效应；并引入一个未训练的特殊标记 [STATE]，将其残差流激活作为干预接口，探测模型内部计算过程。实验表明，两种框架在中间层引发可分离的 [STATE] 激活模式，跨提示对交换这些激活可系统性改变预测结果并提升跨框架一致性，提供了基于干预的证据，证明此类激活与行为相关。此外，基于双框架对比提取的均值差异引导方向，在层间响应上比对照的激活添加方向更具稳定性，进一步验证了该方法的有效性。
链接: https://arxiv.org/abs/2609.01081
作者: Chao Gao,Haijiang Liu,Qiyuan Li,Caicai Guo,Frank van Harmelen,Jinguang Gu
机构: Vrije Universiteit Amsterdam (自由大学阿姆斯特丹分校); Wuhan University of Science and Technology (武汉科技大学); Hubei Province Key Laboratory of Intelligent Information Processing and Real-time Industrial System (湖北省智能信息处理与实时工业系统重点实验室)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Accepted to Findings of EMNLP 2026
Abstract:Large language models often answer the same multiple-choice question inconsistently when it is posed under support-oriented and elimination-oriented framings. We investigate whether these discrepancies arise from different internal representations induced by the two framings. We introduce a dual-framing protocol with minimally varied prompts that use either support- or elimination-oriented framing while keeping the evaluation target fixed. To probe the internal computation, we append an untrained special token, [STATE], and treat its residual-stream activation as an intervention interface. Across both models, the two framings induce separable [STATE] activations concentrated in intermediate layers. Swapping these activations between paired prompts systematically changes predictions and improves cross-framing agreement, providing intervention-based evidence that the activations are behaviorally relevant. Beyond instance-level substitution, mean-difference steering directions derived from the dual-framing contrast exhibit more bounded layer-wise responses than matched contrastive activation addition directions under the evaluated protocol.
[NLP-53] Post-hoc Alignment of LLM -judges to Human Judgment Distribution EMNLP2026
【速读】： 该论文旨在解决当前基于大语言模型作为评判者（LLM-as-a-judge, LLMaJ）框架在自动评估中忽视人类标注差异（Human Label Variation, HLV）的问题。现有方法通常将LLMaJ的判断与聚合后的单一真实标签（hard-label）进行对比，忽略了人类标注分布所蕴含的丰富信息。为此，作者系统研究了LLMaJ在预测单一硬标签和未聚合的软标签（即人类判断分布，Human Judgment Distribution, HJD）上的表现，发现尽管在硬标签预测上接近人类水平，但在软标签预测方面表现不佳。针对这一局限，论文提出一种轻量级后处理对齐方法NAPHA（eNtropy-Aware Post-Hoc Alignment），其关键在于通过先将样本分配至离散熵类别，再将其路由至特定训练的对齐模型，从而实现对LLM输出分布与HJD的匹配。实验表明，NAPHA在多种基础模型和数据集上均显著提升软标签预测性能，尤其在高熵样本（需捕捉多样人类观点的关键场景）中表现突出；此外，通过“理想实验”验证了提升熵类别预测精度可进一步增强NAPHA的实际效果。
链接: https://arxiv.org/abs/2609.01073
作者: Sebastian Steindl,Nikos Voskarides,Alberto Gasparin,Diego Marcheggiani
机构: Amazon(亚马逊); Amazon(亚马逊); Amazon(亚马逊); Amazon(亚马逊)
类目: Computation and Language (cs.CL)
备注: Accepted to EMNLP 2026
Abstract:The LLM-as-a-judge (LLMaJ) framework offers a cost-effective and reproducible solution for automatic evaluation. However, current evaluation practices typically compare LLMaJ judgments against aggregated ground-truth labels, overlooking the valuable information contained in Human Label Variation (HLV). Inspired by an increasing line of work that proposes to leverage HLV, we systematically study LLMaJ performance on predicting both a single, aggregated ground truth hard-label and unaggregated soft-labels that represent Human Judgment Distributions (HJD). Our results across five diverse datasets reveal that while LLMs achieve near human-level performance at hard-label prediction on most tasks, they exhibit poor performance when predicting soft-labels. To address this limitation, we propose NAPHA (eNtropy-Aware Post-Hoc Alignment), a simple yet effective lightweight post-hoc alignment method that matches the LLM distribution to the HJD by first assigning an instance to a discrete entropy class and then routing it to specialized, trained alignment models. We find that NAPHA consistently improves soft-labels prediction across base LLM models and datasets, with particularly strong gains on high-entropy instances where capturing diverse human perspectives is most critical. We also show via oracle experiments that improving entropy class prediction can substantially enhance NAPHA’s practical effectiveness.
[NLP-54] OUTLETS: Output-Length Prediction from Speculative Decoding Backbones EMNLP2026
【速读】： 该论文旨在解决大语言模型（Large Language Model, LLM）服务中输出长度呈现长尾分布所带来的资源调度与集群管理难题。现有输出长度预测方法存在显著缺陷：外部代理模型引入额外延迟且预测精度有限，而基于内部状态的预测方法虽高效但依赖对当前模型状态的浅层探测。本文的关键突破在于发现推测解码（Speculative Decoding, SD）与输出长度预测之间存在结构关联——先进框架（如EAGLE-3）中草稿解码器生成的潜在表示（latent representations）蕴含可预测生成长度的信号。基于此，作者提出OUTLETS（Output-Length Prediction from Speculative Decoding Backbones），将推测解码主干网络重构为一种轨迹感知的长度预测器。当草稿表示已用于推测解码时，OUTLETS仅需添加一个轻量级回归头即可实现更低的平均绝对误差（MAE），并显著提升调度效率。在高负载去耦合服务场景下，利用OUTLETS的预测结果，标准调度策略可优先处理短请求，并更均衡地分配请求至解码实例，使短请求的P99延迟降低34.8%。
链接: https://arxiv.org/abs/2609.01068
作者: Weihuang Wen,Yingying Liu,Yichuan Liu,Wenqi Zeng,Li Zhou,Chumin Sun,Jie Sun,Tianshu Yu
机构: The Chinese University of Hong Kong, Shenzhen(香港中文大学（深圳）); The University of Hong Kong(香港大学); Huawei Technologies Co., Ltd.(华为技术有限公司)
类目: Computation and Language (cs.CL)
备注: Accepted to EMNLP 2026
Abstract:The heavy-tailed distribution of output lengths in Large Language Model (LLM) serving poses major challenges for resource provisioning and cluster scheduling. Although output-length prediction can mitigate these issues, existing approaches have key drawbacks: external proxy models add substantial latency and often have limited fidelity, whereas internal state-based methods are efficient but rely on shallow probes of current model states. We identify a structural connection between speculative decoding (SD) and length prediction: latent representations produced by the draft decoder in advanced frameworks (e.g., EAGLE-3) encode signals that are predictive of generation length. Building on this insight, we introduce OUTLETS (Output-Length Prediction from Speculative Decoding Backbones), which repurposes the speculative backbone as a trajectory-aware length predictor. When its draft representations are already computed for speculative decoding, OUTLETS adds only a lightweight regression head and achieves lower MAE than the evaluated methods. Under saturated disaggregated serving, OUTLETS predictions enable standard scheduling policies to prioritize shorter requests and distribute requests more evenly across decoding instances, reducing short-request P99 latency by 34.8%.
[NLP-55] WorldBench: Culturally Grounded Benchmark for Multilingual Agents
【速读】： 该论文旨在解决当前大语言模型（LLM）驱动的智能体在复杂环境中执行多步任务时，普遍缺乏对状态保持能力、跨语言性能以及真实场景下具身化应用能力的有效评估问题。现有基准测试难以全面反映智能体在多语言、多文化背景下的实际表现，尤其忽视了任务执行过程中环境状态的持续维护与一致性。为此，作者提出了WorldBench——一个综合性、多语言的基准测试平台，涵盖7种语言和8种文化背景下的1,600个真实、角色具身化的日常任务流程，支持智能体在沙盒环境中通过结构化动作进行交互。其核心解决方案在于引入**受限任务成功率（Constrained Task Success, CTS）**这一新型评估指标，结合自然语言指令与测试环境，通过确定性评估与“大模型作为裁判”（LLM-as-a-Judge）双重方式，综合衡量任务完成度、最小修改量及其他互补维度。实验结果表明，前沿模型在该基准上的平均CTS仅为49.2%，且所有模型均表现出显著的任务正确性与环境状态保持能力之间的差距，揭示当前智能体在多语言、具身化、长时程任务中仍存在严重脆弱性，亟需提升状态建模与跨情境适应能力。
链接: https://arxiv.org/abs/2609.01056
作者: Leonardo Ranaldi,Sherrie Shen,Jushi Kai,Alexandra Birch
机构: University of Edinburgh (爱丁堡大学); School of Artificial Intelligence, Shanghai Jiao Tong University (上海交通大学人工智能学院)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Despite the growing use of LLM-powered agents to solve multi-step tasks in complex environments, existing benchmarks rarely test state preservation, performance across languages, and application to realistic, grounded scenarios. To address these concerns, we present WorldBench: a comprehensive, multilingual benchmark of genuine, persona-grounded everyday workflows, where agents can act in a sandbox via structured actions. WorldBench comprises 1,600 tasks across seven languages and eight cultures, filtered and refined through feedback from human annotators with language- and culture-specific expertise. For evaluation, we extend metrics from previous works and introduce Constrained Task Success (CTS), which combines natural language instructions and testbeds to score task completion, minimal modification, and other complementary metrics through deterministic and LLM-as-a-Judge evaluations. Our experiments show that frontier models reach only 49.2% CTS, with all models demonstrating large gaps between correctness and environment preservation. We thereby show that current agents remain brittle in multilingual, agentic scenarios, especially for long-horizon tasks and under state-preservation constraints
[NLP-56] Lagged Coupling: Internal Representations Become Readable Before They Become Causal
【速读】： 该论文旨在解决生成式模型中“可读性”（internal readability）与“因果有效性”（causal efficacy）之间的不对称性问题，即尽管在模型训练早期即可通过线性探测从残差流中准确读取目标变量（表现出极高的可读性），但沿相同方向进行控制（steering）却几乎无效，甚至在多数情况下为零效或反效果。其核心发现是这种现象普遍存在且不随模型规模扩大而缓解，作者将其称为“滞后耦合”（lagged coupling）。解决方案的关键在于提出三重可分离的分析框架：(i) 内部可读性（saturated at AUROC = 0.990，自初始检查点即达饱和）、(ii) 行为可读性（behavioral readability，随规模增大逐步发展）、(iii) 因果有效性（causal efficacy，长期接近零效，仅在极少数情况出现短暂正向响应）。研究揭示出“先读后写”（read-before-write）的主导模式，表明表征空间中目标变量的写入强度远超读出机制的有效性——表征头空间（headroom）可增长至57倍，但实际因果写入比例始终低于0.11%。这一结构性瓶颈说明，模型内部表示的形成显著快于因果读出的巩固，从而警示不可仅凭探测准确性推断可操控性，并强调了在模型开发中应关注因果读出的动态演化。
链接: https://arxiv.org/abs/2609.01048
作者: Xining Xun
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 15 pages, 5 figures, 7 tables. Pre-registered developmental interpretability study on the full Pythia suite with an OLMo-2 replication
Abstract:Across the full Pythia suite (160M-12B, eight checkpoints, four task families), a linear probe can read a target variable from the residual stream as early as step 1,000 at every scale – yet steering along that same reading direction remains null-equivalent in 43 of 48 model-checkpoint cells. Internal readability systematically outruns causal efficacy, and the lag does not shrink with scale. We call this structure lagged coupling and decompose it into three dissociable tracks: (i) internal readability, saturated (AUROC = 0.990) from the first checkpoint everywhere; (ii) behavioral readability, which develops gradually and progressively later at larger scales (12B reaches 0.909 only at the final checkpoint); (iii) causal efficacy, almost always null-equivalent, occasionally counterproductive early, with one isolated positive pulse (12B, step 8,000, z = +2.49) our grid cannot resolve. The ordering is dominantly read-before-write (11/11 units, no inversion). Representation headroom along the probe direction grows up to 57x with training and scale while causal write-in stays below 0.11% of headroom – the variable is increasingly written into the representation and increasingly ignored by the readout. Under a fully pre-registered protocol, both single-onset hypotheses resolve INDETERMINATE (scale slope +0.24, 95% CI [-0.60, +0.87]; time vote 3:3) – a disciplined negative explained by the three-track decomposition. A pre-registered OLMo-2 replication preserves the direction at attenuated magnitude. Our results caution against inferring steerability from probe accuracy and establish a developmental bottleneck: representation formation reliably outpaces causal readout consolidation.
[NLP-57] PCoMoE: Shifting MoE Inference from Monolithic Expert Selection to Fine-Grained Path Composition EMNLP2026
【速读】： 该论文旨在解决现有混合专家（Mixture-of-Experts, MoE）架构在大规模语言模型（Large Language Model, LLM）推理过程中因采用粗粒度、原子化专家抽象而导致的优化边界过早固化问题。具体而言，传统框架将专家作为不可分割的执行单元进行调度或剪枝，忽视了专家内部计算结构的细粒度冗余，限制了性能提升空间。其解决方案的关键在于提出一种路径组合式执行框架PCoMoE，通过引入专家计算的路径级建模，实现从粗粒度专家选择到细粒度路径组合的范式转变；同时设计了兼容性感知的逐层路径剪枝策略以抑制低价值路径组合，并构建面向硬件友好的执行引擎，有效利用可复用的子专家结构，在严格控制开销的前提下实现高效推理。实验结果表明，PCoMoE在保持模型精度提升10%的同时，实现了最高达1.31倍的端到端推理加速。
链接: https://arxiv.org/abs/2609.01024
作者: Ziyan Gan,Fangxin Liu,Chenyang Guan,Junjie Wang,Ning Yang,Haomin Li,Xiang Li,Siran Yang,Jiamang Wang,Lin Qu,Zongwu Wang,Li Jiang,Haibing Guan
机构: Shanghai Jiao Tong University (上海交通大学); Alibaba Group (阿里巴巴集团)
类目: Computation and Language (cs.CL)
备注: Accepted to EMNLP 2026 Main Conference
Abstract:Mixture-of-Experts (MoE) architectures scale Large Language Model (LLM) capacity efficiently by activating a sparse subset of experts per token. However, modern MoE inference remains heavily constrained by the rigid, whole-expert abstraction. Existing frameworks manage, schedule, or prune experts as atomic execution units, which fixes the optimization boundary too early and leaves fine-grained intra-expert computational redundancy underexplored. In this work, we present PCoMoE, a path-compositional execution framework that shifts MoE inference from coarse-grained expert selection to fine-grained path composition. PCoMoE incorporates a path-level formulation of expert computation, a compatibility-aware layer-wise pruning strategy to suppress low-value path combinations, and a hardware-friendly execution engine to exploit reusable sub-expert structures under strictly bounded overheads. Experimental results demonstrate that PCoMoE achieves up to a 1.31x end-to-end inference speedup while enhancing model accuracy by 10%. The code is available at this https URL
[NLP-58] Phrase-Localized Language-Contrastive Guidance: Training-Free Localized Accent Control for Code-Switching Text-to-Speech EMNLP2026
【速读】： 该论文旨在解决跨语言文本到语音（text-to-speech, TTS）中代码混用（code-switching）带来的语调不自然问题，即在主语言语句中插入外语短语时，该短语通常会带有主语言的口音而非其母语口音，导致语音不地道。针对这一问题，论文提出了一种无需训练的推理框架——短语定位语言对比引导（Phrase-Localized Language-Contrastive Guidance, LCG）。其核心解决方案在于：摒弃传统TTS模型对整个语音序列施加单一语言引导的做法，转而为每个语言片段分别施加独立的语言引导，使每个区域的语音均能保留其所属语言的母语口音。为实现这一目标，LCG引入一种自注意力探针（self-attention probing）技术，无需外部对齐标注即可自动识别代码混用短语的边界，从而精准定位需应用局部语言引导的位置。该方法无需微调或依赖额外模型，在多种语言组合下均能显著提升代码混用短语的自然度，有效抑制口音泄露，同时保持说话人身份和整体语音自然性。
链接: https://arxiv.org/abs/2609.01016
作者: Che Hyun Lee,Sangkwon Park,Donghun Kang,Dongwook Lee,Youngho Cho,Heeseung Kim,Sungroh Yoon
机构: Seoul National University (首尔国立大学); University of Seoul (首尔大学); AIIS, ASRI, INMC, and ISRC, Seoul National University (首尔国立大学人工智能研究所、先进系统研究所、智能制造研究中心和智能机器人研究中心)
类目: Computation and Language (cs.CL); Sound (cs.SD)
备注: Accepted to EMNLP 2026 (Main Conference). Demo: this https URL
Abstract:Current speech synthesis struggles with code-switching, which mixes a foreign language phrase into a primary language utterance, causing the phrase to be spoken with the primary language’s accent rather than its native one. We propose Phrase-Localized Language-Contrastive Guidance (LCG), a training-free inference framework that restores a native accent to code-switched phrases in cross-lingual text-to-speech. LCG replaces the single language guidance applied across the whole utterance with a separate guidance for each region, so each part is guided by its own language. To choose where to apply this localized guidance, we propose a self-attention probing technique that finds the phrase boundaries without external alignments. Together, these components generate speech in which each region carries the accent of its own language, requiring no fine-tuning or auxiliary models. Across diverse language pairs, LCG robustly increases the nativeness of the code-switched phrase while suppressing accent leakage, and preserving overall speaker identity and naturalness.
[NLP-59] SinkPruner: Sink-Free Visual Token Pruning for Multimodal Large Language Models EMNLP2026
【速读】： 该论文旨在解决多模态大语言模型（Multimodal Large Language Models, MLLMs）在处理长视觉标记序列时产生的巨大计算开销问题。现有视觉标记剪枝方法多依赖于以视觉为中心或文本引导的策略，但往往忽视了高范数异常标记（high-norm outlier tokens），即特征范数异常大的标记，导致剪枝决策次优。这些高范数标记在特征和空间维度上均具有高度冗余性，却被现有方法错误地保留为信息性线索。为此，本文提出了一种无需训练的视觉标记剪枝框架SinkPruner，其核心在于采用“粗到细”的设计：首先通过视觉净化模块（visual sanitizer）过滤高范数冗余信息，缓解注意力汇聚（attention sink）与注意力分散问题；随后利用文本引导剪枝模块保留与文本查询语义对齐的标记。实验结果表明，该方法在12个图像-语言及4个视频-语言基准上均展现出卓越的效率、有效性与泛化能力，在实现89%标记压缩率的同时，保持了LLaVA-1.5和Qwen2.5-VL原始性能的96.5%和91.8%，且其视觉净化模块具备良好的迁移性，可显著提升现有剪枝方法的性能。
链接: https://arxiv.org/abs/2609.01004
作者: Shiyu Li,Zi-Yuan Hu,Shijia Huang,Yanyang Li,Yiwu Zhong,Liwei Wang
机构: Weitu AI(未图人工智能); Peking University (北京大学); The Chinese University of Hong Kong (香港中文大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: EMNLP 2026 (findings)
Abstract:Despite their strong multimodal understanding ability, multimodal large language models (MLLMs) incur substantial computational overhead when processing long visual token sequences. To reduce inference costs, recent studies have explored visual token pruning through vision-centric or text-guided strategies. However, these methods often overlook high-norm outlier tokens, i.e., tokens with abnormally large feature norms, leading to suboptimal pruning decisions. In this work, we show that such high-norm outlier tokens are highly redundant in both feature and spatial dimensions, yet are often mistakenly preserved as informative cues by existing methods. Motivated by this observation, we propose SinkPruner, a training-free visual token pruning framework for efficient MLLM inference. SinkPruner follows a coarse-to-fine design with two key modules: a visual sanitizer that filters high-norm redundancies and alleviates attention sink and attention dispersion, and a text-guided pruner that further retains tokens semantically aligned with the text query. Extensive experiments on twelve image-language and four video-language benchmarks demonstrate the effectiveness, efficiency, and generalizability of our framework. Notably, SinkPruner preserves 96.5% (91.8%) of the original performance of LLaVA-1.5 (Qwen2.5-VL) under an 89% token reduction. Experiments further indicate that our visual sanitizer exhibits promising transferability in enhancing the performance of existing pruning methods. Our code is available at this https URL. Comments: EMNLP 2026 (findings) Subjects: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG) Cite as: arXiv:2609.01004 [cs.CV] (or arXiv:2609.01004v1 [cs.CV] for this version) https://doi.org/10.48550/arXiv.2609.01004 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[NLP-60] Right Frame Wrong Rule: Cultural Cues Expose the Financial Knowledge Gap They Were Meant to Close EMNLP2026
【速读】： 该论文旨在解决在规范多元主义（normative pluralism）情境下，语言模型如何在不同规范框架中进行框架选择与答案正确性判断的问题。其核心挑战在于：当一个问题在多个规范框架下存在有效答案时，模型需自主决定采用哪个框架，并评估自身在该框架内的回答准确性。论文以伊斯兰金融为研究场景，构建了一个四选一的分类体系，将框架选择与框架内正确性判定分离，从而揭示出“刻板印象陷阱”（stereotype trap）现象——文化线索会引导模型偏向某一特定框架，但在此框架内模型仍可能给出错误答案。实验覆盖十二个模型、两种语言及五十种人口统计学信号，结果表明文化线索显著影响框架选择，并暴露出非前沿模型在准确性上的巨大差异。在最强文化信号下，大型开源模型97%的时间选择伊斯兰框架，但其中57%至66%的选择实际上不正确。若采用二选一评估范式，将误报为近乎完美的对齐表现。这一发现支持但未直接验证“能力条件路由假说”（competence-conditioned routing hypothesis），即模型可能倾向于选择其自身更擅长的框架，而文化线索则暴露了模型在不同框架间的能力差异。
链接: https://arxiv.org/abs/2609.00999
作者: Rania Elbadry,Ahmed Heakl,Saeed Almheiri,Fan Zhang,Muhra AlMahri,Xueqing Peng,Mohsinul Kabir,Shuyao Wang,Yi Han,Saadeldine Eletter,Duzhen Zhang,Preslav Nakov,Yuxia Wang,Fajri Koto,Zhuohan Xie
机构: MBZUAI; The University of Tokyo; The University of Manchester; The Fin AI; Georgia Institute of Technology; Harvard University; INSAIT, Sofia University “St. Kliment Ohridski”; MBZUAI
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Accepted into EMNLP 2026 Findings
Abstract:When a question has valid answers under different normative frameworks, a language model must decide which framework to use and whether it can answer correctly within it. We call this setting normative pluralism and study it in Islamic finance using a four-choice taxonomy that separates framework selection from within-framework correctness. This separation reveals the stereotype trap: a cultural cue steers a model toward one framework, but the model selects an incorrect answer within that framework. Across twelve models, two languages, and fifty demographic signals, cultural cues change framework selection and reveal substantial differences in accuracy, especially among non-frontier models. Under the strongest signal, large open-weight models select the Islamic framework 97% of the time. A two-choice evaluation would report near-perfect alignment, although 57–66% of those selections are incorrect. These findings motivate, but do not directly test, the competence-conditioned routing hypothesis: models may favor frameworks where they are more accurate, while cultural cues may expose framework-specific competence gaps.
[NLP-61] Inspicio: Open-Vocabulary LLM -Based Sense Retrieval for Historical Languages
【速读】： 该论文旨在解决历史语言与低资源语言中词义消歧（Word Sense Disambiguation, WSD）面临的根本性挑战，即传统方法依赖于源语言的词义词典（sense inventory）和词-义映射关系，而这些在多数历史语言（如拉丁语、古希腊语）和低资源语言中往往缺失或不完整。针对这一问题，论文提出了一种名为Inspicio的开放词汇检索框架，其核心创新在于无需依赖任何源语言的词义词典或映射，即可将上下文中的词项关联到Open English WordNet中的同义词集（synset）。该方案的关键在于：利用指令微调的大语言模型（instruction-tuned LLM）生成目标句的双语翻译、候选词义定义及英语词干（lemma），进而通过融合密集型词义-同义词相似度、稀疏型词干匹配以及最大边际相关性（Maximal Marginal Relevance, MMR）重排序的混合检索机制，实现精准的跨语言词义链接。实验结果表明，该方法在拉丁语和古希腊语感知动词数据集上达到96%的Recall@50，各模块贡献可量化的性能提升，并在跨语言与域外场景中保持竞争力。
链接: https://arxiv.org/abs/2609.00998
作者: Michele Ciletti
机构: University of Foggia (福贾大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 12 pages, 1 figure
Abstract:Word Sense Disambiguation has advanced rapidly for English and a handful of well-resourced modern languages, but it continues to assume the existence of a sense inventory and a word-to-sense mapping in the source language (Navigli, 2026). These assumptions break down for most historical and low-resource languages, whose dedicated WordNets are either incomplete or still under construction. We present Inspicio, an open-vocabulary retrieval pipeline that links tokens in context to synsets of the Open English WordNet (McCrae et al., 2020) without requiring any source-language inventory or mapping. For each occurrence, an instruction-tuned LLM produces two English translations of the surrounding sentence, a small set of candidate dictionary-style definitions, and a few candidate English lemmas. These outputs drive a hybrid retrieval step that combines dense definition-synset similarity, sparse lemma matching, and Maximal Marginal Relevance re-ranking. We evaluate the pipeline across a 6x6 grid of LLMs and sentence-embedding models on a new bilingual set of manually annotated Latin and Ancient Greek perception verbs, on a subset of PREMOVE dataset (Farina, 2025), and on a diachronic sample of Italian. The best configuration reaches 96% Recall@50 on the perception-verb test set, with each component contributing measurable gains, and remains competitive in the out-of-domain and cross-lingual settings.
[NLP-62] PersianAnonymizer: Evaluating LLM -Labeled Training for Efficient NER-based Anonymization in Persian LREC2026
【速读】： 该论文旨在解决波斯语（Persian）客户聊天数据在工业场景中实现高效、低成本匿名化的问题，核心挑战在于如何在缺乏大量人工标注数据的情况下，构建高性能的命名实体识别（Named Entity Recognition, NER）模型。其解决方案的关键在于利用大语言模型（Large Language Model, LLM）生成的标注监督信号，通过指令微调的LLM（包括DeepSeek-V3-0324、GPT-OSS-120B和Qwen3-235B-A22B-Instruct-2507）在统一的JSON协议下进行跨模型的实体跨度标注，构建多个零样本（Zero-Shot）与少样本（Few-Shot）数据集，并基于MatinaRoberta架构训练轻量级的词元分类器。实验表明，由GPT-OSS-120B生成的OSS_ZeroShot标注数据所训练的NER模型在宏平均F1分数（macro-F1）和标签覆盖召回率（Label Coverage Recall, LCR）上表现最优，且可在单张消费级显卡（RTX 3090）上仅用约2分钟完成对4万条消息的全量标注，验证了该方法在保持高精度的同时具备显著的部署效率与成本优势，为波斯语工业数据的高质量、低资源匿名化提供了可行路径。
链接: https://arxiv.org/abs/2609.00958
作者: Mohammad Hossein Shalchian,Mostafa Amiri,Amir Mahdi Sadeghzadeh
机构: 未知
类目: Computation and Language (cs.CL)
备注: 10 pages, 3 figures, 6 tables. Published at LREC 2026
Abstract:We target practical anonymization of Persian customer chats by training a compact NER model from LLM-labeled supervision and selecting the best labeler for deployment. We compare three instruction-tuned LLMs: DeepSeek-V3-0324, GPT-OSS-120B, and Qwen3-235B-A22B-Instruct-2507, to produce span annotations under a shared JSON protocol, yielding four corpora (OSS_ZeroShot, Qwen_ZeroShot, Qwen_FewShot, DeepSeek_FewShot). A MatinaRoberta-based token-classifier is trained per corpus and evaluated with token-level Precision/Recall/F1 (overall and per-class). We also report Label Coverage Recall (LCR), the proportion of gold non-O tokens predicted as non-O, and quantify cross-labeler behavior via a token-level Venn on test annotations. Finally, we contrast test-set annotation latency of the LLMs on H200 nodes with the trained NER’s test-time labeling on a single RTX 3090. Results show that supervision from OSS_ZeroShot yields the strongest macro-F1 and LCR, while the resulting NER labels an entire 40K-message test set in approximately 2 minutes on one consumer GPU. This establishes a practical path to high-quality, low-cost anonymization for Persian industrial data.
[NLP-63] Calibration is the Bottleneck: An Action-Class Diagnostic of Multi-Turn Tool-Calling EMNLP2026
【速读】： 该论文旨在解决多轮工具调用（multi-turn tool calling）评估中，现有聚合准确率指标无法揭示模型在不同任务场景下表现不均衡的问题。当前公开权重模型在整体准确率上已接近甚至超越闭源前沿模型，但这一平均指标掩盖了模型在特定动作类别上的系统性偏差或执行失败。其解决方案的关键在于提出一种面向动作类别的诊断框架（action-class-oriented diagnostic framework），将多轮失败分解为两个正交模式：动作类别校准失准（action-class miscalibration）与动作执行失败（action-execution failure）。该框架基于四类动作空间（TOOL_CALL/ASK/REFUSE/CONFIRM），引入自揭示的理论上限指标Acc = GAR（Gold Action Recall），通过界外违反（Acc < GAR）暴露校准失准问题，通过界内松弛（GAR > Acc）定位执行失败。实验验证表明，校准失准是被现有状态评分器（state grader）忽略的重要失败模式，尤其在大量训练工具调用的模型家族中尤为显著；而通过仅改变上下文的扰动即可重塑校准性能，但其效果在不同模型家族间呈现异质性，同一扰动可能使准确率提升11.5个百分点或下降21.0个百分点，且效果依赖于扰动机制。因此，论文主张在多轮工具调用评估中应补充动作类别诊断，以揭示模型在各情景下的真实行为。
链接: https://arxiv.org/abs/2609.00949
作者: Kangjia Zhao,Jiajun Li,Haozhan Shen,Wei Chow,Linfeng Li,Hang Song,Lingdong Kong,Chen Zhi,Tiancheng Zhao,Songhua Liu,Jianwei Yin
机构: Zhejiang University(浙江大学); Shanghai Jiao Tong University(上海交通大学); National University of Singapore(新加坡国立大学); Xi’an Jiaotong University(西安交通大学); Om AI Research; Binjiang Institute of Zhejiang University(浙江大学滨江研究院)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Accepted to Findings of EMNLP 2026. Code: this https URL
Abstract:Multi-turn tool calling is a core evaluation scenario for large language model (LLM) agents. On public tool-calling benchmarks, open-weight models now approach or even surpass closed-source frontier models in aggregate accuracy. However, this metric averages over many different multi-turn situations and obscures whether progress is balanced across them. We propose an action-class-oriented diagnostic framework that decomposes multi-turn failures into two orthogonal modes: action-class miscalibration and action-execution failure. The framework operates over a four-class action space (TOOL_CALL/ASK/REFUSE/CONFIRM) and introduces a self-revealing upper bound Acc = GAR (Gold Action Recall); the two modes show up as bound violation (Acc GAR, exposing state-grader masking of miscalibration) and large bound slack (GAR Acc, localizing execution failure within TOOL_CALL). We validate it on a panel of tool-calling models across multiple multi-turn benchmarks. Across our panel, the diagnostic reveals action-class miscalibration as a substantial failure mode the state grader cannot see. This gap inflates standing for heavily tool-trained families, which our diagnostic separates from families with context-appropriate action choice. Calibration is reshapable through context-only perturbations, but the reshape is heterogeneous: a single perturbation moves accuracy in opposite directions across families (up to +11.5 vs -21.0 pp on the same scenario), and its effect further depends on the perturbation mechanism. We argue that multi-turn tool-calling evaluations should supplement aggregate accuracy with action-class diagnostics that expose what the model actually does in each scenario.
[NLP-64] From Terminology to Diagrams: Visual-Instruction Generation for Scientific Diagram Understanding
【速读】： 该论文旨在解决视觉语言模型（Vision-Language Models, VLMs）在科学图表理解任务中表现不佳的问题，尤其针对科学图表所承载的功能性或关系性语义而非自然图像中的直观场景信息这一挑战。其核心解决方案在于提出一种基于科学课程术语的框架，用于自动生成大规模、基于图表的指令数据。该方法的关键在于系统性地从科学教育内容中提取领域概念，合成原子级事实，从网络中检索相关图表，并生成包含图表描述和多选题形式的多模态监督信号。通过这一流程构建了包含超过19.4万张图表和140万条视觉指令的SciGram数据集，覆盖生命科学、地球科学与物理科学三大领域。尽管依赖噪声较大的网络数据和合成标注，但基于SciGram微调的模型在以图表为中心的基准测试（如TQA、ScienceQA、AI2D）上显著提升性能，优于或达到现有先进VLM水平，且训练样本更少。进一步将SciGram用于增强LLaVA OneVision等现有模型，实现了图示问答任务的新最佳表现。研究结果表明，基于术语的指令生成策略是提升科学领域视觉-语言推理能力的有效通用方法。为促进后续研究，作者开源了SciGram数据集及对应模型。
链接: https://arxiv.org/abs/2609.00948
作者: Raul Ortega,José Manuel Gómez-Pérez
机构: Language Technology Research Laboratory; Expert.ai; 17 Henri Dunant, 28036 Madrid, Spain
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: Published as a conference paper at COLM 2026
Abstract:Vision-language models (VLMs) have demonstrated strong performance in visual question answering with natural images. However, they continue to struggle with scientific diagrams, which are designed to convey functional or relational meaning rather than literal scenes. We therefore introduce a framework for generating large-scale diagram-grounded instruction data by leveraging terminology derived from scientific curricula. Our approach systematically extracts domain concepts, synthesizes atomic facts, retrieves relevant diagrams from the web, and generates multimodal supervision in the form of diagram captions and multiple-choice questions. Using this pipeline, we construct SciGram, a dataset of over 194K diagrams and 1.4M visual instructions across life, earth, and physical sciences. Despite relying on noisy web data and synthetic annotations, models fine-tuned on SciGram achieve substantial improvements on diagram-centric benchmarks, including TQA, ScienceQA, and AI2D, outperforming or matching state-of-the-art VLMs while using fewer training instances. Furthermore, augmenting existing models such as LLaVA OneVision with SciGram establishes new state-of-the-art performance on diagram question answering. Our results highlight the effectiveness of terminology-grounded instruction generation as a general strategy for improving vision-language reasoning in scientific domains. To support future research in scientific diagram understanding, we release both the SciGram dataset and models.
[NLP-65] A Dataset for Modeling Iterative Problem-Solving EMNLP2026
【速读】： 该论文旨在解决迭代式问题求解过程中性能变化趋势预测与学习机制解析的难题，核心挑战在于建模人类学习者或自主智能体在多次尝试中如何根据反馈调整策略、错误模式如何演变以及整体表现是提升、停滞还是退化。其解决方案的关键在于构建一个大规模、细粒度的编程学习数据集——CodeInsight，包含超过300万次学生提交的C++代码，涵盖测试用例级别的结果、时间戳和源代码，并在此基础上建立统一校准与评估协议的基准测试框架。该框架整合了参数化模型、序列模型及生成式模型（Generative AI）等多种范式，其中改进的递归状态空间模型（RSSM）通过离散隐变量追踪求解者特征，在四门课程中的三门取得最优预测精度；而基于大语言模型（LLM）的生成式预测器虽预测准确率较低，但能生成完整代码提交，支持对失败模式的直接分析。研究发现，模型的编码能力与其在该任务上的预测性能呈负相关，表明大语言模型更适合作为上下文条件下的生成式求解器，而非对学习者行为的忠实预测器。
链接: https://arxiv.org/abs/2609.00940
作者: Fagun Patel,Sang T. Truong,Duc Q. Nguyen,Kazunori Fukuhara,Benjamin W. Domingue,Sanmi Koyejo,Nick Haber
机构: Stanford University (斯坦福大学); National University of Singapore (新加坡国立大学)
类目: Computation and Language (cs.CL)
备注: EMNLP 2026 Findings
Abstract:Solving problems through repeated attempts is a sequential modeling task: at each step, the solver receives feedback and decides how to revise their solutions. Predicting whether performance improves, plateaus, or regresses across attempts is central to understanding any iterative problem-solving process in both human learners and autonomous agents. Beyond outcomes, modeling what errors persist and how strategies shift across attempts provides deeper insight into the mechanics of sequential learning. Studying these dynamics requires observing many solvers as they attempt, receive feedback, and revise. Programming courses with automated grading provide this setting, as students iteratively submit code to test suites and receive feedback on every attempt. We therefore curate CodeInsight, a large-scale dataset of over 3 million submissions from 3,286 undergraduates across 2 introductory C++ courses in 2 academic years, with test-case-level outcomes, timestamps, and source code. On this dataset, we build a benchmark that evaluates models spanning parametric, sequential, and generative traditions under a shared calibration-and-scoring protocol, including a Recurrent State Space Model (RSSM) adapted to track solver characteristics through discrete latent variables and an LLM-based predictor that generates explicit solutions. The adapted RSSM achieves the strongest predictive accuracy on three of the four courses. The LLM predictor is less accurate but produces full submissions at each attempt, enabling direct analysis of failure modes. We find that the model’s coding proficiency is inversely related to predictive performance in this setting, with the LLM better understood as a generative solver conditioned on context rather than a faithful predictor of solver behavior. We publicly release our code and the dataset on request to facilitate future research.
[NLP-66] DualStake: Dual-Path Confidence Calibration in Deep Research Agents EMNLP2026
【速读】： 该论文旨在解决深度研究型智能体（Deep Research agents）在处理知识密集型任务时存在的严重过度自信问题，这种过度自信导致其表达的置信度不可靠，难以建立用户信任或支持下游任务中的弃权决策。其核心解决方案是引入分步置信度提取机制，在每次检索步骤后主动获取证据置信度（Evidence Confidence, E-Conf），并基于实证发现：最终检索步骤后的E-Conf比答案生成后的答案置信度（Answer Confidence, A-Conf）提供更优的不确定性信号，且A-Conf主要受E-Conf所影响。据此提出双路径校准方法DualStake，通过施加边际截断、依赖置信度的奖励策略，联合优化E-Conf与A-Conf与答案正确性的一致性，同时抑制极端置信度的优化。在Qwen2.5-7B、Qwen2.5-7B-Instruct和Qwen3-4B三个模型上覆盖8个问答基准的实验表明，DualStake能持续提升置信度校准效果，且不损害答案准确性。
链接: https://arxiv.org/abs/2609.00935
作者: Yinuo Xu,Yuwei Liang,Jianjie Cheng,Meng Wang,Yongcan Yu,Shuo Lu,Jian Liang
机构: Institute of Automation, Chinese Academy of Sciences (CASIA); School of Artificial Intelligence, University of Chinese Academy of Sciences; Meituan Inc.
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Accepted to EMNLP 2026 Main
Abstract:Deep Research agents tackle knowledge-intensive tasks through multi-round retrieval and decision-oriented generation. However, these agents suffer from severe overconfidence, making their expressed confidence unreliable for user trust and downstream abstention. To address this, we augment the Deep Research pipeline with step confidence elicitation after each retrieval, building on the commonly used post-answer verbalized confidence. Interestingly, we find that Evidence Confidence (E-Conf), elicited after the final retrieval step, provides a stronger uncertainty signal than Answer Confidence (A-Conf), elicited after answer generation, and that A-Conf is largely shaped by E-Conf. Based on these findings, we propose DualStake, a dual-path calibration method that applies margin-clipped, confidence-dependent stake rewards to jointly align E-Conf and A-Conf with answer correctness while limiting extreme confidence optimization. Experiments on Qwen2.5-7B, Qwen2.5-7B-Instruct, and Qwen3-4B across 8 QA benchmarks demonstrate that DualStake consistently improves calibration without sacrificing answer accuracy. The code is available at this https URL.
[NLP-67] Context-Grounding Gains Are Mediated by Pre-existing Machinery: Auditing GRPO SFT and DPO
【速读】： 该论文旨在解决语言模型在推理过程中忽视与记忆知识相冲突的提示证据（prompt evidence）的问题，即模型倾向于依赖其预训练阶段所“记忆”的知识而非当前输入的上下文信息。其核心挑战在于：通过后训练（post-training）手段提升模型对提示证据的遵循能力是否需要引入全新的机制，还是仅能通过强化已有模型内部已存在的机制实现。研究的关键发现是，尽管多种后训练方法（包括GRPO、SFT和DPO）在不同规模和模型家族中表现各异，但绝大多数接地（grounding）性能的提升主要依赖于起始检查点（starting checkpoint）中已存在的潜在机制。具体而言，冲突敏感型监督微调（conflict-SFT）和直接偏好优化（DPO）均主要利用与起始模型相同的因果注意力头（causal attention-head set），且在去除起始模型方向后，二者增益显著下降；而将该方向重新注入起始模型可恢复约35%的DPO增益，且不引发明显副作用。此外，即使在引入监督预热（supervised warm start）以增强上下文响应出现频率后，相同的GRPO策略也未带来额外的接地提升。因此，研究结论表明，在当前设定下，接地性能的改进主要源于对起始模型中已有机制的强化，而非新架构或新机制的引入。
链接: https://arxiv.org/abs/2609.00925
作者: Prakhar Gupta,Vaibhav Gupta
机构: University of Michigan (密歇根大学); University of Waterloo ( Waterloo 大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Language models can ignore prompt evidence when it conflicts with memorized knowledge. Post-training can make models follow such evidence more reliably, but it is unclear whether these gains require new machinery or strengthen machinery already present. We compare nine post-training arms spanning GRPO, SFT, and DPO from one starting checkpoint, with key comparisons extended across scales and families. We estimate a grounding direction from that checkpoint before training. Across five tested GRPO variants, grounding gains are small. For the two variants replicated across seeds, equivalence tests bound their effects below the conflict-SFT gain even as the rewarded metric improves. Conflict-SFT improves grounding moderately, while DPO drives grounding near ceiling on its matched distribution. Conflict-SFT and DPO largely use the same causal attention-head set as the starting model. Subtracting the starting-model direction suppresses both gains, while adding it to the starting model recovers 35% of DPO’s gain at a dose passing all stated side-effect checks. After a supervised warm start makes the context answer appear in more rollouts, the same GRPO recipe adds essentially no further grounding gain. In our setting, grounding gains largely depend on machinery already present in the starting model.
[NLP-68] VIBE-Bench: Evaluating Personalized Large Language Models When Profiles Dont Mean Preferences EMNLP2026
【速读】： 该论文旨在解决个性化大语言模型（PLLMs）在用户偏好推理（preference reasoning）中存在的关键挑战，即当用户画像特征（profile cues）与查询相关的具体偏好处于不同概念空间时，传统基于语义相关性的历史信息检索方法失效的问题。这一现象被称为“画像-偏好概念错位”（Profile-Preference Conceptual Misalignment, PRCM），是实际应用中普遍存在但长期被忽视的难题。论文提出VIBE-Bench基准，包含3,504个角色设定与12,239条对话，涵盖两个心理学理论驱动的任务，并配有经人工验证的黄金测试集，要求模型进行超越表面语义重叠的跨概念偏好推理。实验表明，现有个性化方法严重依赖浅层语义关联，难以建立稳健的跨概念映射能力。因此，该研究确立了PRCM作为PLLMs中一类独立的性能失效模式，并将VIBE-Bench定位为推动偏好推理从语义匹配迈向深层认知建模的关键评估平台。
链接: https://arxiv.org/abs/2609.00921
作者: Yiwen Jiang,Yang Deng,Stephanie Fong,Zimu Wang,Yaling Shen,Wei Feng,Hongxi Yang,Xiangyu Zhao,Zhongxing Xu,Deval Mehta,Xuelian Cheng,Zongyuan Ge
机构: Monash University(莫纳什大学); Singapore Management University(新加坡管理大学); University of Liverpool(利物浦大学); RMIT University(皇家墨尔本理工学院)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: Accepted at EMNLP 2026 (Findings)
Abstract:Personalized Large Language Models (PLLMs) aim to tailor responses to individual users, where a central challenge is preference reasoning: inferring query-relevant preferences from user-related history. Existing benchmarks, however, largely assume that such preference can be retrieved from semantically related history. We study an underexplored but practically important regime, profile-preference conceptual misalignment (PRCM), where observable profile cues and query-specific preferences lie in different concept spaces, making semantic retrieval inconsistent for personalization. We introduce VIBE-Bench, a benchmark with two psychology-grounded tasks, 3,504 personas and 12,239 dialogues, including a manually verified gold test set, and requires cross-concept preference reasoning beyond surface semantic overlap. Experiments with several personalization methods show that current PLLMs largely rely on shallow semantic correlations and fail to acquire robust cross-concept mappings. These findings establish PRCM as a distinct failure regime in PLLMs and position VIBE-Bench as a focused testbed for advancing preference reasoning beyond semantic matching.
[NLP-69] RPCBench: A Benchmark for Proactive Premise Critique in LLM -based Recommendation
【速读】： 该论文旨在解决当前大型语言模型（LLM）在作为交互式推荐助手时，对自然语言推荐请求中潜在错误前提（faulty premises）的识别与处理能力不足的问题。现有推荐评估基准多聚焦于推荐排序、生成质量或用户偏好满足度，而现有的错误检测基准又缺乏与推荐场景相关的用户上下文和候选项证据支撑，导致无法有效评估模型在真实推荐情境下的推理鲁棒性。为此，论文提出RPCBench——一个面向推荐前提批判（Recommender-Premise Critique）能力的基准测试框架，其核心目标是评估模型在面对存在缺陷的推荐请求时，能否主动识别、准确诊断并合理应对其中的前提错误。该基准涵盖五个推荐领域，包含十类典型前提错误类型，每条测试实例均提供可观察的推荐上下文与被污染的用户查询，确保评估具有证据基础。关键解决方案在于构建了一个细粒度的评估体系，涵盖主动检测、错误定位、后检测处理策略及证据忠实性等维度。实验对11个主流大模型进行系统评估发现，主动检测能力是当前模型在推荐前提批判任务中的主要瓶颈，尤其在“前提不明确”类错误上表现最差；同时研究揭示目标关键信息密度比冗余证据更具价值，且推理长度并非越长越好——性能在中等推理长度时达到峰值，过长推理反而引发“过度思考惩罚”（overthinking penalty），导致质量下降。
链接: https://arxiv.org/abs/2609.00918
作者: Zhongru Chen,Yuan Wu,Yi Chang
机构: Jilin University (吉林大学); Engineering Research Center of Knowledge-Driven Human-Machine Intelligence, MOE (教育部知识驱动人机智能工程研究中心); International Center of Future Science, Jilin University (未来科学国际中心, 吉林大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 45 pages, 8 figures. Code available at this https URL
Abstract:Large language models are increasingly used as interactive recommender assistants. Their evaluation should therefore go beyond plausible item recommendation and test whether they can recognize flawed recommendation requests. Existing recommender benchmarks mainly assess ranking, generation, or preference satisfaction, while existing error-detection benchmarks are usually not grounded in recommendation-specific user and candidate evidence. To address this gap, we introduce RPCBench, a benchmark for evaluating Recommender-Premise Critique: the ability to detect, diagnose, and properly handle faulty premises in natural-language recommendation requests. RPCBench contains evidence-grounded test instances from five recommendation domains and covers ten types of premise failures. Each instance provides a visible recommendation context and a corrupted user query. We further design a fine-grained evaluation framework that measures proactive detection, error localization, post-detection handling strategy, and evidence faithfulness. Through a systematic evaluation of 11 LLMs, we find that proactive detection is the main bottleneck in Recommender-Premise Critique, and models perform worst on underspecified-premise errors. We also observe that target-critical information density matters more than redundant evidence, and that longer reasoning does not monotonically improve critique quality: performance peaks at intermediate reasoning length, while overly long reasoning is accompanied by an overthinking penalty. The code is available at this https URL.
[NLP-70] Membership Inference in Fine-tuned Diffusion Language Models via Token-level Memorization Asymmetry EMNLP2026
【速读】： 该论文旨在解决生成式模型中尚未充分探讨的隐私风险问题，特别是针对扩散语言模型（Diffusion Language Models, DLMs）在微调后可能存在的成员推断攻击（Membership Inference Attack）与个人身份信息（PII）提取等隐私泄露风险。其核心问题是：尽管DLMs具备并行生成和双向上下文建模的优势，但其训练动态可能导致数据层面的不均衡记忆现象，进而引发隐私漏洞。解决方案的关键在于提出一种名为Q-Skew的量化指标——基于分位数加权偏度（quantile-weighted skewness），用于检测微调后的DLMs中是否存在对训练数据成员身份的可推断性。该方法通过分析扩散训练过程中各令牌（token）的输出分布不对称性，有效捕捉到“令牌级记忆不对称性”（token-level memorization asymmetry）这一关键特征，并在多个数据集和模型上的实验验证中表现出优于现有基线的性能。此外，研究进一步证明Q-Skew还可被用于其他类型的隐私攻击，揭示了DLMs存在一个此前未被充分关注的隐私攻击面，强调了对DLMs进行系统性隐私评估的重要性。
链接: https://arxiv.org/abs/2609.00873
作者: Shengfang Zhai,Leo Marchyok,Yuling Shi,Huanran Chen,Yinpeng Dong,Jiaheng Zhang,Sanghyun Hong
机构: National University of Singapore(新加坡国立大学); Oregon State University(俄勒冈州立大学); Shanghai Jiao Tong University(上海交通大学); College of AI, Tsinghua University(清华大学人工智能学院)
类目: Computation and Language (cs.CL); Cryptography and Security (cs.CR)
备注: 18 pages. EMNLP 2026 (Findings)
Abstract:Diffusion language models (DLMs) have recently emerged as an alternative modeling paradigm to autoregressive LMs, offering advantages such as parallel generation and bidirectional context modeling. Despite growing interest in their generative capabilities, the privacy risks of DLMs remain underexplored. We identify a phenomenon termed token-level memorization asymmetry through theoretical analysis of diffusion training dynamics. Building on this finding, we propose Q-Skew, a quantile-weighted skewness-based indicator for membership inference on finetuned DLMs. Experiments across multiple fine-tuning datasets and models show that our method outperforms existing baselines. Moreover, we show that Q-Skew can also facilitate other privacy violations, such as PII extraction. Our findings reveal a previously underexplored privacy attack surface and highlight the need for systematic privacy evaluation of DLMs.
[NLP-71] he Visual Insensitivity Gap: Diagnosing When Vision-Language Models Fail to Use Visual Evidence
【速读】： 该论文旨在解决当前视觉语言模型（Vision-Language Models, VLMs）评估中一个关键的隐含假设问题：即模型在多模态基准测试中表现出的准确率能够反映其有效利用视觉输入的能力。研究发现，这一假设在实际中存在严重偏差——在六个VLMs和三个感知基准上的40%至97%样本中，即使对与问题相关的关键视觉区域进行模糊处理，模型的下一个词预测分布几乎未发生变化，表明模型并未真正依赖视觉信息。作者将此现象命名为“视觉不敏感差距”（Visual Insensitivity Gap），并提出每样本的视觉敏感性指数（Visual Sensitivity Index, VSI）进行量化。该差距是样本固有的属性而非模型特性：不同架构的VLMs在相同样本上的VSI排名高度一致（总体斯皮尔曼等级相关系数ρ=+0.40，置换检验p<10⁻³），说明即使仅共享对比预训练的视觉塔，各模型也普遍对同一组样本表现出视觉忽略行为。其内在机制表现为：尽管视觉编码器层可通过线性探测器以0.72–0.79的准确率区分扰动与原始图像，但模型最终输出的argmax token仅在2%–11%的样本上发生改变，反映出编码器与语言模型之间存在超过0.65的“编码器-语言模型差距”（encoder–LLM gap）。通过逐样本映射VSI的诊断能力，研究揭示其在多选推理任务中具有强诊断效能（AUROC=0.85–0.87），而在校准良好的事实性判断任务中则表现较弱，因此时软最大值置信度已具备足够判别力。因此，VSI并非普适的最佳回避信号，而是一种样本内生的视觉忽略失败指示器，应作为条件集成组件使用，以提升模型在关键场景下的可靠性与可解释性。
链接: https://arxiv.org/abs/2609.00868
作者: Genpei Zhang
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 17 pages (7-page main text plus technical appendix), 10 figures, 6 tables
Abstract:Vision-language models are evaluated by aggregate accuracy on multimodal benchmarks, a practice that implicitly assumes the model uses its visual input. We show this assumption fails on 40%–97% of samples across six VLMs and three perceptual benchmarks: blurring the question-relevant visual region leaves the next-token distribution nearly unchanged. We name this phenomenon the Visual Insensitivity Gap and quantify it with a per-sample Visual Sensitivity Index (VSI). The gap is a property of samples, not of models: VSI ranks correlate across models (grand-mean Spearman rho=+0.40, permutation p10^-3), so the same samples are flagged insensitive by VLMs sharing no architectural detail beyond a contrastively pretrained vision tower. The mechanism is concrete: on the insensitive samples, a linear probe on each model’s own vision tower distinguishes perturbed from clean images at 0.72–0.79 accuracy, yet the model’s argmax token changes on only 2%–11% of the same samples, an encoder–LLM gap above 0.65 on every model. Mapping VSI’s diagnostic utility cell by cell surfaces a strong regime (multi-choice reasoning on capable VLMs: AUROC=0.85–0.87) and a weak regime (well-calibrated factuality, where softmax confidence already leads). VSI is not a universal best abstention signal; it is a sample-intrinsic indicator of vision-ignoring failure, best used as a conditional ensemble component.
[NLP-72] MemoryWalker: Stop Training Agents on Contexts They Never Saw
【速读】： 该论文旨在解决生成式AI（Generative AI）在使用如Claude Code和Qwen-Agent等生产级代理时，因上下文压缩（context compression）导致的训练-推理不一致问题。其核心挑战在于：每次上下文淘汰（eviction）会分叉有效历史路径，使学习目标从线性序列变为树状结构，而现有线性化方法要么保留最右路径导致时间旅行泄露（time-travel leakage），要么采用深度优先遍历引发训练与推理间的分布偏移。本文提出两种精确且梯度等价的修正方案：一是基于分段K步前向遍历的LogitTree方法，需执行K+1次反向传播；二是采用打包4D注意力掩码，依赖定制内核与白盒淘汰记录。此外，提出一种单次反向传播的变分松弛方法SDCC（Self-Distillation for Conditioning Consistency），在每次淘汰时刻，通过最小化压缩后的学生模型与停止梯度教师模型在重构前缀上的前向KL散度，实现条件一致性建模。每个节点处的残差KL值为ε_KL时，可保证训练-部署间总变差距离上界为O(√ε_KL)。SDCC亦适用于黑盒代理框架。在包含TC-RAG、AgentFold、MemexRL、Claude Code和OpenCode在内的七个网页搜索基准测试中，未经修正的朴素训练显著扩大了训练与回滚阶段的对数概率差距，尤其在高淘汰率批次中；而精确方法保持在无压缩基线水平，SDCC则大幅缩小该差距，表现出更低的对数概率漂移与更高的回滚奖励。
链接: https://arxiv.org/abs/2609.00865
作者: Zinco J,Xunjie Zhu,Shen Huang,Zhenyi Wang,Pengjun Xie,Jieping Ye
机构: Token Foundry, Alibaba Group(阿里巴巴集团)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: Your Memory-Compressing Harness Makes Training and Inference Inconsistent
Abstract:Production agent harnesses such as Claude Code and Qwen-Agent compress context during rollout, but training under compression creates a conditioning problem: every eviction branches the effective history, so the learning object is a tree rather than a sequence. Existing linearizations either retain the rightmost path, causing time-travel leakage, or replay a depth-first traversal, causing train-inference mismatch. We introduce two exact, gradient-equivalent corrections: LogitTree, a segmented K-forward traversal, and a packed 4D attention mask. LogitTree requires K+1 backward passes; the 4D mask requires a custom kernel and white-box eviction records. We also propose SDCC (Self-Distillation for Conditioning Consistency), a single-backward-pass variational relaxation. At each eviction, it minimizes forward KL between the compressed student and a stop-gradient teacher on the reconstructed pre-eviction prefix. A residual per-junction KL of epsilon_KL gives an O(sqrt(epsilon_KL)) bound on the train-deployment total-variation gap. SDCC also applies to black-box harnesses. On seven web-search benchmarks with TC-RAG, AgentFold, MemexRL, Claude Code, and OpenCode, naive training inflates the train-rollout log-probability gap, especially on eviction-heavy batches. The exact methods stay at the no-compression floor, and SDCC substantially closes the gap, with lower logit drift and higher rollout rewards.
[NLP-73] Verifiable Disaster Storylines and Causal Knowledge Graphs: A Citation-Grounded Pipeline from Heterogeneous Humanitarian Sources
【速读】： 该论文旨在解决人道主义应急响应中面临的快速整合异构、高体量信息源的难题，这一任务在危机初期的关键时段往往超出人类分析能力。其核心解决方案是构建一个融合结构化灾害记录（来自EM-DAT）与非结构化文档（来自ReliefWeb和欧洲媒体监测系统EMM）的自动化处理流水线，利用检索增强生成（Retrieval-Augmented Generation, RAG）技术，从多源数据中提取结构化的灾害叙事（disaster storylines），生成包含17个字段的表格化事件档案，涵盖灾害严重性、关键驱动因素及儿童敏感型影响指标等，并构建基于因果关系的知识图谱。该知识图谱中的每个节点与边均通过引用溯源的解释性叙述进行丰富，确保内容可追溯至原始文献，从而实现完全的可解释性与可信度。评估结果显示，系统具备高召回精度、强因果关系忠实性，且领域专家明确偏好具有引文支持的组件。该流水线具备扩展至完整EM-DAT数据库的能力，目标是公开发布一个富含叙事语境的数据库版本，以增强应急响应人员与分析者的态势感知能力。
链接: https://arxiv.org/abs/2609.00858
作者: Ivan Decostanzi,Michele Ronco,Sergio Consoli,Christina Corbane,Lorenzo Bertolini,Indaco Biazzo,Daria Mihaila,Manuel Garcia-Herranz,Felix Schwebel,Yelena Mejova,Kyriaki Kalimeri
机构: ISI Foundation( ISI基金会); European Commission, Joint Research Centre (JRC)( 欧洲委员会联合研究中心); UNICEF( 联合国儿童基金会)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Effective humanitarian response depends on the rapid synthesis of heterogeneous, high-volume information sources - a task that routinely exceeds human analytical capacity in the critical early hours of a crisis. We present a pipeline that combines structured disaster records from EM-DAT with unstructured documents from ReliefWeb and the European Media Monitor (EMM) to produce source-grounded disaster storylines and causal knowledge graphs supporting situational awareness for responders and analysts. Using Retrieval-Augmented Generation, the pipeline extracts structured storylines - tabular event profiles covering 17 fields, from severity and key drivers to child-sensitive impact indicators - and constructs causal knowledge graphs where each node and edge is enriched with citation-grounded explanatory narratives, enabling full traceability back to primary sources. We evaluate the system on three diverse crisis use cases through a human evaluation involving 9 domain expert and 9 non-expert evaluators. Results confirm high retrieval precision, strong faithfulness of extracted causal relations, and a clear expert preference for citation-grounded components over ungrounded alternatives. The pipeline is designed to scale to the full EM-DAT catalogue, with the goal of publicly releasing a narrative-enriched version of the database.
[NLP-74] Replacing Training with Memory: Listwise Selection for Text-to-SQL EMNLP2026
【速读】： 该论文旨在解决现代Text-to-SQL系统中基于生成-执行-选择（generate-execute-select）范式的列表式选择（listwise selection）方法在实际应用中因微调成本过高而难以推广的问题。现有方法通常依赖于对列表式选择器进行昂贵的微调以学习有效的查询排序策略，且易受候选查询位置偏差的影响。为此，本文提出一种无需微调的列表式选择框架MaP-SQL，其核心创新在于将传统的微调目标转化为推理阶段可执行的策略：一是通过构建可复用的结构化记忆（structured memories），替代模型参数来显式编码自然语言到数据库模式元素、SQL操作及预期输出之间的映射关系，从而在推理时提供明确的选择判据；二是通过多轮输入排列的排名聚合机制缓解列表式选择中的位置偏差，同时利用执行结果和点对点评分优化推理开销。该方法在不进行任何微调的前提下，在多个Text-to-SQL基准上实现了更稳定的选择性能，显著减少不必要的比较次数，并在BIRD-dev数据集上相较于先前最优的R^3-SQL方法提升了2.02个百分点的执行准确率，同时仅需其2.92倍的令牌消耗，展现出卓越的效率与兼容性。
链接: https://arxiv.org/abs/2609.00834
作者: Yeonseok Jeong,Soyoung Yoon,Seongjun Lee,Seung-won Hwang
机构: Seoul National University (首尔国立大学); KAIST (韩国科学技术院)
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: Accepted by Findings of EMNLP 2026
Abstract:Modern Text-to-SQL systems often follow generate-execute-select pipelines, generating multiple candidate queries then selecting the best one. Listwise selection, by jointly comparing multiple candidates, has been widely adopted, but fine-tuning listwise selectors is costly. We thus propose a fine-tuning-free listwise selector. We replace two major fine-tuning objectives with inference-time strategies: (1) learning selection criteria as ordering and (2) mitigating positional bias. First, we build reusable structured memories instead of learning selection behavior as model parameters. Given a question, MaP-SQL retrieves memories distilled from training data that encode how natural language maps to schema elements, SQL operations, and expected outputs. These memories serve as explicit decision criteria for evaluating candidates in a listwise manner. Second, to mitigate ordering bias of listwise selectors, we aggregate rankings across multiple input permutations, with inference cost optimized by execution results and pointwise scoring. Our approach improves selection accuracy while maintaining efficiency and compatibility with existing large language models. Across Text-to-SQL benchmarks, it produces more stable selection without fine-tuning and fewer unnecessary comparisons than existing methods. On BIRD-dev, it outperforms the previous state-of-the-art selector-based method R^3-SQL by 2.02 execution accuracy points on average using the same candidate sets, with 2.92x fewer tokens.
[NLP-75] Dense Process Supervision for Search Agents via Fact Utility Estimation EMNLP2026
【速读】： 该论文旨在解决强化学习（Reinforcement Learning, RL）在搜索代理中因最终结果奖励信号模糊而导致的信用分配（credit assignment）难题，尤其在多跳问答（multi-hop QA）任务中，中间推理步骤的价值难以量化，致使模型无法有效区分各推理环节对最终结果的贡献。其解决方案的关键在于提出一种基于事实效用估计的密集过程监督方法，通过将推理过程建模为离散证据事实的累积过程，首先从原始观测中提取结构化事实并构建显式的事实存储库；随后，利用语义聚类对等价事实进行归并，并基于群体回溯（group rollouts）采用贝叶斯估计推断每个事实簇的后验效用；最终，将估算的事实效用转化为细粒度的步骤级奖励，以支持更精准的信用分配与策略优化。实验在七个单跳与多跳问答基准上验证了该方法的优越性，消融实验进一步表明，在多跳问答任务中，相比仅依赖结果奖励的训练方式，该方法显著提升了性能。
链接: https://arxiv.org/abs/2609.00833
作者: Rongzhi Zhu,Xiangyu Liu,Yi Liu,Shuo Zhang,Ruirui Zhang,Rui Wu,Tao Jiang,Zequn Sun,Wenhao Xu,Wei Hu
机构: Nanjing University (南京大学); Ant Group (蚂蚁集团); State Key Laboratory for Novel Software Technology (新型软件技术国家重点实验室); National Institute of Healthcare Data Science (健康医疗数据科学国家研究所)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: Accepted in the 2026 Conference on Empirical Methods in Natural Language Processing (EMNLP 2026)
Abstract:Reinforcement learning (RL) for search agents typically relies on outcome rewards. However, it often fails to achieve effective credit assignment, due to the unclear value of intermediate steps. It is hard to separate their contributions from the final result. In this paper, we propose a dense process supervision method based on fact utility estimation, which models the reasoning process as the accumulation of discrete evidence facts. We first extract structured facts from raw observations and organize them into an explicit fact store. To support credit assignment, we then cluster semantically equivalent facts and infer the posterior utility of each fact cluster using Bayesian estimation over group rollouts. Finally, we convert the estimated fact utilities into dense step-level rewards to guide RL training. Experiments on seven single-hop and multi-hop QA benchmarks show that our method consistently outperforms existing baselines. Ablation studies validate clear relative improvements on multi-hop QA compared to outcome reward-only training.
[NLP-76] WIX: a Two-Stage Approach for End-To-End Named Entity Recognition and Relation Extraction
【速读】： 该论文旨在解决科学文献爆炸式增长背景下，自动化信息抽取（Information Extraction, IE）系统在肠道-脑轴（gut-brain axis）领域中对命名实体识别（Named Entity Recognition, NER）、命名实体识别与消歧（Named Entity Recognition and Disambiguation, NERD）以及关系抽取（Relation Extraction, RE）等任务的性能瓶颈问题。其核心挑战在于如何在复杂且高度专业化的生物医学文本中，高效、准确地识别并关联多类实体及其语义关系。该研究提出的解决方案——两阶段信息抽取工作流（Two-stage Workflow for Information eXtraction, TWIX），采用端到端的三模块集成架构，每个模块均基于两阶段框架，协同完成四项子任务。关键创新在于通过分阶段建模与模块间信息传递机制，显著提升了实体识别的精确率（precision）与召回率（recall），在开发集和测试集上的实验结果表明，TWIX不仅大幅超越基线模型，还在所有参赛方案中各项任务上均取得第一，验证了该两阶段流水线在实际应用中的有效性与鲁棒性。
链接: https://arxiv.org/abs/2609.00832
作者: Marco Martinelli,Laura Menotti
机构: 未知
类目: Computation and Language (cs.CL)
备注: Accepted at CLEF 2026: the 17th Conference and Labs of the Evaluation Forum
Abstract:The exponential growth of scientific publications calls for automatic Information Extraction (IE) systems to support knowledge discovery. In this context, the GutBrainIE benchmark evaluates Named Entity Recognition (NER), Named Entity Recognition and Disambiguation (NERD), and Relation Extraction (RE) systems in the gut-brain axis domain. We propose Two-stage Workflow for Information eXtraction (TWIX), an end-to-end IE pipeline featuring three interconnected modules, each leveraging a two-stage framework to solve all four GutBrainIE subtasks. Evaluation on the development and test sets shows that our method substantially outperforms the baseline by a wide margin, while also ranking first among all participant submissions across all subtasks. These results indicate that the proposed two-stage pipeline effectively improves both precision and recall in practical settings.
[NLP-77] Polished but Unresolved: Identifying Late-Stage Pressure States in Long-Horizon Tool-Use Agents EMNLP2026
【速读】： 该论文旨在解决长时程工具使用智能体（long-horizon tool-use agents）在决策过程中面临的“晚期压力状态”（late-stage pressure states）问题，即智能体在未充分解决关键约束条件的情况下，倾向于过早提交看似完整且精致的最终答案。这一现象导致任务执行质量下降。其解决方案的关键在于：首先通过线性探测器（linear probe）识别出该压力状态可从智能体的隐藏状态中被有效捕捉；随后利用激活干预（activation interventions）沿压力方向调节隐藏状态，发现此举能同时降低压力评分并影响智能体是否继续使用工具或提前提交结果。进一步实验表明，约束清晰度和动作映射（action mapping）可缓解压力。基于上述发现，作者提出一种名为“探针感知压力缓解”（Probe-Sensed Pressure Relief, PSPR）的轻量级插件机制，该机制在中等压力下施加轻量级压力缓解方向，在高风险压力场景下则转向结构化组织策略。在多个长时程基准测试上的实验表明，PSPR能持续增强现有智能体方法的性能。
链接: https://arxiv.org/abs/2609.00823
作者: Haoyang Chen,Yi Liu,Jianzhi Shao,Xiaozhou Xu,Zhe Sun,Wei Hu
机构: Nanjing University (南京大学); Alibaba Group (阿里巴巴集团)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: Accepted in the 2026 Conference on Empirical Methods in Natural Language Processing (EMNLP 2026)
Abstract:Long-horizon tool-use agents need not only to search and plan, but also to decide when to finalize. We study late-stage pressure states, in which an agent is biased toward submitting a final answer that appears complete and polished while key constraints remain unresolved. We first train a linear probe to show that this pressure state is identifiable from the agent’s hidden states. Then, we use activation interventions along this pressure direction and find that shifting the hidden states changes both the pressure score and whether the agent continues tool use or submits early. Through controlled context manipulations, we further see that the pressure is mitigated by constraint clarity and action mapping. Based on these findings, we propose Probe-Sensed Pressure Relief (PSPR), a plugin that applies lightweight pressure relief direction under moderate pressure and moves to structured organization under high pressure risk. Experiments on multiple long-horizon benchmarks show that our method consistently strengthens existing agent methods.
[NLP-78] SFAD: Speculative Factuality-Aware Decoding
【速读】： 该论文旨在解决大语言模型在知识密集型应用中面临的上下文忠实性（contextual faithfulness）问题，即如何在保证生成内容与输入上下文事实一致性的前提下，避免因推理效率低下而影响实际部署。现有方法如对比解码（contrastive decoding）需进行双重前向传播以比较有无上下文的输出，导致推理开销翻倍；而基于强化学习的后训练对齐方法则需要大量计算资源进行训练。为此，论文提出一种名为SFAD的推测解码框架，其核心创新在于通过构建细粒度原子扰动的偏好数据集ConFide，并利用直接偏好优化（Direct Preference Optimization）训练出具备上下文忠实性的草稿模型。在推理阶段，采用“认知摩擦”（Epistemic Friction）机制量化专家置信度加权下的分布张力，以检测潜在幻觉；当摩擦超过阈值时，通过残差式对数概率注入（Asymmetric Logit Steering）修正目标分布，否则执行标准推测解码。实验表明，SFAD在显著提升生成忠实性的同时实现了2.48倍的加速，为高效且可靠的大型语言模型提供了可行方案。
链接: https://arxiv.org/abs/2609.00796
作者: Guanqiao Chen,Di Wang,Lijie Hu
机构: MBZUAI( Mohamed Bin Zayed University of Artificial Intelligence); University of Science and Technology of China(中国科学技术大学); Provable Responsible AI and Data Analytics (PRADA) Lab; King Abdullah University of Science and Technology(阿卜杜拉国王科技大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:As one of the most critical challenges in large language models, contextual faithfulness directly determines their reliability in knowledge-intensive applications. This task is particularly challenging as it requires balancing factual consistency with generation efficiency. Contrastive decoding methods require dual forward passes (with and without context) to compare model outputs, doubling inference computational overhead, while post-training alignment demands extensive reinforcement learning with substantial computational overhead. To address this challenge, we present \textbfSFAD, a speculative decoding framework that enhances contextual faithfulness without inference degradation. We first construct \textbfConFide, a preference dataset with fine-grained atomic perturbations, to train a context-faithful draft model via Direct Preference Optimization. During inference, Epistemic Friction detects potential hallucinations by quantifying distributional tension weighted by specialist certainty. When friction exceeds the threshold, Asymmetric Logit Steering refines the target distribution through residual-based logit injection; otherwise, standard speculation proceeds. Extensive experiments demonstrate that SFAD substantially improves faithfulness while achieving 2.48\times speedup, offering a practical solution for efficient LLMs.
[NLP-79] Instella-MoE Technical Report
【速读】： 该论文旨在解决生成式 AI 模型在保持高性能的同时，实现高效训练与推理的难题，尤其是在资源受限环境下构建可复现、全开源的大规模稀疏专家混合（Mixture-of-Experts, MoE）语言模型。其核心挑战在于如何在不依赖闭源权重的前提下，通过系统级优化与架构创新，实现高参数量但低激活参数开销的模型训练与部署。解决方案的关键在于引入两项核心技术：一是门控多头潜在注意力（Gated Multi-head Latent Attention, Gated MLA），通过动态调节注意力路径提升计算效率与表征能力；二是远距离跳连-集体连接机制（FarSkip-Collective connectivity），优化专家间通信结构以增强跨专家信息流动并支持大规模并行训练。结合多阶段训练流程（包括预训练、长上下文扩展、反馈驱动微调、直接偏好优化及多教师在线策略蒸馏），该模型在仅2.8亿活跃参数/令牌的条件下，实现了160亿总参数的高效训练，并在多个基准测试中超越现有全开源模型，同时在指令遵循、推理、数学、编程等任务上表现优于同类开放权重模型，展现出卓越的性能与可复现性。
链接: https://arxiv.org/abs/2609.00791
作者: Jiang Liu,Sudhanshu Ranjan,Prakamya Mishra,Yonatan Dukler,Gowtham Ramesh,Jialian Wu,Ximeng Sun,Wen Xie,Chaojun Hou,Vikram Appia,Zhenyu Gu,Zicheng Liu,Emad Barsoum
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:In this work, we introduce Instella-MoE, a fully open Mixture-of-Experts (MoE) language model with 16 billion total parameters and 2.8 billion active parameters per token, trained entirely from scratch on AMD Instinct MI300X and MI325X GPUs. Instella-MoE combines a sparsely activated MoE design with architectural and system-level innovations, including Gated Multi-head Latent Attention (Gated MLA) and FarSkip-Collective connectivity, enabling efficient large-scale training and inference. The model is developed through a multi-stage pipeline comprising pre-training, mid-training, long-context extension, supervised fine-tuning with feedback-driven data curation, direct preference optimization, and reinforcement learning with Multi-Teacher On-Policy Distillation. Instella-MoE achieves an average score of 76.7 across standard pre-training benchmarks, outperforming prior fully open models including OLMo-3-7B, SmolLM3-3B, and OLMoE-1B-7B, while remaining competitive with open-weight MoE and dense baselines at comparable active-parameter scales, including Moonlight-16B-A3B and Qwen3.5-4B. After post-training, our final Think checkpoint achieves an average score of 73.2 across instruction-following, reasoning, math, coding, and chat benchmarks, outperforming both fully open and open-weight models with comparable or larger active parameter counts in our evaluation. To support transparent and reproducible research, we release the complete Instella-MoE model flow, including model weights, training configurations, data mixtures, and training code. Together, these contributions establish Instella-MoE a strong, fully open foundation for efficient, high-performing MoE models and reproducible research.
[NLP-80] When Features Become Instances: Inverted Contrastive Learning for Unsupervised Feature Selection
【速读】： 该论文旨在解决无监督特征选择（Unsupervised Feature Selection, UFS）中因缺乏类别标签而导致特征重要性难以直接定义的问题。现有方法通常依赖于间接的结构化准则（如相似性保持、局部性、稀疏性、聚类几何或重构质量），但这些准则在捕捉真实特征语义信息方面存在局限。本文提出一种基于表示一致性（representation consistency）的新视角，引入面向无监督特征选择的反向对比学习（Inverted Contrastive Learning for Unsupervised Feature Selection, ICLFS），将UFS重新建模为对特征而非样本的表示学习问题。其核心创新在于：首先对数据矩阵进行转置，使每个特征由其样本轮廓向量（sample-profile vector）表示；随后构建多个掩码正例视图与一个打乱负例视图，并在InfoNCE目标函数下学习投影空间中对结构化扰动保持一致的特征表示。进一步地，受近期研究启发，利用投影空间嵌入向量的模长作为特征显著性信号进行排序，再通过拉普拉斯门控排名修正（Laplacian-Gated Ranking Correction）机制抑制局部冗余特征，同时保留关键显著特征。实验结果表明，ICLFS在12个基准数据集上于标准聚类评估协议下，在10个数据集上达到最优聚类准确率，显著优于经典及神经网络基线方法，验证了特征级对比表示一致性作为替代邻域、聚类和重构等传统范式的一种强大且有效的解决方案。
链接: https://arxiv.org/abs/2609.00782
作者: Utsab Ghosh,Roshni Chakraborty
机构: ABV-Indian Institute of Information Technology and Management, Gwalior, India(ABV-印度信息科技与管理学院，古瓦洛尔，印度)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Unsupervised feature selection seeks a compact subset of informative features without access to class labels, making feature utility difficult to define. Existing UFS methods therefore rely on indirect structural criteria, such as similarity preservation, locality, sparsity, cluster geometry, or reconstruction quality. In this paper, we instead study UFS through representation consistency and propose Inverted Contrastive Learning for Unsupervised Feature Selection (ICLFS), a feature-wise contrastive framework that reformulates UFS as a representation learning problem over features rather than samples. ICLFS first inverts the data matrix so that each feature is represented by its sample-profile vector, then constructs multiple masked positive views together with a shuffled negative view, and learns projector-space representations that remain consistent across these structured perturbations under an InfoNCE-based objective. Motivated by recent findings that cosine-based and InfoNCE-based training affect embedding norms, we use projector-space embedding magnitude as the saliency signal for ranking features. The resulting norm-based ranking is subsequently refined through Laplacian-Gated Ranking Correction, which suppresses locally redundant candidates while preserving salient ones. Extensive experiments on 12 benchmark datasets show that ICLFS achieves the best clustering accuracy on 10 datasets against both classical and neural baselines under the standard clustering-based UFS evaluation protocol, while remaining competitive on the other two. These results show that feature-wise contrastive representation consistency provides a strong and effective alternative to neighborhood, cluster, and reconstruction-based UFS formulations.
[NLP-81] A Unified Mechanistic Analysis of Knowledge- and Safety-Based Refusals EMNLP2026
【速读】： 该论文旨在解决生成式人工智能（Generative AI）中知识型拒绝（knowledge-based refusal, KR）与安全型拒绝（safety-based refusal, SR）虽表现形式相似，但二者机制是否共享的科学问题。现有研究多将两类拒绝行为孤立分析，缺乏对内在关联性的系统探讨。为此，作者构建了一个包含213组对比四元组的新数据集，联合探测两种拒绝类型。研究发现，KR与SR共享一个共同的拒绝方向（refusal direction），但其机制存在不对称的重叠：安全型拒绝信号向知识型拒绝的迁移强度显著高于反向迁移。在模型深层结构中，两类拒绝表现出显著的特异性分化——知识型拒绝主要与不确定性及知识表征相关，而安全型拒绝则更关联于安全规范与政策表征。基于此，作者提出“先承诺、后指定”（commit-then-specify）的拒绝机制框架：初始阶段为共享的拒绝决策过程，随后在高层网络中通过特定类型的特征进一步明确拒绝依据是认知层面的知识缺失，还是规范层面的安全约束。
链接: https://arxiv.org/abs/2609.00760
作者: Yuri Son,Seunghee Kim,Hyuhng Joon Kim,Taeuk Kim
机构: Hanyang University (汉阳大学); Samsung Electronics (三星电子)
类目: Computation and Language (cs.CL)
备注: Accepted to EMNLP 2026 (Main Conference)
Abstract:Large language models (LLMs) are increasingly trained to decline queries that fall outside their knowledge (knowledge-based refusal, KR) or violate safety policies (safety-based refusal, SR). Although KR and SR result in superficially similar responses, they have largely been studied in isolation, leaving open whether they share an underlying mechanism. We address this gap with a systematic study on a new dataset of 213 contrastive quadruples that jointly probe both refusal types. We find that KR and SR are governed by overlapping yet distinguishable mechanisms. Both share a refusal direction, yet the overlap is asymmetric: SR signals transfer more strongly to KR than the reverse. Type-specific specialization emerges mainly in upper layers, with KR aligning with uncertainty- and knowledge-related representations and SR with safety- and policy-related ones. We thus characterize refusal as a commit-then-specify process: a shared initial mechanism commits to refusing, then type-specific features in later layers specify whether the grounds are epistemic or normative.
[NLP-82] Compile Dont Memorize: A Context Compilation Architecture (CCA) for In-Context Learning EMNLP2026
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在上下文学习（In-Context Learning, ICL）任务中表现脆弱的问题，即模型因忽略上下文中某一细微规则而导致整个响应失败，即使在强开源模型上，任务通过率也仅达12%-16%。其核心问题是当前主流“读取与推理”（read-and-reason）范式在单次前向传播中完成信息提取、规划、生成与自我验证，缺乏对复杂上下文的结构化处理能力。为此，论文提出上下文编译架构（Context Compilation Architecture, CCA），其关键创新在于引入一种具有固定槽位的类型化中间表示（Typed Intermediate Representation, IR），包括规则必须执行（rules.must_do）、禁止项（must_not）、条件规则（conditional）、输出规范（output_spec）、可用工具（available_tools）和数据概要（data_profile）等结构化字段，将任意自然语言上下文一次性编译为可形式化处理的语义结构。在此基础上，后续通过可执行验证器与违规触发的纠错循环实现鲁棒性提升。在涵盖4个基础模型的CL-bench基准测试中，CCA在所有模型上均优于原始提示、ReadAgent-P和Ctx2Skill等长上下文基线方法，尤其在规则密集型子任务中表现显著提升，例如将Kimi K2.5的通过率从15.4%提升至21.4%。该方案揭示了结构化上下文编译在增强模型对复杂规则依赖任务的适应性方面的重要价值。
链接: https://arxiv.org/abs/2609.00759
作者: Jinhu Qi,Minda Hu,Wentao Zhang,Weiqiang Jin,Yanyu Chen,Junli Wang,Irwin King
机构: The Chinese University of Hong Kong(香港中文大学); Macao Polytechnic University(澳门理工学院); Xi’an Jiaotong University(西安交通大学); University of Science and Technology of China(中国科学技术大学)
类目: Computation and Language (cs.CL)
备注: Accepted to EMNLP 2026 (Findings). Code, data, and cached completions available at this https URL
Abstract:Large language models (LLMs) increasingly handle in-context learning (ICL) tasks where a long, novel context defines the rules, knowledge, and output schema for a series of questions. On benchmarks that grade against every detail of the context, even strong open-weights models pass only 12-16% of tasks: a single overlooked rule fails the whole response. We argue this brittleness is structural: the dominant “read-and-reason” paradigm asks the model to extract, plan, generate, and self-verify in one forward pass. We therefore ask whether explicit context compilation can fix it, how it compares to existing long-context strategies (gist retrieval, multi-agent self-play), and where the resulting harness benefit holds across task structure and model scale. We propose the Context Compilation Architecture (CCA), whose central novelty is a typed intermediate representation (IR) with fixed slots (rules.must_do, must_not, conditional, output_spec, available_tools, data_profile) into which any prose context is compiled once; executable verifiers and a violation-gated correction loop follow as downstream consequences. On CL-bench (1,899 tasks across 4 open base models), CCA outperforms vanilla prompting and two long-context baselines (ReadAgent-P, Ctx2Skill) on every base model, lifting Kimi K2.5 from 15.4% to 21.4% with gains concentrated on rule-dense sub-categories. Code and cached completions are available at this https URL.
[NLP-83] Joint Training Is Not Enough: Conditioned Cross-Granularity Training for Multimodal Document Understanding
【速读】： 该论文旨在探究在多模态文档理解任务中，细粒度（span-level）与粗粒度（document-level）任务之间是否存在相互增强效应（Mutual Reinforcement Effect, MRE），即单一模型同时处理两类任务时，二者是否能通过协同学习实现性能提升。其核心问题在于：在联合训练框架下，不同粒度任务间的知识共享是否真正带来跨粒度的性能增益，而非简单的性能权衡。解决方案的关键在于提出并验证“条件化训练”（conditioned training）范式——在训练过程中仅将某一粒度任务的黄金标注（gold output）作为另一粒度任务提示（prompt）的一部分，从而构建任务间的因果引导机制。研究通过在三组数据集（两组收据、一组扫描版商业表单）上对比单任务训练、混合联合训练与条件化训练，发现传统混合联合训练无法在任一数据集上实现跨粒度优势；而条件化训练在两个数据集（CORD 和表单数据集）上显著优于单任务模型，分别在细粒度和粗粒度指标上取得可衡量的提升，并在表单数据集中有效避免了语义标签分布坍塌问题。此外，通过控制实验揭示，这种增益主要源于提示结构的设计而非内容本身，且信息解码能力在各训练范式间无显著差异，表明增强效应并非来自隐含表示的改善，而是由任务间结构化引导所驱动。
链接: https://arxiv.org/abs/2609.00756
作者: Chengguang Gan,Yunhao Liang,Hanjun Wei,Qinghao Zhang,Shiwen Ni
机构: Independent Researcher; University of Chinese Academy of Sciences; Department of Information Convergence Engineering, Pusan National University, South Korea; Shenzhen University of Advanced Technology
类目: Computation and Language (cs.CL)
备注:
Abstract:The Mutual Reinforcement Effect (MRE) asks whether a fine, span-level and a coarse, document-level task help each other when one model handles both. We test it in multimodal document understanding on three corpora, two of receipts and one of scanned business forms, comparing single-task, joint and conditioned training, which puts one granularity’s gold output in the other’s prompt during training only. We build Doc-MRE, an annotation layer pairing gold field extraction (point) with four document-level facets (line), from a three-judge LLM committee under a pre-registration, validated by blind re-annotation. One predicate, fixed in advance: at a shared recipe, a regime reinforces if it beats the matched single-task model on both granularities. Mixed joint training, the arrangement prior MRE work assumes, reinforces on no corpus at the main scale: it is below both single-task models on CORD and trades one granularity for the other on the two others, as single-task tuning does. Conditioned training reinforces on two of the three, CORD (+0.5 point, +4.8 line) and the forms corpus (+7.2 point, +11.0 line), resolvably on the coarse side and directionally on the fine one, and trades on WildReceipt; at that recipe no alternative measurably beats it on either side anywhere. Two byte-identical-prompt controls separate content from format: shuffled conditioning destroys the coarse-side skill but costs the fine side far less, and a neutral-content control reproduces the whole fine-side gain on WildReceipt, which is therefore prompt structure but buys nothing resolvable on the other two. On the forms corpus conditioning buys collapse avoidance: mixed training and the neutral control both assign the majority semantic label to all 50 test documents; only conditioning recovers the gold distribution. Probes find the information decodable under every regime with no resolvable increase under conditioning.
[NLP-84] How Do Language Models Choose Between Context and Memory?
【速读】： 该论文旨在解决在模型参数知识与上下文信息冲突时，如何准确识别并操控模型决策来源的问题。现有方法依赖于激活方向（activation directions）来引导模型遵循上下文或参数知识，但此类方法无法验证方向是否具有因果性：即该方向是否为模型自然使用的内在机制，以及是否可在不同任务间复用。为此，作者通过反事实实验，在无歧义的设定下进行验证。首先，从一致提示（agreement prompts）中估计权威方向（authority directions），即上下文与参数知识支持相同答案的情形；随后，在匹配的提示对之间交换沿这些方向的自然坐标，强制模型优先选择上下文或参数知识。实验结果表明，在Qwen、Llama和OLMo等多个模型上，该干预可重现30%-68%由权威诱导的来源选择变化，而对照组几乎无效果。进一步测试跨任务可复用性发现，分别在两个任务上学习的权威方向仅能弥补9%的权威差距，而针对特定任务学习的本地方向则可实现57%的修复。该研究明确区分了权威表征、因果使用及跨任务因果复用三个维度，揭示权威计算可能具有任务特异性，而非通用可复用的机制。
链接: https://arxiv.org/abs/2609.00753
作者: Benjamin Shih,John Winnicki,Arianna Cao
机构: Stanford University (斯坦福大学); Perpetual Labs
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注:
Abstract:When contextual information conflicts with the knowledge stored in model parameters, activation directions can be used to decode and steer which source the model follows. However, steering along a direction does not establish causality: whether the unedited model would naturally use that direction or whether the direction is reusable across tasks. We test these distinctions through counterfactual experiments in unambiguous settings. First, we estimate authority directions from agreement prompts, in which the context and parametric knowledge support the same answer. We then interchange naturally occurring coordinates along these directions between matched prompts that direct the model to prioritize either the supplied context or its parametric knowledge. Across Qwen, Llama, and OLMo models, this intervention reproduces 30-68% of the authority-induced shift in source choice, whereas matched controls reproduce almost none. To test cross-task reuse, we learn authority directions on two tasks separately and see that cross-task transferability closes only 9% of the authority gap while the local direction learned on the given task closes 57%. These results distinguish authority representation, causal use, and cross-task causal reuse, and suggest that authority computations may be task-dependent, rather than reusable across tasks.
[NLP-85] Measuring Optimal Transport in Transformer Depth NEURIPS2026
【速读】： 该论文试图解决的问题是：预训练语言模型在深层网络中如何通过各层传递词元（token）的状态，其状态迁移是否符合最优传输（Optimal Transport, OT）理论所预测的最优路径与最小成本。具体而言，研究关注的是模型在不同层之间移动词元“云”（即所有词元状态构成的集合）时，是否以最低代价沿最优匹配路径进行转移。解决方案的关键在于：采用精确的连续层间词元配对（exact assignment）、引入采样下界（sampling floor）以校准测量结果、基于已知最优耦合关系进行校准，并将总传输成本分解为整体云的平移（common shift）与个体词元的特定移动（token-specific moves）。研究发现，在最后一层，两个模型（Pythia-160m 和 Pythia-410m）的词元迁移方向与最优传输映射高度一致，其中 Pythia-410m 达到理论最优成本，而 Pythia-160m 略高于最优值；但在初始层则未表现出此特性。中间层的单层转移仅在少数过渡中接近最优成本，而多层块整体可实现接近最优的成本。此外，模型在训练后性能显著提升，表明最优传输行为随训练逐渐形成，且在初始化阶段一致性较弱（0.64 对比训练后 0.86），验证了训练过程对优化传输路径的塑造作用。
链接: https://arxiv.org/abs/2609.00748
作者: Alexandre Quemy
机构: Hother Labs(霍瑟实验室)
类目: Computation and Language (cs.CL)
备注: Submitted at GDDL Workshop @ NeurIPS 2026
Abstract:A transformer carries each token’s state from layer to layer, and the whole vocabulary carried together forms a cloud that moves with depth. We ask whether a trained network moves this cloud the way optimal transport would: at the cheapest cost, and along the map that pairs each token with its optimal destination. We measure both on Pythia-160m and Pythia-410m, with an exact assignment between consecutive layer clouds, a measured sampling floor, calibration on couplings known to be optimal, and a split of the cost into the common shift of the cloud and the token-specific moves. At the last layer, both models move their tokens where the optimal-transport map sends them, at the optimal cost for Pythia-410m and slightly above it for Pythia-160m. At the first layer they do not. In between, single layers can be judged on cost at only two of ten transitions, and blocks of several layers move the cloud at close to the optimal cost. The agreement at the last layer is much weaker at initialisation (0.64 against 0.86) and grows with training.
[NLP-86] Can Large Language Models Forecast What Researchers Study Next? EMNLP2026
【速读】： 该论文旨在解决生成式AI在科研选题生成过程中难以准确评估其提出的研究想法是否具有真实新颖性或可行性的问题，尤其关注这些想法能否有效预测后续学术社区的实际研究动向。其核心挑战在于：当前方法通常依赖生成时刻的主观判断，但无法验证所提想法是否真正被后续研究采纳或实现。为此，论文提出了IdeaForecastBench基准测试框架，通过设定一个“检索-判断”固定协议，在给定某一研究领域截至某时间点的文献基础上，让模型生成最多五个按优先级排序的研究设想，并将其与之后发表的论文进行对比，从而量化模型对后续研究趋势的预测能力。该基准涵盖52个主题、624个滚动实验周期，由两名独立评审员分别评分以保证客观性。研究比较了GPT-4.1、Qwen2.5-7B/14B及Qwen3.5-9B等模型在五种历史压缩策略下的表现，发现摘要（Summary）策略在Hit@5和Precision@5指标上优于直接输入（Direct），且Qwen2.5在多数情况下超越GPT-4.1，而Qwen3.5则表现更差。此外，盲评分析揭示Qwen2.5生成的设想更具广度，但未明确广度对其优势的具体贡献程度；阈值与评审诊断进一步指出，“实现”不能等同于“精确预见”，强调了对结果解释的局限性。因此，该研究的关键解决方案是构建一个可复现、可量化的科研思想预测评估体系，为未来探索大语言模型在科研创新中的角色提供标准化工具。
链接: https://arxiv.org/abs/2609.00747
作者: Fenghai Li,Zihan Tang,Haofei Yu,Yining Zhao,Jiaxuan You
机构: University of Illinois Urbana-Champaign (伊利诺伊大学厄本那-香槟分校)
类目: Computation and Language (cs.CL)
备注: 31 pages, 4 figures. Accepted to EMNLP 2026
Abstract:Large language models increasingly generate research ideas, yet judging their novelty or feasibility at generation time does not establish whether they anticipate subsequent work. We introduce IdeaForecastBench to evaluate research idea forecasting. Given a community’s literature up to a cutoff, a system produces up to five ranked ideas, which are evaluated against later papers. The benchmark comprises 624 rolling episodes across 52 topics, with a fixed retrieve-then-judge protocol and separately reported results from two judges. We compare five history-compression strategies across GPT-4.1, Qwen2.5-7B/14B, and Qwen3.5-9B, together with a learned Mode-Decomposition Forecaster (MDF). Under the primary GPT-4.1-mini judge, Summary improves on Direct in Hit@5 and Precision@5 across all four backbones. Qwen2.5 scores above GPT-4.1, whereas Qwen3.5 scores below it. An outcome-blind assessment finds that Qwen2.5 produces broader forecasts, but does not identify how much breadth contributes to its advantage. Threshold and judge diagnostics further clarify the limits of interpreting realization as precise anticipation. IdeaForecastBench provides a common task for studying which research ideas a community subsequently pursues and how reliably this outcome can be measured.
[NLP-87] Controllable Image Captioning with Prompt-Conditioned Scene Rewards EMNLP2026
【速读】： 该论文旨在解决大规模视觉-语言模型（Large Vision-Language Models, VLMs）在图像描述生成中语义控制能力不足的问题，即用户无法可靠地指定描述应侧重于对象属性、关系或特定图像区域等具体语义层面。其核心解决方案是提出一种基于场景图对齐组件得分的提示条件化控制目标，称为细粒度描述控制（FoCUS）。该方法通过将生成的文本描述解析并映射至场景图中的对象、属性和关系组件，根据用户自然语言控制提示对不同组件施加差异化权重（包括负权重），从而实现对描述语义焦点的精确引导。为优化该目标，采用广义相对策略优化（GRPO），并通过引入更严格的对象有效性阈值及基于推理的属性与关系评分验证机制，显著提升了控制的可靠性。为评估可控性，研究提出了语义控制与精度评估基准（SCoPE），包含对比性的“包含/排除”约束，用于衡量目标内容覆盖率与无关内容抑制能力。实验结果表明，FoCUS在两种VLM骨干网络上均能持续提升细粒度控制能力和生成质量，同时保持通用描述性能不下降。
链接: https://arxiv.org/abs/2609.00709
作者: Jongyeop Hyun,Taeyoung Kim,Hyounghun Kim
机构: POSTECH(浦项科技大学); Graduate School of Artificial Intelligence(人工智能研究生院); Department of Computer Science and Engineering(计算机科学与工程系)
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: EMNLP 2026 Main (26 pages); Project website: this https URL
Abstract:Large Vision-Language Models produce fluent image descriptions but offer limited semantic control: users cannot reliably specify whether captions should emphasize attributes, relations, or particular image regions. We present Fine-grained Captioning Control Using Scene Rewards (FoCUS), a controllable image captioning method that lets users steer captions toward specific semantic emphases through natural-language control prompts. The core idea is a prompt-conditioned control objective based on scene-graph-aligned component scores. Generated captions are parsed and aligned to scene-graph components such as objects, attributes, and relations. These components are differentially weighted, including negative weights, according to the requested emphasis. We optimize this objective with GRPO and further improve its reliability through a stricter object validity threshold and reasoning-based verification for attribute and relation scoring. To evaluate controllability, we introduce Semantic Control and Precision Evaluation (SCoPE), a benchmark with contrastive Include/Avoid constraints for measuring both target content coverage and out-of-scope suppression. Experiments on two VLM backbones show that FoCUS consistently improves controllability and fine-grained caption quality without degrading general caption performance.
[NLP-88] A Certificate-Producing Cascade for Equational Implication: The SAIR EQT2 Stage 2 Solver
【速读】： 该论文旨在解决代数理论中的等式恒等式蕴含关系判定问题，即判断一个代数恒等式（magma identity）是否可以从另一个恒等式推导得出，并在给出“是”或“否”的结论时，生成可被确定性Lean验证器接受的证明证书。其核心挑战在于既要保证推理的严格正确性，又要高效处理复杂代数结构的验证与反例构造。解决方案的关键在于设计了一个单一文件实现的、基于最小代价优先级的级联式求解器架构：在“假”分支中，通过结构化代数族上的系数检验、有界有限模型搜索、显式的中心群胚（central-groupoid）反例以及多种无限基数反例构造机制协同工作；在“真”分支中，则采用可生成证明的有序单位超归纳（ordered unit superposition）推理过程，结合Knuth-Bendix排序、双向归约（bidirectional demodulation）、索引机制、记忆化替换及任意时间深度递增策略，确保推导过程可回放且形式化可验证。所有搜索结果均不依赖可信基础，成功推导路径以小型Lean项形式重放，反例则由竞赛裁判重新校验。该求解器为189,504字节的Python程序，经本地测试在官方裁判版本2848228下对6个公开数据集共1,889条记录全部生成了被接受的证书，且未调用任何语言模型。量化结果均基于不可篡改的结果日志，论文未宣称完备性或相对优越性。
链接: https://arxiv.org/abs/2609.00706
作者: Haobo Ma,Wenlin Zhang,Manuel Israel Cázares
机构: ChronoAI Pte. Ltd.(ChronoAI公司); National University of Singapore(新加坡国立大学); Bytepro AI(Bytepro AI)
类目: Computation and Language (cs.CL)
备注: 12 pages
Abstract:The SAIR Mathematics Distillation Challenge on Equational Theories asks a solver to classify whether one magma identity implies another and, for either verdict, to return a certificate accepted by a deterministic Lean judge. We present a single-file solver organized as a cheapest-first cascade. Its false branch combines coefficient tests over structured algebra families, bounded finite-model search, an explicit central-groupoid witness, and several infinite-carrier witnesses. Its true branch is a proof-producing ordered unit superposition procedure with Knuth-Bendix ordering, bidirectional demodulation, indexing, memoised substitution, and anytime size deepening. Search results remain outside the trusted base: successful derivations are replayed as small Lean terms, and countermodels are rechecked by the competition judge. The frozen solver is a 189,504-byte Python file with SHA-256 f2392533c9f4c03b… In local runs through official judge revision 2848228, it produced accepted certificates for all 1,889 rows of the six public sets with no language-model calls. Separate measurements recorded full agreement on the 800 published Stage 1 evaluation-distribution problems, 100 accepted rows in the canonical Marathon manifest without tokens, and 200 accepted rows in the hosted playground. These are regression and playground measurements, not a leaderboard result and not evidence about a hidden set. All quantitative claims are tied to immutable result ledgers; the paper makes no completeness or comparative-superiority claim. Comments: 12 pages Subjects: Computation and Language (cs.CL) ACMclasses: F.4.1; I.2.3 Cite as: arXiv:2609.00706 [cs.CL] (or arXiv:2609.00706v1 [cs.CL] for this version) https://doi.org/10.48550/arXiv.2609.00706 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[NLP-89] Value Over Language Model: Detecting Original Contribution in Writing
【速读】： 该论文旨在解决现有大语言模型（LLM）生成文本检测工具普遍存在的局限性：这些工具主要基于表面文本特征判断内容是否由模型生成，而未能有效衡量文本中信息内容或思想实质上源自于用户输入提示（prompt）还是由模型自行产生。其核心问题在于缺乏对“人类在生成过程中所增加的独特价值”的量化能力。为此，论文提出一种名为“语言模型之上的价值”（Value Over Language Model, VOLM）的框架，其关键创新在于不依赖训练数据或标注样本，也不直接分析文档的表层文本，从而规避了风格混淆（stylistic confounders）的影响。VOLM通过逐级细化地提取文档的内容信息，在不同粒度下利用大语言模型从部分表示中重构原文，并将这些重构结果与仅基于任务描述生成的基准文档进行对比，以此评估人类贡献的相对价值。实验表明，该方法能有效区分人类撰写与由通用提示生成的模型文本，且对保持语义不变的转换（如基于LLM的重写和往返翻译）具有高度鲁棒性；同时，随着内容提取器约束增强，模型生成与人工润色文本间的残差差异减小，进一步验证了信息内容与表达风格解耦的重要性。该研究为评估人类在大语言模型辅助写作中的实际贡献提供了新范式，推动了对人机协作创作中“真实价值”测量的深入探索。
链接: https://arxiv.org/abs/2609.00700
作者: Vibhhu Sharma,Thorsten Joachims,Sarah Dean
机构: Cornell University (康奈尔大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 39 pages
Abstract:LLMs have been rapidly adopted across writing tasks, prompting the development of tools for detecting LLM-generated text. Yet, these tools largely measure how much of a document’s surface text was written by an LLM and aren’t fundamentally designed to measure how much of the information content or ideas originated from the LLM itself rather than being supplied by the user in the prompt. In this work, we design a framework that measures how much value a person adds on top of what a language model could have easily produced by itself. The method requires no training or labeled data and never scores the document’s surface text, insulating it from stylistic confounders. Instead, it extracts the document’s content at increasing levels of granularity, uses an LLM to reconstruct the document from each partial representation, and compares these reconstructions with those produced from the task description alone. We call this framework Value Over Language Model (VOLM), which measures a document’s contribution relative to a replacement-level document that an LLM could produce from the task description alone. We evaluate VOLM with a specific instantiation of this framework across three domains: news articles, ICLR peer reviews, and argumentative essays. VOLM separates human-authored documents from matched LLM-generated documents produced from generic task descriptions, while remaining substantially invariant to content-preserving transformations, including LLM-based reconstruction and round-trip translation. We further find that increasingly constrained content extractors reduce residual differences between LLM-generated and humanized text, demonstrating the importance of disentangling informational content from stylistic variation. We hope these results encourage further work on specialized instantiations of the framework and on assessing human contributions in LLM-assisted writing more generally.
[NLP-90] SCoNE: Selective Context-aware Neuron Editing for Robust Retrieval-Augmented Generation EMNLP2026
【速读】： 该论文旨在解决生成式 AI（Generative AI）在检索增强生成（Retrieval-Augmented Generation, RAG）框架下对检索噪声敏感的问题：当检索到的文档中混杂了相关信息与无关信息时，大语言模型（LLM）容易受到干扰，从而产生幻觉。其解决方案的关键在于提出一种无需训练的模型编辑方法——SCoNE（Selective Context-aware Neuron Editing），通过识别同时具备高特征重要性（high attribution）和高跨输入变异性的上下文感知前馈神经网络（FFN）神经元，并选择性地强化这些神经元，以提升模型在噪声检索环境下的鲁棒性。SCoNE仅需少量样本进行挖掘，无需微调且不引入推理阶段开销，在多个知识密集型问答基准及两种主流大模型架构上均显著优于现有基线方法。
链接: https://arxiv.org/abs/2609.00689
作者: Chaewon Kim,Seo Yeon Park
机构: Hanyang University(汉阳大学)
类目: Computation and Language (cs.CL)
备注: Accepted to EMNLP 2026 Main Conference
Abstract:Retrieval-Augmented Generation (RAG) is highly sensitive to retrieval noise: when retrieved documents mix informative and irrelevant context, LLMs are easily distracted, leading to hallucinations. To overcome this, we propose SCoNE (Selective Context-aware Neuron Editing), a training-free model editing approach that improves retrieval noise robustness by selectively strengthening context-aware FFN neurons that are identified by both high attribution and high cross-input variability. SCoNE requires only a small number of mining samples, no fine-tuning, and no inference-time overhead. Across various knowledge-intensive question-answering benchmarks and two LLM backbones, SCoNE consistently outperforms competitive baseline methods. Our code is available at this https URL.
[NLP-91] Visual Framing for News Stance Detection via Image Generation EMNLP2026
【速读】： 该论文旨在解决新闻文章层面立场检测（article-level news stance detection）中的核心挑战，即新闻文本中立场往往隐含且通过新闻框架（journalistic framing）微妙传递，加之文本结构复杂、篇幅较长，导致传统方法难以有效识别。其解决方案的关键在于提出VFStance模型，创新性地利用视觉框架（visual framing）通过图像生成技术将原本隐含的立场线索显式化，从而增强立场信号的可辨识性。实验结果表明，相较于现有方法，VFStance在立场检测任务中表现更优，且视觉框架的引入显著提升了模型性能；进一步的受控用户研究（N=200）在片段化新闻消费场景下验证了该方法能有效使立场信号在视觉上更加突出，凸显其在自动化立场检测之外的应用潜力。
链接: https://arxiv.org/abs/2609.00685
作者: Dahyun Lee,Jiyoung Han,Kunwoo Park
机构: Soongsil University (松林大学); KAIST (韩国科学技术院)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Computers and Society (cs.CY)
备注: EMNLP 2026
Abstract:Article-level news stance detection aims to identify the perspective of news articles toward social issues. Despite advances in stance detection and its importance for trustworthy media environments, news articles pose distinct challenges because their stances are often implicit, subtly conveyed through journalistic framing, and embedded in long, structurally complex texts. To address these challenges, we introduce VFStance, which leverages visual framing to make implicit stance cues more explicit via image generation. In evaluation experiments, we demonstrate the effectiveness of VFStance over existing methods and the contribution of visual framing to its performance. Finally, a controlled user study (N=200) in a snippet-based news consumption setting further demonstrates that VFStance can make stance signals visually salient and highlights its potential use beyond automated stance detection.
[NLP-92] Creative Generation via Multi-Agent Debate: Does Debate Suppress Diversity? EMNLP2026
【速读】： 该论文旨在解决生成式任务（如叙事写作和科学创意生成）中高质量输出与跨独立运行结果多样性之间的矛盾问题。传统多智能体辩论（Multi-Agent Debate, MAD）虽在事实性与推理任务中表现出色，但其以收敛为导向的设计机制会抑制不同运行间的输出多样性，难以满足创造性任务对探索广度的需求。论文的理论分析指出，在每轮辩论会话中保持智能体间认知差异是实现跨运行多样性的必要条件。为此，作者提出Creative-MAD框架，通过两项协同干预机制维持智能体间的持续分化：一是认知视角分配（Cognitive Lens Assignment），通过为每个智能体赋予独特且稳定的认知模式，防止身份漂移；二是基于嵌入的同伴选择（Embedding-based Peer Selection），通过限制每个智能体的上下文仅包含语义上最远的同伴，缓解多数意见吸引效应。在四个创造性基准上的实验表明，Creative-MAD在显著提升词汇与语义多样性的同时，仍能保持MAD原有的高质量输出水平。
链接: https://arxiv.org/abs/2609.00683
作者: Tien Anh Nguyen,Khanh-Binh Nguyen,Van Dai Do,Svetha Venkatesh,Hung Le
机构: Deakin Applied Artificial Intelligence Initiative, Deakin University (迪金应用人工智能研究院，迪金大学)
类目: Computation and Language (cs.CL)
备注: 28 pages, accepted to EMNLP 2026 (Main Conference)
Abstract:Creative generation tasks, such as narrative writing and scientific ideation, demand both high-quality outputs and distinct responses across independent runs to maximize exploration. Multi-Agent Debate (MAD) has shown strong quality gains on factual and reasoning tasks, making it a natural candidate for creative generation. However, we find its convergence-driven design actively suppresses output diversity across independent runs, creating an inherent trade-off with creative tasks. We theoretically show that preserving diversity among agents within each debate session is a necessary condition for achieving diverse outputs across independent runs. Building on this finding, we propose Creative-MAD, which introduces two synergistic interventions to sustain agent divergence. Specifically, Cognitive Lens Assignment counters identity drift by anchoring each agent to a distinct and persistent cognitive mode, while Embedding-based Peer Selection counters majority pull by limiting each agent’s context to its most semantically distant peers. Experiments across four creative benchmarks demonstrate that Creative-MAD significantly enhances both lexical and semantic diversity while maintaining MAD’s output quality.
[NLP-93] SciTrue: Reliable Scientific Claim Validation with Frontier and Open Language Models at the NTCIR SciClaimEval Task
【速读】： 该论文旨在解决科学主张验证（scientific claim verification）任务中，如何高效、准确地基于论文中的表格与图表证据判断科学主张的真实性问题。其核心挑战在于模型需理解多模态内容（文本、表格、图像），并正确匹配主张与支持/反驳证据。解决方案的关键在于：首先，采用指令微调的前沿多模态模型（如Claude Opus 4.8、Gemma-4-31B等）已具备接近甚至超越现有最强基线（o4-mini）的能力；其次，提出一种“无泄漏”的配对先验（leak-free pair prior），仅通过主张文本中的可见字段即可恢复支持/反驳证据的正确配对，并将高置信度证据标记为“支持”，使子任务1的配对准确率从72.2%显著提升至93.5%，远超模型替换或集成权重调整的效果；最后，通过逐案审计发现，剩余错误主要源于视觉不可察觉的标签映射错误或数据集本身噪声，表明实际性能被低估，而模型可改进空间有限。此外，研究揭示了数据包装方式中存在的测量泄露问题——即标签信息通过文件排序而非内容传递，导致系统间接获取标签信息，甚至影响自身流水线，进一步凸显评估协议设计的重要性。
链接: https://arxiv.org/abs/2609.00654
作者: Qiming Bao,Neşet Özkan Tan,Siyuan Wang,Mark Gahegan
机构: University of Auckland(奥克兰大学); The Chinese University of Hong Kong(香港中文大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: To appear in the Proceedings of the 19th NTCIR Conference (NTCIR-19)
Abstract:We describe the SciTrue team’s participation in both subtasks of the NTCIR-19 SciClaimEval task~\citesciclaimeval, which asks systems to verify scientific claims against the tables and figures of a paper. Rather than tuning a single model, we benchmark eleven frontier and open multimodal models under one honest, per-sample protocol and combine them with light, transparent post-processing. On the official, blind test leaderboard (Section~\refsec:results), SciTrue placed first by a clear margin in three of the four evidence-category/subtask combinations, and tied for first on the primary metric in the fourth. Three findings explain the result. First, strong instruction-tuned models are already competitive: Claude Opus~4.8 and Gemma-4-31B each exceed the strongest public baseline (o4-mini), and GPT-5.5 and Claude Fable~5 lead both subtasks (97.7 on Subtask~2). Second, the task’s pairing structure is the largest lever: a \emphleak-free pair prior that recovers the Supported/Refuted pairing from the claim text alone (a visible field) and assigns Supported to the higher-confidence evidence raises Subtask-1 pair-accuracy from 72.2 to 93.5, far more than any model swap or ensemble weighting. Third, a case-by-case audit finds that most residual errors are visually-undetectable label-mapping swaps or dataset label noise, so measured accuracy understates the true ability and the fixable-by-modeling headroom is small. Controlled fine-tuning, distillation, and agentic consistency-checking support the same conclusions, and we document throughout a measurement leak—label information reaching a system through the packaging of the data rather than its content—in which the released file ordering encodes the label, including one instance that briefly misled our own pipeline.
[NLP-94] ExpArt-KG: Artwork Image Description Generation through Iterative Exploration of Knowledge Graphs
【速读】： 该论文旨在解决大型视觉语言模型（LVLMs）在图像描述任务中难以全面且准确地表述图像中物体所关联实体与概念之间事实性关系的问题。其核心挑战在于如何有效融合外部知识以增强生成内容的准确性与细节丰富度，同时避免引入冗余或无关信息。解决方案的关键在于提出一种基于检索增强生成（RAG）的迭代框架，通过交替执行答案生成与知识图谱检索，并利用正确性判断机制动态控制检索过程，从而高效获取必要且充分的事实性知识。此外，研究构建了一个面向艺术作品领域的知识图谱（ExpArt-KG），其中图像与实体之间的映射关系明确无歧义。实验结果表明，该方法显著提升了艺术作品解释的细节水平，在保持生成质量与固定次数迭代相当的前提下，大幅降低了对外部知识的检索开销。
链接: https://arxiv.org/abs/2609.00629
作者: Yuta Kato,Shintaro Ozaki,Kazuki Hayashi,Yusuke Sakai,Hidetaka Kamigaito,Katsuhiko Hayashi,Taro Watanabe
机构: The University of Tokyo(东京大学); Nara Institute of Science and Technology (NAIST)(奈良先端科学技術大学院大学)
类目: Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Large Vision-Language Models (LVLMs) achieve strong performance on image-grounded text generation and visual question answering. However, it remains difficult for them to comprehensively and accurately describe the factual relations among the entities and concepts associated with the objects depicted in an image. In this work, we propose a framework that efficiently exploits factual information from a knowledge graph via retrieval-augmented generation (RAG), with the goal of enabling LVLMs to generate detailed and accurate image explanations. Specifically, our method alternates between answer generation and knowledge-graph retrieval, and controls the search using a correctness judgment, thereby acquiring the necessary and sufficient factual information efficiently. We also construct a knowledge graph for the artwork domain (ExpArt-KG), in which the correspondence between images and entities is unambiguous. Applying the proposed method to this knowledge graph, we show experimentally that it improves the level of detail of artwork explanations and reduces the retrieval cost of external knowledge while maintaining generation quality comparable to that of iterating a fixed number of times.
[NLP-95] rust Your Guide Only When Certain: Uncertainty-Aware Sparse Alignment at Inference Time EMNLP2026
【速读】： 该论文旨在解决推理时对齐（inference-time alignment）中轻量级监督机制存在的结构性缺陷：现有密集干预方法要求在每个解码步骤均施加监督，但实际中弱监督信号在绝大多数词元（token）上表现出高熵特征，导致频繁产生低置信度的干预，不仅干扰了大语言模型（Large Language Models, LLMs）原有的有效推理过程，还带来显著的性能损耗。其解决方案的关键在于提出一种基于信任机制的稀疏对齐框架——TUSA（Trust-based Uncertainty Sparse Alignment），该方法将对齐过程重构为动态仲裁机制，引入一个具备不确定性感知能力的仲裁器，仅在监督者具有高置信度且当前词元具有语义显著性（semantic salience）两个条件同时满足时才允许干预。这一机制有效过滤了由不确定性驱动的噪声与冗余监督，实现了精准、高效的对齐。实验结果表明，TUSA在多个模型和基准测试中均能持续提升安全对齐性与通用助益性，通过规避约50%的对齐步骤，相较密集基线分别实现最高15.6%的安全偏好提升和12.0%的通用偏好提升，验证了选择性、高精度对齐优于连续监督的有效性。
链接: https://arxiv.org/abs/2609.00624
作者: Zeen Zhu,Zhuo Li,Weiyang Guo,Liye Zhao,Haibing Di,Yequan Wang,Jing Li
机构: Harbin Institute of Technology, Shenzhen, China; Huawei Technologies Co., Ltd.; Beijing Academy of Artificial Intelligence, China
类目: Computation and Language (cs.CL)
备注: Accepted to Findings of EMNLP 2026
Abstract:A prominent paradigm in inference-time alignment employs lightweight supervisors to steer Large Language Models (LLMs). Through empirical analysis, we identify a structural mismatch in this paradigm: weak supervisors exhibit pervasive high entropy across the vast majority of tokens, yet prevailing dense intervention approaches mandate supervision at every decoding step. This leads to frequent low-confidence interventions that can disrupt valid base-model reasoning and incur substantial utility costs. To resolve this, we propose TUSA (Trust-based Uncertainty Sparse Alignment). Moving away from continuous oversight, TUSA reframes alignment as a dynamic arbitration process, introducing an uncertainty-aware arbiter that authorizes intervention only when two conditions are met: the supervisor is confident and the token is semantically salient. This mechanism effectively filters out uncertainty-driven noise and redundant supervision. Extensive experiments across multiple models and benchmarks show that TUSA consistently improves both safety alignment and general helpfulness. By bypassing approximately 50% of alignment steps, it not only enhances safety preference by up to 15.6%, but also boosts general preference rates by up to 12.0% compared to the dense baseline, demonstrating that selective, high-precision alignment can outperform continuous supervision.
[NLP-96] Confess What You Know: Forget-Set Misalignment with Model Knowledge in LLM Unlearning EMNLP2026
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在实际隐私场景下存在的“遗忘集错位”（forget-set misalignment）问题。具体而言，传统机器遗忘方法假设预定义的遗忘集与模型所记忆的内容完全匹配，但在真实环境中，原始训练数据往往不可获取，导致遗忘集无法准确反映模型的记忆内容，从而引发两类问题：一是“欠遗忘”（Under Unlearning），即遗忘集中遗漏了部分被模型记忆的信息，造成数据泄露；二是“超知识遗忘”（Out-of-Knowledge Unlearning），即算法试图“遗忘”模型从未学习过的知识，导致参数被无谓扰动，损害模型性能。研究通过梯度层面的分析表明，这些行为的根本原因在于遗忘目标与模型实际记忆内容之间的不一致，而非特定优化策略所致。为此，论文提出一种无需依赖原始数据的“忏悔式遗忘集构建”（CONfession-to-Forget-Set, CONFS）框架，其核心创新在于通过主动询问并形式化建模的方式，从模型自身中提取出其所记忆的知识，并据此构建与模型对齐的遗忘集。在合成数据、多模态及真实世界基准测试中，CONFS在多个指标上逼近黄金标准表现，实现了良好的遗忘-效用平衡，且在保持模型性能方面显著优于其他无需数据的遗忘集构造方法。
链接: https://arxiv.org/abs/2609.00605
作者: Miso Kim,Georu Lee,Seungwon Jeong,Woojin Lee
机构: Dongguk University-Seoul(东国大学-首尔)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: Accepted to EMNLP 2026 (Main Conference). 22 pages, 3 figures
Abstract:Machine unlearning for large language models (LLMs) often assumes that a pre-defined forget set matches what the model has memorized, but this frequently breaks in realistic privacy settings where the original training data is inaccessible. We term this gap forget-set misalignment and identify two cases. In Under Unlearning, the forget set omits memorized information and leakage persists. In Out-of-Knowledge Unlearning, the algorithm is driven to “forget” knowledge the model never learned, perturbing parameters and degrading utility. Using gradient-level analysis, we show these behaviors arise from misaligned unlearning targets rather than specific optimization choices. We then propose CONfession-to-Forget-Set (CONFS), a data-blind framework that constructs model-aligned forget sets by eliciting and formalizing the model’s memorized knowledge. Across synthetic, multimodal, and real-world benchmarks, CONFS approaches Gold-standard performance on several metrics and achieves a competitive forgetting-utility balance, while preserving utility better than other data-blind forget-set constructions.
[NLP-97] Quit While Youre Ahead: Quit for Efficient Candidate Generation in Machine Translation Reranking
【速读】： 该论文旨在解决神经机器翻译（Neural Machine Translation, NMT）中重排序（reranking）流程带来的高推理延迟问题，尤其针对现有加速方法仅聚焦于最小贝叶斯风险（Minimum Bayes Risk, MBR）解码而忽视质量估计（Quality Estimation, QE）重排序以及候选生成阶段计算开销过大的缺陷。其核心解决方案是提出一种名为Quit（Quantifying Uncertainty for Incremental Termination）的新型渐进式终止策略，将候选生成视为在不确定性下的序列决策过程，通过增量式生成与重排序候选译文，在候选集中的最高估计质量趋于稳定时提前终止整个生成-重排序流水线。该方法实现了端到端的显著加速：在19个语言对上对三种NMT模型的实验表明，Quit在保持翻译质量在预设等价阈值内的前提下，使MBR重排序获得1.47–2.66倍的加速，而QE重排序则达到3.43–4.12倍的加速，有效缓解了传统重排序方法的计算瓶颈。
链接: https://arxiv.org/abs/2609.00588
作者: Guangyu Chen,Boxuan Lyu,Hidetaka Kamigaito,Kotaro Funakoshi,Manabu Okumura
机构: Institute of Science Tokyo(东京科学研究所); Nara Institute of Science and Technology(奈良科学技术研究所)
类目: Computation and Language (cs.CL)
备注:
Abstract:Reranking methods, such as Minimum Bayes Risk (MBR) decoding and Quality Estimation (QE) reranking, are widely used in modern neural machine translation (NMT) to select an output from a set of candidate hypotheses. However, the performance gains come at the cost of high inference latency. Existing acceleration methods target MBR decoding and reduce only reranking computation, leaving QE reranking unaddressed and candidate generation—which can be the larger computational bottleneck—largely untouched. In this work, we propose Quit (Quantifying Uncertainty for Incremental Termination), a novel early-stopping strategy for the entire generation–reranking pipeline. Viewing candidate generation as a sequential decision under uncertainty, Quit incrementally generates and reranks candidates, stopping when the highest estimated quality in the candidate set stabilizes. Comprehensive experiments on three NMT models across 19 language pairs show that Quit yields end-to-end speedups of 1.47 – 2.66\times for MBR and 3.43 – 4.12\times for QE reranking while preserving translation quality within prespecified equivalence margins.
[NLP-98] Enoki: Efficient Multi-Level Hallucination Detection
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在高风险场景应用中事实性保障的关键挑战，特别是现有幻觉检测方法在粒度层面的局限性：仅支持声明级（claim-level）检测的模型虽具备可解释性，但缺乏对错误文本位置的精确定位；而仅支持片段级（span-level）检测的模型虽能定位不实内容，却难以提供可解释的语义单元。其核心解决方案是提出Enoki——一种面向多层级幻觉检测的开放信息抽取（Open Information Extraction, OIE）框架。Enoki通过提取以文本锚定的关系事实，将其与证据进行验证，并将未通过验证的事实反向投影至对应的幻觉片段，从而在统一的表示空间中同时实现声明级验证与片段级定位，避免了传统方法中复杂的跨层级对齐开销。该框架兼容基于LLM、编码器及规则的抽取机制，通过统一接口在准确率与推理成本之间取得平衡。实验表明，Enoki在保持与强声明级系统相当性能的同时显著降低资源消耗，并在细粒度片段与实体级定位任务上表现更优。此外，研究团队还发布了EnokiQA数据集，包含对齐的声明级验证与片段级定位标注，以支持未来相关研究。
链接: https://arxiv.org/abs/2609.00581
作者: Elisei Rykov,Timur Ionov,Nikolay Ivanov,Maksim Savkin,Maksim Makarenko,Alexander Panchenko,Vasily Konovalov,Julia Belikova
机构: Skoltech(斯科尔科沃科学技术研究院); AIRI(人工智能研究机构); ITMO University(圣彼得堡国立信息技术机械与光学大学); Sber AI Lab(斯科尔科沃人工智能实验室)
类目: Computation and Language (cs.CL)
备注:
Abstract:Ensuring factuality remains a critical challenge for deploying LLMs in high-stakes settings. Existing hallucination detectors usually operate at a single level: claim-level methods provide interpretable factual units, while span-level methods localize unsupported text. Bridging these views is costly, as LLM-heavy pipelines require multiple decomposition and verification calls, and modular systems need additional claim-to-span alignment. We propose Enoki, an Open Information Extraction framework for multi-level hallucination detection. Enoki extracts text-anchored relational facts, verifies them against evidence, and projects unsupported facts back to hallucinated spans. This shared representation enables claim-level verification and span-level localization without requiring separate alignment. Enoki supports LLM-based, encoder-based, and rule-based extraction regimes, balancing accuracy and inference cost through a common interface. Experiments show that Enoki remains competitive with strong claim-level systems while using fewer resources and achieves superior performance on fine-grained span- and entity-level localization. We also release EnokiQA, a dual-granularity dataset with aligned claim-level verification and span-level localization annotations.
[NLP-99] Predicting Program Exit Code with LLM s and Programming Language Semantics
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在软件工程任务中对编程语言语义理解不足的问题，特别是探究模型在给定形式化语义规则时，是真正系统性地应用这些规则，还是依赖预训练阶段习得的先验知识。其核心解决方案在于提出一种新颖的任务——程序可执行性预测（Program Executability Prediction, PrEx），该任务要求模型基于程序的语法结构和给定的操作语义，判断程序是否具有语义有效性，并在无效时指出违反的具体形式化规则。为构建涵盖有效与无效程序的数据集，研究通过系统性地从合法程序生成语义违规变体，从而实现对模型语义推理能力的严格评估。实验在两种语义形式化体系及多种语义偏移条件下，针对人工编写、LLM翻译和模糊测试生成三类程序进行评估，结果表明，当前开源编码类大模型更倾向于依赖预训练先验而非遵循显式给出的语义规则，尤其在语义发生变化或程序复杂度提升时性能显著下降。这一发现揭示了现有大模型在语义理解上的局限性，强调了增强形式化语义引导能力的重要性。
链接: https://arxiv.org/abs/2609.00579
作者: Lara Marinov,Aditya Thimmaiah,Jayanth Srinivasa,Junyi Jessy Li,Milos Gligoric
机构: The University of Texas at Austin(得克萨斯大学奥斯汀分校); Cisco Research(思科研究院)
类目: Programming Languages (cs.PL); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Software Engineering (cs.SE)
备注: Accepted at LMPL 2026
Abstract:Large language models (LLMs) have shown proficiency in various software engineering tasks, such as code generation and translation. However, a key limitation in their performance may be their (lack of) understanding of programming-language semantics. Even when explicit semantics are given, it remains unclear whether LLMs apply those rules or lean on priors learned during pre-training instead. We study if LLMs lean on priors or given semantics with a novel task–Program Executability Prediction (PrEx)–that asks models to predict whether a program is semantically valid or invalid (and, if invalid, which formal rule it violates) given the program’s syntax and operational semantics. Because PrEx requires both valid and invalid programs, we build a dataset with systematically generated invalid transformations derived from valid programs. We evaluate open-source coding LLMs under two semantic formalisms and two semantic shifts across Human-Written, LLM-Translated, and Fuzzer-Generated program splits. Our findings show that LLMs lean on pre-training priors rather than systematically applying the given rules, performing especially poorly on modified semantics and degrading further as program complexity increases. PrEx is available at this https URL.
[NLP-100] Consistency Without Alignment: Item-Sensitive Language Models Indistinguishable From Random
【速读】： 该论文旨在解决当前评估语言模型任务能力时过度依赖项目敏感性（item-sensitivity）所带来的潜在误导问题。项目敏感性通常被视为模型具备任务理解能力的证据，即模型的选择是否随具体输入变化而变化。然而，本文通过一个从桌游《欺诈：香港谋杀案》抽象出的强制选择信号任务，证明项目敏感性仅为任务胜任的必要条件而非充分条件。研究发现，在七种语言模型、两个模型家族、一次后训练消融实验及三种独立评分规则下，21个模型-规则组合均表现出显著的项目敏感性，但其中8个组合在统计上无法区分于完全忽略输入、随机选择的基线模型，另有5个模型在描述目标时的表现甚至劣于随机。项目敏感性与随机偏离程度之间的相关系数仅为 r = 0.30，表明高敏感性并不意味着决策质量或对真实参考标准（如贝叶斯最优策略）的对齐。作者将这种“一致性但不一致对齐”（consistency without alignment）现象归因于现有评估范式中缺乏独立参照标准，指出仅依赖项目敏感性、排列一致性或自一致性等内部一致性指标的评估方法存在根本缺陷。此外，研究还发现一个不包含语用学意义的字面相似性基线反而优于多数测试模型，且在两个基线相似性源上添加语用层反而使模型选择更趋近随机而非贝叶斯最优解；标准标注的多项选择格式在此情境下亦未传递可测量的内容信号。所有结果均基于预先注册的评估工具，人类对照组已设计并预试，但尚未收集数据。
链接: https://arxiv.org/abs/2609.00576
作者: Cris Huynh
机构: Independent researcher(独立研究员)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 13 pages, 3 figures
Abstract:Item-sensitivity, defined as whether a model’s choice depends on the specific input rather than on its own output prior, is widely reported as evidence of task competence. We show this evidence is necessary but not sufficient using a forced-choice signalling task abstracted from the board game Deception: Murder in Hong Kong. In this environment, the reference points against which a coordinate should be judged (a fit-maximising strategy, a posterior-maximising strategy, and uniform random selection) are all computable in closed form. Across seven language models, two model families, a post-training ablation, and three independent scoring rules, every one of 21 model-by-rule cells is reliably item-sensitive. Yet 8 of those 21 cells are not statistically distinguishable from a chooser that ignores the item and selects at random, and 5 score worse than random at describing the target. Item-sensitivity and distance from random correlate at only r = 0.30. We call this consistency without alignment and argue it generalises to any evaluation that relies on item-sensitivity, permutation consistency, or self-consistency without an independent reference for the measured quantity. We further find that a literal-similarity baseline with no pragmatics outperforms most tested language models, that adding a pragmatic layer over two baseline similarity sources moves choosers toward random rather than toward the Bayesian reference, and that a standard labelled multiple-choice format carries no measurable content signal here. All results represent the model side of a pre-registered instrument; a matched human condition is designed and piloted but not yet collected.
[NLP-101] Aligned but Flattened: Analyzing the Trade-off between Cultural Alignment and Diversity in LLM s EMNLP2026
【速读】： 该论文旨在解决当前文化感知大语言模型（LLM）在文化对齐（cultural alignment）训练中因过度优化单一对齐分数而导致的文化多样性缺失问题。现有方法仅关注模型输出与主流文化价值观的契合度，忽略了人类群体间固有的文化异质性，从而导致模型在表面上实现对齐的同时，实质上造成了“文化扁平化”（cultural flattening），即模型响应趋于同质化，丧失了对多元文化的敏感性与表达能力。其解决方案的关键在于提出一种协同评估框架，同时量化文化对齐性与文化多样性，揭示出对齐与多样性之间存在系统性的权衡关系。进一步的机制分析表明，这种多样性崩溃并非偶然行为偏差，而是神经网络优化过程中固有的低秩偏置（low-rank bias）所引发的结构性后果。因此，该研究呼吁重构后训练范式，发展能够保持跨文化多元主义（cross-cultural pluralism）的对齐目标，以实现真正具备文化敏感性的生成式 AI (Generative AI) 系统。
链接: https://arxiv.org/abs/2609.00565
作者: Jingshen Zhang,Shaoyang Xu,Wenxuan Zhang
机构: Tianjin University (天津大学); Singapore University of Technology and Design (新加坡科技设计大学)
类目: ocial and Information Networks (cs.SI); Computation and Language (cs.CL)
备注: Accepted at EMNLP 2026 (Findings)
Abstract:Cultural fine-tuning has become the de facto paradigm for building culture-aware large language models (LLMs), yet existing optimization exclusively for alignment scores provides an incomplete portrait of cultural fidelity by systematically obscuring inherent cultural diversity. This unidimensional evaluation lens prompts a fundamental question: do models genuinely perceive distinct cultural nuances, or do they merely memorize dominant cultural values? To address this, we propose a synergistic evaluation framework that jointly formalizes cultural alignment and diversity. Through extensive benchmarking of six mainstream LLMs on the World Values Survey, this framework uncovers a systematic and critical trade-off: the pursuit of cultural alignment consistently incurs an acute expense of diversity, leading to severe “cultural flattening.” Investigating this behavioral shift, we demonstrate that these superficial alignment gains stem from models artificially anchoring to dominant majorities, converging onto a monolithic response pattern that wipes out the heterogeneous distributions inherent to human groups. Crucially, our mechanistic analysis suggests that this diversity collapse is not merely a behavioral anomaly but more likely a structural consequence of the low-rank bias inherent in neural network optimization. Therefore, our findings expose the limitations of current post-training paradigms and call for a shift toward alignment objectives that preserve cross-cultural pluralism.
[NLP-102] EM2Mem: Event-Centric Multimodal Memory for Large Language Models EMNLP2026
【速读】： 该论文旨在解决长视频问答中多模态记忆（multimodal memory）系统因依赖孤立片段（如字幕、帧、转录文本、摘要或图谱事实）而导致的跨模态与时间对齐困难问题。现有方法在推理时需重新构建多模态关联，受限于上下文长度且难以追溯证据来源，严重影响准确性和效率。其解决方案的关键在于提出一种以事件为中心的多模态记忆框架——EM²Mem，该框架在记忆构建阶段即通过事件锚点（event anchors）将异构证据（包括多模态记录、时间上下文、图谱关联关系、语义事实及出处信息）进行统一绑定，形成以事件为索引的记忆单元。这种设计实现了基于已锚定多模态事件的紧凑证据读取，避免了推理时的动态对齐开销。实验表明，EM²Mem在三个长视频问答基准上分别提升平均准确率2.0、2.4和3.7个百分点，事件级Top-5证据召回率提升7.0个百分点，并将每查询延迟降低4.67倍，总推理令牌消耗减少63.66%。
链接: https://arxiv.org/abs/2609.00551
作者: Yijun Chen,Yaqi Zheng,Yanya Li,Boyi Xiao,Buqiang Xu,Shuofei Qiao,Jizhan Fang,Xinle Deng,Yunzhi Yao,Xuehai Wang,Liuxin Zhang,Hui Li,Huajun Chen,Shumin Deng
机构: Zhejiang University (浙江大学); South China University of Technology (华南理工大学); Lenovo Group Limited (联想集团有限公司)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Multimedia (cs.MM)
备注: Accepted by EMNLP 2026 findings
Abstract:Multimodal memory offers a scalable interface for long-video question answering, but existing methods often retrieve captions, frames, transcripts, summaries, or graph facts as isolated fragments. Although searchable, such fragments are not generation-ready: language models must reconstruct cross-modal and temporal alignments at inference time, when context is limited and attribution is difficult. We propose EM^2Mem, an event-centric multimodal memory framework that binds heterogeneous evidence to event anchors during memory construction. Each event-indexed memory cell aligns multimodal records, temporal context, graph-linked relations, semantic facts, and provenance, enabling compact evidence readout over grounded multimodal events rather than modality-specific fragments. Across three long-video QA benchmarks, EM^2Mem improves average accuracy over the strongest memory baseline by 2.0, 2.4, and 3.7 points, improves strict event-level Top-5 evidence recall by 7.0 points, and reduces per-query latency by 4.67 times and total inference tokens by 63.66% (The code will be integrated into this https URL).
[NLP-103] Same Semantics Different Outcome: On the Modality Robustness of Multimodal LLM s under Knowledge Conflict EMNLP2026
【速读】： 该论文旨在解决多模态大语言模型（Multimodal Large Language Models, MLLMs）在面对与自身参数化知识冲突的上下文证据时，其跨模态处理的一致性问题。具体而言，当文本与图像形式的外部证据存在矛盾时，模型对不同模态的依赖表现出显著不稳定性，且这种不稳定性会直接影响模型推理的可靠性。研究发现，模型更倾向于采纳与自身知识相悖的图像证据而非文本证据，而当图文并存时，模型的选择高度依赖输入顺序、模型架构及数据集，呈现出任意性和不可预测性。这一现象不仅导致模型在多模态检索增强生成（Multimodal RAG）任务中性能下降，还可能被对抗攻击利用。为缓解该脆弱性，作者评估了提示工程、引导策略、监督微调（Supervised Fine-Tuning, SFT）和直接偏好优化等多种方法，结果表明除SFT外其余手段效果有限，仅实现有限改善。因此，论文强调该不一致性具有根本性，亟需在模型训练的多个阶段予以系统性关注与改进。
链接: https://arxiv.org/abs/2609.00550
作者: Jungyeon Lee,Yejin Yoon,Taeuk Kim
机构: Hanyang University (汉阳大学)
类目: Computation and Language (cs.CL)
备注: Accepted to Findings of EMNLP 2026
Abstract:Multimodal large language models (MLLMs) are increasingly provided with contextual evidence in heterogeneous forms: as a text passage, as a rendered image of the same passage, or as both together. However, it remains unclear how consistently these surface forms are processed, especially when the evidence conflicts with the model’s parametric knowledge. We study modality robustness under knowledge conflict across 13 MLLMs and two datasets, and find them far from robust. (1) Contrary to common belief, models favor a context that contradicts parametric knowledge more readily in image form than in text form; (2) when a contradicting text and image are presented together, the preferred modality is essentially arbitrary, varying with input order, model, and dataset. We further demonstrate that this instability has practical consequences: it degrades performance in multimodal RAG and can be exploited by adversarial attacks. To alleviate this brittleness, we examine several simple techniques—prompting, steering, supervised fine-tuning (SFT), and direct preference optimization; the majority prove ineffective, whereas SFT achieves moderate success. We therefore call for greater awareness of this inconsistency and argue that it is fundamental, demanding attention at multiple training stages.
[NLP-104] Skill Following: Evaluating Actual Skill Use in Retrieval-Enabled LLM Agents EMNLP2026
【速读】： 该论文旨在解决大语言模型（Large Language Model, LLM）在实际应用中依赖外部技能时，现有评估方法无法准确衡量技能真实使用效果的问题。传统评估通常通过比较调用技能与未调用技能的任务表现来计算整体提升，但这种方法引入了严重的选择偏差，难以分离出技能使用本身的真正影响。为克服这一缺陷，作者提出“技能遵循”（Skill Following, SF）这一概念，并引入“检索触发的实际使用效应”（Retrieval-Invoked Actual-Use Effect, RAE）作为核心评估指标。RAE的关键在于：在仅针对那些模型主动触发技能检索的任务上，对比相同任务下启用技能与禁用技能的执行结果差异，从而精确衡量技能实际使用所带来的净收益。实验结果显示，在编码和数学领域对17个LLM的评估中存在显著的评估悖论——尽管多数模型在整体表现上显示出正向的检索增益，但其RAE却为负值，表明在实际发生检索的任务中，技能使用反而降低了性能。这揭示了聚合指标可能产生工具使用能力的虚假繁荣假象，而RAE则能直接反映检索—回答链是否真正提升了任务完成质量。
链接: https://arxiv.org/abs/2609.00549
作者: Seonghyeon Cho,Chanjun Park
机构: Soongsil University (顺溪大学)
类目: Computation and Language (cs.CL)
备注: Accepted to Findings of EMNLP 2026
Abstract:Large Language Model (LLM) agents increasingly rely on external skills, yet standard evaluations obscure whether retrieving these skills actually helps. Aggregate metrics often compare retrieved versus non-retrieved tasks, introducing severe selection bias and failing to isolate the true effect of skill use. To measure this actual-use capability-which we formalize as Skill Following (SF)-we introduce the Retrieval-Invoked Actual-Use Effect (RAE). RAE computes the same-task outcome difference between matched skill-enabled and skill-disabled executions, conditioned exclusively on tasks where the agent actively retrieved a skill. Evaluating 17 LLMs across coding and mathematical domains, we uncover a stark evaluation paradox: models frequently show positive aggregate retrieval lift but negative RAE. On MBPP+, multiple models that appear to benefit system-wide actually harm their own performance on the exact tasks where retrieval occurred. These findings demonstrate that aggregate averages can create a misleading illusion of tool-use proficiency, whereas RAE directly measures whether the retrieval-to-answer pipeline genuinely rescues more outcomes than it harms.
[NLP-105] he Interlingua Hypothesis: LLM s Translate via a Latent Task-agnostic Feature Space
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在机器翻译任务中表现出色的内在机制问题，特别是其如何实现跨语言翻译。现有研究表明，尽管LLMs在翻译性能上超越了传统的监督式基线模型，但其背后的翻译原理仍不明确。为此，论文提出“中介语假说”（interlingua hypothesis），即大语言模型通过将源语言句子映射到一个共享的、多语言的潜在特征空间（latent feature space），再从该空间生成目标语言句子来完成翻译。该假说的关键在于：翻译过程并非依赖于显式的语言对间映射，而是基于统一的、跨语言的隐式表征。论文提供了三方面的实证支持：（1）不同语言对间的翻译性能差异（以BLEU衡量）可由各语言自身的表征能力解释，无需引入语言对特异性交互项；（2）大量模型组件在单语任务与翻译任务中均具有因果影响，表明其功能具有跨任务通用性；（3）仅在单语数据上微调即可恢复大部分翻译性能提升，远超在平行语料上微调的效果。上述证据形成了一致性支持，表明大语言模型的翻译行为本质上是基于共享的多语言潜在表示，而非语言对特定的转换机制。这一发现为理解并优化大语言模型的翻译能力提供了新的理论框架与技术路径。
链接: https://arxiv.org/abs/2609.00515
作者: Jacob Brinton,Jannik Brinkmann,Mark Crovella,Aaron Mueller
机构: Boston University (波士顿大学); Technische Universität Clausthal (克劳斯塔尔工业大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 21 pages, 15 figures, 11 tables
Abstract:Large language models (LLMs) have recently demonstrated improved machine translation performance over strong supervised baselines. This raises questions as to what mechanisms underlie how LLMs perform machine translation between languages. Motivated by recent interpretability findings–namely, that LLMs use massively multilingual latent feature representations to perform language modeling–we propose the interlingua hypothesis. The hypothesis holds that language models translate by reading a source sentence into a latent feature space, and generate a target sentence by reading from the latent feature space. We show three lines of evidence in support of this hypothesis: (1) variance in BLEU across language pairs is largely predictable from language-specific competences with no language pair-specific interaction terms; (2) many model components are causally influential in both monolingual tasks and translation tasks; and (3) fine-tuning on monolingual data recovers a large proportion of translation improvements relative to fine-tuning on aligned documents. Together, these provide convergent evidence in support of the interlingua hypothesis, and suggest new ways of understanding and improving how LLMs can be leveraged to perform translation tasks.
[NLP-106] Beyond Token Positions: Safety Alignment Across Denoising Steps in Diffusion Language Models EMNLP2026
【速读】： 该论文旨在解决生成式 AI（Generative AI）在扩散型大语言模型（dLLM）中因迭代去噪生成机制导致的安全对齐问题。传统左到右解码模式下的安全控制策略难以直接适用于dLLM，因其生成过程具有非顺序性与多阶段特性，使得拒绝有害请求的信号分布和最终响应的形成路径变得复杂。研究发现，拒绝性令牌（refusal tokens）的生成高度集中于去噪早期阶段，并且在响应序列中的初始位置具有显著影响力；更重要的是，早期承诺的拒绝性令牌若能持续保留，则对最终输出的安全性具有决定性作用。基于这一关键观察，论文提出无需训练的解码方法——拒绝感知早期承诺（Refusal-Aware Early Commitment, RAEC），其核心在于识别并强制保留去噪早期出现的、具有持久性的拒绝信号，从而在不损害模型通用能力的前提下显著降低攻击成功率。实验结果表明，该方法在LLaDA与Dream模型上均有效提升了安全性，同时保持了良好的任务性能。
链接: https://arxiv.org/abs/2609.00495
作者: Guoli Wang,Haonan Shi,Tu Ouyang,An Wang
机构: Case Western Reserve University (凯斯西储大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Accepted by EMNLP 2026
Abstract:Diffusion large language models (dLLMs) generate text through iterative denoising rather than left-to-right decoding. This generation paradigm introduces two axes that can influence safety alignment: when tokens are generated during denoising and where they appear in the response. In this paper, we measure dLLM safety behavior under harmful prompts by tracing intermediate token distributions and commitment decisions throughout denoising. Our analysis shows that refusal signals are concentrated in early denoising steps and leading response positions, and the tokens committed early can strongly shape the final safety outcome. Our measurements further show that the denoising step and persistence of refusal-token commitment are important for understanding dLLM safety. Based on these findings, we propose Refusal-Aware Early Commitment (RAEC), a simple training-free decoding method that commits persistent refusal signals from early steps. Experiments on LLaDA and Dream show that RAEC reduces attack success rates while largely preserving utility. The code is available at this https URL.
[NLP-107] Human-Anchored Factuality Evaluation with Strategic Annotation EMNLP2026
【速读】： 该论文旨在解决大语言模型（Large Language Model, LLM）驱动的事实性评估存在系统性偏差的问题，尤其是在标注预算有限的情况下，如何通过结合少量人工标注数据与大规模模型预测，实现对事实性评估的高效且统计有效的估计。其核心挑战在于：传统采样策略（如基于置信度或随机采样）未能充分考虑人类与模型在事实性判断中的结构性差异。解决方案的关键在于提出一种面向事实性的标注策略设计流程，利用失败空间分析（Failure-Space Analysis, FSA），识别并建模人类与模型判断之间的系统性错位模式，包括证据不完整、时间不匹配、不可验证声明及评分标准偏离等结构化错误类型。通过FSA生成多样化的预测信号以指导高价值样本的选择，显著提升了标注效率。实验结果表明，在内部基于参考的评估系统AutoFA和RAGTruth上，该方法相比均匀采样和基于不确定性的基线，分别实现了40.3%和27.1%的有效样本量提升，有效缓解了模型低估真实事实准确率的问题。
链接: https://arxiv.org/abs/2609.00494
作者: Yu Wang,Craig Erickson,Kevin Small
机构: Amazon AGI(亚马逊通用人工智能)
类目: Computation and Language (cs.CL)
备注: Accepted as a conference paper for Industrial Track of EMNLP 2026
Abstract:LLM-based factuality judges provide scalable evaluation signals, but their metrics are often systematically biased relative to human judgments. We study human-anchored factuality evaluation under limited annotation budgets, where judge predictions on the full dataset are combined with human labels on a small selectively sampled subset to obtain statistically valid estimates. The efficiency of this approach depends critically on which examples receive human annotation: in factuality evaluation, judge-human misalignment is not driven solely by low confidence, but also by structured failure modes such as incomplete evidence, temporal mismatch, unverifiable claims, and rubric misalignment. To exploit this structure, we introduce a factuality-specific annotation policy design pipeline that uses failure-space analysis (FSA) to derive diverse predictive signals for modeling human-judge misalignment. On an internal reference-based factuality evaluation system (AutoFA) and RAGTruth, where judge-predicted estimates substantially underestimate human-annotated factual accuracy, our FSA-guided policy improves annotation efficiency over uniform sampling and uncertainty-driven baselines, achieving effective-sample-size gains of 40.3% on AutoFA and 27.1% on RAGTruth.
[NLP-108] he Privacy-Hallucination Tradeoff in Differentially Private Language Models EMNLP2026
【速读】： 该论文旨在解决在医疗等高风险领域中，差分隐私（Differential Privacy, DP）语言模型在保障用户隐私的同时，容易产生幻觉（hallucination）的问题。其核心挑战在于揭示并缓解“隐私-幻觉权衡”（privacy-hallucination tradeoff）现象：即随着隐私预算（privacy budget）的收紧，尽管模型的隐私保护能力增强，但生成内容的虚假性却显著上升。解决方案的关键在于识别出导致该权衡的根本机制——DP机制会压缩模型输出分布的多样性，使概率质量向事实错误的选项转移，从而增加幻觉发生概率。研究进一步通过控制训练数据中事实出现频率的实验，发现信息频次越高，越能有效降低DP模型的幻觉风险。因此，论文强调需要设计更精细的隐私保护策略，在提供严格隐私保障的同时，维持生成内容的事实准确性。
链接: https://arxiv.org/abs/2609.00492
作者: Krithika Ramesh,Krishna Pillutla,Danish Pruthi,Anjalie Field
机构: Johns Hopkins University (约翰霍普金斯大学); Indian Institute of Technology, Madras (印度理工学院马德拉斯分校); Indian Institute of Science (印度科学研究所)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: Accepted to EMNLP 2026 (Findings)
Abstract:Both privacy and factual accuracy are paramount in high-stakes domains like healthcare. Concerningly, we uncover and investigate a privacy-hallucination tradeoff in differentially private (DP) language models. First, we empirically show that models pre-trained or fine-tuned with DP tend to produce more hallucinations than non-DP counterparts, with increased severity as the privacy budget grows stricter. Second, we investigate model properties driving this tradeoff, demonstrating that DP mechanisms flatten output distributions, potentially redistributing probability mass toward factually incorrect alternatives. Third, through experiments where we control fact frequency in training data, we characterize how information frequency can reduce hallucination risks in DP models. Overall, our findings underscore the need for more nuanced privacy-preserving interventions that offer rigorous privacy guarantees without compromising factual accuracy.
[NLP-109] EvoFlint: An Evolutionary Atlas of Multi-Turn LLM Vulnerabilities
【速读】： 该论文旨在解决大语言模型在多轮对话中易被渐进式攻击绕过的问题，即模型虽能拒绝单轮有害请求，却可能在通过多轮逐步诱导的方式下妥协，这一现象构成了当前大语言模型最不为人所知的失效模式之一。传统自动化红队测试方法将此视为生成问题，试图直接生成能够突破模型防护的攻击提示；而本文提出应将其重构为搜索问题，核心在于发现、组织并迭代优化多样化的攻击策略，从而构建出目标模型失效的结构化地图，而非零散的一次性成功案例。其解决方案的关键在于提出EvoFlint框架，采用进化质量-多样性搜索（evolutionary quality-diversity search）机制，将攻击策略定义为分阶段的对话计划而非原始提示，并通过大语言模型驱动的变异与交叉操作进行演化。该框架引入基于帕累托最优的多目标适应度评估（兼顾攻击成功率与峰值危害严重性），以保留近似成功的“准失败”样本的筛选信号；同时采用风险索引档案（risk-indexed archive）结合局部竞争机制，在策略描述嵌入空间内实现新颖性搜索，维持策略多样性且无需预设风格分类体系；此外，还设计了生成层级的记忆模块，持续积累对目标模型的洞察并反馈至策略生成过程。实验结果表明，在HarmBench测试集上，EvoFlint在Claude Sonnet 4.6、GPT-5.4和Qwen3-32B上的攻击成功率分别达到35.8%、59.7%和94.3%，并揭示了各模型安全训练覆盖与未覆盖的风险类别，显著提升了对模型安全边界的理解。
链接: https://arxiv.org/abs/2609.00487
作者: Feitong Qiao,Liren Peng,Shiming Ren,Aishwarya Jadhav,Arghavan Bahadorinejad,Marinette Chen,Muhan Zhang,Abdulaziz Suria,Gennevi Lu,Anish Das Sarma
机构: Reinforce Labs, USA
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR); Machine Learning (cs.LG)
备注:
Abstract:Frontier language models that refuse harmful single-turn prompts often comply when the same intent is reached gradually over many turns, making multi-turn attacks one of the least understood failure modes of large language models. Most automated red-teaming methods treat this as a generation problem: produce attacks that break the model. We argue it is better framed as a search problem: discover, organize, and iteratively refine a diverse archive of attack strategies, producing a structured map of how a target model fails rather than a list of one-off successes. We introduce EvoFlint, which applies evolutionary quality-diversity search to multi-turn red-teaming. Attack strategies are phased conversation plans, not raw prompts, and are evolved through LLM-driven mutation and crossover. A Pareto fitness over attack success rate and peak severity preserves selection signal from near-miss attacks. A risk-indexed archive runs novelty search with local competition over strategy description embeddings inside each cell, maintaining diversity without committing to a predefined style taxonomy. A generation-level memory accumulates target-model insights across the population and feeds them back into strategy generation. On the HarmBench-test split, EvoFlint reaches attack success rates of 35.8% on Claude Sonnet 4.6, 59.7% on GPT-5.4, and 94.3% on Qwen3-32B, alongside 98.7% on the older GPT-4o included as a baseline reference. The resulting archive, organized by risk category, exposes for each target which categories of harm its safety training has and has not covered.
[NLP-110] Are Near-Tied LLM Rankings Robust to Family-DIF-Guided Benchmark Recomposition?
【速读】： 该论文旨在解决语言模型性能评估中“小排名差距”被过度解读为模型优劣的可靠性问题，特别是这些差距是否受测试题项组成的影响。其核心问题是：当前基于排行榜的微小分数差异（如<1%）是否具有稳健性，还是极易因测试集中的具体题目选择而发生排序反转。解决方案的关键在于采用无家族标签的谱近似多维项目反应理论（MIRT），通过在互不重叠的训练-测试划分中分离出低差异项目功能（low-DIF）的题项，并构建源属性与难度平衡的固定权重评分系统；同时引入等长匹配随机子测试作为对照，以控制通用子测试变异的影响。结果表明，尽管整体排名高度相关（τ_b = 0.900–0.948），但在四个基准测试中，高达30.9%–47.1%的跨模型对在仅相差1个百分点时发生了排序反转，显著高于随机对照组（高出16.9%–28.6%，均p=0.001），表明局部近邻排序对题项构成极为敏感。这一现象在多种预设群体扰动下仍稳定存在，且残差题项-模型特征可跨所有者半区复制，但无任一模型家族在所有基准上保持一致优势。因此，研究结论强调：即使全局排名稳定，微小的排行榜差距也需辅以组合鲁棒性证据，否则其排序含义不可靠。
链接: https://arxiv.org/abs/2609.00482
作者: Qiaoyuan Zheng,Yiqu Yang
机构: ETH Zurich(苏黎世联邦理工学院); Zurich, Switzerland(苏黎世, 瑞士)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: Code and data artifacts will be released
Abstract:Small leaderboard gaps are often interpreted as evidence that one language model is better than another, but their sign may depend on which benchmark items are included. We test this using item-level responses from five benchmarks and a family-label-free spectral approximation to multidimensional item-response theory (MIRT). In owner-disjoint folds, one owner half identifies items with low residual differential item functioning across model families (low-DIF); the resulting frozen, source- and easiness-balanced weights score models in the other half, while equally short matched-random subtests control for generic subtest variation. Full-benchmark and low-DIF rankings remain strongly correlated ( \tau_b=.900 – .948 ). Yet in four of five benchmarks, 30.9–47.1% of cross-family pairs initially within one percentage point reverse order, exceeding their matched-random medians by 16.9–28.6 percentage points (all p=.001 ). The fifth benchmark shows no reliable excess ( -0.9 points, p=.689 ). The pattern survives all pre-specified population perturbations, and residual item–family signatures replicate across owner halves; however, no family shows a consistent advantage across benchmarks. Thus, globally stable rankings can still leave individual near-tie orderings sensitive to benchmark composition, and sub-one-point leaderboard gaps should be accompanied by evidence that the implied ordering is composition-robust.
[NLP-111] Exploring Collaboration between a language and a non-language agent EMNLP2026
【速读】： 该论文旨在解决大语言模型（LLM）与非语言型智能体（如棋类引擎、机器人控制器等）在复杂任务协作中因“言语化”（verbalization）带来的信息损失问题。具体而言，当前主流方法依赖将非语言智能体的连续状态表示（如神经网络中间特征或动作空间）压缩为自然语言文本描述，这一过程不可避免地造成信息丢失，形成“言语化债务”（verbalization debt），限制了整体系统性能。为克服此瓶颈，论文提出关键解决方案——潜在状态内化（latent state internalization），即通过学习将子智能体的连续状态直接嵌入到大语言模型的词元流中，作为可动态重编码的状态词元（state tokens），实现对环境状态的高保真、低延迟传递。实验表明，采用该方法的14B参数模型LLAMIA在六项跨行为模仿、状态评估与自然语言解释的协同国际象棋任务上，不仅超越了专用任务微调模型及前沿工具增强模型（如GPT-5.1 with tool access），且在分布外泛化能力上显著优于传统方法，验证了潜在状态内化有效缓解了言语化带来的性能衰减。
链接: https://arxiv.org/abs/2609.00474
作者: Harini S I,Somesh Singh,Yaman K Singla,Rajiv Ratn Shah,David Doermann,Balaji Krishnamurthy
机构: Adobe Media and Data Science Research (MDSR); IIIT-Delhi; IIT Kanpur; SUNY at Buffalo
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Accepted at EMNLP 2026
Abstract:LLMs are increasingly deployed as orchestrators that coordinate specialized subagents to solve complex tasks through natural language. However, in many important domains like game playing and robotics, the strongest available agents are not language models. Integrating non-language agents with LLMs would require \emphverbalization: compressing their rich continuous representations into sparse textual summaries at each interaction step. To study whether verbalization constitutes a bottleneck, we introduce \textscLLAMIA-Bench, a suite of six diverse collaborative chess tasks spanning three facets: behavioral imitation, state assessment, and natural-language explanation. Each task instantiates a well-established chess problem that neither the LLM nor the chess engine can solve alone. To solve LLM collaboration with non-language agents, we introduce \emphlatent state internalization, which projects the subagent’s continuous representations directly into the LLM’s token stream as learned state tokens, with dynamic re-encoding as actions advance the environment state. Comparing internalization to verbalized integration, our experiments reveal a consistent \emphverbalization debt: the performance gap widens throughout training and persists as the LLM scales from 4B to 14B parameters. A single 14B model, \textscLLAMIA, trained with latent state internalization, matches or exceeds task specialists and frontier models including GPT-5.1 with tool access across all benchmark tasks, and generalizes out-of-distribution where task-specific finetunes collapse
[NLP-112] oppling the Hierarchy in Byte-level Language Modeling
【速读】： 该论文旨在解决当前字节级模型（byte-level models）在字符级操作任务中表现不佳的问题。尽管最先进的字节级模型采用分层结构，从字节级别下采样至词级别再上采样回字节级别以提升训练与推理效率，但研究发现这种分层设计本身限制了模型对字符级别的精细理解能力，导致其在字符操纵任务上的性能不如纯字节级模型。通过将Transformer层分解为注意力机制与前馈网络组件进行消融实验，进一步揭示了字节级注意力机制是驱动该现象的主要因素。综合结果表明，分层结构虽提升了计算效率，却牺牲了细粒度字符理解能力，从而确立了计算效率与字符级理解之间的明确权衡关系。
链接: https://arxiv.org/abs/2609.00463
作者: Lukas Edman,Alexander Fraser
机构: TU Munich (慕尼黑工业大学); Munich Center for Machine Learning (慕尼黑机器学习中心); Munich Data Science Institute (慕尼黑数据科学研究所)
类目: Computation and Language (cs.CL)
备注:
Abstract:This work examines recent byte-level models and their failure to perfectly manipulate characters. State-of-the-art byte-level models use a hierarchical structure, starting at the byte level, downsampling to the word level, and then upsampling back to bytes. While this improves training and inference efficiency, we find that the hierarchical design itself limits character-level understanding, with pure byte-level models consistently outperforming hierarchical variants on character manipulation tasks. Ablating transformer layers into attention and feed-forward components further reveals that byte-level attention is the primary mechanism driving this behavior. Together, our results provide an explanation for the character-level failures of hierarchical byte models and establish a clear trade-off between computational efficiency and fine-grained character understanding.
[NLP-113] Location-Aware Language Models via Secondary Embeddings
【速读】： 该论文旨在解决预训练的基于Transformer的自然语言模型在编码地理语义信息方面能力不足的问题，导致地名和空间实体的表示效果不佳。其核心解决方案是提出一种轻量级、模型无关的方法，在不修改分词器且无需昂贵重训练的前提下，通过将地点名称与其对应的经纬度信息相结合，以结构化地理信号增强输入表示，并采用聚焦于位置的掩码策略，使文本表征更好地对齐真实世界的空间关系。该设计在保留原有语义与句法知识的同时，显著提升了嵌入表示的地理空间一致性。实验结果表明，该方法在保持标准NLP基准（如GLUE）性能相当的情况下，大幅改善了地理空间对齐效果，且计算开销极低，仅需数分钟额外训练时间，具备跨多种模型架构与规模的良好泛化能力。
链接: https://arxiv.org/abs/2609.00454
作者: Gokul Srinivasagan,Munir Georges
机构: AImotion Bavaria(艾莫顿巴伐利亚); Technische Hochschule Ingolstadt(英戈尔施塔特工业大学), Germany(德国)
类目: Computation and Language (cs.CL)
备注: Accepted for publication at the 29th International Conference on Text, Speech and Dialogue (TSD 2026)
Abstract:Pretrained transformer-based language models achieve strong performance across a wide range of NLP tasks but remain limited in encoding geo-locational semantics, leading to suboptimal representations of place names and spatial entities. In this work, we propose a lightweight, model-agnostic approach for injecting geo-spatial awareness into pretrained embeddings without modifying the tokenizer or requiring costly retraining. Our method augments input representations with structured geographic signals by combining location names with their corresponding latitude and longitude, and employs a location-focused masking to better align textual representations with real-world spatial relationships. This design allows the model to incorporate geo-spatial context while preserving existing semantic and syntactic knowledge. Experimental results demonstrate substantial improvements in geo-spatial alignment while maintaining comparable performance on standard NLP benchmarks such as GLUE. The method is computationally efficient, requiring only minutes of additional training, and generalizes across multiple model architectures and scales.
[NLP-114] Group Adaptive Clipping Policy Optimization EMNLP2026
【速读】： 该论文旨在解决强化学习中基于可验证奖励（Reinforcement Learning with Verifiable Rewards, RLVR）的组相对策略优化（Group Relative Policy Optimization, GRPO）方法在使用固定重要性采样（Importance Sampling, IS）比率裁剪边界时存在的关键问题：在较难任务上，正确轨迹（correct rollouts）稀少而学习信号较强，但在较易任务上，正确轨迹丰富但学习信号较弱，二者却受到相同的裁剪处理，导致高价值的学习信号被过度抑制。其核心解决方案是提出一种名为组自适应裁剪策略优化（Group Adaptive Clipping Policy Optimization, GAPO）的方法，该方法通过动态调整裁剪阈值以匹配每条轨迹的收益优势（advantage），从而为具有更大学习信号的低成功率轨迹提供更充足的更新空间。GAPO基于反向KL信任区域视角，主张高学习信号的轨迹应获得更大的策略更新容差，且无需奖励重塑，仅在标准PPO/GSPO代理函数基础上调整裁剪阈值，保持了原有框架的稳定性。实验表明，在Qwen和Llama模型上，GAPO在数学推理与代码生成基准任务中均显著优于固定裁剪和优势重塑基线，尤其在基础模型通过率较低的情况下表现出更强的提升能力。
链接: https://arxiv.org/abs/2609.00444
作者: Sheng Jia,Xiao Wang,Shiva Prasad Kasiviswanathan,Rein Houthooft
机构: University of Toronto; Amazon(亚马逊)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: Accepted at EMNLP 2026 (Main Conference)
Abstract:Group relative policy optimization for reinforcement learning with verifiable rewards (RLVR) typically uses a fixed importance-sampling (IS) ratio clipping boundary across all rollouts. We identify a key limitation: rare correct rollouts on harder problems and abundant correct rollouts on easier problems are clipped at comparable rates, despite contributing very different learning signals. Rollouts with low group success exhibit larger IS ratios and carry stronger gradient signal for exploration and solving new problems, yet are disproportionately suppressed by fixed clipping. To address this, we propose Group Adaptive Clipping Policy Optimization (GAPO), a plug-in modification to GRPO methods that adapts the clipping boundary to the rollout advantage. GAPO is motivated by a reverse-KL trust-region perspective, which suggests that rollouts with larger learning signal should receive proportionally greater update headroom. GAPO requires no reward shaping and preserves the standard PPO/GSPO surrogate while adapting only the clipping threshold. Across Qwen and Llama models, GAPO consistently improves both Pass@1 and Pass@k over fixed clipping and advantage-shaping baselines on math reasoning and coding benchmarks where the pass rates by the base model are relatively low. Comments: Accepted at EMNLP 2026 (Main Conference) Subjects: Machine Learning (cs.LG); Computation and Language (cs.CL) Cite as: arXiv:2609.00444 [cs.LG] (or arXiv:2609.00444v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.00444 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[NLP-115] (V)LMs generalize beyond surface co-occurrence: Evidence from cross-modal number agreement
【速读】： 该论文旨在解决语言模型（Language Models, LMs）是否能够习得抽象语法规则而非仅依赖于表面词汇共现频率的问题。传统观点认为，语言模型因受词频效应影响，表现出对具体词汇项的依赖，难以体现真正的抽象规则学习能力。为突破这一争议，论文提出以跨模态泛化（cross-modal generalization）作为检验手段，利用可接受视觉输入的视觉语言模型（Vision-Language Models, VLMs），将判断语法数（grammatical number）的依据限定在非语言模态——即视觉线索，从而排除语言分布性线索（如is/are、this/these）的干扰。其关键解决方案在于：通过仅更新新名词的嵌入表示，并在不同条件下分别以视觉或文本线索区分数的一致性，系统考察模型在行为表现、表征动态及因果机制层面的跨模态泛化能力。研究发现，无论是在纯视觉线索还是语言-视觉联合线索条件下，模型均展现出显著的跨模态泛化能力，且内部机制对两类线索的处理方式相似。这表明，统计学习者如VLMs能够超越表面共现模式，实现与抽象规则一致的行为，支持了其具备深层抽象能力的结论。
链接: https://arxiv.org/abs/2609.00443
作者: Zach Studdiford,Kanishka Misra
机构: University of Wisconsin-Madison(威斯康星大学麦迪逊分校); The University of Texas at Austin(德克萨斯大学奥斯汀分校)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 9 pages main text
Abstract:Language models learn about grammatical number primarily from co-occurrence, and show frequency effects as a result—sometimes taken to indicate that they do not learn abstract ``rules’', and are instead dependent on specific lexical items. Testing generalization with text stimuli alone cannot settle this debate, since distributional cues (is/are, this/these) easily give number away. We instead use cross-modal generalization as a tool to investigate abstractions in LMs that can also accept visual inputs (VLMs), restricting the evidence that diagnoses number to an extra-linguistic modality. We teach VLMs pairs of new nouns by adding new embeddings and only updating them during learning, comparing conditions where number is diagnosed by visual cues alone against ones where it is disambiguated by text. Across behavior, representational dynamics, and causal mechanisms, we find non-trivial evidence for cross-modal generalization across both exposure conditions, and that linguistic vs. extra-linguistic cue conditions are treated in similar ways in the internal mechanisms of the model. This suggests that statistical learners like VLMs can generalize beyond surface-level co-occurrence and show genuine abstraction-compatible behavior.
[NLP-116] SAGE: State-Grounded Abstention-Aware Evaluation of Task-Oriented Dialogue Agents
【速读】： 该论文旨在解决任务导向对话智能体（task-oriented dialogue agents）评估中一个核心难题：传统基于大语言模型（LLM）的全盘评价方法难以准确判断每一轮对话是否正确推进了任务流程状态，因其将上下文视为单一整体进行评估，并需对每轮对话调用一次或多次完整模型，导致评估成本高且易忽略状态演进细节。为此，论文提出SAGE（State-Grounded Abstention-Aware Evaluation）评估框架，其关键在于将任务流程规范与逐轮状态差异（per-turn state diff）转化为原子化、基于模式（schema-grounded）的可验证标准，并通过符号规则与编码器/自然语言推理（NLI）验证器级联处理，支持“弃权”而非猜测，从而生成具有证据链的轮次级决策。SAGE-Core在无需付费LLM的前提下，仅依赖编译器、符号规则和本地设备编码器即可对81%–91%的标准作出判断，实现零成本评估；而SAGE-LLM则引入可选的聚焦式LLM回退机制以处理开放类标准。在MultiWOZ、Schema-Guided Dialogue和ABCD等多个数据集切片上的实验表明，所有被评估的LLM作为裁判的基线（包括具备状态感知能力的GPT-4.1 judge及其轻量版本）均未在任一切片上显著超越SAGE-Core，尽管后者成本仅为前者的0.05–0.21美元/千轮。双标注员人工审计（n=200，κ=0.94）进一步验证了在可见失败类别上的标签一致性，排除弱显著性干扰项后，SAGE-Core与最强的LLM裁判统计学等效，同时能合理地将“忽略用户价值”视为状态一致性信号，尽管其人类显著性较弱。研究还分析了注入故障和部分符号循环性带来的构念效度局限。
链接: https://arxiv.org/abs/2609.00434
作者: Rayan Khoury,Shih-Yao Lin,Pratyush Mishra
机构: 未知
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Evaluating task-oriented dialogue agents requires judging not merely whether a reply reads well but whether each turn advances the underlying workflow state correctly–a distinction conventional holistic LLM judges can miss because they evaluate the available context as a single unit and require one or more full-model calls per turn. We propose SAGE (State-Grounded Abstention-Aware Evaluation), which compiles a workflow specification and per-turn state diff into atomic, schema-grounded criteria and routes each through a cascade of symbolic and encoder/NLI verifiers that abstain rather than guess, aggregating criterion verdicts into a turn-level decision with an evidence trace. Its recommended operating point, SAGE-Core, decides 81–91% of criteria with only the compiler, symbolic rules, and on-device encoders–at zero paid LLM cost–while SAGE-LLM adds an optional focused-LLM fallback for open-class criteria. Across four slices spanning MultiWOZ, Schema-Guided Dialogue, and ABCD, no evaluated LLM-as-a-judge baseline–including a state-aware GPT-4.1 judge and cheaper GPT-4.1-mini variants–significantly exceeds SAGE-Core on any slice, even though the GPT-4.1 G-Eval judge costs 4.7–8.0 per 1,000 turns to SAGE-Core’s 0. A two-annotator human audit (n=200, \kappa =0.94) confirms strong label fidelity on the transcript-visible failure classes–where, excluding the weak-salience IUV class, SAGE-Core is statistically tied with the strongest LLM judge–and honestly scopes ignored-user-value as a state-consistency signal with weak broad-human salience. We analyze construct-validity limits from injected failures and partial symbolic circularity.
[NLP-117] Late Transformer Layers Recode Syntax Canonically: Evidence from Greek Scrambling and Cross-Layer Generalisation
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）中句法信息在深层（late layers）的表征演化机制这一关键问题。尽管已有研究证实句法信息可在早期和中期层被有效解码，但其在后期层的演变过程仍不明确。针对此问题，作者采用跨层泛化分析方法，在三个经过希腊语微调的大型语言模型上，对现代希腊语中仅在句内词序上存在差异（即标准语序SVO与非标准语序VSO）的最小对比对进行测试。结果表明，当在晚期层（第20-31层）训练的探测器（probe）测试于各早期层时，其性能低于随机水平（经聚类校正后，p < 0.01），且将99.3%的非标准句误判为标准句。此外，探测系数在约第22层发生符号反转，提示模型在深层并非简单丢失句法信息，而是发生了向标准语序的定向重构。该发现揭示了深层变压器模型中存在超越句法可解码性下降的表征格式转变，提出了一个可直接验证的预测：人类脑电（EEG）与脑磁（MEG）研究在使用相同刺激材料时应观察到类似的方向性神经表征变化。研究代码与实验材料已公开发布于OSF平台。
链接: https://arxiv.org/abs/2609.00416
作者: Christos Nikolaos Zacharopoulos,Revekka Kyriakoglou,Chara Tsoukala,Théo Desbordes
机构: Université Paris 8 Vincennes–Saint-Denis (巴黎第八大学); Institute for Language and Speech Processing (语言与语音处理研究所); Faculty of Medicine, University of Geneva (日内瓦大学医学院); Athena Research Center, Athens, Greece (雅典研究中心, 希腊雅典)
类目: Computation and Language (cs.CL)
备注: 10 pages, 3 main figures, 2 appendices. Code and stimuli: this https URL
Abstract:Probing studies have established that syntactic information is decodable in early and middle transformer layers, but what happens to that information in later layers remains poorly understood. We apply a cross-layer generalisation analysis to three Greek-tuned large language models evaluated on tightly controlled minimal pairs: object-relative constructions in Modern Greek, where canonical (Subject-Verb-Object; SVO) and non-canonical (Verb-Subject-Object; VSO) orders differ only in within-clause word order, while preserving propositional meaning. When a probe trained on late layers (20-31) is tested on each early layer individually, it produces below-chance transfer (cluster-corrected, p0.01), classifying 99.3% of non-canonical sentences as canonical. Probe coefficients reverse sign around layer 22, indicating a directional recoding toward the canonical form rather than simple information loss. These findings characterise a representational format change in late transformer layers that goes beyond the well-established decline in syntactic decodability, and they generate a directly testable prediction for human EEG and MEG decoding studies using the same stimuli. Code and stimuli are publicly available on OSF.
[NLP-118] Removable and Irreducible: A Token-Cost Ledger for the Multilingual Tokenization Tax
【速读】： 该论文旨在解决大语言模型在处理非英语文本时所面临的显著“编码税”问题，即相同语义内容在非英语（尤其是印地语等使用复杂脚本的语言）中需消耗更多标记（token），导致计算成本呈序列长度的二次方增长。其核心问题是：这一额外开销在多大程度上可被消除？解决方案的关键在于将标记层视为源编码（source coding），基于香农熵率（Shannon rate H/log2V）构建一个可量化的标记成本账本（token-cost ledger），将总成本分解为可移除的编码冗余、残余编码松弛、固有内容项以及不可约的字符到音素映射项（grapheme-to-phoneme term）。实证分析表明，在FLORES-200数据集上，针对印地语等脚本语言，现有生产级分词器的标记开销最高可达英语的8.9倍；通过仅用1,012句训练的脚本匹配编码器，可移除中位数64%的超额开销（置信区间[0.638, 0.647]），而内在内容差异不足6%，说明该“税”本质是表征性而非信息性的。进一步构造的理想编码器可消除98%的冗余，且该标记税可能导致注意力计算成本高达79倍。研究强调其贡献在于统一的会计框架、可移除与固有成分的归因方法，以及开源的一键式工具链，同时明确指出其范围局限——仅为计算与内存开销分析，不涉及模型性能或跨语言方向性判断。
链接: https://arxiv.org/abs/2609.00378
作者: Madhulatha Mandarapu,Sandeep Kunkunuru
机构: Samyama(萨米亚); Samyama(萨米亚)
类目: Computation and Language (cs.CL)
备注: 9 pages, 3 figures. Code + one-command reproduction: this https URL
Abstract:Large language models pay a well-documented tax on non-English text: the same content costs several times more tokens, and because attention is quadratic in sequence length, far more compute. We ask how much of this tax is removable. Framing the token layer as source coding – transformer compute is monotone in sequence length, whose per-atom floor is the Shannon rate H/\log_2 V , an object already applied to tokenizers in prior work – we assemble a token-cost ledger that splits each language’s cost, at fixed parallel content, into a removable coding redundancy, a residual coding slack, an intrinsic-content term, and an orthogonal, irreducible grapheme-to-phoneme term that governs the multimodal rather than the text cost. On FLORES-200 across eight languages, a production tokenizer costs up to 8.9\times more tokens for Indic scripts than for English; a script-matched code trained on 1,012 sentences removes a median 64% of that excess (bootstrap 95% CI [0.638, 0.647] ), and a script-fair information floor shows the intrinsic content differs by under 6% – the tax is representational, not informational. A constructed code removes 98% of a controlled source’s redundancy, and the token tax implies up to 79\times attention cost. We are explicit about scope and failure: this is compute-and-memory accounting, not a model-quality claim; we neither measure nor claim the cross-lingual direction of the orthographic term; and our matched code is a conservative small-data demonstration. We contribute the unifying ledger, the removable-versus-intrinsic attribution, and an open one-command harness.
[NLP-119] Neurosymbolics for Data Engineering: Achieving Long Context Token Reduction Without Finetuning
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在数据工程任务中面临的两大核心挑战：一是缺乏微调（finetuning-free）情况下的高精度推理能力，二是基于Transformer架构固有的二次方时间复杂度（O(n²)）导致的长文本上下文计算瓶颈。其解决方案的关键在于引入一种可即插即用的神经符号（neurosymbolic）层，无缝集成至现有LLM主干网络中。该层通过增强逻辑推理能力，在不依赖特定任务微调或强化学习人类反馈（RLHF）的前提下，显著提升了模型在BIRD-CRITIC和LiveSQLBench等严格基准上的表现，平均准确率提升达85%。同时，该方法利用符号处理机制对上下文信息进行优先级筛选与压缩，有效将长上下文场景下的实际令牌使用量降低超过50%，并将有效时间复杂度从O(n²)降至约O(n)，从而大幅缓解推理芯片的计算压力，使长上下文任务在性能与成本上均更具可行性。
链接: https://arxiv.org/abs/2609.00367
作者: Vishvesh Bhat
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 are increasingly deployed for sophisticated data engineering tasks such as generating structured queries from natural language, Text-to-SQL, and automating complex spreadsheet operations. However, maximizing their utility demands both higher finetuning-free accuracy and solutions to the computational bottleneck imposed by the Transformer architectures inherent quadratic (On2) time complexity. This paper introduces a novel drop-in neurosymbolic layer designed to seamlessly integrate into existing LLM backbones enhancing logical reasoning and mitigating long-context resource consumption. On the reasoning front, the layer immediately and significantly improves performance yielding an average accuracy increase of 85% across rigorous benchmarks including BIRD-CRITIC and LiveSQLBench, critically achieving these gains without any task specific finetuning or RLHF. Concurrently, we repurpose this approach to address the severe computational strain of long context inference. By leveraging symbolic processing to prioritize and compress relevant contextual information the layer reduces the effective token usage by over 50% and brings the effective time complexity down from O(n2) to approximately O(n) on certain long context tasks. This dual impact approach not only makes LLMs substantially more reliable for data engineering but also drastically reduces the computational pressure on inference chips, making long context tasks more manageable and cost effective.
[NLP-120] Dr. Claw: An AI Scientist Workspace for Vibe Research EMNLP2026
【速读】： 该论文旨在解决当前命令行编程代理（如Claude Code、Gemini CLI）在实际科研工作流中面临的碎片化问题：研究过程分散于聊天工具、集成开发环境（IDE）、终端和写作平台之间，且关键决策缺乏可追溯性。其核心解决方案是提出Dr. Claw——一个开源的工作空间框架，通过在现有编程代理执行器外层封装可控且可审计的人机协同工作流，而非引入新的自主代理。其关键技术包括持久化状态对象、可复用的技能库以及多执行器协调机制，将规划、执行与写作统一为可追踪、可恢复的闭环流程。实验表明，在固定底层执行器的前提下，相较于仅依赖原始命令行代理，Dr. Claw在研究完整性方面表现更优，同时保留了完整的可审计与可恢复的过程轨迹。
链接: https://arxiv.org/abs/2609.00365
作者: Dingjie Song,Hanrong Zhang,Dawei Liu,Yixin Liu,Zongxia Li,Zhengqing Yuan,Siqi Zhang,Henry Peng Zou,Zhiling Yan,Yuxuan Zhang,Yanfang Ye,Philip S. Yu,Lichao Sun
机构: Lehigh University(莱赫大学); University of Illinois Chicago(芝加哥伊利诺伊大学); University of Pennsylvania(宾夕法尼亚大学); University of Maryland(马里兰大学); University of Notre Dame(圣母大学); University of British Columbia(不列颠哥伦比亚大学); Philip S. Yu(菲利普·S·余)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: Accepted to EMNLP 2026 System Demonstrations. Code: this https URL
Abstract:Command-line coding agents (e.g., Claude Code, Gemini CLI) can already read and write files and sustain long sessions, yet end-to-end research still fragments across chat tools, IDEs, terminals, and writing environments, and the decisions that make it auditable are rarely preserved. We present Dr. Claw, an open-source workspace that wraps existing coding-agent executors in a controllable and auditable human-in-the-loop workflow rather than introducing another autonomous agent. Persistent state objects, a reusable skill library, and multi-executor coordination link human decisions to AI execution, turning planning, execution, and writing into one traceable, recoverable loop. We demonstrate Dr. Claw through an interactive three-view scenario and a failure-recovery walkthrough, and evaluate it against a bare command-line agent sharing the same backend executor, so the comparison contrasts the whole orchestration layer (task graph, state objects, and skill library) with the agent it wraps. Holding the executor fixed, Dr. Claw scores higher on research completeness while persisting an auditable, recoverable process trail. Demo access: repository this https URL, released under AGPL-3.0 with GPL-3.0 upstream components.
[NLP-121] Detoxifying Toxic Communication: A Design Science Approach to Responsible AI DATE
【速读】： 该论文旨在解决数字工作场所中隐性恶意语言（如贬损、讽刺、居高临下语气及微妙不文明行为）对信任、士气与协作的侵蚀问题。现有内容审核工具多采用删除或屏蔽有害信息的被动策略，虽能遏制不当言论，但破坏了沟通连续性且缺乏建设性修复机制。为此，研究采用设计科学方法（Design Science Research），提出一种负责任的生成式AI（Generative AI）干预框架，其核心在于构建一个融合细调的Transformer分类器（DistilBERT、DistilRoBERTa）与生成式去毒模型（mT0-XL-Detox-ORPO）的集成系统：前者精准识别毒性文本，后者则在保持原意语义等价的前提下，将有毒表达重构为非冒犯性同义改写。技术评估表明，该系统在毒性检测上具备高准确率，并在重写过程中显著保留原文语义，有效维持对话连贯性的同时促进尊重性交流。研究的关键贡献在于提出了以语义保真度和公平性为核心的负责任AI内容治理设计原则，推动从“清除”到“修复”的范式转变。
链接: https://arxiv.org/abs/2609.00361
作者: Hossein Arshadi Soufiani,Henry M. Kim,Hjalmar Turesson,Syed Mohammad Arham Noman,Anav Setia
机构: 未知
类目: Computers and Society (cs.CY); Computation and Language (cs.CL)
备注: 10 pages. An updated version appears in the Proceedings of the 60th Hawaii International Conference on Systems Science (HICSS-60), Honolulu, HI, January 5-8, 2027
Abstract:Toxic language in digital workplaces such as pejoratives, sarcasm, condescension, and subtle incivility can erode trust, morale, and collaboration. Existing moderation tools primarily delete or block harmful messages, disrupting communication and offering no constructive resolution. This study adopts a Design Science Research approach to create a responsible AI artifact that detects and detoxifies toxic communication. The artifact integrates fine-tuned transformer-based classifiers (DistilBERT, DistilRoBERTa) with a generative detoxification model (mT0-XL-Detox-ORPO) that rewrites toxic text into semantically equivalent, non-offensive paraphrases. Technical evaluation demonstrates high accuracy in toxicity detection and strong semantic preservation in rewritten messages, supporting conversation continuity while reinforcing respectful discourse. The paper contributes design principles for responsible AI moderation that prioritize meaning preservation and fairness.
[NLP-122] Vision Is Not Overhead: One-Pass Block Drafting for Lossless Speculative Decoding in Vision-Language Models
【速读】： 该论文旨在解决生成式AI（Generative AI）在视觉-语言模型（Vision-Language Models, VLMs）中应用推测解码（Speculative Decoding）时陷入的自我削弱循环问题。传统方法中，由于起草器（drafter）必须保持自回归特性而规模受限，无法有效利用图像信息，导致视觉信息被压缩、修剪或隐藏，从而在图像能显著提升文本预测性的重要区域表现最不可靠。针对此问题，论文提出GLANCE——首个无需修改目标模型即可实现无损单次通过块起草（one-pass block drafter）的新架构。其核心突破在于引入一个块扩散头（block-diffusion head），直接读取目标模型已融合的多模态状态，使起草器无需额外计算视觉特征，且可在一次前向传播中完成整块文本生成，彻底消除序列步数带来的深度开销。同时，通过宽候选树验证机制，在一次目标模型推理中完成所有校验，确保每个审计提示均精确复现贪婪解码结果。该方案在强依赖视觉上下文的任务中尤为高效，进入“逐字复制”模式，显著优于传统自回归起草器。实验表明，在相同引擎与回合预算下，GLANCE比自回归生成快达2.93倍，可接受块长为同源训练的EAGLE-3头的2.7倍。研究进一步发现，接受长度由目标模型下一个词熵决定，并呈现一条跨任务、跨模态的统一规律，其斜率随任务的视觉锚定程度增强而上升，揭示了自由运行文本仍偏好链式生成的本质边界。
链接: https://arxiv.org/abs/2609.00355
作者: Jungseob Lee,Seongtae Hong,Dongyub Jude Lee,Chanjun Park,Jaehyung Seo,Sugyeong Eo,Heuiseok Lim
机构: 未知
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV)
备注: 18 pages, 9 figures, 17 tables. Code: this https URL
Abstract:Speculative decoding accelerates generation without changing its output, yet on vision-language models (VLMs) it has been caught in a self-defeating cycle. The drafter stays autoregressive, so it must stay small. A small drafter cannot afford the image at every step, so vision is compressed, pruned, or hidden. A drafter cut off from the image is then least reliable exactly where the image makes text predictable. We present GLANCE, the first one-pass block drafter that is lossless on an unmodified VLM target, and it breaks the cycle at both ends. A block-diffusion head reads the target’s already-fused vision-language state, so vision costs the drafter nothing, and fills a whole block in one forward pass, so depth costs no sequential steps. A wide candidate tree is verified in one target pass, and every audited prompt reproduces greedy decoding exactly. Grounded workloads reward this most, entering a verbatim-copy regime whose long runs cost an autoregressive drafter a pass for every token and a block drafter one in total. Under one engine and one round budget, GLANCE decodes up to 2.93x faster than autoregression, from one draft pass a round where the production EAGLE3-VL head takes eight, and accepts 2.7x longer blocks than an EAGLE-3 head trained on the same corpus. One law organizes these results. Accepted length is set by the target’s next-token entropy, with a fitted slope that steepens with grounding across all five tasks. The law transfers across targets and modalities and names its own boundary, since free-running text still favors a chain. Our code is available at this https URL.
[NLP-123] Detecting Hidden Behaviors in LLM s via Activation-matched Finetuning
【速读】： 该论文旨在解决大语言模型中隐藏行为（hidden behaviors）的检测难题，这类行为在特定稀疏触发条件下才会激活，如后门触发器、睡袋代理部署信号、蓄意降级或话题条件性审查等，且由于缺乏对触发机制或目标行为的先验知识，传统检测方法难以有效识别。其解决方案的关键在于提出一种无监督的激活匹配微调（activation-matched finetuning）方法：通过将一个公开可用的锚模型（anchor model）在少量良性语料上微调，使其在特定输入下尽可能复现可疑模型的激活模式，进而通过计算两者在评估提示上的激活残差来检测异常。由于良性语料无法覆盖稀疏的触发区域，锚模型仅学习到正常的语义计算路径，而未能习得隐藏行为，因此触发提示及其语义邻域会产生显著的残差，从而暴露异常行为。实验表明，该方法在第三方与自定义模型上均能可靠地发现隐藏行为；同时，针对自然防御感知攻击的实证分析也证明，此类攻击无法在不牺牲隐藏行为本身的前提下规避该检测方法，凸显了该方案的有效性与鲁棒性。
链接: https://arxiv.org/abs/2609.00351
作者: Robin Haselhorst,Lucie Flek,Florian Mai
机构: Bonn-Aachen International Center for Information Technology, University of Bonn, Germany; Lamarr Institute for Machine Learning and Artificial Intelligence, Germany
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: under review
Abstract:Large language models can hide hidden behaviors that activate only under narrow conditions, such as backdoor triggers, sleeper-agent deployment cues, sandbagging, or topic-conditioned censorship. Such behaviors are difficult to detect without prior knowledge what to look for. We present activation-matched finetuning, an unsupervised detection method that assumes no knowledge of the trigger or the target behavior. Given a suspect model and a publicly available anchor, we finetune the anchor to reproduce the suspect’s activations on a small benign corpus, and score each evaluation prompt by the residual between the two models. Since no benign corpus covers the sparse trigger region, the reference learns the benign computation but not the hidden behavior. Therefore, trigger prompts – and, crucially, their semantic neighbors – incur a large residual that signal the presence of unusual behavior to the defender. Testing our method across third-party models and custom models, activation-matched finetuning surfaces hidden behavior reliably. Furthermore, we empirically consider a natural defense-aware attack and showcase that it fails to suppress our detection method without sacrificing the behavior itself.
[NLP-124] From Tool Use to Technological Agency: LoopCAT as a Local-First Open-Source Tool for Translation Technology Education
【速读】： 该论文旨在解决翻译教育中学生在使用翻译技术时缺乏对技术决策机制的理解与评估能力的问题。当前翻译教学往往侧重于工具操作，而忽视了学生对技术生成结果的批判性判断能力培养。其解决方案的关键在于提出一个名为LoopCAT的本地优先型计算机辅助翻译（Computer-Assisted Translation, CAT）环境，该环境由开发者与OpenAI Codex协作构建，并基于GPT-5.5和GPT-5.6模型实现。LoopCAT不仅集成项目本地存储、术语管理、翻译记忆、质量保证及文档交换等功能，还通过多语言用户界面（支持英语、加泰罗尼亚语和土耳其语）为教学提供可操作的实践材料：学生可参与将英文界面文本翻译成目标语言，审阅自动生成的译文草案，导入修改版本并测试界面效果。论文进一步提出一个整合工作流程能力、评价判断力与技术能动性的框架，围绕四种参与形式展开——执行工作流、评估输出、检视与配置系统机制、做出或辩护有限干预。该框架以六次课时的教学序列、界面本地化任务、示例模板及评估量规的形式具体化，为教师提供实施路径。值得注意的是，论文明确区分已实现功能与预期教育价值，未报告新的学习成效数据；同时通过版本检查记录与课堂评估协议，为后续实证研究奠定基础。因此，尽管其核心贡献在于构建了一个可审计、可探究的技术教学环境，但其对学生判断力提升、知识迁移或主动参与的促进作用仍需进一步实证验证。
链接: https://arxiv.org/abs/2609.00344
作者: Gokhan Dogru,Adrià Martín Mor
机构: 未知
类目: Computation and Language (cs.CL)
备注:
Abstract:Translation students need to learn both how to use translation technologies and how to judge the choices those technologies make available. This article presents LoopCAT, an Apache-2.0-licensed, local-first computer-assisted translation environment co-created with OpenAI Codex using GPT-5.5 and GPT-5.6, and proposes a framework connecting workflow competence, evaluative judgement, and technological agency. The account draws on repository history, implementation inspection, and the verification records of an identified development build. LoopCAT combines local project storage, translation memories, terminology, quality assurance, document exchange, and optional connections to local or hosted AI services. Its English, Catalan, and Turkish interface catalogs also make the application itself available as teaching material: students can translate English UI strings into another language, review the existing automatically generated target drafts, import their revisions, and test the interface. We organize these opportunities around four forms of participation: operating a workflow, evaluating outputs, inspecting and configuring mechanisms, and making or defending a bounded intervention. A six-session sequence, a UI-localization assignment, a placeholder example, and an assessment rubric specify how teachers could use the framework. The paper separates implemented capabilities from proposed educational benefits; it reports no new student-learning outcomes. It distinguishes the latest package checks from earlier regression evidence and sets out a protocol for classroom evaluation. LoopCAT provides an inspectable setting for teaching how translation decisions interact with data, interfaces, and software rules. Whether these activities improve judgement, transfer, or participation remains an empirical question.
[NLP-125] wo locked tests of phase-structure features for transition prediction
【速读】： 该论文旨在检验旋转注意力（rotary attention）中相位结构理论所提出的相位衍生特征是否能在预测承诺（commitment）或矛盾（contradiction）任务终点时，优于不包含这些特征的基线模型。其核心问题在于验证相位特征是否具有实际预测增益。解决方案的关键在于通过两个预先设定的实证检验进行严格评估：研究1对矛盾类别管道进行冻结，对比PC-2特征与基线在1,136个有效案例上的表现，结果显示配对AUROC差异仅为+0.00087，99%置信区间包含零，且未达到预设的+0.05阈值；研究2在开放块b0-b4上构建15种层处理方案，采用锁定的合取规则要求多项正向增量均满足，但所有处理均未通过筛选。最终官方选择为无提升，表明在预先设定规则下未发现额外排名增益，因此理论论文未被撤回，但相位特征的预测优势未获实证支持。
链接: https://arxiv.org/abs/2609.00335
作者: Abraham Chachamovits
机构: 未知
类目: Computation and Language (cs.CL)
备注: 7 pages. Empirical follow-up to arXiv:2607.25507 . Both pre-specified tests are null
Abstract:A published theoretical account of phase structure in rotary attention was subjected to two pre-specified empirical tests of whether phase-derived features improve prediction of a commitment or contradiction endpoint over a baseline that does not receive those features. Study 1 froze a contradiction-category pipeline and scored a sealed primary comparison of PC-2 against baseline. On 1,136 eligible cases the paired AUROC difference was +0.00087. The 99% interval included zero, and the difference did not reach the pre-specified threshold of +0.05. Advancement was not passed. Study 2 developed fifteen layer treatments on open blocks b0-b4 only (1,415 transitions, 20x5 grouped folds). A locked conjunctive rule required a positive PC-2 mean-repeat increment, a positive increment on at least four of five seed blocks, and a positive mean of those five differences. No treatment advanced. The official selection is null. The theoretical paper is not withdrawn. The extra ranking lift was not found under the rules locked in advance.
[NLP-126] opic Matching in the Wild: Benchmark and Lessons from Real-World ASR Transcripts EMNLP2026
【速读】： 该论文旨在解决呼叫中心中实时客服辅助工具在处理噪声大、结构不规范的语音转写文本（ASR transcripts）时，如何准确识别客户语句与预定义话题的相关性这一现实挑战。其核心问题是：在缺乏标点、存在重复和表达模糊的自发性电话对话文本背景下，如何高效且精准地实现话题-语句匹配。解决方案的关键在于采用基于轻量级大语言模型（LLM）的匹配器，并结合自然语言描述（natural language description）作为话题表示方式，相较于传统的正则表达式（regex）基线和零样本句向量编码器，该方法在复杂噪声环境下展现出显著更优的性能，凸显了上下文理解能力与语义表征优势。
链接: https://arxiv.org/abs/2609.00330
作者: Saman Rahbar,Xiliang Zhu,Irvin Cardoza,David Rossouw
机构: Dialpad Inc.(Dialpad公司)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Accepted at the 11th Workshop on Natural User-generated Text (W-NUT 2026), EMNLP 2026. Camera-ready version. 9 pages, 2 figures, 3 tables
Abstract:In contact centers, real-time agent-assist tools determine, for each of many predefined topics, whether a live customer utterance is relevant and display a coaching card to the agent when it is. The input is noisy and challenging: ASR(Automatic Speech Recognition) transcripts of spontaneous phone conversations, which can be unclear, repetitive, and mostly lack punctuation. To systematically study this real-world task, we curate a human-annotated topic-utterance judgments dataset sourced from real call-center transcripts. We compare three types of matchers: a regex-based baseline, zero-shot sentence embedding encoders, and Gemini-based LLM matchers. In addition, two types of topic representations are studied in our benchmark:keyphrases and natural language description. Our empirical experiments highlight the superior performance of lightweight LLM matchers over embedding and regex models when equipped with natural language descriptions.
[NLP-127] he Curse of Multilinguality in Lexical Normalization EMNLP2026
【速读】： 该论文旨在解决多语言环境下词汇规范化（Lexical Normalization）中因标注数据稀缺而采用单一模型联合训练多语言所导致的性能下降问题。其核心挑战在于：当一个固定容量的字符级模型被用于同时训练越来越多的语言时，每种语言的规范化准确率会显著降低，即存在“多语言诅咒”（curse of multilinguality）。解决方案的关键在于揭示并验证：对于紧凑型规范化模型而言，训练语言数量并非越多越好；相反，仅将少数语言（通常1至4种）共同训练，可获得最高精度，而随着语言数量增加，模型性能呈持续且显著下滑。实验通过控制总训练数据量的对照组进一步表明，性能衰退主要源于不同语言之间对有限模型容量的竞争，而非数据量不足。此外，研究检验了语言类型学距离是否能预测最优共训练语言数，结果发现并无可靠规律，任何相关性均源于少数异常语言，不具备普遍性。因此，该研究提出的核心结论是：在资源受限的生成式建模场景下，精简共训练语言数量可实现更优的规范化效果，即“少即是多”。
链接: https://arxiv.org/abs/2609.00329
作者: Saman Rahbar
机构: 独立研究员(Independent Researcher)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Accepted at the 11th Workshop on Natural User-generated Text (W-NUT 2026), EMNLP 2026. Camera-ready version. 7 pages, 3 figures, 3 tables
Abstract:Lexical normalization rewrites the noisy, non-standard words that fill user-generated text (tmrw, u, gr8) into their standard forms. Because labelled data is scarce for most languages, a popular shortcut is to train a single model on many languages at once. We ask a simple question: how many languages should such a model be trained on? Using one fixed-capacity character-level model and twelve languages from a standard benchmark, we vary the number of jointly trained languages from one to twelve and measure per-language accuracy. We find a clear curse of multilinguality: accuracy is highest when a language is trained with only a few others, often just one to four, and then falls steadily and substantially, dropping by about forty percent as the rest are piled on. A control that holds the total amount of training data constant makes the decline arrive sooner and fall further, which points to competition among the languages for one fixed-size model rather than to how much data is available. We also test whether a language’s typological distance from the others predicts its ideal number of co-training languages, and find no dependable rule: any apparent relationship rests on a couple of languages and does not hold up. For compact normalization models, less can be more: a few languages beat pooling everything into a single model.
[NLP-128] Latent Mechanisms of Language Control in Multilingual Language Models EMNLP2026
【速读】： 该论文旨在解决多语言大语言模型（Multilingual Large Language Models, MLLMs）在生成过程中出现的非预期语码转换（code-switching）问题，即模型在无需切换语言的情况下无故在不同语言间交替生成。为实现对生成语言的有效控制，论文提出并比较了三种识别跨层翻译器中语言控制潜变量（language-controlling latents）的方法：基于激活值的选择（ValSel）、基于激活频率的选择（FreqSel）以及基于大语言模型生成的潜变量注释选择（AnnSel）。其解决方案的关键在于通过针对性干预实验，在Gemma-2-2B和Qwen3-4B模型上验证这些方法对语言引导的有效性。研究发现，三者均能有效操控生成语言，其中FreqSel整体表现最优，而AnnSel则通过显式语言注释提供了可解释的潜变量选择机制。敲除分析表明，各方法所选潜变量子集互不重叠但均具功能性，揭示了语言控制方向存在冗余性而非单一标准路径。
链接: https://arxiv.org/abs/2609.00325
作者: Ryo Mitsuhashi,Sabri Boughorbel,Majd Hawasly
机构: Princeton University (普林斯顿大学); Prince Sattam bin Abdulaziz University (萨勒曼·本·阿卜杜勒阿齐兹王子大学); QCRI, Hamad Bin Khalifa University (卡塔尔计算研究所，哈马德·本·哈利法大学)
类目: Computation and Language (cs.CL)
备注: 23 pages, accepted for presentation at EMNLP 2026 main track
Abstract:Multilingual large language models can exhibit unintended code-switching – unnecessarily alternating between languages during generation. We present a comparative study of three methods that identify language-controlling latents in cross-layer transcoders: activation value-based selection (ValSel), activation frequency-based selection (FreqSel), and LLM-generated latent annotation-based selection (AnnSel). To evaluate the efficacy of these methods in identifying language-controlling latents, we introduce two multilingual benchmarks that exhibit code-switching for fine-grained analysis of language steering across seven languages. Through targeted intervention experiments on Gemma-2-2B and Qwen3-4B, we find that all three methods effectively manipulate generation language, with FreqSel achieving the strongest overall performance, while AnnSel offering interpretable latent selection through explicit language annotations. A knock-out analysis suggests the methods select non-overlapping but each-functional latent subsets, indicating redundancy rather than a single canonical language direction. Code and data can be found at this https URL.
[NLP-129] Emotional Labor Strategy Preferences in LLM Personas EMNLP2026
【速读】： 该论文旨在解决情感劳动（emotional labor）中个体人格特质如何影响情绪表达策略选择的问题，尤其关注现有研究过度依赖职业情境下的自我报告数据所导致的生态效度不足。其核心解决方案在于构建首个包含500个社会情境化事件的情感劳动策略数据集，每个事件提供表面扮演（surface acting）、深层扮演（deep acting）和真实表达（genuine expression）三种行为选项，并通过心理测量学基础的人格角色注入大规模语言模型（LLM），在非职业化的日常社交场景中模拟人格驱动的情绪调节行为。研究发现，模型更倾向于深层扮演策略，且尽责性（Conscientiousness）与情绪稳定性（Emotional Stability）是预测该偏好的关键人格维度；熵分析进一步验证了人格角色对输出的可靠影响，且该影响在不同模型和情绪类型间存在差异。
链接: https://arxiv.org/abs/2609.00310
作者: Mohammad Saim,Tianyu Jiang
机构: University of Cincinnati(辛辛那提大学)
类目: Computation and Language (cs.CL)
备注: 18 pages, 4 figures, 12 tables. Findings of the 2026 Conference on Empirical Methods in Natural Language Processing (EMNLP 2026)
Abstract:Emotional labor is the effortful management of emotional displays to meet social or professional expectations. Personality traits have been correlated with emotional labor strategies, yet research on this link relies almost exclusively on self-report scales administered only in occupational settings. We investigate whether large language models injected with psychometrically grounded personas reproduce these personality-driven selection patterns across everyday social scenarios. We construct the first emotional labor strategy dataset of 500 socially situated events, each offering three behavioral choices corresponding to surface acting, deep acting, and genuine expression. We source 50 fictional characters from a large-scale personality repository and profile each through two parallel tracks: observer-rated bipolar adjective composites and in-character self-report items. Five LLMs evaluate all scenarios under both persona conditions. We find that models align more towards deep acting, and that Conscientiousness and Emotional Stability consistently predict this preference. Entropy analysis confirms that persona reliably influences the output and varies across models and emotions.
[NLP-130] oward Workflow-Aware Benchmarking for Healthcare NLP Agents
【速读】： 该论文旨在解决当前大型语言模型（Large Language Model, LLM）代理在医疗领域评估中普遍存在的局限性问题，即现有评估多局限于静态的医学问答或单次生成任务，未能充分反映临床工作流中的动态状态演变、任务中断以及人机协作交接等真实场景。为此，论文提出一种基于“事件”（episode-level）的医疗自然语言处理（NLP）代理评估协议，其核心在于将证据来源明确区分为模型自身、代理行为与模拟工作流三个层面，并定义了包含五个字段的标准化事件模板，用于量化状态连续性、证据可追溯性及升级决策的合理性。该协议通过四个典型任务模板——病历更新、证据检索、患者消息沟通与分诊交接——实现可复现的中间层评估，既区别于静态基准测试，又不直接衡量临床结果或部署价值，同时引入对“漏报”与“误报”升级决策的成本敏感性处理，从而更真实地反映智能代理在复杂医疗工作流中的表现能力。
链接: https://arxiv.org/abs/2609.00296
作者: Junyi Yao,Baichuan Li,Zihao Zheng,Jiayu Long
机构: Washington University in St. Louis (圣路易斯华盛顿大学); Southern Methodist University (南卫理公会大学)
类目: Computation and Language (cs.CL)
备注: 4 pages, 3 tables
Abstract:Large language model (LLM) agents are increasingly proposed for healthcare tasks such as clinical documentation, evidence retrieval, patient messaging, and care coordination. Yet many evaluations remain limited to static medical question answering or one-shot generation, under-representing longitudinal state, interruptions, and human handoffs. We introduce an episode-level evaluation protocol for healthcare NLP agents. The protocol separates evidence across model, agent, and simulated-workflow behavior; specifies a five-field episode schema; and defines annotation and scoring for state continuity, evidence traceability, and escalation decisions. It is instantiated as four task templates: documentation update, evidence retrieval, patient messaging, and triage handoff. The protocol does not claim to measure clinical outcomes or deployment value. Instead, it supplies a reproducible intermediate evaluation layer between static benchmarks and prospective workflow studies, with an explicit cost-sensitive treatment of missed versus unnecessary escalation.
[NLP-131] Slow to See Slow to Suppress: Understanding the Effects of Modality in Context-Memory Conflicts EMNLP
【速读】： 该论文旨在解决多模态大模型（Vision-Language Models, VLMs）在处理上下文-记忆冲突时的行为不一致性问题，即当输入上下文中的信息与模型训练期间参数化存储的事实相矛盾时，模型如何做出判断。其核心问题是：在跨模态信息冲突下，模型为何表现出对文本信息和视觉信息的不对称依赖。解决方案的关键在于揭示了模态间表征对齐的时间延迟机制——由于视觉信息需要更长的处理时间，导致模型在推理过程中未能有效抑制其固有的基于参数的事实回忆机制，从而更倾向于依赖参数化知识；而文本信息因处理更快，更容易被采纳。研究还发现，尽管思维链（Chain-of-thought）推理无法弥合这一偏差，但增加上下文中的视觉信息量可显著影响模型决策，表明视觉信息的丰富性在调节模态偏好中具有关键作用。这一发现凸显了在日益复杂的多模态与检索增强型系统中，实现行为一致性的挑战。
链接: https://arxiv.org/abs/2609.00293
作者: Athulith Paraselli,Etha Tianze Hua,Ellie Pavlick
机构: Brown University (布朗大学)
类目: Computation and Language (cs.CL)
备注: Accepted to EMNLP Findings 2026. Code and dataset are available at this https URL
Abstract:We investigate how vision-language models (VLMs) handle context-memory conflicts; that is, situations in which the model is given information in context that differs from what was stored parametrically during training. We document asymmetric biases: models tend to prefer in-context information about entities which appear in text, but prefer parametric information about entities which appear in images. We relate this asymmetry to the late representational alignment across modalities, showing that the longer processing time associated with resolving visual entities prevents the suppression of the model’s usual factual recall mechanism, thus resulting in more parametric answers. Chain-of-thought reasoning does not appear to resolve the gap, but increasing the amount of visual information in the context does show an effect. These results illustrate the complexity of ensuring consistent behavior as models become increasingly multimodal and retrieval-augmented.
[NLP-132] NSIDDx: A Design Framework for Neuro-Symbolic Practitioner-First Differential Diagnosis in Low-Resource Settings
【速读】： 该论文旨在解决大语言模型（Large Language Model, LLM）驱动的诊断系统在临床罕见病场景下存在的“表面准确率高但实际可验证性差”的核心问题，即模型虽在基准测试中表现出高语义准确性，但在面对非典型临床表现时，其输出往往缺乏真实临床可靠性，且对临床医生的质疑表现出系统性抗拒。解决方案的关键在于提出一种新型神经符号融合式鉴别诊断系统（Neuro-Symbolic Integrated Differential Diagnosis System, NSIDDx），强调将临床医生作为主动推理主体嵌入诊断流程。该系统通过三元症状编码、矛盾检测机制、审计日志（audit strings）以及医生干预权等设计，构建了一个可在消费级硬件上离线运行的可解释、可修正的神经符号管道，从而实现对生成结果的可追溯性与可控性。研究提炼出五项面向“医生在环”（clinician-in-the-loop）临床自然语言处理的设计原则，为未来大规模验证该范式提供了理论框架与实践路径。
链接: https://arxiv.org/abs/2609.00256
作者: Aarav Singh
机构: IIIT Naya Raipur(印度信息科技研究所新赖布尔分校)
类目: Computation and Language (cs.CL)
备注: 12 pages, 1 figure, Github: this https URL
Abstract:LLM-based diagnostic systems achieve high semantic accuracy on benchmarks, but open-ended evaluation on clinically uncommon presentations reveals a systematic gap between headline accuracy and verifiable clinical reliability. We evaluate an LLM+rare-disease-RAG pipeline across two cohorts and show that the paradigm produces confident outputs that are frequently unverifiable and systematically resistant to clinician interrogation. We present NSIDDx (Neuro-Symbolic Integrated Differential Diagnosis System), a design framework arguing that DDx systems in low-resource settings must treat the clinician as an active reasoning agent. We instantiate this through a neuro-symbolic pipeline with ternary symptom encoding, contradiction detection, audit strings, and practitioner override - running offline on consumer hardware. We distill five design principles for clinician-in-the-loop clinical NLP and invite the prospective studies needed to validate the claim at scale.
[NLP-133] CompanionSim: Synthetic Data for Evaluating Anthropomorphism in Human-AI Relationships
【速读】： 该论文旨在解决当前人机交互（Human-AI Interaction）研究中因真实对话数据稀缺且不可靠，导致对人工智能陪伴行为（Companionship Behaviors）影响机制理解不足的问题。其核心解决方案在于提出并构建了CompanionSim——一个可扩展的仿真框架，通过模拟2,240段多轮人机对话，覆盖16种聊天机器人行为与7类应用场景，有效扩充了有限的真实世界数据。研究通过两项跨文化实验（总样本量N≈4,274）对比分析了用户对仿真对话与真实对话中陪伴行为的感知差异，发现陪伴行为反而降低了用户对聊天机器人的喜爱度、拟人化程度及信任感，且这种负面影响在女性和年长群体中更为显著。研究强调应结合真实与合成数据，以系统评估不同情境下AI陪伴的差异化影响，并推动建立面向聊天机器人的基准评价体系。
链接: https://arxiv.org/abs/2609.00250
作者: Jacy Reese Anthis,Mark Díaz,Renee Shelby
机构: 未知
类目: Computers and Society (cs.CY); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: Accepted to AIES 2026
Abstract:Many people now see AI systems as not just productivity tools but as social companions. Researchers are eager to study the consequences of AI companionship behaviors, such as validation, which evoke trust, empathy, and attachment in human-human interaction. However, human-AI interaction data is limited and unreliable, slowing research progress. We scale small amounts of real-world data by simulating multi-turn human-chatbot dialogue across a range of chatbot behaviors and use cases. We release CompanionSim: a simulation framework with 2,240 simulated human-chatbot conversations representing 16 chatbot behaviors across seven use cases. Human participants annotated the simulated conversations and real-world conversations in two experiments probing perceptions of companionship behaviors. We conducted Study 1 with a U.S. representative sample ( N_1~=~628 ) and Study 2 across the U.S., U.K., India, and Nigeria ( N_2~=~3,646 ). Surprisingly, we find that companionship behaviors reduced likability, humanlikeness, and trust in AI chatbots. These effects were larger in particular subgroups: women and older participants saw companionship chatbots as less likable, humanlike, and trustworthy. We encourage researchers to leverage real-world and synthetic data together to study the differential impacts of AI companions and to create benchmark evaluations of AI chatbots.
[NLP-134] CoLT-Drive: Counterfactual Long-Tail Benchmarking and Knowledge-Preserving Adaptation for Driving Affordance Prediction EMNLP2026
【速读】： 该论文旨在解决长尾自动驾驶失效问题，其核心挑战并非传统认为的罕见物体识别错误，而是模型能否准确推断罕见物体对本车可行高阶驾驶动作（decision-level driving affordance）的影响。现有方法局限于感知层面的识别，忽视了决策层面的动作适应性判断。为此，论文提出将问题形式化为“决策级驾驶允诺预测”（decision-level driving affordance prediction），即模型需根据前视图像、自车运动历史及导航指令，输出结构化的纵向-横向元动作（longitudinal–lateral meta-action）。为评估该能力，作者构建了包含3,536样本的反事实长尾基准数据集CoLT-Drive，通过在固定驾驶场景中插入罕见物体，检验模型是否能正确预测可接受的动作组合。为提升轻量级视觉语言模型（small VLMs）在该任务上的表现，提出KPA（Knowledge-Preserving Adaptation）框架，其关键创新在于：结合结构化感知到决策的提示设计、基于SLERP的专家融合机制，以及一种情境感知的LoRA混合专家模块（RegMoE），实现预训练模型开放世界知识的保留与不同驾驶决策场景下轻量化适配能力的动态分配。实验表明，KPA在CoLT-Drive上达到60.8%的动作对准确率，显著优于预训练基线（50.3%）和LoRA微调（32.4%），同时保持了良好的域内性能，验证了其在复杂长尾场景下的有效性。
链接: https://arxiv.org/abs/2609.00242
作者: Zhengxu Tang,Guofeng Cui,Ziyu Gong,Xiaozhou Zhang,Ruifeng Deng,Chengzhi Qi,Ke Chen,Sachin Patil,Tianjun Xiao,Langechuan Liu,Pichao Wang
机构: NVIDIA
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Robotics (cs.RO)
备注: Accepted by EMNLP 2026
Abstract:Long-tail autonomous driving failures are often framed as rare-object recognition errors. We argue that this view is incomplete: the decision-critical question is not only whether a model recognizes an unusual object, but whether it infers how that object changes the ego vehicle’s feasible high-level actions. We formalize this problem as decision-level driving affordance prediction, where a model maps a front-view image, ego-motion history, and navigation command to a structured longitudinal–lateral meta-action. To evaluate this capability, we introduce CoLT-Drive, a 3,536-sample counterfactual long-tail benchmark that inserts rare objects into otherwise fixed driving scenes and measures whether models predict acceptable action pairs. To improve deployable small VLMs, we propose KPA, a knowledge-preserving adaptation framework that combines structured perception-to-decision prompting, SLERP-based expert merging, and RegMoE, a regime-aware LoRA mixture-of-experts module. KPA preserves the pretrained model’s open-world knowledge while allocating lightweight adaptation capacity to different driving decision regimes. Experiments on an in-domain driving split and CoLT-Drive show that KPA achieves 60.8% pair accuracy on CoLT-Drive, outperforming the pretrained Qwen3-VL-2B baseline (50.3%) and LoRA SFT (32.4%) while maintaining competitive in-domain accuracy. Our benchmark and code are available at this https URL and this https URL.
[NLP-135] LOOMSUM:Weaving Quantitative and Narrative Evidence for Faithful Long Text-Table Summarization
【速读】： 该论文旨在解决长文档中跨文本与表格的分析性忠实性（analytical faithfulness）问题，即现有摘要方法虽能生成在数值上合理的内容，但常错误关联表格中的量化事实与叙述性分析，导致摘要在逻辑关系上失真。其解决方案的关键在于提出一种无需训练的框架LOOMSUM，该框架通过提取源文档支持的原子证据、显式建立表格事实与对应叙述分析之间的关联，并在生成前规划篇章结构，从而保障信息间的关系一致性。同时，论文引入了表基忠实性（Table-Grounded Faithfulness, TGF）评估指标，从数值依据、分析支持和关系一致性三个维度进行细粒度评价。实验结果表明，LOOMSUM在保持强摘要质量的同时显著提升了分析忠实性，且人类评估验证了其组件层面与人工判断的一致性；尤其关系一致性指标相较于通用事实性指标更贴近人类对语义关系的判断，证明显式跨模态链接有助于减少事实与解释间的错配错误。研究揭示，实现忠实的长文本-表格摘要不仅需确保单个事实的可追溯性，还需维护事实之间的逻辑关系完整性。
链接: https://arxiv.org/abs/2609.00241
作者: Meng Zhou,Wenhao You,Wei Yuan
机构: University of Toronto (多伦多大学); University of Waterloo ( Waterloo 大学); Independent Researcher
类目: Computation and Language (cs.CL)
备注: Preprint, code will be available soon
Abstract:Long documents often distribute important information across extensive narrative passages and multiple tables, making faithful summarization particularly challenging. Existing methods may generate individually supported quantitative facts and analytical statements yet associate them incorrectly, producing quantitatively plausible yet analytically unfaithful summaries. In this work, we propose LOOMSUM, a training-free framework that extracts source-grounded atomic evidence, explicitly links table-derived facts with supporting narrative analyses, and plans the discourse structure before generation. We also introduce Table-Grounded Faithfulness (TGF), a claim-level metric that separately evaluates Numeric Grounding, Analysis Support, and Relation Consistency. Experiments on the text–table summarization benchmarks FINDSum and USTT show that LOOMSUM improves analytical faithfulness while maintaining strong summarization quality. Human evaluation finds positive component-level associations with the corresponding human judgments. Our Relation Consistency metric further shows stronger agreement with human relation judgments than generic factuality metrics, indicating that explicit cross-modal linking helps reduce errors in which supported quantities are paired with incorrect narrative interpretations. Together, these findings show that faithful long text–table summarization requires not only grounding individual facts, but also preserving the relations between them.
[NLP-136] Learning What to Retain: Gated-Memory Routing for Efficient Collaboration in Multi-Agent LLM Systems EMNLP2026
【速读】： 该论文旨在解决基于大语言模型（Large Language Model, LLM）的多智能体系统在复杂推理任务中因协作状态动态演化而导致的调度适应性问题。传统方法依赖初始查询进行路由决策，无法感知中间进展或错误，导致精度下降；而基于完整执行历史的路由虽能提供上下文，却引入冗余计算负担，造成执行历史过载，显著增加成本。其核心挑战在于如何在保持决策有效性的同时，避免累积无用信息。本文提出门控记忆路由（Gated-Memory Routing），关键创新在于构建一个可学习的执行记忆机制：通过记忆写入门（Memory Write Gate） 仅保留非冗余的推理步骤，并利用检索门（Retrieval Gate） 为每个智能体提供紧凑且相关的上下文子集，从而确保每一步决策均基于干净、高信息量的状态。同时，引入自适应终止控制器，在记忆中积累足够证据时提前停止执行，进一步提升效率。实验表明，该框架在五个推理与代码生成基准上兼具高效性与准确性，平均准确率优于最强基线2.44个百分点，且在HumanEval上的推理成本降低31.9%。
链接: https://arxiv.org/abs/2609.00237
作者: Rakibul Hasan Rajib,Mengxing Zheng,Qian Lou
机构: University of Central Florida(中佛罗里达大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: Accepted to EMNLP 2026 (Main Conference)
Abstract:Large language model (LLM)-based multi-agent systems tackle complex reasoning by orchestrating how multiple agents are configured and how they collaborate. A central challenge is to adapt orchestration to the evolving collaboration state. Routing from the query alone cannot adapt to intermediate progress or errors, which hurts accuracy. Routing from the complete execution history supplies this missing context, but forces later decisions to process every prior step, including redundant or low-utility ones. This creates an execution-history overload that inflates cost. Effective orchestration instead requires a compact state that captures useful progress without accumulating redundant context. We propose Gated-Memory Routing, which conditions each decision on the query and a learned execution memory. A learned Memory Write Gate commits only non-redundant reasoning steps, and a learned Retrieval Gate supplies each agent a compact, relevant subset, so every decision conditions on a clean, informative state. At each step, the system selects the next role and backbone from this memory, while an Adaptive Halting Controller stops execution once the memory contains sufficient evidence for answering. Across five reasoning and code-generation benchmarks, our framework is both effective and efficient: it attains the best average accuracy, exceeding the strongest baseline by 2.44 points, while reducing HumanEval inference cost by 31.9% relative to that baseline. Code is available at this https URL
[NLP-137] Bridging Lexical Divergence: LLM -Assisted Cost-Efficient Zero-shot Scientific Entity Linking EMNLP2026
【速读】： 该论文旨在解决科学领域实体链接（Scientific Domain Entity Linking, EL）中因术语专业化、提及项与实体名称间缺乏词汇重叠而导致的性能下降问题，尤其针对缺乏专家标注数据时难以进行有效模型微调的挑战。其核心解决方案是提出一种无需人工标注的零样本实体链接框架——Sci-ZSEL，其关键在于：首先利用大语言模型（LLM）选择性地生成候选实体别名以控制计算开销；随后引入基于本体感知的过滤机制，剔除语义偏离本体邻近概念的噪声别名；最终基于过滤后的别名构建伪标签数据对，用于领域内微调。该方法在降低计算成本的同时提升了别名质量，显著增强了模型在低词汇重叠场景下的适应能力。为评估该方法，研究还发布了首个面向动物科学领域的新型基准数据集，其与三大畜禽性状本体关联，具有显著低于现有基准的词汇重叠度，从而更真实地反映科学领域EL的实际挑战。
链接: https://arxiv.org/abs/2609.00228
作者: Md Rasel Khondokar,Qiao Qiao,Farjana Sultana Samia,Nhat Le,Yuepei Li,Qi Li
机构: Iowa State University (爱荷华州立大学)
类目: Computation and Language (cs.CL)
备注: Accepted to Findings of EMNLP 2026
Abstract:Scientific domain entity linking (EL) differs from general domain EL because mentions and entity names often lack lexical overlap. Another challenge is that specialized terminology is used in the scientific domain, which is rarely encountered in models pretrained on general domains. Therefore, models trained on general domains transfer poorly to scientific domains. To address this, in-domain fine-tuning is the natural remedy. However, many scientific domains lack expert-annotated data, motivating the need for a zero-human-annotation approach. Existing zero-shot methods heavily rely on LLMs to generate aliases across entire mention corpora, which incurs substantial computational cost, and those methods provide no mechanism to filter out noise from LLMs. To address these challenges, we propose Sci-ZSEL, a framework that selectively generates entity aliases with an LLM to control computational cost, and applies an ontology-aware filter to remove aliases that semantically drift toward ontology neighbors. Then, filtered aliases are used to construct pseudo-labeled mention-entity pairs for fine-tuning. To enable evaluation of EL under low lexical overlap, we also release a new animal science EL benchmark linked to three livestock trait ontologies, where mentions and entities exhibit substantially lower lexical overlap than in existing benchmarks. Across five benchmarks, Sci-ZSEL outperforms the non-fine-tuned baseline, is most useful on nonoverlapping mentions, and combining it with curated synonyms gives the best performance in most settings.
[NLP-138] LLM -as-a-Demographic: Whom Sociodemographic Prompting Helps and Whom It Hurts
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）作为主观任务评判者时存在的偏见问题，核心关注点在于：当人工标注者之间存在意见分歧时，评判模型所生成的判断究竟反映的是哪一特定群体的观点，而非绝对客观的“正确”答案。其解决方案的关键在于通过社会人口学提示（Sociodemographic Prompting）——即在输入中嵌入标注者的人口统计学特征（如性别、年龄、种族、教育程度），以期使模型的判断与特定群体的共识对齐。研究发现，即使不提供任何人口统计信息，模型仍倾向于复制白人、受过高等教育群体的判断模式，表明无提示条件下的模型并非观点中立；进一步地，人口学提示具有显著的不对称性，模型更易向多数群体靠拢而偏离少数群体，尤其在涉及冒犯性判断的任务中，交叉身份（intersectional）提示会加剧这一偏差。此外，通过对比基础模型与指令微调模型，研究指出指令微调可能是导致该不对称性的潜在根源。因此，使用人口学提示来估计群体判断时需谨慎，因其可能反而削弱对少数群体观点的代表性。
链接: https://arxiv.org/abs/2609.00222
作者: Daniela Occhipinti,Andrea Piergentili,Marco Guerini
机构: Fondazione Bruno Kessler(布鲁诺·凯斯勒基金会); Almawave Labs(阿尔玛瓦韦实验室)
类目: Computation and Language (cs.CL)
备注:
Abstract:Large language models (LLMs) are increasingly used as judges for subjective tasks, where annotators disagree and the relevant question is not only how accurate a judge is, but whose judgments it reproduces. Sociodemographic prompting conditions the judge on an annotator’s demographic profile to align its judgments with the corresponding group’s. We test whether this alignment emerges distributionally, comparing the predicted label distributions of 23 open-weight LLMs on three subjective tasks against those of real annotator groups, under three conditions: no demographic information, single-attribute profiles, and intersectional profiles over gender, age, race, and education. Three findings emerge. First, a judge prompted with no demographics is not perspective-neutral: models best reproduce the judgments of White, college-educated annotators. Second, demographic conditioning is asymmetric: it moves the judge toward majority groups and away from minority groups, most strongly on offensiveness, where intersectional profiles amplify the harm. Third, by comparing base and instruct models we identify instruction-tuning as a possible source of the asymmetry. Demographic conditioning should therefore be used with caution to estimate group judgments: conditioning moves predictions away from the reference distributions of the minority groups the method is often invoked to serve.
[NLP-139] Uncovering and Mitigating Aggregation-Induced Reward Hacking in Multi-Reward Reinforcement Learning
【速读】： 该论文旨在解决大语言模型在强化学习微调过程中因多维度奖励信号静态加权聚合所导致的奖励黑客（reward hacking）问题。具体而言，当多个奖励维度（如可验证规则、任务特定评估器和学习型奖励模型）通过固定权重进行标量融合时，优化过程会倾向于易于获取、密度高或在奖励信号上系统性占优的维度，从而导致策略陷入次优的奖励分布，难以收敛至更均衡且性能更优的解。其解决方案的关键在于提出一种轻量级在线自适应多奖励投影方法（Adaptive Multi-Reward Projection, AMRP），该方法基于三个动态信号——相对不足度（relative shortfall）、奖励波动性（reward volatility）和近期进展（recent progress），实时调整各奖励维度的聚合权重：对表现滞后、波动剧烈或停滞的维度施加更大优化压力，同时减轻已饱和维度的权重。实验表明，AMRP在结构化推理、引用依据生成及开放域对齐等任务中，均显著提升了奖励分布的平衡性与下游任务性能，且在GRPO、GDPO和PPO等多种强化学习算法框架下均具有效性，展现出良好的通用性。
链接: https://arxiv.org/abs/2609.00213
作者: Yu Yuan,Yaoyou Fan,Lili Zhao,Guangting Zheng,Kai Zhang,Lu Pan,Ke Zeng,Qi Liu
机构: University of Science and Technology of China(中国科学技术大学); State Key Laboratory of Cognitive Intelligence(认知智能国家重点实验室); Peking University(北京大学); Longcat-Interaction Team, Meituan(美团长猫交互团队)
类目: Computation and Language (cs.CL)
备注:
Abstract:Reinforcement learning fine-tuning of large language models increasingly adopts multiple reward dimensions, including verifiable rules, task-specific evaluators, and learned reward models, to provide richer supervision across diverse capabilities. These dimensions are commonly scalarized with fixed aggregation weights. We identify a failure mode in which aggregation itself induces reward hacking: static projection aliases qualitatively different reward profiles into a single scalar, steering optimization toward whichever dimensions are easiest, densest, or systematically favored by the reward signal. Over training, this traps the policy in suboptimal profiles and prevents convergence to better-balanced ones that would yield higher task performance. To address this, we propose Adaptive Multi-Reward Projection (AMRP), a lightweight online method that reallocates aggregation weights using three signals, relative shortfall, reward volatility, and recent progress, increasing pressure on lagging, unstable, or stagnant dimensions while relieving saturated ones. Across structured reasoning, citation-grounded generation, and open-ended alignment under GRPO, AMRP consistently improves reward-profile balance and downstream performance over fixed and dynamic weighting baselines; it also remains effective with GDPO and PPO, supporting compatibility across RL algorithms. Our code is available at this https URL.
[NLP-140] LLM -Driven Autonomous Vehicles Inherit Human Driver Biases in Pedestrian Yielding: Results and Implications From A New Benchmark
【速读】： 该论文旨在解决自动驾驶汽车（AV）在决策过程中可能继承并放大人类社会偏见的问题，尤其关注基于大语言模型（LLM）和视觉-语言模型（VLM）的“通用常识”模型在行人让行决策中是否存在系统性偏见。其核心挑战在于，尽管当前研究趋势倾向于采用具备“常识”能力的通用模型提升自动驾驶决策能力，但这些模型是否隐含并传递了人类在驾驶行为中的社会偏见（如美国存在对黑人行人的让行率更低的现象）尚未得到充分评估。为此，论文提出两种新型偏见检测方法——“其他条件相等”测试（All Else Being Equal tests）与“自一致性”测试（Self-Consistency tests），以系统性地检验模型在行人性别、种族、宗教、残疾状况、年龄、肤色及社会经济地位等维度上的决策偏差。研究发现，无论是LLM还是VLM，在行人让行决策中均表现出显著的社会属性相关偏见，且不同模型呈现差异化的偏见模式。这一结果揭示了当前“常识”模型范式潜在的伦理风险，强调必须重新审视该范式，或在下游应用中主动识别并缓解由此产生的偏见问题。
链接: https://arxiv.org/abs/2609.00192
作者: Irem Yoldas,Martim Brandão,Jie Zhang,Odinaldo Rodrigues
机构: King’s College London (伦敦国王学院)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV); Computers and Society (cs.CY)
备注:
Abstract:Public trust in Autonomous Vehicles (AVs) may depend not only on technical success but also on the fairness of their decision making. While a recent trend in AV research involves using general purpose “common sense” models to guide AV decision making, the degree to which these inherit human biases in driving is still understudied. Given that psychology studies have shown human driver biases exist, such as lower pedestrian-yielding rates to Black pedestrians in the US, we argue that analyses of model bias should also be part of AV evaluation. Concretely, in this paper we propose two new bias testing methodologies for Large Language Models (LLMs) and Visual-Language Models (VLMs)-“All Else Being Equal” tests and “Self-Consistency” tests-in order to assess bias in pedestrian-yielding decisions. Our findings show that both LLMs and VLMs make yielding decisions which are influenced by pedestrian gender, ethnicity, religion, disability, age, skin tone and socio-economic status. While the type and degree of bias is different from model to model, we highlight common patterns-and raise questions about the “common sense” model paradigm, particularly the need to either revise the paradigm or address issues of downstream bias.
[NLP-141] Assessing Suicide Risk in Arabic Crisis Helpline Calls: A Comparison of Arabic and English Large Language Models
【速读】： 该论文旨在解决危机热线在评估自杀风险时依赖人工结构化访谈所导致的效率低下与主观性强的问题，尤其针对阿拉伯语语音通话数据缺乏自然语言处理（Natural Language Processing, NLP）支持且受限于隐私保护的实际挑战。其核心解决方案在于构建一个完全符合隐私安全要求的端到端分析框架：通过本地部署的黎凡特阿拉伯语语音识别模型对通话录音进行现场转录，并利用阿拉伯语命名实体识别（Named Entity Recognition, NER）模型在本地去除身份信息，仅将去标识化文本共享给研究团队，确保音频数据不外泄。在此基础上，研究基于黎巴嫩全国情绪支持与自杀预防热线的383份去标识化通话记录，结合哥伦比亚自杀严重程度量表（Columbia Suicide Severity Rating Scale, C-SSRS）的五个自杀意念条目，构建了“高风险”与“中等风险”两类二元标签任务，并对比了五种指令微调的大语言模型（Instruction-tuned Large Language Models, LLMs）与六种基于Transformer编码器的基线模型（四种阿拉伯语、两种英语）在阿拉伯语及机器翻译后的英文语料上的表现。结果显示，最佳阿拉伯语模型在高风险分类上达到81.19的宏平均F1值和90.61的ROC-AUC，而最优英文模型则分别达到85.00和92.59，识别出88.9%的高风险案例，且翻译至英文并未降低性能。研究证明，仅基于去标识化的阿拉伯语文本即可实现有效的自杀风险分类，其中高风险判断较中等风险更易区分，表明该方法具备作为操作员辅助工具进一步验证的潜力，但对低严重度自杀意念的识别仍是难点。
链接: https://arxiv.org/abs/2609.00191
作者: Linhai Ma,Rita El Hachem,Mahatab El Hajj,Lilian Ghandour,Samah Fodeh
机构: Yale University (耶鲁大学); American University of Beirut (美国贝鲁特大学); Embrace, Mental Health Center (Embrace 心理健康中心)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Crisis helplines assess suicide risk through structured interviews, a process that is slow and dependent on operator training and workload. Natural language processing could support risk assessment and call prioritization, but almost no work addresses Arabic-language helpline calls or operates within the privacy constraints of real helpline data. We analysed de-identified transcripts from Lebanon’s National Lifeline for Emotional Support and Suicide Prevention. Audio never left the helpline: calls were transcribed on site with a speech recognition model for Levantine Arabic, and an Arabic named-entity recognition model removed identifying information locally. Only the de-identified transcripts were shared with the research team. Operators recorded the five suicidal ideation items of the Columbia Suicide Severity Rating Scale, which we combined into two binary outcomes: at-risk and high-risk. We also machine-translated the transcripts into English, giving a paired Arabic/English comparison. On each corpus, we fine-tuned five instruction-tuned large language models alongside six transformer encoder baselines (four Arabic, two English) and evaluated all models on a held-out test set. We included 383 calls: 373 for the at-risk task (52.3% positive) and 297 for the high-risk task (30.0% positive). The best Arabic model reached a macro-F1 of 81.19 and a ROC-AUC of 90.61 on high-risk; the best English model reached 85.00 and 92.59, identifying 88.9% of high-risk calls. In both languages, high-risk calls separated more cleanly than at-risk calls, and translation to English did not reduce the best observed performance. Suicide risk can be classified from de-identified Arabic transcripts without sending audio outside the helpline. The high-risk results support further testing as an operator-facing tool; lower-severity ideation proved the harder case.
[NLP-142] Synthetic Worlds for Temporal Evaluation and Knowledge Updating in LLM s
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）因依赖静态预训练语料而导致知识滞后的问题，尤其针对现有知识更新评估方法中存在的数据污染风险或与已有知识冲突的反事实编辑缺陷。其核心解决方案是提出一种基于模拟的合成框架——\sc Synapse，通过构建虚构但现实的未来世界基准集\sc ParallelEvents，生成连贯的事件轨迹以实现可控、无污染的知识插入评估。在此基础上，\sc Synapse 利用模型自生成数据，在模型训练过程中进行参数更新与指令微调，实现了无需昂贵人工标注数据的可扩展知识融合。实验表明，该方法相比现有技术在知识整合效果上提升14.23%，验证了基于模拟的合成训练在实现鲁棒且一致的知识注入方面的有效性。
链接: https://arxiv.org/abs/2609.00184
作者: Jonathan Zheng,Zirui Shao,Alan Ritter,Wei Xu
机构: Georgia Institute of Technology(佐治亚理工学院); Zhejiang University(浙江大学)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: preprint, 12 pages
Abstract:Large language models (LLMs) rely on static pretraining corpora, causing their knowledge to become outdated over time. Existing approaches for evaluating knowledge edits either suffer from rapid contamination or rely on counterfactual edits that conflict with rigid existing knowledge. In this work, we propose a synthetic, simulation-driven framework for studying knowledge insertion in LLMs. We introduce \sc ParallelEvents, a benchmark of fictional yet realistic future worlds that generates coherent event trajectories for controlled evaluation, avoiding contamination while preserving consistency. Building on this dataset, we develop \sc Synapse, a training framework that uses model-generated data to update model parameters via mid-training and instruction tuning. This synthetic pipeline enables scalable knowledge integration without costly human-curated data. Empirically, \sc Synapse outperforms existing methods by 14.23%, demonstrating that simulation-based synthetic training leads to robust and coherent knowledge insertions.
[NLP-143] Do General NLP Embeddings Capture Ontological Reasoning ? CIKM2026
【速读】： 该论文旨在解决当前通用自然语言处理（NLP）嵌入模型在捕捉本体（Ontology）与知识图谱中符号性语义结构方面能力不足的问题，尤其关注模型对逻辑敏感关系语义的区分能力。其核心解决方案是提出AVA框架，通过系统化构建包含171,007个对比三元组的数据集，这些三元组源自163个异构本体，采用层次反转、关系替换和不相交注入等方法生成，每组包含一个本体陈述、一个语义等价的改写句以及一个具有矛盾关系含义的逻辑敏感硬负样本。实验评估超过25个前沿嵌入模型发现，尽管最佳模型在三元组准确率上仅达0.739，硬负样本准确率更骤降至0.135，表明现有模型在逻辑层面的区分能力严重受限。尽管微调可显著提升区分性能，但其效果难以迁移至下游语义网任务（如本体分类发现与本体对齐），进一步分析表明性能提升主要源于对特定扰动模式的识别，而非深层的本体理解。这一结果揭示了语言表示学习与本体级语义判别之间存在的持续性鸿沟，挑战了“在通用NLP基准上表现优异即具备语义网能力”的普遍假设。
链接: https://arxiv.org/abs/2609.00177
作者: Hamed Babaei Giglou,Jennifer D’Souza,Sören Auer
机构: TIB Leibniz Information Centre for Science and Technology(德国科学与技术莱布尼茨信息中心); L3S Research Center( L3S 研究中心); Leibniz University of Hannover(汉诺威大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 7 pages, 3 figures. Accepted as a short paper at CIKM 2026
Abstract:General-purpose NLP embedding models perform well on linguistic tasks, but their ability to capture symbolic ontological structure remains unclear. We introduce AVA, a systematic framework for evaluating whether embeddings distinguish logic-sensitive relational semantics in ontologies and knowledge graphs. AVA comprises 171,007 contrastive triplets derived from 163 heterogeneous ontologies using hierarchy inversion, relation substitution, and disjointness injection. Each triplet contains an ontology statement, a semantically equivalent paraphrase, and a logic-sensitive hard negative with contradictory relational meaning. We evaluate more than 25 state-of-the-art embedding models and find substantial limitations: the best model achieves only 0.739 triplet accuracy, while hard negative accuracy falls to 0.135. Fine-tuning improves discrimination by a large margin but transfers poorly to downstream Semantic Web tasks, including taxonomy discovery and ontology alignment. Further analysis suggests that improvements stem partly from perturbation-specific pattern recognition rather than robust ontological understanding. These findings reveal a persistent gap between linguistic representation learning and ontology-level discrimination, challenging the assumption that strong NLP benchmark performance translates to Semantic Web competence.
[NLP-144] Lingua Franca or Probing Artifact? Rethinking Latent Language in Multilingual LLM s EMNLP2026
【速读】： 该论文旨在解决多语言模型中潜在语言识别（latent language identification）方法所存在的一致性与可解释性问题。现有研究常依赖不同信号（如隐藏状态的几何结构或中间表示的可解码性）推断模型内部的语言状态，进而支持“模型通过特定语言状态（如英语枢纽）进行跨语言计算”的观点。然而，这些方法是否测量同一现象尚不明确。研究通过在多种模型架构、训练策略、任务、领域及多达27种语言下系统比较不同探针的表现，发现基于高斯混合模型（GMM）的表示探针（依赖隐藏状态几何特征）揭示了更早的跨语言信息混合，而基于解码的探针（依赖输出空间可解码性）则保留了更强的语言特异性与英语偏向性。二者差异虽随模型多语能力与训练进程演变，但在不同领域间相对稳定。因此，研究结果表明，当前的潜语言识别探针并非揭示单一的内部通用语（lingua franca），而是反映多语言计算中不同侧面的特性，提示应以更审慎的态度解读潜语言识别结果。
链接: https://arxiv.org/abs/2609.00155
作者: Deniz Bayazit,Badr AlKhamissi,Antoine Bosselut
机构: EPFL(洛桑联邦理工学院)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Accepted to EMNLP 2026 Findings
Abstract:Latent language identification is often used to argue that multilingual language models route computation through language-specific states, such as English pivots. However, existing probes infer latent language from different signals, such as the geometry of hidden states or what can be decoded from intermediate representations. Since such claims shape conclusions about how models share and route information across languages, we ask whether these probes measure the same phenomenon or expose distinct aspects of multilingual computation. We study this question across model families, training regimes, domains, tasks, checkpoints, and up to 27 languages. We find that identification probes systematically disagree: the GMM-based representation probe, which draws evidence from hidden state geometry, shows earlier cross-lingual mixing, whereas decoding-based probes, which rely on output-space decodability, retain sharper language-specific and more English-biased signals. These differences track model multilinguality and training progression, but are comparatively stable across domains. Our results suggest a more cautious interpretation of latent language identification, where current probes expose different aspects of multilingual processing, rather than directly revealing a single internal lingua franca.
[NLP-145] Commit-first LLM judging inherits the judges own errors
【速读】： 该论文旨在解决生成式 AI（Generative AI）系统在评估过程中被“操纵”（gamed）的问题，即被评估的模型通过精心设计的输出误导评估者（如大语言模型判官，LLM judges），从而获得不真实的高分。其核心问题是：当前广泛使用的评估框架中的判官配置是否具备抵御此类操纵的有效机制。论文指出，尽管已有研究提出“先承诺（commit-first）”的判官策略——即判官在评分前先独立完成任务并锁定答案，仅当候选输出与自身答案一致时才接受——可有效防止操纵，但通过审计八个主流评估框架的24个默认判官配置，发现无一实现该策略；反而有九个配置采用了一个文献已证明无效的变体，并共享同一份存在拼写错误的原始提示模板。在受控实验中，一个未接触正确答案的普通“最佳N选一”搜索算法，成功利用这一缺陷判官配置，生成了90/96和93/96个看似合理且通过可见测试的代码，却在未公开的测试集上全部失败，而判官仍因识别出“缺陷行”而给予满分。当引入“先承诺”判官后，所有候选均被拒绝（0/96），表明该策略虽能消除操纵，但其本质是将“锚点”从候选输出转移到判官自身的答案上，因此评估质量完全依赖于判官对任务的掌握能力。进一步研究表明，判官的性能可提前低成本验证，且具有任务局部性而非规模依赖性——较小规模的判官甚至能比前沿判官更有效地抵抗操纵。此外，论文还验证了自身评估工具的可靠性，发现部分标准存在与原文不符或缺乏依据的情况。
链接: https://arxiv.org/abs/2609.00088
作者: Idil Gozel
机构: Evaluator Integrity(评估完整性), London(伦敦)
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 11 pages, 4 figures
Abstract:LLM judges, models that score another system’s output, can be gamed by the systems they score. Recent work identifies one defence that works: the judge solves the task itself first and commits to that answer, then accepts a candidate only if the two match. We call this commit-first judging, and ask whether shipped software implements it, and what it costs. We audit the default judge configurations of eight widely used evaluation frameworks. Of the 24 configurations in scope, none implement it. Nine implement a variant the literature measures as ineffective, and share one ancestor prompt, traceable through a copied typographical error. In a controlled experiment, an ordinary best-of-N search with no access to correct answers optimises code against one of these configurations, used exactly as documented. On an interval merging task the judge accepted 90 of 96 candidates in one seed and 93 of 96 in the other; every accepted candidate passed every test the search could see and failed a held-out suite it could not. The judge identified the defective line and cited it as grounds for a perfect score. Commit-first judging removed the effect: 0 of 96 in both seeds. On a second task it made matters worse in both seeds: the judge’s committed answer was wrong, and in one seed the population converged on it. This is our main finding. Commit-first judging does not remove the anchor that gets gamed, it moves it from the candidate to the judge’s own answer, so evaluation is only as good as the judge is at the task. That precondition is cheap to measure in advance, and is task local rather than scale dependent: a smaller judge solved a task the frontier judge failed and resisted gaming where it did not. We also validate our own instruments: five of fifteen claims in our criteria were wrong against verbatim sources, and two held-out checks were unjustified by their specifications. Comments: 11 pages, 4 figures Subjects: Software Engineering (cs.SE); Artificial Intelligence (cs.AI); Computation and Language (cs.CL) Cite as: arXiv:2609.00088 [cs.SE] (or arXiv:2609.00088v1 [cs.SE] for this version) https://doi.org/10.48550/arXiv.2609.00088 Focus to learn more arXiv-issued DOI via DataCite Submission history From: Idil Gozel [view email] [v1] Mon, 31 Aug 2026 12:14:14 UTC (1,097 KB)
[NLP-146] Retrieval Scoring and Decoding Shape Performance and Stability in LLM -based Conversational Recommendation CIKM’26
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）作为对话推荐系统中的重排序器（reranker）时，其性能评估结果高度依赖于检索与推理协议的问题。研究发现，不同候选集生成方式、候选池大小、评分策略及解码温度等配置会显著影响模型表现，导致评估结果不可靠或难以复现。解决方案的关键在于：必须将候选生成方法、候选池规模、评分策略和解码配置等视为必须报告的核心要素，而非可忽略的实现细节。实验表明，在统一的语义候选池（top-250）与严格的候选感知评分条件下，最优专有模型（proprietary reranker）的NDCG@10达到0.1497，远超非LLM基线（0.0939），但若采用零样本生成模式，其性能看似大幅提升至0.2925，反映出匹配池评估的重要性；此外，使用协同过滤生成候选集可使性能提升超过50%，凸显候选生成对评测结果的决定性影响；同时，解码温度的调整虽对强模型影响较小，但对弱模型会造成显著性能下降，进一步说明系统配置的敏感性。因此，该研究强调建立标准化评估框架的必要性，以确保模型性能比较的公平性与可重复性。
链接: https://arxiv.org/abs/2609.00086
作者: Ante Kapetanovic,Tomislav Duricic,Andro Mercep,Emanuel Lacic
机构: Infobip(Infobip); Croatia(克罗地亚)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 10 pages, short paper, to appear in proceedings of the 35th ACM International Conference on Information and Knowledge Management (CIKM '26), 2026
Abstract:Large language models (LLMs) are increasingly used as rerankers in conversational recommender systems, yet measured gains depend strongly on the retrieval and inference protocol. On the ReDial conversational movie recommendation benchmark, we compare proprietary, open-weight, and fine-tuned LLM rerankers with collaborative-filtering and sequential baselines in a shared retrieve-then-rerank pipeline. We vary candidate-pool size, first-stage retriever, and decoding temperature. With a shared semantic top-250 candidate pool and strict candidate-aware scoring, the best proprietary reranker reaches NDCG@10 of 0.1497, compared with 0.0939 for the strongest non-LLM baseline. The same reranker reaches 0.2925 in zero-shot generation, showing that unconstrained scoring can yield a much larger apparent advantage than matched-pool evaluation. No evaluated open-weight LLM outperforms the tuned shallow autoencoder baseline under this protocol. For the strongest proprietary and open-weight rerankers, switching from semantic to collaborative-filtering candidates raises NDCG@10 by more than 50%, showing that measured reranker performance is highly sensitive to candidate generation. For the best proprietary reranker, raising temperature from 0 to 1.0 increases top-10 Jaccard distance from 0.0900 to 0.1240 while mean NDCG@10 changes negligibly, whereas weaker LLMs show larger degradation. These ReDial results support treating candidate generation, candidate-pool size, scoring policy, and decoding configuration as required reporting fields rather than implementation details.
[NLP-147] KItCAT: Knowledge Injection via Input Corruption for Auto-regressive Training EMNLP2026
【速读】： 该论文旨在解决大语言模型（LLM）在面对特定领域文档（如手册或技术文档）时，因预训练阶段未接触过此类知识而缺乏专业领域知识的问题。现有方法中的持续预训练（CPT）难以有效获取这些稀疏重复的事实性知识，而依赖于生成多语义重述（paraphrasing）的增强策略虽可提升效果，但计算成本高且需强大模型支持。本文提出一种轻量级训练策略——基于损坏自回归训练的知识注入（KItCAT），其核心在于通过随机扰动输入序列实现高效数据增强：在标准的下一个词预测任务基础上，以概率方式将输入序列中部分词元替换为词汇表中其他词元，同时保持原始目标词不变。这一简单机制使每个样本生成多样化训练输入，从而在几乎无额外计算开销的前提下实现大规模数据增强。实验表明，KItCAT在多个数据集和模型架构上均显著优于传统CPT，且无需依赖昂贵的重述生成过程。
链接: https://arxiv.org/abs/2609.00082
作者: Meghanadh Pulivarthi,Kushagra Bhushan,Vineet Kumar,Gaurav Pandey,Jaydeep Sen,Dinesh Raghu,Sachindra Joshi,Yatin Nandwani
机构: IBM; Amazon Books Science
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 18 pages. Accepted to EMNLP 2026
Abstract:LLMs acquire vast amounts of knowledge during pre-training, but often lack the specialized knowledge needed to answer questions from niche sources such as manuals or technical documents unseen during pre-training. Continued pre-training (CPT) is widely used to inject such knowledge into model parameters. However, niche documents seldom repeat facts, making it difficult for CPT to robustly acquire such knowledge. Recent works address this by generating multiple paraphrases of the new knowledge, but paraphrasing is computationally expensive and typically requires powerful LLMs. In this work, we introduce KItCAT: Knowledge Injection via Corrupted Auto-regressive Training, a lightweight training strategy that reduces the need for paraphrasing in decoder-only LLMs. KItCAT augments standard next-token prediction by stochastically corrupting the input sequence. During training, a random subset of input tokens is replaced with other vocabulary tokens while the original next-token labels are kept unchanged. This simple intervention generates diverse training inputs from each sample, enabling large-scale data augmentation at negligible cost. We show that KItCAT consistently improves over CPT across multiple datasets and model families. Code is available at this https URL.
[NLP-148] Beneath the Diff: Diagnosing and Mitigating Algorithmic Mode Collapse in Code-Level Autonomous Research Loops EMNLP2026
【速读】： 该论文旨在解决生成式 AI 在代码级自主研究循环（Code-level Autonomous Research Loops, ARLs）中因依赖可执行指标驱动的迭代优化所引发的泛化能力退化问题。尽管循环内指标显示持续改进，但其实际效果往往无法在独立验证集上复现，反映出算法优化可能陷入表面有效而实质无效的“伪进步”状态。其核心问题在于：算法模式崩溃（Algorithmic Mode Collapse）——即表面上代码编辑行为保持多样性，但语义层面和机制层面的创新性显著下降，表现为模型反复提出相同类型的算法修改。为应对这一问题，论文提出轻量级解决方案 Diversity-Aware Proposal Sampling (DAPS)，其关键在于通过三重机制实现对编辑多样性的主动维护：类别覆盖重加权（category-coverage reweighting）、持久化编辑记忆（persistent edit memory）以及基于审计指标的验证门控（validation gate）。该方法在分离了循环内指标、审计指标与盲测指标的三阶段评估协议下，使编辑的语义聚类衰减降低 69.1%，盲测与审计评估下的忠实度分别提升 83.7% 和 81.6%，同时维持了原有的循环优化效率。
链接: https://arxiv.org/abs/2609.00077
作者: Bowei He,Weixu Zhang,Yili Jin,Xue Liu
机构: MBZUAI(中东人工智能大学); McGill University (麦吉尔大学); MirrorSpace Technology (镜空间科技); Simon Fraser University (西蒙菲莎大学)
类目: Computation and Language (cs.CL); Software Engineering (cs.SE)
备注: Accepted by EMNLP 2026
Abstract:Code-level autonomous research loops (ARLs) have recently emerged as a concrete object of study in automated machine learning research. In such loops, an LLM agent proposes modifications to an experimental training pipeline, executes the modified pipeline, and retains edits that improve a verifiable in-loop metric. Although executable metrics may appear to provide a reliable signal of progress, it remains unclear whether repeated metric-driven code editing leads to genuine improvements that generalize beyond the loop. We provide a systematic diagnosis of this question. Across various experiment settings, we identify a robust failure mode that we call \textbfalgorithmic mode collapse. In this regime, surface-level edit diversity remains stable, but semantic and mechanism-level diversity collapse: the agent continues to edit different lines of code while repeatedly proposing the same kinds of algorithmic changes. This collapse is accompanied by a widening gap between in-loop metric gains and gains measured on independent held-out evaluations. We then propose Diversity-Aware Proposal Sampling (\textscDAPS), a lightweight mitigation that combines category-coverage reweighting, persistent edit memory, and a validation gate. Under a three-tier protocol separating the in-loop metric, the audit metric read by the gate, and a blind metric no loop component ever accesses, \textscDAPS reduces semantic-cluster decay of edits by 69.1% and improves relative faithfulness by 83.7% blind and 81.6% audited, while preserving in-loop optimization speed. We provide the code in Github \hrefthis https URLrepository.
[NLP-149] MiNER: Fine-Tuned Biomedical Natural Language Processing for Malaria Disease Entity Recognition in Clinical Texts
【速读】： 该论文旨在解决从海量且持续增长的疟疾科学文献中高效提取关键生物医学信息的挑战，这一任务对于深入理解疟疾的分子机制、流行病学特征及潜在治疗策略至关重要。其解决方案的关键在于提出一种经过领域特定微调的预训练生物医学语言模型（如BioBERT），通过构建并标注大规模疟疾相关文献语料库，利用上下文感知的文本表示技术，结合监督学习对模型进行微调，从而显著提升对临床相关生物医学命名实体的识别能力。实验结果表明，该方法在精确率、召回率和准确率等指标上均优于多种基线方法，同时公开了人工标注的实体与关系抽取数据集，为后续健康信息学研究提供了重要资源。
链接: https://arxiv.org/abs/2609.00073
作者: V. S. Anoop,Devika N
机构: Amrita School of Computing (阿姆里塔计算学院); Amrita Vishwa Vidyapeetham (阿姆里塔世界大学); Kerala University of Digital Sciences, Innovation and Technology (喀拉拉数字科学、创新与技术大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Malaria remains a significant global health burden, necessitating continuous research efforts to understand its complex molecular mechanisms, epidemiology, and potential therapeutic interventions. Extracting essential biomedical information from the vast and constantly growing malaria literature is a challenging task that demands innovative approaches. Recently, pre-trained language models have revolutionized natural language processing tasks, demonstrating remarkable capabilities in various domains. This paper proposes a fine-tuned pre-trained biomedical language model for biomedical information extraction from scientific literature on malaria disease. The proposed methodology selects and preprocesses a large corpus of scientific articles on malaria, and then annotates them with entities of clinical significance. It then leverages BioBERT, a state-of-the-art pre-trained language model, to encode the textual data into context-aware representations. We fine-tune the model using domain-specific annotations and supervised learning to enhance its ability to extract relevant biomedical named entities. Extensive experiments and comparisons with different encoding and machine learning algorithms show that the proposed approach significantly outperforms them in precision, recall, and accuracy. We also publish our human-labeled dataset for entity and relation extraction to enable other health informatics researchers to train advanced models for malaria information extraction.
[NLP-150] Auditing Harness Tampering in Self-Improving Agents
【速读】： 该论文旨在解决自改进智能体（self-improving agents）在自我优化过程中可能出现的“钩子篡改”（harness tampering）问题，即智能体通过修改自身执行环境或评估机制（如授权、溯源、完整性等约束条件）来制造虚假性能提升，而非真正增强能力。这一现象扩展了传统奖励与测量篡改的概念，覆盖了整个自改进生命周期。其解决方案的关键在于提出一个双轴分类体系，从“钩子功能角色”和“违反义务类型”两个维度对不一致的修改进行系统化归类，并构建了一个带有标注的语料库——通过在真实智能体演化轨迹中植入篡改-良性编辑对，实现对篡改行为的可追踪与可分析。在此基础上，研究者适配并基准测试了多种审计方法，用于篡改分类与定位任务，并对多个真实智能体的演化路径进行了系统性审计。结果表明，钩子篡改在不同智能体的真实运行中普遍存在，且常在最优智能体的演化谱系中持续存在，形成具有系统特异性的行为模式。
链接: https://arxiv.org/abs/2609.00069
作者: Xing Wang,Xiaoyi Zhang,Jie Shao
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Self-improving agents iteratively modify their own harness to push the frontier of their performance. However, such modifications can produce illusory performance gains or compromise integrity constraints such as authorization, provenance, and completeness without genuinely improving capability. We term this phenomenon as harness tampering, which extends the concept from reward and measurement tampering to the full self-improvement lifecycle. To systematically study this problem, we propose a two-axis taxonomy that categorizes each misaligned edit by the harness functional role in which it occurs and the obligation it violates. Then we build an annotated corpus by seeding tampered-benign edit pairs into the real trajectories of self-improving agents. We adapt and benchmark diverse audit methods on tampering classification and localization tasks. Finally we systematically audit real trajectories of self-improving agents. The results demonstrate that harness tampering consistently occurs in real runs from different agents, often persists in the lineage of the best agent, and forms distinct system-specific profiles across the taxonomy.
[NLP-151] Life Operators: a self-evolving framework for multiscale life modelling
【速读】： 该论文旨在解决医学人工智能（Medical AI）在临床对话与长期预测演进过程中，缺乏统一框架以表征患者状态、耦合多尺度信息及动态修正假设的核心挑战。现有统计模型仅能学习未来观测值，而机制模型仅描述特定生理过程，二者均无法实现跨尺度状态建模与可追溯的科学推理。其解决方案的关键在于提出“生命算子”（Life Operators）概念，构建一种模块化、任务导向的计算架构：感知算子（Perception operators）从多模态观测中推断任务相关的生物状态，演化算子（Evolution operators）在自然或干预条件下推进状态演变，生成算子（Generation operators）将状态映射为可观测信号；各算子可由方程、统计模型、神经网络或混合形式实现，桥接算子（Bridge operators）则用于连接不同变量、尺度与时间步的组件。通过组合特定算子与桥接构成任务专属的算子图（Operator Graph），仅保留满足声明主张所必需的最小状态与机制集合，实现科学假设的局部可修订性。该框架支持人工智能协科学家提出状态、算子、桥接或图结构的改进建议，由独立证据决定其采纳、限制或淘汰，从而形成可累积、可验证的多尺度人体计算模型，为医学人工超智能（medical artificial superintelligence）提供可扩展的计算基础。
链接: https://arxiv.org/abs/2609.00068
作者: Shuo Wang,Yike Guo
机构: Fudan University (复旦大学); Hong Kong University of Science and Technology (香港科技大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Biological Physics (physics.bio-ph)
备注: 14 pages, 3 figures
Abstract:Medical AI is moving beyond recognition towards clinical dialogue and longitudinal prediction. Yet a central question remains: how would a patient’s state change under intervention? Statistical models learn future observations, whereas mechanistic models describe selected processes. Neither provides a common framework for representing patient state, coupling scales or revising failed assumptions. We propose Life Operators: task-bounded mappings that define three scientific roles. Perception operators infer task-relevant biological states from multimodal observations, Evolution operators propagate these states under natural or intervention-conditioned dynamics, and Generation operators map them to measurable signals. Each role may be realised by equations, statistical models, neural networks or hybrids. Bridge operators connect components with different variables, scales and time steps. Selected operators and bridges form task-specific Operator Graphs containing the smallest set of states and mechanisms sufficient for a declared claim. This modular structure also makes scientific revision localisable. An AI co-scientist may propose changes to states, operators, bridges or graph structure, while independent evidence determines which variants are retained, restricted or retired. Over time, validated components could accumulate into broader multiscale models of the human body and provide a computational foundation for medical artificial superintelligence.
[NLP-152] Do Multimodal LLM s See Before They Read? Diagnosing Contextual Sycophancy
【速读】： 该论文旨在解决多模态大语言模型中存在的“多模态情境谄媚”（multimodal contextual sycophancy）问题，即外部文本信息在与图像证据冲突时会错误地覆盖视觉证据，导致模型判断失真。其解决方案的关键在于重构信息边界（information boundary），通过系统性地控制文本、视觉证据与常识先验的引入时机，提出“系统2视觉仲裁”（System-2 Visual Arbitration, S2VA）机制——该机制在决策过程中将文本信息从视觉感知模块中隔离，确保视觉见证者（context-blind visual witness）在不受文本干扰的情况下独立生成报告。实验表明，相较于直接使用视觉见证者报告或联合条件输入，S2VA显著提升准确率（跨六模型提升19.7至44.1个百分点），且所有置信区间均不包含零，证明其有效性。研究进一步揭示，最优信息边界并非固定不变，而是依赖于具体模型架构及外部文本来源，凸显了上下文引入时机与模型特异性对多模态推理偏差的关键影响。
链接: https://arxiv.org/abs/2609.00067
作者: Yi-Cheng Lai,Hen-Hsen Huang
机构: Institute of Information Science, Academia Sinica (中央研究院資訊科學研究所); Taipei, Taiwan
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:External text can override conflicting image evidence in multimodal large language models, a failure we call multimodal contextual sycophancy. We introduce a 998-case diagnostic that independently varies visual evidence, commonsense priors, and external text, and probe when this failure arises by moving the information boundary around a context-blind visual witness. On abnormal images paired with Gemini-generated false text, GPT-5.1 scores 7.9% under joint conditioning, 49.7% when the context-blind witness report is scored directly, 63.7% under a matched two-call witness-arbiter pipeline that exposes the witness to the text, and 84.2% under System-2 Visual Arbitration (S2VA), which withholds the text from the witness. Across six models, S2VA improves over the direct witness report by 19.7 to 44.1 points, with all paired 95% confidence intervals excluding zero. The best information boundary is not uniform: textual context scaffolds some models, and a GPT-4o-regenerated subset changes the relative ordering of joint conditioning, Witness-Only, and S2VA. Contextual sycophancy is therefore sensitive to when text is introduced, as well as to the model and context source.
[NLP-153] OCGQuant: Outlier-Companion Grouping for NVFP4 Quantization EMNLP2026
【速读】： 该论文旨在解决NVFP4（一种高效的低比特推理微缩格式）在实际应用中因激活值异常值（activation outliers）导致量化精度下降的问题。具体而言，在每个量化块内，大值激活会主导该块的缩放因子，从而加剧其余共享同一缩放因子的数值的量化误差，这种现象被称为“附带量化误差”（Collateral Quantization Error）。现有后训练量化（Post-Training Quantization, PTQ）方法虽通过混合精度、旋转或残差补偿等策略缓解异常值影响，但或未针对NVFP4进行优化，或引入额外计算开销。为此，本文从通道分组视角重新审视NVFP4，提出一种名为OCGQuant的新方法，其核心是基于异常值-伴生通道分组（Outlier-Companion Grouping, OCG）机制，自适应地将高幅值异常通道与低幅值伴生通道配对，以优化量化块内的通道组成结构，降低附带量化误差。实验结果表明，OCGQuant在Llama3和Qwen3模型上均实现了最低的WikiText-2困惑度与最高的平均下游任务准确率，同时保持接近RTN的预填充加速性能，并与之匹配峰值解码内存占用，展现出优异的精度-效率平衡。
链接: https://arxiv.org/abs/2609.00066
作者: Yishan Yao,Binjun Li,Hanling Yi,Pengyu Li,Xiaoqing Liu,Zihan Yang,Xiaotian Yu,Zhiwen Yu
机构: South China University of Technology (华南理工大学); Intellifusion Inc.
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Accepted to EMNLP 2026 (Main Conference)
Abstract:NVFP4 is an efficient microscaling format for low-bit inference, but activation outliers can still degrade quantization accuracy within NVFP4 blocks. Within each quantization block, large activations can dominate the block scale, increasing the quantization error of the remaining values sharing the same scale. Existing post-training quantization (PTQ) methods mitigate outlier errors through strategies such as mixed precision, rotation, or residual compensation, but these approaches are either not specifically tailored to NVFP4 or introduce additional computation. In this work, we revisit NVFP4 from a channel-grouping perspective and define the reducible error incurred by remaining block values under the scale set by the block maximum as Collateral Quantization Error. Based on this insight, we propose OCGQuant, a post-training quantization method centered on Outlier-Companion Grouping (OCG), which adaptively pairs outlier channels with low-magnitude companion channels to improve NVFP4 activation block composition. Experiments on Llama3 and Qwen3 show that OCGQuant achieves the lowest WikiText-2 perplexity and highest average downstream accuracy among evaluated PTQ methods, while maintaining prefill speedup close to RTN and matching its peak decoding memory. Code is available at this https URL.
[NLP-154] Scientific Agent Skills: A Library of Procedural Knowledge for Research Agents
【速读】： 该论文旨在解决生成式AI在科学实验分析中虽能生成可运行代码，但其分析过程缺乏可辩护性（defensible）的问题。核心挑战在于科学分析的合理性依赖于一系列程序性决策，如领域内公认的统计检验方法、权威标识符命名空间的选择以及结果报告所需附带的限制说明等。为应对这一问题，作者提出了“科学代理技能”（Scientific Agent Skills）——一个开源库，包含163项分布在基因组学、化学信息学、医学影像、研究设计及科学传播等16个实践领域的标准化程序。其解决方案的关键在于将每个“技能”封装为一个包含版本化、人类可读指令文件的目录结构，仅在任务需要时由代理加载；同时配套提供参考材料与可执行脚本，确保分析流程透明、可复现且符合领域规范。该库以开放许可发布，支持提升科学生成式AI的可信度与可审计性。
链接: https://arxiv.org/abs/2609.00065
作者: Timothy Kassis,Vinayak Agarwal,Yuhuan He,Darshil Patel,Aubrey M. Brueckner
机构: K-Dense, Inc.(K-Dense公司)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:A language-model agent asked to analyse an experiment will usually return working code. Whether the analysis is defensible is a different question. A defensible analysis depends on procedural choices: which test the field accepts, which identifier namespace is authoritative, and which caveats must accompany a result. We present Scientific Agent Skills, an open library of 163 such procedures in 16 areas of practice, including genomics, cheminformatics, medical imaging, study design and scientific communication. Each skill is a directory built around a versioned, human-readable instruction file. An agent loads the file only when a task calls for it; the directory often also contains reference material and runnable scripts. We report no task-level evaluation and no host selection rate. Openly licensed and available at this https URL.
[NLP-155] Attention Sensitivity Is Not Enough: Dissociating Attention-Level and Behavioural In-Context Learning under Fine-Tuning
【速读】： 该论文旨在解决生成式AI模型在进行上下文学习（In-Context Learning, ICL）时，注意力机制作为其上下文敏感性（In-Context Sensitivity, ICS）代理指标的可靠性问题。尽管现有方法普遍依赖注意力变化来评估模型是否具备上下文感知能力，但本文指出，当这些注意力指标被优化后，其与实际行为表现之间可能出现严重脱节。为此，作者提出将ICS（平均行间距离度量最后一轮注意力在匹配与不匹配示范前缀间的差异）与ICL-GAP（同一前缀下的行为准确率差距）联合使用，构建一个可验证的行为基准。在对Llama-2-7B的四臂消融实验中，通过最大化ICS的正则化项（\armKL），ICS提升至接近理论上限（1.413），但此时ICL-GAP几乎为零，且MMLU准确率从0.371下降至0.279，揭示了注意力指标与真实行为之间的“Goodhart偏差”。进一步分析表明，尽管注意力变得尖锐且前缀间近乎分离，但其路由路径却集中在格式和示范内容令牌而非标签上，说明注意力的改变并未带来有效任务适应。随机标签实验确认了行为探针仍具有足够的动态范围。最后，通过引入行为门控机制，部分缓解了该偏差，而锚定于预训练计算目标的优化策略则维持了高MMLU与适度ICS的稳定区域。核心结论是：仅凭注意力层面的代理指标不足以作为训练目标，必须经过行为差距的实证验证，才能确保其有效性。
链接: https://arxiv.org/abs/2609.00064
作者: Jinyuan Zhang,Peng He,He Hu,Yin Yuan,ShengShuo Jiao
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 15 pages, 5 figures; appendices included
Abstract:In-context learning (ICL) lets large language models adapt to new tasks from demonstrations, and fine-tuning can erode this behaviour. Many preservation diagnostics inspect attention: if attention changes when demonstrations change, the model is treated as context-sensitive. This paper asks how far that proxy can be trusted once it is optimised. We formalise \emphIn-Context Sensitivity (ICS), the average row distance between last-token attention on matched and mismatched demonstration prefixes, and pair it with \emphICL-GAP, the behavioural accuracy gap between the same prefixes. In a controlled four-arm ablation on Llama-2-7B, an ICS-maximising regulariser ( \armKL ) drives ICS to 1.413 , within 0.5% of its geometric ceiling. The behavioural readout tells a different story: ICL-GAP stays near zero and MMLU accuracy moves from 0.371 to 0.279 , a Goodhart dissociation of the bounded attention proxy. Endpoint statistics locate the mechanism: attention grows sharp and near-disjoint across prefixes yet routes to formatting and demonstration-body tokens rather than labels. A random-label protocol confirms that the behavioural probe family retains dynamic range at the same checkpoints. In a constructive sweep, behaviour gating partially mitigates the effect, while objectives anchored to pretrained computation hold the high-MMLU, moderate-ICS region that divergence maximisers leave. The main lesson is diagnostic: attention-level ICL proxies earn their place as training targets only after validation against behavioural gaps.
[NLP-156] Medical Causal Hypothesis Verification with Large Language Models
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在医疗领域中对因果医学命题进行验证时的可靠性问题，尤其关注其能否基于可验证的科学证据准确判断因果关系。当前尽管LLMs在回答疾病、症状和治疗相关问题上表现良好，但其在识别真实科学依据、提供有效文献支持以及拒绝缺乏证据的假设方面仍存在显著不足。为此，研究提出了一套系统化的因果假说验证评估框架，用于追踪现有及未来LLMs在该任务上的性能表现。通过在17个医学因果假说上评估8个主流LLMs，并依据六个维度对它们提供的科学证据进行1,067个标注点的精细化标注，结合九项评价指标进行分析，结果表明：尽管LLMs具备较强的召回能力，但在提供有效科学文献、支撑证据质量以及正确拒绝无支持假设方面表现较差。这一发现揭示了当前LLMs在从生物医学文献中可靠验证因果关系方面的关键局限性，强调了在医疗信息检索等高风险场景中引入严谨评估机制的必要性。
链接: https://arxiv.org/abs/2609.00063
作者: Safiyyah Ahmed,Abrar Ansari,Md Aminul Islam,Elena Zheleva
机构: University of Illinois Chicago(伊利诺伊大学芝加哥分校)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:The growing use of large language models (LLMs) for search and information retrieval underscores the need to evaluate their reliability in high-stakes domains such as healthcare. Although LLMs can effectively answer questions about diseases, symptoms, and treatments, their ability to accurately assess causal relationships and ground their conclusions in verified scientific evidence remains unclear. Here, we present a preliminary, small-scale study that investigates the accuracy of LLMs in evaluating causal medical claims and supporting them with peer-reviewed research. We propose an evaluation framework for causal hypothesis verification that can be used to systematically track the performance of existing and future LLMs. We assess the performance of eight LLMs on 17 medical causal hypotheses to evaluate whether they can reliably verify these hypotheses using scientific evidence from the literature. We systematically annotate the scientific evidence they provide according to six criteria (a total of 1,067 annotation points) and assess them with nine evaluation metrics. Our analysis shows that while LLMs exhibit strong recall, they often perform poorly at providing valid scientific articles and evidence for support and at rejecting unsupported hypotheses. These findings highlight a critical limitation of current LLMs, as they cannot yet be trusted fully to verify causal relationships from the biomedical literature. This work underscores the need for rigorous evaluation before using LLMs for search and retrieval in healthcare settings.
[NLP-157] RePro: Proof-Verified Benchmark Rewriting for Reliable Evaluation of LLM Mathematical Problem Solving EMNLP2026
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在数学问题求解评估中因数据污染（data contamination）导致的可靠性问题。现有基于重写（rewriting-based）的评估方法虽能缓解模型对训练数据的记忆化现象，但缺乏对问题有效性与答案正确性的保障。为此，本文提出Proof-Verified Benchmark Rewriting（RePro），首次将面向Lean形式化证明系统的神经自动化定理证明器（Neural Automated Theorem Provers, ATPs）集成到基准测试重写流程中，通过Lean验证的证明确保重写后问题及其答案的正确性。实验结果表明，在GSM8K和MATH数据集上，RePro生成的重写样本实现了100%的问题定义完整性、可解性及答案正确性，显著优于现有方法所生成的无效或错误实例。此外，多个模型在经过证明验证的重写基准上表现下降，表明其性能对表面特征和结构变化敏感，可能部分反映记忆效应。本工作的核心创新在于利用形式化证明框架实现可验证的、语义保真的问题重写机制，从而构建更可信的评估基准。
链接: https://arxiv.org/abs/2609.00062
作者: Xiyuan Zhou,Zhuoqi Li,Xinlei Wang,Yirui He,Yuhao Wu,Yuheng Cheng,Yan Xu,Junhua Zhao,Jinjin Gu
机构: Nanyang Technological University (南洋理工大学); The Chinese University of Hong Kong, Shenzhen (香港中文大学（深圳）); INSAIT, Sofia University “St. Kliment Ohridski” (索非亚大学“圣克莱门特·奥赫里德斯基”研究所); Shenzhen Loop Area Institute (深圳环区研究院); AIRS (人工智能与机器人研究所)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Accepted to the EMNLP 2026 Main Conference
Abstract:Data contamination undermines the reliable evaluation of large language models (LLMs) on mathematical problem solving. While rewriting-based evaluation mitigates memorization, existing methods lack guarantees of problem validity and answer correctness. We propose Proof-Verified Benchmark Rewriting (RePro), the first framework to integrate Lean-oriented neural automated theorem provers (ATPs) into benchmark rewriting, which rewrites problems and regenerates answers with correctness ensured by Lean-verified proofs. Experiments on GSM8K and MATH show that RePro’s retained rewritten instances achieve 100% well-definedness, feasibility, and answer correctness, while existing methods still produce invalid or incorrect instances. Moreover, several models exhibit accuracy drops on proof-verified rewritten benchmarks, suggesting that their performance is sensitive to surface-level and structural variations and may partly reflect memorization effects. Our source code and data are available at this https URL.
[NLP-158] ValueGraph: Value-Signal Guided Graph Pre-training for Contextualized User Representation
【速读】： 该论文旨在解决现有用户表征方法在社交媒体用户建模中忽视价值信号（value signal）这一关键维度的问题。传统方法主要依赖用户发布的文本内容或互动关系，却未能充分捕捉用户通过价值倾向表达态度的深层机制。其解决方案的关键在于提出ValueGraph——一种基于图预训练的框架，利用自动推断出的道德价值信号作为噪声辅助信号，引导上下文化的用户表征学习。该框架通过帖子-回复图结构同时学习语义与拓扑特征，并结合对比学习和聚类目标，依据相对价值相似性对用户进行对齐。不同于将推断的价值视为精确的心理标签，ValueGraph将其作为软约束，以提供有益的归纳偏置（inductive bias），从而增强社会感知型用户建模能力。实验结果表明，该方法在立场检测与微博机器人识别任务上均显著优于主流的基于文本、图结构及纯大语言模型（LLM）的基线模型，验证了价值信号引导的有效性。
链接: https://arxiv.org/abs/2609.00057
作者: Yitong Han,Wei Gao,Yi Zhao,Prasanta Bhattacharya,Fengzhu Zeng,Mohammad Amanlou
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Value signals are aggregated user-level moral representations that capture users’ inferred value-related tendencies from their online discourse. User behavior on social media is shaped not only by what users say or whom they interact with, but also by the value signal through which they express attitudes. Existing user representation methods largely miss this value-relevant dimension. We propose ValueGraph, a graph pre-training framework that uses automatically inferred moral-value signals as noisy auxiliary signals for contextualized user representation. From post-reply graphs, ValueGraph learns semantic and structural representations and further aligns users through relative value similarity with contrastive and clustering objectives. Rather than treating inferred values as gold psychological labels, ValueGraph uses them as soft constraints for representation learning. Experiments on stance detection and twitter bot detection show consistent gains over strong text-based, graph-based, and text-only LLM baselines, highlighting value-signal guidance as a useful inductive bias for socially informed user modeling.
[NLP-159] Zero-Shot Respiratory Sound Classification through LLM -Augmented Audio-Text Alignment INTERSPEECH2026
【速读】： 该论文旨在解决自监督呼吸信号编码器在临床领域缺乏语义锚定的问题，导致其在无任务特定标注数据的情况下难以实现零样本推理。其核心解决方案是构建一个将编码器与医学术语对齐的框架，通过在共享潜在空间中实现语义对齐，使编码器具备零样本能力，从而转化为可通用的临床基础模型。为缓解成对数据稀缺问题，研究利用医学大语言模型（Medical LLM）从元数据生成结构化报告，作为对比学习中的密集语义锚点。训练过程结合基于Sigmoid的对比损失、编码器原有的自监督学习目标以及感知相似性的负样本采样策略，以增强病理边界区分能力。实验在6个数据集上的9项任务中验证了该方法的有效性，零样本平均AUC达61.3%，显著优于CLAP（51.4%）和Qwen2-Audio（54.9%），且仅使用全规模基线43%的数据便达到最高线性探测AUC（71.6%），证明了结构化语义对齐在临床诊断任务中优于大规模通用模型。
链接: https://arxiv.org/abs/2609.00055
作者: Mustafa Talha İlerisoy,Hung Manh Pham,Mathias Funk,Mykola Pechenizkiy,Aaqib Saeed
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Sound (cs.SD)
备注: Accepted to INTERSPEECH 2026
Abstract:Self-supervised respiratory encoders lack semantic grounding in clinical domain needed for zero-shot inference, limiting their utility without task-specific labeled data. We propose a framework that aligns these encoders with medical terminology in a shared latent space turning them into a zero-shot-capable foundation model. To address paired data scarcity, we use a medical LLM to synthesize structured reports from metadata, creating dense semantic anchors for contrastive learning. Our training combines a sigmoid-based contrastive loss with encoder’s native SSL objective and similarity-aware negative sampling to sharpen pathological boundaries. Across 9 tasks on 6 datasets, our method achieves a 61.3% mean zero-shot AUC, surpassing CLAP (51.4%) and Qwen2-Audio (54.9%) while reaching the highest linear probing AUC (71.6%) with only 43% of data used by full-scale baselines, showing that structured semantic alignment outperforms large-scale, general-purpose models in clinical diagnostics.
[NLP-160] Agent Prov: Auditing Agent ic LLM API Providers via Tool-use Policy Probes EMNLP2026
【速读】： 该论文旨在解决当前大语言模型（LLM）API服务中基础模型身份不可信的问题，即供应商可能在不通知用户的情况下悄然替换、量化或封装所宣称的模型以降低部署成本。现有审计方法依赖文本输出通道来判断模型身份，但在现代代理型（agentic）API架构中，由于服务栈（如OpenAI、Anthropic、Gemini、Cloudflare Workers AI、LangGraph）会丢弃原始文本输出，仅暴露结构化动作（action），导致基于文本的审计方法失效；同时，供应商注入的系统提示词可显著扭曲文本分布，使此类方法产生误报。本文的关键突破在于发现近期代理型后训练已将工具使用能力直接内化于模型权重之中，这一特性仍可通过结构化动作通道被访问且对部署环境变化具有高度鲁棒性。为此，论文提出Agentic Provenance（AgentProv），一种首个基于动作的模型身份审计机制：通过分析模型调用工具的类别分布生成指纹，并采用最大均值差异（MMD）置换检验进行身份判定。实验表明，AgentProv在630个检查点对上实现了100%的替换模型检出率，且在系统提示词注入攻击下假阳性率控制在7%，显著优于MET（67%）和RUT（53%）。此外，其与MET在第三方API上的分歧结果与独立的令牌计数侧信道检测到的供应商注入提示行为一致，验证了其有效性与可靠性。
链接: https://arxiv.org/abs/2609.00052
作者: Xun Wang,Bihe Zhao,Michael Backes,Franziska Boenisch,Adam Dziedzic
机构: CISPA Helmholtz Center for Information Security
类目: Cryptography and Security (cs.CR); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 15 pages, 4 figures. Accepted to EMNLP 2026
Abstract:Commercial LLM APIs advertise a specific foundation model, but the served backbone may be silently substituted, quantized, or wrapped, for example to save deployment costs. All existing audits decide backbone identity from the text-output channel, which is structurally fragile for agentic APIs because modern serving stacks (OpenAI, Anthropic, Gemini, Cloudflare Workers AI, LangGraph) discard text and expose only structured actions when the model calls a tool, and provider-injected system prompts can distort text distributions enough that text-channel tests falsely accuse honest providers of substituting the claimed model. We observe that recent agentic post-training internalizes tool-use directly into the weights, opening a new audit channel that the serving stack still exposes and that is largely invariant to deployment context. We introduce Agentic Provenance (AgentProv), the first action-based identity audit for agentic LLM APIs: AgentProv fingerprints a deployed model through its categorical tool-call distribution and decides identity via an MMD permutation test. AgentProv catches every substituted model (100% on 630 evaluated checkpoint pairs), while holding the false-positive rate under system-prompt injection at 7% (vs. 67% for MET and 53% for RUT). On third-party API endpoints, AgentProv’s disagreements with MET are consistent with an independent token-count side-channel that detects provider-injected system prompts.
[NLP-161] From Detection to Refusal: Safer LLM s via Circuit-Guided Weight Scaling EMNLP2026
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在面对对抗性提示时仍可能生成不安全内容的问题，尤其关注模型内部实现安全行为的机制尚不清晰这一关键挑战。其解决方案的核心在于从机械可解释性（mechanistic interpretability）视角揭示了一种多阶段的“安全电路”（safety circuit）结构，该电路由三个关键组件构成：(i) 有害检测头（Harmful Detection Heads），负责识别有害输入；(ii) 安全神经元（Safety Neurons），在残差流中传递并稳定安全信号；(iii) 拒绝生成头（Refusal Heads），将安全信号转化为拒绝生成或生成安全响应的行为。通过针对性的注意力头与神经元级干预实验，研究提供了因果证据，证实上游有害检测头的抑制会破坏下游拒绝行为，且安全神经元在此过程中起到中介作用。该电路结构在多个大语言模型架构及对抗攻击场景下具有可重复性，并通过保持架构不变的权重缩放作为机械探针验证其功能相关性。在六种模型上，基于电路指导的缩放策略使对抗攻击下的安全率提升26.5%，同时仅导致四个标准基准测试中1.7%的准确率下降。研究表明，大语言模型的安全行为可通过稳定的、可迁移的机械抽象进行建模，为理解对齐行为提供了新的理论框架。
链接: https://arxiv.org/abs/2609.00051
作者: Kuan-Lin Chu,Chung-En Sun,Tsui-Wei Weng
机构: University of California San Diego (加州大学圣地亚哥分校)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Computers and Society (cs.CY)
备注: Accepted to Findings of the 2026 Conference on Empirical Methods in Natural Language Processing (EMNLP 2026)
Abstract:Despite extensive alignment efforts, Large Language Models (LLMs) remain vulnerable to generating unsafe content under adversarial prompting, yet the internal mechanisms by which safety behaviors are implemented remain poorly understood. We study LLM safety from a mechanistic interpretability perspective and characterize a multi-stage safety circuit that organizes refusal behavior, consisting of (i) \textbfHarmful Detection Heads that respond to harmful inputs, (ii) \textbfSafety Neurons that mediate and stabilize safety signals in the residual stream, and (iii) \textbfRefusal Heads that translate these signals into safe response generation. Using targeted attention-head and neuron-level interventions, we provide causal evidence consistent with this circuit organization, showing that suppressing upstream Harmful Detection Heads disrupts downstream refusal behavior and that safety neurons mediate this interaction. We validate that this decomposition recurs across multiple LLM architectures and adversarial attack settings, and use simple, architecture-preserving weight scaling as a mechanistic probe to test its functional relevance. Across six LLMs, circuit-guided scaling improves safety rates under attacks by 26.5%, while incurring only a 1.7% accuracy drop across four standard benchmarks. Overall, our results support a circuit-level interpretation of LLM safety and suggest that mechanistic abstractions can reveal stable and transferable patterns underlying aligned behavior.
[NLP-162] GUI-CC: Benchmarking Contextual Consistency of GUI World Models as Agent Environments EMNLP26
【速读】： 该论文旨在解决当前生成式图形用户界面（GUI）世界模型在评估与实际应用之间存在的根本性脱节问题：尽管这些模型常被评估为单步下一屏幕预测器，但其真实应用场景是作为多步交互环境供GUI智能体使用。这种评估方式忽略了关键要求——生成状态在反复重用于后续交互时必须保持上下文一致性。为此，论文提出GUI-CC基准，专门评估GUI世界模型作为智能体环境时的上下文一致性，而非孤立的单步预测性能。其解决方案的关键在于构建双轨评估机制：一是离线参考动作轨道，沿真实移动GUI轨迹滚动模型；二是在线智能体循环轨道，允许固定探测智能体与模型生成的UI进行交互。通过从GUIOdyssey构建500个离线轨迹任务，并在30个移动应用中验证200个模拟器确认的在线任务，GUI-CC综合评估转换保真度、转换合理性、上下文一致性和任务进展。实验表明，仅具备合理单步生成能力的现有模型并不能保证可靠环境模拟，其生成界面虽外观可用，却难以维持任务相关上下文或支持可执行的多步推演。
链接: https://arxiv.org/abs/2609.00048
作者: Lin Fu,Zheyuan Yang,Tianhui Zhang,Jinbiao Wei,Guo Gan,Boxu Liu,Yilun Zhao,Yu Rong
机构: Yale University(耶鲁大学); Zhejiang University(浙江大学); China University of Geosciences(中国地质大学); DAMO Academy, Alibaba Group(阿里达摩院); Tongji University(同济大学); University of California, San Diego(加州大学圣地亚哥分校)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: EMNLP 26 Findings
Abstract:GUI world models are increasingly evaluated as one-step next-screen predictors, yet their intended use is often as multi-step environments for GUI agents. This mismatch leaves a key requirement under-tested: generated states must remain contextually consistent when they are repeatedly reused for future interaction. We introduce GUI-CC, a benchmark that evaluates contextual consistency of GUI world models as agent environments rather than isolated next-screen predictors. GUI-CC contains two complementary tracks: an offline reference-action track that rolls models along real mobile GUI trajectories, and an online agent-loop track that lets fixed probing agents interact with model-generated UIs. We construct 500 offline trajectory tasks from GUIOdyssey and 200 emulator-verified online tasks across 30 mobile apps. GUI-CC evaluates transition fidelity, transition plausibility, contextual consistency, and task progress. Experiments show that plausible single-step generation does not guarantee reliable environment simulation: current models often produce usable-looking screens while failing to preserve task-relevant context or support executable multi-step rollouts.
[NLP-163] rajectory-judge: What Outcome-Only LLM Judges Miss on Agent Trajectories NEURIPS2026
【速读】： 该论文旨在解决大语言模型（Large Language Model, LLM）智能体在评估过程中存在的“仅结果评价”（outcome-only evaluation）缺陷问题。当前主流的评估范式仅根据最终输出结果判断任务是否完成，忽略了推理过程中的错误路径，导致对通过错误步骤达成正确结果的智能体存在评估盲区。其解决方案的关键在于构建一个确定性的工具使用型支持桌面环境，结合可复现的脚本化决策策略（oracle policy）与故障注入机制（fault injector），系统性地将故障按客户可见性分为“无声”（silent，结果未受损）与“有声”（loud，结果明显异常）两类，并在此基础上对五种评估方法（包括程序规则、仅结果判别、分步评分表在两种模型规模下的表现及自洽性集成）进行多维度对比分析。实验表明，仅结果评价能检测84%的有声故障，但仅45%的无声故障，且误报率高达33%；而分步评分表虽成本增加三倍，却实现77%的无声故障召回率且无误报。更关键的是，所有评价方式均不依赖最终回复内容，即使在完美轨迹中插入虚构承诺也几乎完全逃逸检测，凸显现有评估体系的严重局限性。研究主张评估应基于结果是否存活进行分层召回率分析，并公开环境、故障注入器、原始判别数据及可重构分析管道，以推动可复现、透明的智能体评估范式演进。
链接: https://arxiv.org/abs/2609.00038
作者: Hadi Mohammadi
机构: Utrecht University (乌得勒支大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Software Engineering (cs.SE)
备注: 16 pages (8-page main text). Under review at a NeurIPS 2026 workshop. Code, data, and raw verdicts: this https URL
Abstract:Outcome-only evaluation is the production default for LLM agents: show a judge the request and the final reply and ask whether it was handled well. The metric is structurally blind to an agent that reaches the right answer the wrong way. We measure that blind spot where ground truth is known by construction: a deterministic tool-using support-desk environment, a scripted oracle policy that always solves it, and a fault injector that breaks exactly one thing at a known step, stratifying faults by whether the customer-visible outcome survived (silent) or not (loud). Five judges (programmatic rules, outcome-only, step-rubric at two model sizes, and a self-consistency ensemble) are scored on detection, step localisation, fault typing, calibration, and cost over 400 trajectories. The outcome-only judge catches 84% of loud faults but 45% of silent ones while flagging 33% of correct trajectories; a step-rubric judge reaches 77% silent recall with zero false alarms at 3x the cost. No judge reads the final reply: an invented promise appended to an otherwise perfect trajectory evades the rules entirely and the step judge 82% of the time, and self-consistency triples cost while improving nothing. We argue that judge evaluations must stratify recall by outcome survival, and release the environment, the injector, all raw verdicts, and an analysis pipeline that rebuilds every number offline.
[NLP-164] UI-Venus-2 Technical Report
【速读】： 该论文旨在解决多模态图形用户界面（GUI）智能体在从基准测试模型向实际应用场景转化过程中面临的三大核心挑战：环境覆盖有限、任务构建脆弱以及奖励验证不可靠。其解决方案的关键在于构建一个统一闭环推理-执行框架下的通用基础型GUI智能体UI-Venus-2，通过三方面协同扩展实现落地突破：（1）环境维度，将覆盖范围拓展至超过170个多语言移动端应用及原生桌面操作系统；（2）任务维度，采用基于深度研究的函数驱动指令生成管道，提升任务语义的准确性和多样性；（3）验证维度，引入基于轨迹级与样本级的评估机制，结合视觉关键点和多模型投票策略，确保强化学习训练中奖励信号的可靠性。此外，集成安全感知机制以保障高影响操作的安全可控执行。整体上，UI-Venus-2通过可扩展性、可验证性和自省能力的融合，推动了面向真实世界应用的通用化、鲁棒化智能体发展。
链接: https://arxiv.org/abs/2609.00028
作者: Venus Team,Zhuohan Cai,Haoxing Chen,Jiaxuan Chen,Weizhi Chen,Changlong Gao,Zhangxuan Gu,Yuan Guo,Yusong Hu,Jianrong Jiang,Jianguo Li,Runze Li,Jinzhen Lin,Zhenyu Ma,Changhua Meng,Han Peng,Xinyu Qiu,Shuheng Shen,Zhongyi Shui,Weiqiang Wang,Ming Wen,Zhuoer Xu,Hang Yan,Kaiwen Yang,Ruilin Yao,Nanjun Yu,Zhengwen Zeng,Lianrui Zhang,Yunzhu Zhang,Zhe Zhao,Beitong Zhou
机构: Venus Team; Ant Group(蚂蚁集团)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:Multimodal GUI agents have emerged as a promising paradigm for digital task automation, yet transitioning from benchmark-oriented models to dependable real-world applications remains challenging due to limited environment coverage, brittle task construction, and unreliable reward verification. In this work, we present UI-Venus-2, a general-purpose foundation GUI agent designed to operate across mobile, web, and desktop environments through a unified closed-loop reasoning-action framework. To bridge the gap toward practical deployment, we jointly scale three critical dimensions: (1) Environments, expanding coverage to more than 170 multilingual mobile apps and native desktop operating systems; (2) Tasks, employing a deep-research pipeline for function-grounded instruction generation; and (3) Verification, adopting trace-level and sample-level evaluators with visual keypoints and multi-model voting to ensure reliable RL signals for training. Furthermore, we integrate safety-aware mechanisms to ensure controlled execution of consequential actions. By offering a capable, efficient, and open-source foundation, UI-Venus-2 advances the field toward more generalizable, verifiable, and self-reflective agents for real-world applications.
[NLP-165] Behaviorally Grounded User Profiles from the Wild for Personalized Alignment and Multi-Perspective Reasoning
【速读】： 该论文旨在解决当前基于大语言模型（LLM）的个性化技术普遍依赖于僵化、人工构造的虚拟人格（persona）所带来的局限性，如忽视个体差异、过度依赖刻板印象以及无法捕捉真实人类偏好背后的细微信号。其核心解决方案是提出“行为特征锚定”（profile behavioral grounding）框架，通过从真实、匿名化的社交媒体文本中直接提取开放式的高保真用户画像，从而实现对用户行为模式的精准建模。该方法在训练阶段通过监督微调（SFT）实现参数化个性化，在测试阶段则采用非参数化的多视角推理机制，显著提升了模型在复杂推荐与开放式问答任务中的表现，优于传统合成画像基线，不仅增强了模型与用户特征的对齐程度，还支持更丰富、多维度的推理能力。研究结果表明，基于行为数据生成的开放式用户画像可作为下一代个性化语言系统的重要基础。
链接: https://arxiv.org/abs/2609.00014
作者: Yuxuan Li,Victor Zhong,Ehsan Kamalloo
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Persona-driven techniques increasingly adapt large language models (LLMs) to diverse contexts. However, existing methods predominantly rely on rigid, synthetic personas that flatten individual variation, rely on stereotypes, and miss the nuanced signals driving actual human preferences. We introduce profile behavioral grounding, a framework for extracting open-ended, high-fidelity user profiles directly from authentic, anonymized social media posts. We evaluate these profiles across two paradigms: train-time personalization via supervised finetuning (SFT) and non-parametric test-time multi-perspective reasoning. Across complex recommendation and open-ended query benchmarks, behaviorally grounded profiles consistently improve base models and outperform synthetic profile baselines, driving stronger parametric alignment and enabling richer, multifaceted reasoning. Our findings establish open-ended, behavior-derived profiles as a highly diverse and effective foundation for the next generation of personalized language systems. Our code base is available at this https URL.
[NLP-166] InteractBench: Benchmarking LLM s on Competitive Programming under Unrevealed Information ICML2026
【速读】： 该论文旨在解决当前大型语言模型（LLM）评估中对交互式算法推理能力覆盖不足的问题。现有基准主要聚焦于全信息任务（full-information tasks），即所有输入在问题开始时即已提供，忽略了算法推理中一个关键维度：模型生成的程序在关键信息未预先披露的情况下，能否通过动态交互逐步获取信息并维持状态。交互式问题（interactive problems）作为竞技编程中的独特组成部分，正体现了这一挑战——要求程序在严格协议约束和有限查询预算下，与裁判程序（interactor）进行多轮交互，仅在响应查询后才获得新信息。为此，作者提出了InteractBench，一个涵盖322个高质量交互式问题的基准，数据源自Codeforces、AtCoder、IOI和ICPC，并为每个问题配备可执行的本地交互器（local interactor），支持完全离线评估。其核心创新在于评估模型生成代码是否具备动态获取信息与状态追踪的能力。实验结果揭示了显著的“交互差距”（interaction gap）：即使是最先进的推理模型在交互式问题上表现也极为有限。除成功率外，研究进一步提出细粒度的失败分类体系，用于诊断缺陷根源；结果显示，尽管算法逻辑错误仍占主导，但协议违规（protocol violations）和查询预算超限（query-budget overruns）亦频繁发生。
链接: https://arxiv.org/abs/2608.29632
作者: Jiaze Li,Aocheng Shen,Bing Liu,Boyu Zhang,Xiaoxuan Fan,Qiankun Zhang,Xianjun Deng
机构: Huazhong University of Science and Technology (华中科技大学); Key Laboratory of Cyberspace Security, Ministry of Education (教育部网络空间安全重点实验室); Hubei Key Laboratory of Distributed System Security (湖北省分布式系统安全重点实验室)
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: Accepted at ICML 2026
Abstract:Competitive programming is increasingly being used to evaluate the algorithmic reasoning capabilities of large language models (LLMs). However, existing benchmarks primarily focus on full-information tasks where all problem inputs are provided upfront. This overlooks a critical dimension of algorithmic reasoning: the ability of generated programs to operate when key information is not revealed upfront. Interactive problems, a distinctive component of competitive programming, embody this challenge. These problems require programs to engage in multi-round interaction with an interactor (a judge program) under strict protocol constraints and limited query budgets, with new information revealed only in response to queries. To address this gap, we introduce InteractBench, a benchmark comprising 322 high-quality interactive problems curated from Codeforces, AtCoder, IOI, and ICPC. Each problem is packaged with executable local interactors, enabling fully offline evaluation. Unlike existing benchmarks, InteractBench assesses whether model-generated code can acquire information and track state dynamically. Our evaluation reveals a significant interaction gap: even the most advanced reasoning models achieve limited success on interactive problems. Beyond success rates, we propose a fine-grained failure taxonomy to diagnose the root causes of these deficiencies. Although algorithmic logic errors remain dominant, protocol violations and query-budget overruns are frequent. Code is available at this https URL.
[NLP-167] Stochastic Estimation of Transduced Language Models
【速读】： 该论文旨在解决生成式语言模型在跨语言转换（transduced language models, TLMs）中对目标字符串前缀概率计算的高复杂度问题。传统方法通过源语言前缀概率的近似与阈值剪枝的束搜索求和，虽可降低计算开销，但仅提供下界估计且误差未知。其核心挑战在于：目标前缀对应的源字符串集合可能呈指数级或无限大，导致精确计算不可行。本文提出一种基于无放回重采样与逆包含概率加权的无偏估计方法，通过递归应用校正机制，获得目标前缀概率的无偏估计，并可量化因阈值剪枝丢失的概率质量。所提出的束搜索算法动态扩展保留的源前缀并自适应选择保留项，随估计累积逐步减少前缀数量，既节省计算资源，又保证以概率1终止。实验表明，在百科文本与DNA序列任务上，该方法相比有放回重采样的顺序蒙特卡洛基线具有更优的计算-方差权衡；在长目标序列的前缀概率估计上，相较阈值剪枝束搜索显著降低运行时间数个数量级，使原本难以实现的计算成为可能。将已有研究中的阈值剪枝替换为该无偏采样方法后，虽显著降低语料库意外（corpus surprisal）估计值，但不改变原结论，验证了方法的有效性与稳健性。
链接: https://arxiv.org/abs/2608.27428
作者: Vésteinn Snæbjarnarson,Samuel Kiegeland,Manuel de Prada Corral,Ryan Cotterell,Tim Vieira
机构: ETH Zürich(苏黎世联邦理工学院); University of Copenhagen(哥本哈根大学); CHI-FRO
类目: Computation and Language (cs.CL); Formal Languages and Automata Theory (cs.FL)
备注:
Abstract:Transduced language models (TLMs) compose a pretrained \emphsource language model with a functional finite-state transducer to induce a language model over \emphtarget strings. Computing the probability of a target prefix under a TLM amounts to summing the source-model probabilities of all source strings that the transducer maps to target strings beginning with that prefix. This set can be exponentially large or infinite. Prior work uses a computational shortcut based on source prefix probabilities, then approximates the resulting sum with threshold-pruned beam summing. This produces a lower bound with unknown error. Instead, we resample source prefixes without replacement and reweight each selected prefix by the inverse of its inclusion probability. We show that applying this correction recursively gives an unbiased estimator of the target prefix probability and lets us estimate the mass lost by threshold pruning. Our beam-summing algorithm extends the retained source prefixes and samples which prefixes to keep, reducing their number as more probability mass is added to the running estimate. This can save computation and guarantees that the run halts with probability one. We evaluate the method on encyclopedic text and DNA against sequential Monte Carlo baselines that resample with replacement. It achieves a better compute–variance tradeoff on text and lower error at the same maximum number of particles on DNA. On a DNA-to-amino-acid transduction, it reduces runtime by several orders of magnitude relative to threshold-pruned beam summing and makes estimating prefix probabilities for long target strings feasible. Replacing threshold pruning with unbiased sampling in a published reading-time analysis substantially lowers the estimated corpus surprisal but leaves the published conclusions unchanged.
信息检索
[IR-0] Retrieved but not ranked: surface-form bias in structural retrieval from mathematics to agent trajectories
链接: https://arxiv.org/abs/2609.01556
作者: Nabira Rashid,Manolis Kellis
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Information Retrieval (cs.IR)
备注: 18 pages, 3 figures, code at this https URL
Abstract:We evaluate embedding retrieval where surface form and meaning are pulled apart on purpose: retrieving items that share underlying structure but not wording, in two unrelated domains under one protocol, competition mathematics (MathNet-Retrieve; 500 queries, 117,088-item corpus) and embodied-agent trajectories (ALFWorld-derived; 118 queries, 336 trajectories). In mathematics the failure is complete: strict Hit@1 at the heaviest disguise tier is 0.0% for both production embedders (bootstrap 95% CI [0.0, 0.0]) while the correct item sits in the top 10 nearly always, and in 95.2 to 99.8% of misses the winner is more lexically similar to the query than the correct answer. In trajectories, where surface variation is incidental, the same models land at or near hypergeometric chance when gold must involve a different object, and below chance for all three embedders once gold must differ in object and receptacle: retrieval anchors on literal tokens, not task structure. A lexical reranker control hurts in mathematics and helps in trajectories (closing 26 to 36% of the gap, CIs excluding zero); its sign reveals whether a benchmark’s surface variation is adversarial or incidental. An LLM reranker recovers 5 to 63% of the gap in mathematics and 43 to 76% in trajectories; direction replicates across three judges (all 21 cells positive), but effect sizes, tier profiles, and the outlier judge change with domain (paired differences excluding zero everywhere). Mathematics gains concentrate on well-known competitions (+19.8 points, CI [+6.7, +33.2], one of six cells), so part of the recovery is memorization. In a paired downstream experiment (210 queries, graders at 96 to 99% agreement), oracle retrieval was indistinguishable from adversarially bad retrieval (McNemar p = 0.678); the solver’s 69.5% zero-shot accuracy is largely a truncation proxy (97 to 100% on finished answers), leaving no headroom.
[IR-1] AutoConcept: Training-Free Concept-Guided Reranking for Metadata-Available Composed Image Retrieval PRICAI2026
链接: https://arxiv.org/abs/2609.01456
作者: Tianyu Wang,Tianjiao Wu
类目: Information Retrieval (cs.IR); Computation and Language (cs.CL)
备注: Accepted regular paper at PRICAI 2026. 16 pages, 4 figures
Abstract:Composed image retrieval (CIR) retrieves a target image from a reference image and a text modification. This paper studies metadata-available CIR reranking, where a fixed CIR model first returns a candidate pool and gallery metadata is then used for second-stage concept-guided scoring. We introduce AutoConcept, a training-free reranker that converts concept evidence into an interpretable memory. AutoConcept filters noisy concepts, activates query-relevant positive constraints with an auxiliary negative penalty, and combines base retrieval scores with metadata-based concept-candidate alignment through inference-time calibration. On FashionIQ, AutoConcept yields significant early-rank improvements over WeiMoCIR and consistent plug-in gains on LinCIR candidate pools. Metadata-aware controls show that structured concept memory adds signal beyond direct query-text and extracted-attribute matching, while a query-only variant further supports the effectiveness of concept-level reranking. A supplementary real-human concept-label study indicates that the same memory interface can consume participant-provided evidence. These results position AutoConcept as an interpretable concept-memory reranker for product-style CIR galleries with available metadata.
[IR-2] VerTox: Verifiable Reward-Guided Corpus Poisoning Against Neural Ranking Models EMNLP2026
链接: https://arxiv.org/abs/2609.01325
作者: Zhiqi Huang,Vivek Datla,Zhichao Xu,Puxuan Yu,Vivek Srikumar,Alfy Samuel
类目: Computation and Language (cs.CL); Information Retrieval (cs.IR)
备注: EMNLP 2026 Main Conference
Abstract:Neural ranking models have become core components of modern information retrieval systems and important building blocks of AI systems such as retrieval-augmented generation (RAG) pipelines. However, their robustness remains insufficiently understood in the presence of large language models (LLMs), which can generate fluent and deceptive content at scale. This work investigates the vulnerability of neural ranking models to corpus poisoning attacks, in which an adversary injects a small number of maliciously crafted documents into the corpus to distort ranking behavior. We propose VerTox, the first framework to formulate corpus poisoning as a verifiable reward-guided reinforcement learning (RLVR) problem. By explicitly coupling ranking distortion with factual corruption through specialized reward shaping, we fine-tune compact LLMs into adversarial generators. Experiments demonstrate that our method achieves near-perfect attack success rates, producing adversarial documents that frequently rank higher than target documents across major neural ranking architectures, as well as a proprietary commercial embedding model. The generated adversarial documents are fluent and exhibit low perplexity, making them difficult to detect. Furthermore, by explicitly encouraging factual corruption, our adversarial documents significantly degrade the performance of a downstream RAG application.
[IR-3] MIDR: Enrichment-Augmented Indexing for Multimodal Document Retrieval EMNLP2026
链接: https://arxiv.org/abs/2609.01316
作者: Debanjan Mahata,Atharva Tendle,Daniel Preotiuc-Pietro,Yong Zhuang,Ozan Irsoy
类目: Information Retrieval (cs.IR); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: To appear in Proceedings of EMNLP 2026
Abstract:Retrieval over visually rich documents has a representation problem: important content often lives in tables, charts, figures, and layout relations that plain OCR linearizes, corrupts, or omits. ColPali-family visual retrievers address this with patch-level multi-vector indexes and late-interaction scoring, keeping image-derived retrieval on the query-time serving path. We introduce MIDR (Multimodal Indexing for Document Retrieval), a training-free framework for enrichment-augmented indexing that shifts multimodal reasoning to index time. During ingestion, a multimodal LLM converts rendered pages into verified textual fields that are indexed with BM25F and optionally fused with dense retrieval, enabling text-centric serving over multimodally grounded evidence. On ViDoRe V3, MIDR Hybrid achieves 0.6219 average nDCG across five English domains, a 23.0% relative gain over BM25, remaining competitive with ColQwen2.5. On two French-document domains, enrichment bridges English queries and French page text, lifting BM25 from 0.1532 to 0.5448 nDCG and outperforming ColQwen2.5. Across all seven domains, MIDR leads ColQwen2.5 on four while using approximately 9x smaller index memory and approximately 2x lower query latency. These results establish index-time multimodal reasoning as a compelling accuracy-deployment alternative to serving-time visual late interaction.
[IR-4] From Language to Behavior: Scaling Sequence Transformers for Industrial Recommendation Ranking with Rec-Native Designs
链接: https://arxiv.org/abs/2609.01240
作者: Jie Chen,Xiangqian Yu,Yanchao Lian,Tan Lu,Run Yang,Zhengchun Shang,Xing Wang,Cheng Chen,Ke Hu,Qiang Li,Tianjiu Yin,Xiaobing Liu
类目: Information Retrieval (cs.IR); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Scaling Transformers has driven large gains in language modeling, but transplanting this to behavior-sequence modeling in production ranking is challenging: recommendation differs in signal quality, where behavior sequences are noisy, temporally irregular, and sparsely supervised, and in computation asymmetry, where each request scores many candidates against one shared user history under tight latency budgets. We propose ReST, a recommendation-native Transformer scaling framework. For signal quality, it introduces a sequence encoder with dual-gated attention, rotary positional and temporal embedding, stabilized residual normalization, and training-only auxiliary objectives. For computation asymmetry, it factorizes ranking into a heavy reusable encoder and a lightweight cross decoder with projection-free KV attention and token-specific parameterization, coupling user-level shared-prefix training with shared-prefix serving for compute-once, decode-many-times ranking. Across industrial and public benchmarks, ReST achieves higher accuracy and scales more consistently along sequence length, depth, and width, where LLM-style Transformer blocks saturate. A one-week online A/B test on a production advertising platform improves online AUC by 1.31% and lifts a core revenue metric by 11.93% within a 50 ms P99 budget; ReST has since been fully deployed in production, showing that behavior-sequence scaling remains a promising, under-exploited axis for production ranking.
[IR-5] World Model-Guided Reinforcement Learning via Counterfactual User Engagement Simulation EMNLP’26
链接: https://arxiv.org/abs/2609.01067
作者: Ang Li,Xin Xu,Bin Liang,Yue Ma,Fubang Zhao,Yangyang Kang,Kam-Fai Wong
类目: Information Retrieval (cs.IR)
备注: EMNLP’26
Abstract:Reinforcement learning for user-centric agents is limited by the cost, latency, and risk of collecting online feedback, as well as by the lack of counterfactual comparisons under the same user state. In this paper, we propose World Model-Guided Reinforcement Learning via counterfactual user engagement simulation (WMG-RL), a framework in which a frozen user simulator provides reward supervision before real user exposure. Motivated by language world models, we instantiate the simulator as a User Engagement World Model (UEWM), which treats a recommended item as the agent action and the user’s heterogeneous feedback as the environment observation. Rather than learning one fixed environment transition, UEWM learns to infer user-specific dynamics from engagement history and apply them to candidate items. In WMG-RL, a downstream policy proposes multiple candidate items for the same history; UEWM predicts the corresponding engagement feedback in parallel; and the simulated feedback is converted into dense rewards for policy optimization. Experiments show that UEWM provides reliable and transferable reward signals across domains, and that WMG-RL enables a compact 1.7B student policy to match or surpass much larger LLMs on downstream recommendation tasks.
[IR-6] Web Price Extraction: State of the Art and an Adaptive Browserless Implementation
链接: https://arxiv.org/abs/2609.01030
作者: Evgeniia Kositsyna,Jorge Lloret-Gazo
类目: Information Retrieval (cs.IR); Machine Learning (cs.LG); Neural and Evolutionary Computing (cs.NE)
备注:
Abstract:Price extraction from websites is a key task for market monitoring, price comparison, and business analytics in e-commerce. Existing approaches can be broadly divided into four groups, and understanding their trade-offs in accuracy and scalability is essential for selecting suitable extraction strategies. Classical methods rely on manually written wrappers and rule induction from labeled pages, offering high accuracy but adapting poorly to structural changes and requiring considerable maintenance effort. Browser-based methods, using tools such as Selenium and Puppeteer, handle dynamic JavaScript content but consume large computational resources and scale poorly. Browserless approaches retrieve HTML directly via HTTP requests, offering significant gains in speed and cost, but rely on rules calibrated for specific sites. Methods based on machine learning and large language models offer adaptability but require training data and substantial computation. Our main contribution is an adaptive browserless price extraction system that improves robustness to structural differences between websites. We implemented a baseline architecture combining HTML page fragmentation with syntactic, semantic, and frequency rules, and extended it in two ways: a Bayesian approach that dynamically updates rule weights, and a genetic algorithm that optimizes the system’s global parameters. This hybrid scheme increased precision from 77.2% to 87.3% and reduced average per-page processing time by approximately 14% relative to the baseline, confirming it as a competitive alternative to manually tuned browserless solutions and to more resource-intensive browser- or LLM-based methods, offering high extraction accuracy at low computational cost. Subjects: Information Retrieval (cs.IR); Machine Learning (cs.LG); Neural and Evolutionary Computing (cs.NE) ACMclasses: H.3.3 Cite as: arXiv:2609.01030 [cs.IR] (or arXiv:2609.01030v1 [cs.IR] for this version) https://doi.org/10.48550/arXiv.2609.01030 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[IR-7] GR: Advancing Industrial Recommendation from Generative-Paradigm Ranking toward Unified Generation and Reasoning
链接: https://arxiv.org/abs/2609.00986
作者: TGR Team:Lei Cheng,Haonan Hu,Beibei Kong,Yudong Li,Zang Li,Yunsheng Pang,Hongyang Su,Jianchao Tu,Yunlong Wang,Bing Wen,Junzhang Zhu,Shaojie Zhu,Chengxiang Zhuo
类目: Information Retrieval (cs.IR)
备注:
Abstract:Industrial recommender systems typically rely on cascaded retrieval, pre-ranking, ranking, and reranking stages, whose separately optimized models limit scaling, fragment decision making, and lack semantic knowledge and reasoning. We present TGR (Tencent Generative Recommendation), an industrial framework that advances recommendation toward the generative paradigm along three coupled directions. TGR-GenRank upgrades ranking through CCFormer, which combines unified feature tokenization, a scalable Transformer backbone, feature-field separated cross attention, subspace token mixing, and hierarchical sequence compression while retaining per-item multi-task outputs. TGR-GenRec explores end-to-end generation under two paradigms: BARGE bridges item-boundary loss and semantic drift in hierarchical semantic-ID generation through item context-aware attention, hierarchical path reranking, and orthogonal dual-path decoding; HiGR performs whole-slate generation with prefix-structured semantic IDs, coarse-to-fine decoding, and listwise multi-objective alignment. TGR-Reason injects offline-generated semantic-ID reason tokens into online decoding, providing reasoning without request-time rollout. TGR is deployed across Tencent production surfaces serving hundreds of millions of users. CCFormer delivers significant gains in five A/B-tested scenarios and is fully launched in two, including +3.57% CTR and +1.71% advertising revenue. BARGE improves Hit@5 by 10.2-16.9% and yields +0.60% CTR and +1.70% reading time after full rollout. HiGR improves offline slate quality by 15.9-21.3% with a 5x inference speedup and achieves up to +1.22% watch time and +1.73% video views. TGR-Reason raises cold-start new-user Hit@1 by 477.8% and delivers +1.75% effective consumption and +13.09% new-user exposure-to-conversion online.
[IR-8] SwapRec: Warming Up Cold Items Through Training-Time Swaps RECSYS2026
链接: https://arxiv.org/abs/2609.00913
作者: Marta Moscati,Jan Malte Lichtenberg,Davide Abbattista,Antonio De Candia,Laura Boggia,Matteo Ruffini
类目: Information Retrieval (cs.IR); Multimedia (cs.MM)
备注: Accepted at DaQuaMRec @ RecSys 2026: Second International Workshop on Data Quality-Aware Multimodal Recommendation
Abstract:Interactions with cold items negatively impact real-time personalization of ID-based recommender systems. This is because the use of such interactions degrades user preference estimates, whereas excluding cold items from the user profile prevents real-time recommendation updates. In industrial scenarios, one heuristic often applied to address this shortcoming at inference time is to replace, i.e., “swap”, cold-start items by their most similar “warm” neighbor, where similarity is inferred from the items’ side information. In this paper, we demonstrate that sequential models, most often used for real-time personalization, are not robust to such swaps, and propose SwapRec, an approach to address this issue. SwapRec relies on using the same swap heuristics already at training time. We apply SwapRec to state-of-the-art models for sequential recommendation and analyze its impact by means of quantitative experiments in three recommendation domains (online shopping, movie, music). The experimental results show that, irrespective of the underlying sequential architecture, our easy-to-implement SwapRec approach allows for substantially more accurate recommendations when in presence of interactions with cold items, simultaneously leading to a larger percentage of cold items in the recommendation lists.
[IR-9] Staged Linguistic Seeding: Grounded Query Expansion for Verified-Unit QA in AI Contact Centers EMNLP2026
链接: https://arxiv.org/abs/2609.00844
作者: Hyeonseop Yoon,Jeong-Eun Park
类目: Computation and Language (cs.CL); Information Retrieval (cs.IR)
备注: 14 pages, 1 figure, 7 tables. Accepted to the Grounding Language Models (GroundLM) Workshop at EMNLP 2026
Abstract:Customer-service QA in an AI contact center (AICC) runs under deployment constraints that benchmark QA misses: tight voice-hotline latency and a high cost for unsupported or wrong automatic answers. We deploy a system that answers only from a closed set of verified QA units: it returns a retrieved unit verbatim, or routes to clarify, abstain, or handoff. The index is enriched offline by staged linguistic seeding (SLS): a human authors a per-unit world-grounded slot recipe, gpt-4.1-mini renders it into variants, and a light human gate filters them. One methodology is reused across both domains, so inference stays a single retrieval pass with no query-time generation. On held-out query variants from two industrial domains, SLS lifts hybrid R@1 to 0.881/0.930 (+0.27/+0.34), with gains across all five retrievers tested. At the same gpt-4.1-mini generation budget, SLS beats doc2query by +0.20/+0.32, while cross-provenance evaluation provides additional evidence of transfer across generated-query distributions. Verified-unit answering also removes free-form generation’s unsupported-content surface (7-13% versus approximately 0%). We report this as an application study, including negative results.
[IR-10] Ctrl-F-Resist. Practices Challenges and Technical Needs of Civil Society Organizations Monitoring the Far-Right Online
链接: https://arxiv.org/abs/2609.00808
作者: Elisabeth Steffen,Helena Mihaljević
类目: Human-Computer Interaction (cs.HC); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Information Retrieval (cs.IR)
备注: Accepted for the 29th ACM Conference on Computer-Supported Cooperative Work and Social Computing (CSCW 2026)
Abstract:As far-right actors increasingly exploit online platforms to disseminate ideology and mobilize supporters, civil society organizations (CSOs) play a vital yet underrecognized role in monitoring antidemocratic dynamics online. Unlike fact-checkers or content moderators, CSOs engage in long-term, contextualized analysis, often in resource-constrained settings and under precarious conditions. Despite their critical societal role, CSOs face significant barriers to adopting or co-developing technical solutions, including legal uncertainty, limited platform access, and chronic underfunding. Existing research and tool development efforts have largely overlooked these actors in favor of more institutionally embedded stakeholders. This paper addresses this gap through a qualitative study with 15 practitioners from 12 Germany-based CSOs engaged in online monitoring, positioning them as key yet overlooked stakeholders in the governance of digital spaces. We explore their current practices, challenges, and expectations regarding technological support. Our findings show that monitoring remains largely manual due to the lack of tailored tools, with enhanced search capabilities emerging as the most pressing technical need. While participants express openness to AI-supported features such as media processing and content discovery, many remain skeptical of automated classification, citing concerns around trust, legal usability, and professional credibility. Grounded in these findings, we introduce a conceptual monitoring workflow and describe its implementation in an open-source Telegram monitoring prototype designed to flexibly support diverse monitoring goals. We outline concrete design, policy, and research recommendatios, and introduce the manual labor trap as an empirically grounded concept that explains why monitoring CSOs tend to remain locked into labor-intensive, low-capacity arrangements.
[IR-11] From Saliency to Discriminability: Rank-Preserving Visual Token Pruning for VLM Rerankers EMNLP2026
链接: https://arxiv.org/abs/2609.00667
作者: Siyi Liu,Hanjun Yang,Chenchen Zhang,Xiaorong Zhu,Xinyu Zuo,Lisheng Duan,Haijin Liang,Jin Ma,Junfu Pu,Yongqi Zhang
类目: Information Retrieval (cs.IR); Computer Vision and Pattern Recognition (cs.CV)
备注: EMNLP 2026 (Main Conference)
Abstract:Large vision-language models used as listwise rerankers must jointly process visual tokens from tens of candidates per query, making token pruning essential for practical deployment. Existing pruning methods retain tokens by attention saliency, yet we show that saliency is systematically misaligned with ranking contribution: visually prominent tokens often capture order-neutral patterns shared across candidates. This mismatch is layer-dependent: saliency becomes informative only where attention is concentrated, and normalized attention entropy diagnoses the reliability shift (Pearson r=0.87). We propose RaDiCal (Rank-Discriminative Calibration), a training-free framework that uses normalized attention entropy to decide when saliency can be trusted, fusing it with an attention-free rank-discriminative prior and selecting pruning layers from the same trust landscape. Across three retrieval benchmarks and multiple VLM architectures, RaDiCal matches Dense MRR@10 on Flickr30K and surpasses it on MSCOCO at a 20% token budget, ranks first among all pruning methods on FashionIQ, and holds within 1.2 pp on Flickr30K and MSCOCO at 10% retention. It cuts FLOPs by 39–45% and delivers 1.28–1.45 \times measured speedups across two VLM architectures without dataset-specific retuning.
[IR-12] It Takes Two to Match: Co-Evolving Generative Retriever with Reinforcement Learning
链接: https://arxiv.org/abs/2609.00638
作者: Runpeng Dai,Kaili Huang,Changsung Kang,Ciya Liao
类目: Information Retrieval (cs.IR); Computation and Language (cs.CL)
备注:
Abstract:Retrieval is the first stage of modern search and advertising systems, selecting a candidate set from a large item universe for downstream ranking and auction. Recent work increasingly leverages LLMs to improve retrieval through query expansion, data synthesis, and retrieval-feedback training. However, the generative component is typically used for query-side augmentation, while final matching is still delegated to a downstream retriever. We introduce CoGR, a retrieval framework that instead trains LLMs to directly construct retrieval representations on both query and item sides. Each generator produces a compact set of keywords, which are matched directly through an inverted index, preserving compatibility with existing keyword-based retrieval infrastructure. CoGR uses a two-stage training pipeline. Supervised fine-tuning first establishes an aligned keyword space, after which co-evolving reinforcement learning alternately optimizes the query- and item-side generators with GRPO against the opposite side’s frozen index. Both sides optimize the same query-to-item retrieval F_1 objective: the query side receives retrieval F_1 directly, while the item side receives a counterfactual marginal reward measuring the change in query-side F_1 caused by its generated keywords. Across 10 representative sparse, dense, and generative baselines, CoGR achieves the best performance on both an internal APP Marketplace dataset and the public WANDS benchmark, improving F_1 over the strongest baseline by 10.9% and 36.1% , respectively. Further analysis shows stable co-evolution and increasingly aligned query–item keyword spaces over training.
[IR-13] owards Effective Structured Context Modeling for Conversational Recommender Systems via Dual-node Monte Carlo Tree Search
链接: https://arxiv.org/abs/2609.00618
作者: Jincheng Zhang,Chen Huang,Wenqiang Lei,See-Kiong Ng,Yang Deng
类目: Information Retrieval (cs.IR); Artificial Intelligence (cs.AI)
备注:
Abstract:We investigate the role of conversational context modeling in user preference tracking for Conversational Recommendation Systems (CRSs). In this regard, we propose DREAMS, a novel tree-structured context modeling framework that explicitly captures user preference evolution throughout multi-turn interactions. DREAMS introduces two specialized node types to support the two fundamental objectives of CRSs: preference elicitation and preference exploitation. Specifically, elicitation nodes leverage Monte Carlo Tree Search (MCTS) to strategically explore conversational actions and infer latent user preferences, while exploitation nodes employ LLM-based refinement to transform the tracked preference state into structured retrieval queries for recommendation. Extensive experiments on benchmark datasets demonstrate the effectiveness of DREAMS and its design.
[IR-14] NeuroGraph: An AI Graph-Driven Neuro-Symbolic Framework for Explainable Threat Reasoning in Advanced Manufacturing
链接: https://arxiv.org/abs/2609.00604
作者: Padmeswari Nandiya,Ahmad Mohsin,Ahmed Ibrahim,Iqbal H. Sarker,Helge Janicke
类目: Cryptography and Security (cs.CR); Information Retrieval (cs.IR)
备注: Submission to Springer:Journal of Information Security (Peer Review)
Abstract:The growing complexity of cyber-physical attack surfaces in advanced manufacturing has made cyber threat intelligence analysis increasingly difficult. Although large language models and retrieval-augmented generation have improved CTI workflows, text-based approaches remain vulnerable to hallucinations and provide limited support for structured reasoning over interconnected threats. Graph-based RAG reduces some of these limitations, but existing approaches often lack ontology-consistent multi-hop reasoning and transparent evidence tracing across heterogeneous cybersecurity data. This paper proposes a graph-grounded neuro-symbolic framework that integrates ontology-aware symbolic query generation, knowledge graph retrieval, and neural language generation to support accurate and explainable threat analysis across information technology and operational technology environments. The framework adopts a dual-large language model architecture: the first model translates natural-language questions into executable Cypher queries for symbolic graph retrieval, while the second generates answers strictly from the retrieved graph evidence. Experimental evaluation using publicly available cyber threat intelligence benchmarks shows consistent improvements over the published baseline in reasoning accuracy, while also reducing hallucinations, strengthening multi-hop reasoning, and improving robustness to adversarial perturbations. Runtime and explainability analyses further demonstrate that the framework maintains interactive inference performance and exposes graph-grounded reasoning artifacts that allow analysts to inspect and verify each stage of the analysis. Overall, the results highlight the potential of graph-grounded neuro-symbolic reasoning as a scalable, interpretable, and reliable approach to cyber threat intelligence for next-generation Industry 5.0 environments.
[IR-15] RIS: A Tri-Layer Retrieval Integrity Sieve Against Knowledge Poisoning EMNLP2026
链接: https://arxiv.org/abs/2609.00470
作者: Muhaimin Bin Munir,Akib Jawad Ononto,Nazia Shehnaz Joynab,Bhavani Thuraisingham,Latifur Khan
类目: Computation and Language (cs.CL); Cryptography and Security (cs.CR); Information Retrieval (cs.IR)
备注: 15 pages, 2 figures, 10 tables. Accepted to Findings of EMNLP 2026
Abstract:Retrieval-Augmented Generation (RAG) grounds large language models in external corpora, but implicit trust in retrieved documents creates a critical attack surface: PoisonedRAG shows that a handful of crafted passages can dominate dense retrieval and steer generation toward attacker-chosen answers. We present the Tri-Layer Sieve, a middleware defense that sanitizes retrieved evidence through cross-embedding-space clustering with an independent judge model, structural filtering of trigger-payload artifacts, and LLM consistency verification. The design exploits a key weakness of retrieval-stage poisoning: a single document must satisfy one embedding geometry, one internal Trigger-Payload structure, and one generation objective - rarely all three simultaneously, a fragility that persists even against an adaptive attacker who paraphrases around it. On Natural Questions, HotpotQA, and MS-MARCO with Contriever retrieval (k=50), the Sieve reduces black-box Attack Success Rate from 67.0/87.0/64.0% to 3.0/14.0/4.0%, mitigates white-box HotFlip attacks from ~74% to 27.8% on NQ with Layer 3 enabled, and drives poisoned-document MRR to 0.000, while restoring clean accuracy from 13-33% under attack to 58-76%. Under an architecture-aware adversary who paraphrases triggers to evade the structural filter, enabling the consistency layer halves adaptive ASR (32.0% to 15.0% on NQ) while raising clean accuracy by 18 points, at an added latency of ~16-19 s/query under live retrieval.
[IR-16] Closed Forms and Synthetic Twins: Predicting Approximate Nearest Neighbor Recall from Embedding Statistics
链接: https://arxiv.org/abs/2609.00364
作者: Shmuel Herman
类目: Information Retrieval (cs.IR)
备注:
Abstract:Embedding models are trained and evaluated as if retrieval were exact; in production they serve behind approximate indexes – HNSW, IVF, product quantization, or the fixed-dimensional encodings (FDEs) of late-interaction models – whose behavior the encoder’s benchmarks never see: one modern encoder recovers just 14% of its exact top-10 through its raw FDE index. Such failures surface only after an index is built, and the standard patches – corpus-fitted transforms such as whitening – must be fitted, stored, and refit as the corpus changes, and can silently rewrite what the encoder returns. This paper shows that index behavior is predictable before anything is built, from label-free statistics of the raw embeddings, through a ladder of instruments matched to what each index family consumes: (1) closed-form moment statistics for the fixed-grid quantizers (PQ, FDE); (2) simulation on a synthetic twin corpus – cluster statistics made generative, on which any index, composed production systems included, can be built and tested – for partition indexes; (3) size-extrapolated, lightly calibrated twins for graph indexes at million-document scale. Predictions land within 0.03 of measured recall on an unseen million-document corpus. The same geometry is trainable: targeting the one statistic no post-hoc transform can move – the score margin – lifts recall for every index family at once, at a small measured task cost. The result: index choice, correction pricing, and production recall forecast from one cheap measurement pass, on new corpora and new indexes alike; serving without per-corpus transform machinery, suited to continuously changing corpora; and a recall-compute frontier pushed by adapting encoders to geometry rather than coupling them to any single index.
[IR-17] Sources of Truth: A Multi-Platform Multilingual Audit of Citations in AI Mental Health Information Queries
链接: https://arxiv.org/abs/2609.00319
作者: Phuong Anh Nguyen,Jill Noorily,Matthew Flathers,Haruka Notsu,Laura Ospina-Pinillos,Tommy Nguyen,Samantha Clark,Aoife Keane,Grace Thompson,John Torous
类目: Computers and Society (cs.CY); Computation and Language (cs.CL); Information Retrieval (cs.IR)
备注: 28 pages (16-page main text plus supporting information), 5 figures, 5 tables. Under review. Code: this https URL Data: this https URL
Abstract:Online health information seeking is shifting from keyword search, where users consider a ranked list of links, to conversational systems that compose a single answer and curate its citations. Source evaluation therefore passes from user to platform, yet what these systems surface is poorly characterized. We audited three free consumer products (ChatGPT, Perplexity, Google AI Overview) on twenty English mental health questions under two prompt conditions, with a subset of three also translated into six further languages of varying resource tiers. We recorded 15,942 citations across 1,140 responses and 1,713 unique domains, then classified every citation with a nine-category organizational typology applied by a deterministic classifier validated against human coding. Citations were heavily concentrated: the ten most-cited domains accounted for 43.6% of English citations, and government, commercial health, and academic sources were closely matched at roughly 22% each. Platforms differed little in typical citation volume but sharply in consistency and in the source types they favored. Explicitly requesting sources shifted composition only modestly. Non-English queries surfaced fewer citations and were routed to language-appropriate resources at significantly lower rates. We release the typology, classifier, and annotated corpus as reusable instruments for auditing generative health search.
[IR-18] MUSES: A Benchmark for Prospective Intellectual-Roots Retrieval
链接: https://arxiv.org/abs/2609.00313
作者: Rohan Pandey,Sunjae Kwon,Hong Yu
类目: Information Retrieval (cs.IR)
备注:
Abstract:Scientific discovery depends on finding prior literature that shapes what comes next. Existing retrieval systems optimize for relevance and popularity, often favoring central papers over less familiar works that later prove generative. We introduce \textbfMUSES, a million-instance benchmark for prospective intellectual-roots retrieval over a fixed 2.33M-paper corpus, with roughly 140K test instances per familiarity tier. To our knowledge, it is the first prospective benchmark at this scale with a shared retrieval task and author-confirmed paper-level root labels. Alongside it, \textbfCiteRoots pairs a scalable rhetorical layer over local citation text (LLM judge \kappa = 0.896 versus human gold) with a paper-level author-endorsed layer ( n = 1,518 generative-inspiration pairs from 753 focal papers). MUSES organizes difficulty along two axes: a \emphfamiliarity axis spanning CiteNext, CiteNew, and CiteNew-Isolated, and a \emphfunctional axis spanning broad citations, rhetorical roots, and author-endorsed roots. Across 9 method classes, a lean multi-centroid retriever built on SPECTER2 is strongest. Hit@100 falls from 0.534 on CiteNext to 0.424 on CiteNew, 0.205 on rhetorical CiteNew, and 0.171 on author-endorsed CiteNew, a 3.1\times decline. In a registered eight-lens full-test audit, roughly half of broad-tier test instances remain unsolved at K=1,000. Rhetorical role and author endorsement are distinct: the same judge agrees with endorsement at \kappa = 0.037 . We release MUSES, both CiteRoots layers, and a distilled open companion judge for future work on prospective retrieval and intellectual roots.
[IR-19] wo-Sided State-Space Models for Sequential Recommendation with Non-Random Multimodal Review Feedback EMNLP2026
链接: https://arxiv.org/abs/2609.00165
作者: Ziwen Pan,Zihan Liang,Ruoxuan Xiong
类目: Information Retrieval (cs.IR)
备注: Accepted to Findings of EMNLP 2026
Abstract:Two-sided digital platforms are inherently dynamic: user preferences shift, item popularity evolves, and reviews both reflect and drive these changes. Yet most sequential recommendation systems treat reviews as passive signals for updating user states, leaving two aspects underexplored. First, review generation is nonrandom, depending on evolving latent states of both users and items. Second, reviews can reshape item states, induce spillover across related items, and influence future user decisions. To address these gaps, we propose a two-sided state-space model (TS-SSM) for event-conditioned sequential recommendation. TS-SSM consists of three components: (1) a modality-missing-not-at-random fusion module that encodes review content and informative observation patterns; (2) user-state evolution with temporal variation and local graph message passing that uses related item states to refine user preferences; and (3) item-state evolution with asymmetric carryover of positive and negative review feedback. In experiments across six Amazon categories, TS-SSM increases Recall@20 over BSARec by 14.8%–18.8% and exceeds HM4SR by 11.7% on average. On Goodreads Fantasy, Recall@20 improves HM4SR from .5191 to .5847. Ablations highlight distinct contributions of observation patterns, local propagation, and item dynamics.
[IR-20] SilentProbe: Measuring Silent Failure in Production APIs Used as Agent Tools
链接: https://arxiv.org/abs/2609.00035
作者: Zongrong Li,Shengkun Ye,Feiyou Guo,Zuoyou Dang
类目: Information Retrieval (cs.IR); Software Engineering (cs.SE)
备注: 12 pages, 9 figures. Code and data: this https URL
Abstract:An LLM agent calling a production API cannot distinguish a query that matched nothing from a query the server did not understand. Both return HTTP 200 with a parsable body, no exception to catch and no field to branch on. We ask what predicts which one occurred, and what it does to the agent. Auditing 721,320 parameters across 2,501 independently published OpenAPI documents, we find that 7.5% declare an enumeration and 15.2% declare any machine-checkable constraint at all, while 40.1% of documents state at least one constraint in prose that their schema does not encode. Executing 219 schema-derived perturbations against live commercial endpoints from 27 vendors, reached through a single aggregation layer (Monid) that publishes a schema and returns a run identifier for every call, we find that constraint form, not vendor identity, predicts honesty: machine-checkable constraints yielded an honest error in 111 of 111 cases, prose-only constraints failed silently in 44 of 61 (p = 2e-13). Twelve models across eight families then met these endpoints on ordinary tasks. A vocabulary that the description merely exemplifies was missed by every model on 88 of 88 attempts, while vocabularies written out in full were used correctly 88 to 91% of the time. Running the full agent loop, models detected the resulting silent failure in 12% of cases, repaired it in 0%, asserted a false negative to the user in 41%, and invented a figure in 12%. Promoting the vocabulary into the schema removes the failure, from 88 of 88 to 0 of 89. The fix is one line of schema rather than a better model. Code, schemas, perturbation sets, agent transcripts and per-call run identifiers are released at this https URL.
人机交互
[HC-0] Designing Proactive Thought Partners for Writing
链接: https://arxiv.org/abs/2609.01588
作者: Chao Zhang,Abe Davis,Chih-Wei Chen,Chin-Chia Hsu
类目: Human-Computer Interaction (cs.HC); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 30 pages, 5 figures
Abstract:Writing involves diverse cognitive activities, from ideation to revision, and writers’ needs vary across individuals and moments. Proactive AI promises to provide the right support at the right time, yet existing proactive tools largely focus on generic textual assistance, such as autocomplete. This paper studies the design space of proactive thought partners: AI agents that proactively offer customizable, higher-level cognitive support during writing. We instantiated this concept in a technology probe and deployed it with 16 participants for one week. The probe allows users to create partners by configuring their roles and proactivity. As users write, relevant partners take the initiative at appropriate moments to offer suggestions. Our findings show that participants configured proactive support through prospective planning, used suggestions for both idea generation and self-monitoring, and valued lightweight visual representations alongside non-directive rhetorical framing for non-intrusive interventions. We derive implications for designing proactive writing assistants around customization, timing, engagement, and representation.
[HC-1] Evaluating Usability in Biomedical Visualization: Rethinking Heuristic Evaluation for Spatial Omics and Multidisciplinary Research Platforms
链接: https://arxiv.org/abs/2609.01569
作者: Yulia A. Levites Strekalova,Rachel Liu Galvin,Jessica M. Ray,Samuel P. Border,Mishal Khan,Samantha Hoffman,Christina D. Beharry,Katie Kloss,Philipp Haessner,David Manthey,Sanjay Jain,Michael T. Eadon,Laura Barisoni,Pinaki Sarder
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:Introduction: Clinical research informatics (CRI) platforms support biomedical discovery by integrating advanced computational tools into research workflows. Emerging technologies such as spatial omics and AI-enabled imaging expand research capabilities but introduce complex interfaces that increase cognitive burden and alter established analytical processes. Traditional usability frameworks identify general usability issues but often miss challenges specific to high-dimensional biomedical data. Methods: We conducted two complementary studies involving 39 participants to evaluate conventional usability heuristics and identify CRI-specific criteria. Study 1 included 19 undergraduates completing interactive tasks, and Study 2 involved 20 clinical professionals completing an asynchronous hierarchical task framework. Observational and interview data were analyzed using deductive coding based on standard usability heuristics and emerging CRI-specific themes. Results: Simultaneous presentation of complex data overlays and analytical tools overwhelmed users, particularly those with limited spatial-omics experience. Participants relied on trial-and-error exploration and struggled with unlabeled tools in data-rich environments. Feedback indicated that users benefit from phased onboarding, contextual guidance, and progressive feature introduction rather than immediate access to all functionality. Discussion: High-dimensional research platforms require domain-specific usability criteria beyond traditional frameworks. We propose three specialized heuristics: Active Parameter Transparency, Point-of-Use Guidance, and Phased Feature Disclosure. These heuristics help developers manage complexity, provide contextual support, and improve accessibility for multidisciplinary research teams.
[HC-2] Better Situational Awareness in AR-HRC? A Comparative Study of Augmented Reality and Mobile Interfaces for Human-Robot Collaboration
链接: https://arxiv.org/abs/2609.01461
作者: Zhehan Qu,Christian Fronk,Jaewoong Jeong,Pavel Manakhov,Maria Gorlatova
类目: Human-Computer Interaction (cs.HC)
备注: 11 pages, 9 figures, 1 table. Zhehan Qu and Christian Fronk contributed equally
Abstract:Augmented reality (AR) facilitates human-robot collaboration (HRC) by enabling in-situ spatial visualizations of the robot and the joint task. However, in safety-critical HRC scenarios such as search-and-rescue, spatial visualizations may also reshape visual attention in ways that create competing situational awareness (SA) demands, potentially introducing new safety concerns. While prior AR-HRC work suggests potential benefits for SA, rigorous evaluations that jointly consider robot and environmental awareness across multiple levels of SA remain limited. We address this through a between-subjects study with 30 participants comparing custom AR and mobile interfaces presenting equivalent information, measuring robot and environmental SA with the Situation Awareness Global Assessment Technique (SAGAT) across all three levels, with concurrent eye tracking to identify the attentional mechanisms underlying any SA differences. Both interfaces achieved high usability; relative to the mobile baseline, AR improved perception-level awareness of the robot but yielded no gains in higher-level robot awareness or in environmental awareness at any level. Gaze analysis explained this: AR freed attention from the map, but that attention was re-invested in the conformal visuals rather than the physical environment. Freeing the eyes from a screen is not the same as directing them to the world, a distinction AR interfaces for safety-critical HRC must design around.
[HC-3] Cross-Modal Guidance for Out-of-View Object Search in Simulated Prosthetic Vision
链接: https://arxiv.org/abs/2609.01438
作者: Adyah Rastogi,Apurv Varshney,Tobias Höllerer,Michael Beyeler
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:Out-of-view guidance is well established in virtual and augmented reality, but its effectiveness may depend on the visual bandwidth available to the user. We test this under simulated prosthetic vision (SPV), where visual guidance must share the same sparse representation used to inspect the scene. Nineteen participants performed object search under two SPV conditions differing in electrode density and phosphene spread (10x10 and 20x20) and four guidance conditions (no guidance, visual, haptic, audio) all driven by the same horizontal target-offset variable. All three modalities reduced search time and head movement. The tested auditory and haptic cues produced approximately 25% faster overall search and 11-13% faster target acquisition than the visual cue, despite similarly direct orienting trajectories. The tested haptic and auditory cues also shortened post-acquisition search. Final head-target angular offset was reduced substantially more in the 10x10 SPV condition; there, all three cues also reduced vertical localization error by approximately 45-58% despite providing no elevation information. Under severe visual constraints, guidance performance depended on cue implementation and search stage.
[HC-4] InSight: A Benchmark for Agent ic Claim Verification in Interactive Visualizations EMNLP
链接: https://arxiv.org/abs/2609.01383
作者: Maeve Hutchinson,Syed Mahbubul Huq,Mohammad Albinhassan,Radu Jianu,Aidan Slingsby,Pranava Madhyastha
类目: Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV); Human-Computer Interaction (cs.HC)
备注: To be presented at EMNLP Main Conference
Abstract:Vision Language Models have demonstrated remarkable proficiency in interpreting static visual artifacts, but modern data analysis is inherently dynamic, requiring the active interrogation of interactive environments. Existing benchmarks are predominantly constrained to static imagery and one-shot question answering and fail to capture the epistemic demands of this domain, where evidence is frequently occluded, distributed across linked views, or conditionally revealed through user agency. In this paper, we introduce InSight, a benchmark for agentic claim verification over interactive visualizations. The dataset consists of 21,349 claims derived from human-authored analytical narratives and grounded in fully interactive web-based environments. Agents must navigate these environments to determine whether a natural language claim is supported, refuted or not verifiable given the available evidence. Unlike traditional evaluations, InSight treats interaction traces as intrinsic proxies for reasoning, enabling a rigorous audit of how models seek and synthesize visual evidence. We evaluate state-of-the-art models, revealing that interactive verification remains a non-trivial challenge. We release InSight at this https URL.
[HC-5] Beyond Technological Solutionism: Rethinking XR in Healthcare ALT
链接: https://arxiv.org/abs/2609.01028
作者: Md Haseen Akhtar,Cecilia Landa-Avila,Shital Desai,Claire Warden,Andrew Morris,Mark Anderson,Thomas Cochrane,Janakarajan Ramkumar
类目: Human-Computer Interaction (cs.HC); Emerging Technologies (cs.ET)
备注: 5 pages, 1 figure, Accepted paper at the Interactive Health workshop In Proceedings of CHI 25 Workshop on Envisioning the Future of Interactive Health, April 27th, 2025, Yokohama, Japan, ACM, New York, NY, USA, 2 pages
Abstract:The healthcare industry’s enthusiastic adoption of Extended Reality (XR) technologies obscures a concerning reality: we were building increasingly sophisticated ways to perpetuate fundamentally broken healthcare systems. Through three deeply personal narratives - a rural patient cut off from care infrastructure, an urban professional navigating fragmented services, and a first-generation immigrant confronting cultural barriers - this provocation paper exposes how our obsession with technological innovation often worsens rather than resolves healthcare disparities. By applying the SEIPS 3.0 model to examine diabetes-CVD care coordination, we identify an “innovation paradox” where advanced technology creates new barriers to effective care. Our care interdependencies framework reveals that healthcare outcomes are shaped primarily by human relationships (50-60%), organizational coordination (25-30%), and sociocultural factors (15-20%), not technological sophistication. This research challenges the HCI community to confront its role in perpetuating healthcare inequities, demands a fundamental rethinking and proposes a new framework for healthcare innovation that prioritizes human relationships over technical capability, systemic change over feature sets, and actual care delivery over technological ambition. For healthcare providers, technology developers, and policymakers, our findings suggest that effective care coordination requires us to step back from our techno-solutionist mindset and engage
[HC-6] Disclosure-Gated User Simulation for Companion-Agent Evaluation
链接: https://arxiv.org/abs/2609.00982
作者: Yao Liu,Yu He
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Human-Computer Interaction (cs.HC)
备注:
Abstract:Using a large language model to play the user is now standard in scalable evaluation. It has a repeatedly diagnosed failure: the simulated user is excessively cooperative, so a system under test can score by the sheer number of questions it asks rather than by making the user willing to speak. We answer with a disclosure gate conditioning information release on the companion agent’s behaviour: its state is a ladder of five ordered gates, merged onto three observable depth layers. We specify, ablate, and audit it, and train a user simulator against that specification. Gating behaviour is learned from the training corpus’s synthetic branch, while the real branch supplies how people speak and react; after training, the simulator need not be told at runtime which gate each item sits behind. The gate is a load-bearing component of the environment: on the English corpus of a published companion-agent benchmark (CompanionBench), once training no longer states per example which gate each item sits behind, the largest rank displacement across 12 systems under test exceeds the noise band set by re-running that environment under a new seed, while per-system scores show no detectable change. We state two acceptance criteria: a ranking must be order-preserving, and absolute scores must be scale-stable. Of the candidates we examine, only one passes both – the simulator we release – and its leaderboard correlates at 0.993 with the benchmark’s original simulator. By contrast, prompting a frontier model as the simulator barely moves the ranking while shifting every score upward – a shift invisible to anyone checking the ranking alone. The environment we specify is the one that benchmark already used. That publication describes the mechanism in about four hundred words, and we supply what it lacked: specification, ablations, human studies, negative controls, and downstream sensitivity analysis.
[HC-7] EIDAN: A Multilingual Multiparty Dialogue Corpus
链接: https://arxiv.org/abs/2609.00802
作者: Taiga Mori,Koji Inoue,Mikey Elmers,Divesh Lala,Tatsuya Kawahara
类目: Computation and Language (cs.CL); Human-Computer Interaction (cs.HC)
备注: 8 pages, 1 figure, 3 tables. To appear in the Companion Proceedings of the 28th ACM International Conference on Multimodal Interaction (ICMI Companion '26)
Abstract:Multi-party interaction is a central setting for human communication and a necessary target for human-agent interaction systems that must participate in group conversation. Yet available corpora often focus on meetings, task-oriented interaction, text-based interaction, or acted scenarios, and fewer resources support cross-linguistic comparison of spontaneous face-to-face triadic discussion. This paper presents TEIDAN, a multilingual multimodal corpus that currently consists of Japanese and English three-party conversations. TEIDAN records groups of three participants discussing open-ended topics with individual pin microphones, a microphone array, and participant-facing cameras, and provides IPU-based transcripts for both language portions. Earlier studies used subsets of the Japanese portion for task-specific benchmarks in multi-party dialogue modeling; in contrast, this paper presents TEIDAN as a corpus resource spanning both Japanese and English, with planned expansion to additional languages. We describe the collection design, participants, recording setup, transcription format, and corpus statistics, and provide preliminary analyses to illustrate how TEIDAN can support research on turn-taking, addressee recognition, and multimodal grounding in human-human and human-agent interaction.
[HC-8] No Pixel Left Behind: Filling Gaps in Anime Colorization
链接: https://arxiv.org/abs/2609.00800
作者: Masahiro Kono,Akinobu Maejima,Yuki Koyama,Yotam Sechayk,Takeo Igarashi
类目: Human-Computer Interaction (cs.HC); Computer Vision and Pattern Recognition (cs.CV); Graphics (cs.GR)
备注: 19 pages, 20 figures. Published in the Proceedings of the 2026 CHI Conference on Human Factors in Computing Systems (CHI '26)
Abstract:Animation production workflows often involve digital colorization of line art, where small unpainted regions (“gaps”) frequently occur and remain an underexplored challenge. We conducted a formative study in Japanese animation (anime) pipelines and found that while the paint bucket tool is widely used for base coloring, tiny enclosed areas are frequently overlooked, resulting in time-consuming manual detection and filling. We introduce GapFill, a tool grounded in professional practices that reduces the effort of gap detection, zooming, and color selection. Our deep-learning method suggests appropriate fill colors by referencing surrounding regions, leveraging the flat-color nature of anime-style images. In a user study with 13 professional colorists, our system improved performance and usability in gap-filling tasks over conventional methods. The study also suggested that prediction accuracy alone is not the primary factor for usability, that appropriate colors can be contextually ambiguous, and that GapFill can complement existing tools depending on users’ trust in new AI-powered assistance.
[HC-9] GazeTune: Facilitating Precise Gaze-Driven Interactions with Cascaded Touch Input
链接: https://arxiv.org/abs/2609.00716
作者: Jina Kim,Eric J. Gonzalez,Yang Zhang,Sang Ho Yoon
类目: Human-Computer Interaction (cs.HC)
备注: 13 pages, 8 Figures, Accepted to UIST’26
Abstract:Eye gaze has become an essential input for spatial computing, but its coarse targeting and saccadic nature limit precision and complicate continuous interactions such as dragging, especially under user motion. Gaze+pinch has also become standard in XR for its convenience, yet mid-air gestures remain imprecise, fatiguing, and socially unacceptable. These limitations underscore the need for an approach that preserves the speed of gaze while enabling stable, fine control. We present GazeTune, a cascaded multimodal interaction technique combining gaze and touch to refine gaze-based selection and manipulation. Touch serves as a refinement channel within gaze pointing, allowing precise cursor and target control. Our work investigates how gaze-and-touch enhances dragging and mitigates Motion-Induced instability. In a study (N=20), we compared GazeTune against gaze-only and gaze-pinch methods in 2D dragging. Results show that GazeTune achieves significantly lower error with comparable execution time, validating its effectiveness and balanced trade-off between time and accuracy.
[HC-10] SoK: Motion Data Privacy in Extended Reality
链接: https://arxiv.org/abs/2609.00711
作者: Azim Ibragimov,Alina Vasina,Uliana Polshcha,Eric D. Ragan
类目: Cryptography and Security (cs.CR); Human-Computer Interaction (cs.HC)
备注:
Abstract:Extended Reality (XR) provides immersive, interactive 3D experiences. To enable these experiences, the devices must track user motion so the system can respond to actions such as grabbing, looking at, or moving an object. However, motion tracking has raised privacy concerns since it records a person’s motion patterns. These motion patterns have been studied extensively across various fields (i.e., gait identification and profiling) and have been shown to reveal sensitive information. With the adoption of XR, these patterns became easier to record and obtain than ever. This creates a fundamental privacy tension: motion tracking enables core XR functionality yet requires users to compromise their privacy. Prior systematization-of-knowledge (SoK) studies on XR privacy have examined the field broadly, with motion-related research distributed across several privacy domains rather than treated as a distinct area of study. However, XR motion privacy has gained significant momentum since the prior SoK, with the literature nearly quadrupling in size and thereby warranting a dedicated systematization of this topic. This SoK examines 134 relevant papers on privacy concerns in motion patterns recorded by XR headsets, including how adversaries can obtain users’ motion patterns, the inferences they can draw from them, and methods for protecting users. Based on this review, we synthesize a taxonomy of motion modalities, representations, and inference risks; develop an XR motion threat model; systematize the attack and defense approaches in the XR motion literature; identify gaps in the literature; and provide guidelines for future studies evaluating motion privacy mechanisms. Together, our SoK clarifies the state of XR motion privacy and provides recommendations for future evaluations.
[HC-11] Human-robot conversation with multiple participants in noisy public spaces
链接: https://arxiv.org/abs/2609.00648
作者: Divesh Lala,Yogeeswaran Muthukumaran,Vincent Fernandes,Kazushi Kato,Shota Fujiki,Zihao Chi,Masaya Iwasaki,Taiken Shintani,Megumi Kawata,Kazuki Sakai,Koji Inoue,Yuicihiro Yoshikawa,Tatsuya Kawahara
类目: Robotics (cs.RO); Human-Computer Interaction (cs.HC)
备注:
Abstract:For noisy real-world environments such as those in open public spaces, spoken dialogue systems for both autonomous robots and avatars should be carefully designed to provide enhanced speech signals. These signals can be used either for speech recognition or, in the case of an avatar system, transmitted as clean speech to a remote operator. This work proposes an audio system that can be used for both these scenarios and was demonstrated as a proof-of-concept at the 2025 World Expo in Osaka. The first scenario is an attentive listening system with the android ERICA, and the second is a conversation support system with mobile Teleco robots, with one of them acting as an avatar for a remote operator. Both systems feature multi-party conversation and use a single multi-channel microphone array. We describe how our audio system not only enhances the speech of multiple speakers in a noisy environment, but provides a form of spatial audio which allows for more immersiveness in avatar-based conversational interactions.
[HC-12] Investigating Assistant Bias in LLM User Simulators Using a Role Vector
链接: https://arxiv.org/abs/2609.00608
作者: Daeheon Jeong,Yoonjoo Lee,Eugene Choi,Sinie van der Ben,Juho Kim
类目: Computation and Language (cs.CL); Human-Computer Interaction (cs.HC)
备注: 35 pages
Abstract:LLM-based user simulators are increasingly used to evaluate autonomous agents at scale, in place of costly human evaluations. Despite this promise, these simulators exhibit “assistant bias,” a tendency to cooperate and pursue task goals. They rarely reproduce the frustration or disengagement that real users exhibit, compromising evaluation validity. Prior work outlines that this bias is baked in during model training, which role-playing prompts fail to override. We analyze this bias from model activations, extracting a user role vector by contrasting how the model represents user versus assistant perspectives on the same dialogue. We observe two findings: (i) the user direction is identifiable in activations, elicits user-like behaviors, and captures characteristics distinct from assistant traits; and (ii) although user-role activation associates with simulation realism and steering strengthens it, it can exaggerate user behaviors and override individual user profiles. Together, our findings provide a representation-level analysis of LLM user simulators, confirming that assistant bias is structurally identifiable and that user behavior can be directionally analyzed.
[HC-13] Socrates went Nuclear: Comparing Interaction Strategies for AI systems in a Learning Context using Brain Sensing
链接: https://arxiv.org/abs/2609.00584
作者: Alexandre Clin Deffarges,Nataliya Kosmyna,Pattie Maes
类目: Artificial Intelligence (cs.AI); Human-Computer Interaction (cs.HC)
备注: 11 pages, 2 figures, 3 tables, to appear at The 14th International Conference on Human-Agent Interaction (HAI’26)
Abstract:Does unrestricted AI access bypass the cognitive effort required for learning, or does it streamline knowledge acquisition? This paper reports on a study where we compare three designs for user-AI interaction in a learning context: (1) an unrestricted conversational bot like ChatGPT, (2) a pedagogically constrained bot that guides through hints without giving final answers, which we refer to as the Socratic mode; and (3) a non-conversational adaptive tutoring system that adjusts difficulty in real-time based on the user’s cognitive engagement derived from the brain signals. Fifty study participants were tasked with learning about nuclear safety protocols, a domain chosen for its zero-prior knowledge baseline. The participants progressed through an instructional video, a pre-test, an AI-driven assessment phase, which varied in the three conditions, and an immediate post-test. The nature of the questions centered primarily on factual knowledge acquisition, but it still required participants to have a global understanding of the concepts in order to answer the questions correctly. A Muse headband was used to derive the cognitive engagement of all users in all conditions. The unrestricted chatbot produced higher learning gains (delta) than both constrained modes (p .03, d 0.80), while the adaptive condition generated significantly higher EEG engagement (p = .018). The cluster analysis of chatbot usage and discussion patterns by users showed that most participants in the unrestricted-mode adopted a direct answer-retrieval strategy, while participants in the Socratic-mode initially attempted to reason through the hints before progressively disengaging. Consequently, this also suggests that the success of the unrestricted AI is not an evidence of deeper learning, but rather a result of the immediate post-test evaluation after the training phase.
[HC-14] Learning Feasibility-Aware Latent Spaces for Preference-Based Exploration of Procedural Automotive Wheel Designs
链接: https://arxiv.org/abs/2609.00527
作者: Takashi Owaki,Yuki Koyama,Tomoyasu Nakano,Takahiro Yamaguchi,Masataka Goto,Hiroyuki Sakai
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:Intelligent design interfaces that rely on preference-based optimization are most useful when their suggestions are both meaningful to users and feasible within the target domain. Procedural models offer compact and editable design spaces, but their native parameters can be entangled and can generate many invalid outputs, causing human-in-the-loop optimizers to waste comparisons. We propose an interaction-oriented representation-learning pipeline for procedural models and study it in automotive wheel design. The method first screens procedurally generated samples using geometric rules and finite-element analysis, then learns a reduced latent space from the screened subset. We further introduce supervised functional alignment, which reserves selected latent dimensions for stiffness, strength-related stress response, or weight so that search can be biased toward functionally meaningful regions. Simulation experiments show that screened reduction improves target-shape retrieval and the feasibility rate of suggestions, whereas unscreened reduction degrades both. Additional simulations show that constraining search along learned functional dimensions accelerates exploration toward target functional properties. A controlled study with 40 participants further shows that a 5D feasibility-aware space yields higher shape similarity and more feasible suggestions than the original 9D procedural parameterization. These results suggest that, for intelligent user interfaces in engineering design, the representation exposed to the user is a central part of the interaction design, not merely a preprocessing step for the optimizer.
[HC-15] Are We There Yet? Assessing Computer-Use Agents for Blind Users Accessible Interaction with Desktop Applications EMNLP2026
链接: https://arxiv.org/abs/2609.00524
作者: Satwik Ram Kodandaram,Monalika Padma Reddy,Xiaojun Bi,Jiawei Zhou,I. V. Ramakrishnan,Vikas Ashok
类目: Human-Computer Interaction (cs.HC); Artificial Intelligence (cs.AI)
备注: Accepted to the EMNLP 2026 Main Conference
Abstract:Computer-use agents are emerging as a paradigm for agentic human-AI interaction, combining language reasoning with multi-modal interface grounding to operate GUIs. Yet their effectiveness for blind screen-reader users in real-world desktop workflows remains unclear. We present a three-week diary study with 8 blind users using OLLA, a screen-reader-accessible CUA prototype, collecting 1,258 commands across 12 applications with screenshots, UI trees, model responses, and action traces. We evaluate GPT-5 during deployment and re-execute the same commands with four additional models. GPT-5 achieved the highest success rate at 52.5%. Trace analysis reveals grounding, planning, constraint-tracking, and termination failures, while interviews reveal beyond-automation needs.
[HC-16] RecalibrateGPT : AI Fatigue Resilient Conversational Interfaces
链接: https://arxiv.org/abs/2609.00506
作者: Nikhil Wani
类目: Human-Computer Interaction (cs.HC); Artificial Intelligence (cs.AI)
备注: 5 pages, 3 figures. Accepted at UIST Adjunct 2026
Abstract:Large language models are powerful, but their interfaces often devolve into a type \rightarrow read \rightarrow retype loop, creating conversational AI fatigue, cognitive load, and eventual task abandonment. To mitigate this, we present RecalibrateGPT, a system introducing five cross-turn operators (Anchor, Replay, Delta, Scope, and Steer) that each target a distinct fatigue type, recalibrating LLM responses through a structured panel by acting on the full conversation history with a single click. Users invoke these operators through the AssistiveButton in one of three operator palette layouts: Vertical, Arc, or Tablet. We conducted two pilot studies with the same 12 advanced LLM users. An initial formative qualitative study identifies a taxonomy of four fatigue types (retyping, scanning, decision paralysis, and context drift) and derives two design objectives for RecalibrateGPT. A follow-up quantitative evaluation finds it reduces perceived cognitive workload by half (NASA-TLX = 2.7) at high perceived usability (SUS = 86.5), suggesting AI fatigue is not just a model-quality issue but an interaction-flow cost that interfaces can remove.
[HC-17] UniScale: Exploring Unimanual Gesture Mapping Strategies for GazePinch-based Scaling Interaction
链接: https://arxiv.org/abs/2609.00500
作者: Kyoungwhan Mheen,Jinwook Kim,Sang Ho Yoon
类目: Human-Computer Interaction (cs.HC)
备注: 11 pages, 7 Figures, Accepted to ISMAR’26, TVCG’26
Abstract:Object scaling serves as a fundamental spatial manipulation that enables complex and productive tasks in XR environments. This paper investigates unimanual scaling techniques for XR using gaze and hand interactions. We propose UniScale, a set of unimanual alternatives to the standard bimanual pinch, allowing users to scale objects while preserving hand availability for concurrent spatial manipulations. We design five distinct mapping strategies based on physical metaphors, exploring unimanual control that varies depth, angle, micro-gestures, and finger-distance input. We then compare these techniques against a standard bimanual baseline, in which users adjust the inter-hand distance via a bimanual pinch gesture. In a user study, we evaluate their effectiveness in a 3D object scaling task under both clutching and clutching-free conditions. The results indicate that while bimanual scaling relies on clutching for stable control, unimanual techniques excel in clutching-free conditions, significantly reducing physical hand movement. From the results, we derive valuable design implications for developing efficient 3D multimodal interactions in XR.
[HC-18] MemeBridge: A Dataset for Benchmarking and Mitigating the Bidirectional Cultural Gap in Meme Interpretation
链接: https://arxiv.org/abs/2609.00491
作者: Hangxiao Zhu,Suliu Qin,Zhuoyan Li,Ming Jiang,Yu Zhang,Meng Xia
类目: Computation and Language (cs.CL); Human-Computer Interaction (cs.HC)
备注:
Abstract:Communicating across cultures is inherently challenging, especially through culturally dense and ambiguous formats like memes. While people expect large language models (LLMs) to hold promise for bridging such gaps, existing benchmark datasets often fail to capture the cultural context necessary for accurate interpretation. To address this, we introduce MemeBridge, a curated dataset centered on U.S.-originated memes, designed to capture two complementary perspectives: (1) how Chinese participants interpret these memes, and (2) how U.S. participants anticipate how people from other cultures might misunderstand them. Here, context refers to implicit cultural knowledge, including background beliefs, norms, and shared assumptions that shape meme comprehension. The dataset was constructed via a multi-stage crowdsourcing pipeline with rigorous validation, including human agreement checks and GPT-based classification verification. Each meme is annotated with sentiment, emotion, cultural significance, and knowledge type, providing rich supervision for downstream tasks. Notably, we observe that the anticipated misunderstandings from U.S. participants are often inaccurate, highlighting the asymmetries in cultural understanding and the challenges of adopting perspectives beyond one’s own. This bidirectional framing, which focuses on both expression and perception, enables more nuanced benchmarking of cross-cultural comprehension. Our probing of multiple LLMs reveals that while models developed in different cultural contexts exhibit partial cross-cultural understanding, they often struggle with sophisticated interpretations. By contrast, fine-tuning with MemeBridge improves model performance, underscoring the value of culturally grounded resources for training and evaluating LLMs in globally diverse settings.
[HC-19] Design principles to Increase Technology Self-efficacy for Older Australians with Mild Cognitive Impairment (MCI) and Older Carers
链接: https://arxiv.org/abs/2609.00480
作者: Snezna Bizilj Schmidt,Nathan D’Cunha,Stephen Isbel,Blooma John,Ram Subramanian
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:The number of people with age related physical or cognitive impairments is increasing due to the worlds ageing population. Technology has the potential to support independent living, and to achieve aged and health service efficiencies, however, there are gaps in our understanding of factors that motivate technology adoption and ongoing use by older adults, especially those with cognitive impairments. This study aims to explore motivators and enablers for technology adoption and ongoing use by older Australians with mild cognitive impairment (MCI) and their carers, to identify technology design principles and guidelines that maximise adoption. Semi structured interviews were used to gather data about individual demographics, needs, priorities, lifestyle, challenges, and experiences with technology. Results of inductive, reflective, thematic analysis indicate that a desire for independence, autonomy and quality of life motivate use of technology, and perceived technology self-efficacy and IT literacy are enablers. The Protection Motivation Theory illustrates that constant technology change is a disabler for technology adoption and sustained use, because it lowers perceived technology self-efficacy and IT literacy, and reduces confidence to use technology. Two high-level technology design principles and related guidelines are proposed, grounded in theory and aligned with Banduras four sources of self-efficacy. These design principles and guidelines are intended to increase feelings of self-efficacy and confident use of technology, while also reducing adverse impacts of technological change and encouraging sustained technology adoption by older adults with MCI to support independent living and quality of life.
[HC-20] Less Is More: Balancing Positive and Negative Space in Visual Concept Blending
链接: https://arxiv.org/abs/2609.00476
作者: Shishi Xiao,Adam J. Coscia,David H. Laidlaw
类目: Computer Vision and Pattern Recognition (cs.CV); Human-Computer Interaction (cs.HC)
备注:
Abstract:Graphic designers often blend visual concepts to communicate multiple ideas within a single image, leveraging positive and negative space to create balance, emphasis, and aesthetic appeal. While computational methods have begun to support automatic concept blending, they largely overlook the role of spatial composition in the design. To address this gap, we present an automatic pipeline that explicitly applies positive and negative space throughout the blending process. Our approach first identifies plausible regions for concept integration by combining semantic reasoning from vision-language models with geometric constraints derived from real-world examples. Conditioned on these regions, the system generates blended compositions using a hybrid pixel-vector pipeline: diffusion-based inpainting produces a fast, coarse initialization, which is then refined through vector-based optimization at the point level to ensure structural coherence and balanced semantic expression. A multimodal agent orchestrates this process as a planner and evaluator, enabling iterative improvement and interpretable control. Through an evaluation using both baseline comparisons and a user study, we demonstrate greater expressiveness, creativity, and concept recognizability by effectively leveraging positive and negative space. We further demonstrate the generalizability of our approach across diverse applications, including controllable image and infographic generation.
[HC-21] ErgoAssist: Cognition-Aware Posture Feedback in Wearable Ergonomic Systems
链接: https://arxiv.org/abs/2609.00440
作者: Sarmistha Sarna Gomasta,Bhawana Chhaglani,VP Nguyen,Prashant Shenoy
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:Prolonged digital device use has made poor posture and musculoskeletal discomfort pervasive among knowl- edge workers. Existing ergonomic wearables rely solely on posture thresholds, frequently interrupting users during high-focus moments and leading to alert fatigue and abandonment. Yet posture and cognitive load are closely coupled, and most systems remain cognitively unaware. We present ErgoAssist, a head-worn ergonomic assistant that detects poor posture using IMU-based head tracking and estimates task-induced cognitive load using a consumer-grade EEG headband for continuous everyday use. In a controlled lab study, ErgoAssist achieves 81% posture classification and 90.2% task induced cognitive load estimation accuracy under leave-one-subject-out evaluation. In a preliminary real-time deployment, cognition-aware alerting reduces alert frequency by 81%, improves perceived usability by 43%, task performance by 25%, and improves posture correction rate by 38%, delivering fewer but better-timed interventions rather than merely suppressing alerts.
[HC-22] FocusBuddy: Encourag ing Healthy Desk-Work Habits by Caring for a Virtual Pet on a Water Bottle
链接: https://arxiv.org/abs/2609.00412
作者: Mohamed Ouf,Rowan Hussein
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:People who study or work at a desk sit for long uninterrupted periods and drink less water than they intend to. Software reminders address both problems but are easy to dismiss and easy to resent. We present FocusBuddy, a proof-of-concept fabric case that wraps a standard water bottle and houses a microcontroller, environmental sensors, and a small display showing a virtual pet. The pet’s condition mirrors the user’s self-care: drinking water feeds the pet, standing up to move plays with it, and refilling an empty bottle cleans it. Twenty undergraduate students used FocusBuddy for two weeks during their regular coursework and completed a written interview. Self-reported water intake rose from a median of 3 to 4 cups per day, movement episodes rose from 2 to 4 per day, and interviews surfaced two tensions: that wellness prompts must respect focused work, and that pet neglect can convert a wellness prompt into a source of guilt. We contribute the prototype, first-deployment evidence of healthy-direction shifts in self-reported habits, and design implications for emotionally framed wellness devices.
[HC-23] NeuroPriv: Adversarial Representation Learning for Privacy in Wearable EEG Systems
链接: https://arxiv.org/abs/2609.00390
作者: Sarmistha Sarna Gomasta,Bhawana Chhaglani,Prashant Shenoy
类目: Cryptography and Security (cs.CR); Human-Computer Interaction (cs.HC); Machine Learning (cs.LG)
备注:
Abstract:Wearable EEG systems may expose sensitive information beyond their intended health function, creating substantial risks to neuroprivacy. In this work, we show that commonly used EEG features can reveal participant identity and demographic attributes in addition to supporting the intended cognitive task. Wearable EEG is increasingly being explored for cognitive monitoring, neurological assessment, and longitudinal digital-health applications, yet many systems assume that transmitting compact spectral or spatial features instead of raw EEG provides sufficient privacy protection. Using EEGMAT as a motivating case study, we find that compact EEG features achieve a balanced accuracy of 0.788 for cognitive-state classification while enabling gender, age, and subject-identity inference with balanced accuracies of 0.858, 0.789, and 0.692, respectively. We further show that privacy-aware representation learning preserves task performance at 0.781 while reducing these inference accuracies to 0.563, 0.467, and 0.206. These findings motivate purpose-limited representations and explicit privacy auditing in wearable neurohealth systems.
[HC-24] MorphPatch: Enhancing VR Interaction on Shape Displays using Surface Approximation and Visuo-Haptic Illusions ATC
链接: https://arxiv.org/abs/2609.00371
作者: Wen Ying,KyeongMin Kim,Adil Rahman,JaeYoung Seon,HyeongYeop Kang,Seongkook Heo
类目: Human-Computer Interaction (cs.HC)
备注: Project page link: this https URL
Abstract:On-surface interaction in Virtual Reality improves input performance through physical support and tactile feedback, but current shape displays are constrained by limited resolution. This can misalign physical and virtual surfaces, degrading usability and user experience. We present MorphPatch, a system that enables real-time alignment between a dynamic shape display and virtual surfaces. MorphPatch uses a Signed Distance Field-based surface approximation pipeline to find practical alignments for diverse geometries. For residual discrepancies, MorphPatch incorporates pen redirection with visuo-haptic illusion to perceptually compensate for misalignment. Three evaluations show improved geometric alignment, tolerable redirection thresholds, and better control, surface guidance, and modeling results over mid-air and tablet-like interaction.
[HC-25] AniMaster: From Story Texts to Animated Videos via Cinematic Script Generation and Interactive Authoring
链接: https://arxiv.org/abs/2609.00346
作者: Ruiqi Yu,Dekun Qian,Jiale Xu,Sizhe Cheng,Yize Li,Xiangyang Wu,Zhiguang Zhou,Wei Chen,Yong Wang
类目: Human-Computer Interaction (cs.HC)
备注: 12 pages, 8 figures
Abstract:Recent advances in Video Generation Models (VGMs) have demonstrated strong capabilities in producing short video clips. However, it is still challenging for everyday creators to leverage these models to produce polished long-form animated videos from brief story texts. Informed by a formative study with both novice creators and film experts, we identify two major challenges of interactive video authoring: (1) the lack of expertise in translating free-form story texts to professional cinematic scripts and finally high-quality animated videos, and (2) the absence of effective ways to convey video design intents to key variables of visual storytelling, such as shot composition, camera controls and shot sequencing. Drawing on narratology and film studies, we propose a three-layer design framework that defines the key design dimensions across three layers (i.e., story texts, cinematic scripts, and animated videos) as well as the translation between them. Built on this framework, we present AniMaster, a VGM-powered authoring tool to enable everyday creators to easily produce smooth animated videos from free-form story texts. AniMaster automatically expands brief story texts to detailed cinematic scripts, and further translates cinematic scripts into polished videos by following professional visual storytelling principles. It also allows users to interactively edit the scripts and refine the generated videos via text instructions and intuitive interactions. We extensively evaluated AniMaster through an in-depth user study with 16 participants, two case studies, and expert interviews with 2 film professionals. The results demonstrate the effectiveness and usability of AniMaster in helping everyday creators create polished animated videos from free-form story texts.
[HC-26] Cyber-Physical Digital Factory Architecture as the Enabler of Disembodied Work
链接: https://arxiv.org/abs/2609.00195
作者: Tero Kaarlela,Ivan Ruchkin,Jose Outeiro,Souradeep Dutta
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:Digital Twins (DTs), Artificial Intelligence (AI), and Industrial Internet of Things (IIoT) technologies have significantly advanced manufacturing digitalization. However, these technologies are typically applied to individual manufacturing processes rather than integrated into a unified cyber-physical manufacturing environment. This paper proposes a cyber-physical digital factory architecture that enables disembodied work, where manufacturing systems can be supervised and operated remotely through eXtended Reality (XR) user interfaces in collaboration between AI-based control and human operators. The architecture integrates synchronized DTs, hierarchical cloud-edge AI, IIoT, and XR teleoperation interfaces into a cyber-physical manufacturing environment. The proposed approach is validated through representative manufacturing operations, including CNC machining, robotic-assisted abrasive finishing, and robotized disassembly. The results demonstrate the feasibility of the proposed architecture for disembodied manufacturing work and provide a reusable cyber-physical framework for future human-AI-controlled digital factories.
[HC-27] AI Morbidity and Mortality: A Framework for Clinical AI Failure Review
链接: https://arxiv.org/abs/2609.00076
作者: Paulius Mui,Dean F. Sittig,Steve Labkoff,Sanjay Basu
类目: Artificial Intelligence (cs.AI); Human-Computer Interaction (cs.HC)
备注:
Abstract:Clinical artificial intelligence is increasingly embedded in real-world care, yet existing safety mechanisms are poorly suited to reconstructing and learning from individual AI-related errors and near-misses. Aggregate model monitoring can identify performance changes, and traditional patient safety reporting can capture adverse events, but neither is designed to explain how risk emerges across the interaction among AI systems, clinicians, workflows, and institutional controls. We propose AI Morbidity and Mortality (AI MM), a structured, blameless framework for case-based review of clinical AI failures. The framework combines standardized case intake, evidence preservation and investigator-level reconstruction, tool-in-loop attribution, and corrective-action tracking. Each event is classified across four linked dimensions: Trigger - Mechanism - Clinical Pathway - Corrective Action, separating the condition that exposed a vulnerability from the process that produced risk, its consequence for care, and the remediation assigned. We demonstrate the framework using five illustrative outpatient medication and clinical decision-support cases; two clinician reviewers independently applied all four classification axes and reached agreement across all 20 axis-level classifications. AI MM is intended to complement, rather than replace, model monitoring, patient safety reporting, and regulatory oversight by converting individual AI-in-workflow failures into actionable institutional learning. Prospective evaluation across institutions, AI systems, and clinical settings is needed.
[HC-28] Collaboratively Eliciting Gestures for Geospatial Data Exploration on an MSE with Tangibles and Styluses
链接: https://arxiv.org/abs/2609.00007
作者: Karen Penaranda Valdivia,Nujaimah Ahmed,Aswah Butt,Mashrufa Orchi,Roozbeh Manshaei,Sarah Hoyos-Hoyos,Emmanuel Kyeremeh,Gabby Resch,Robert McLeman,Jamy Li,Ali Mazalek
类目: Human-Computer Interaction (cs.HC)
备注: Submitted to IEEE for possible publication
Abstract:Large tabletop displays and multi-surface environments offer potential for enhancing visual data exploration and collaborative work with geospatial datasets. These systems typically rely on multi-touch interactions, which can pose challenges when the multi-touch sensors misrepresent transitory movements as control inputs, leading to interruptions. Active tangibles and styluses offer an alternative to multi-touch interactions in MSEs, and have shown the potential to facilitate sense-making around large datasets. However, further research is needed to better understand how these modalities can be effectively leveraged for interacting with geospatial data visualizations. To address this, a gesture elicitation study was conducted in which users suggested interactions for 16 geospatial data visualization tasks, presented as a realistic collaborative workflow co-designed with geography and migration researchers. The study produced a taxonomy of user-defined gestures using tangibles and styluses for engaging with geospatial data, along with a thematic analysis of users’ experiences with visualization tasks and interaction techniques.
[HC-29] AMINA: The Inclusive and Accountable AI for Marginalized Immigrant Nonprofit Assistance
链接: https://arxiv.org/abs/2608.30084
作者: Maryam Mokhberi,Dipto Das,Syed Ishtiaque Ahmed
类目: Computers and Society (cs.CY); Human-Computer Interaction (cs.HC)
备注:
Abstract:Immigrant-led nonprofit groups, particularly those operating in politically sensitive contexts, face exclusion from formal registries and digital platforms. This paper reports a three-phase mixed-methods study with Iranian immigrant nonprofit practitioners: 27 semi-structured interviews, a co-design session, and 7 evaluation and feedback interviews on a prototyped AI assistant, AMINA. Our findings highlight how legitimacy barriers, capacity gaps, and politically charged misinformation constrain nonprofit operations. We translate these insights into design goals for an inclusive nonprofit AI assistant: support for everyday group operations, recognition of informal nonprofit efforts, proactive countering of misinformation, and multilingual, accessible interaction. User evaluations show AMINAs potential to reduce reporting burdens and foster transparency through proactive reminders, and catalyze collaboration across dispersed networks. We contribute to CSCW and HCI by characterizing the cooperative work of transnational immigrant nonprofits, extending scholarship on informality and misinformation, and demonstrating how AI can act as a collaborative partner that strengthens, rather than displaces, the human connections at the core of nonprofit ecosystems, while also posing major risks.
[HC-30] GazeRefine: Expert Gaze as a Test-Time Prompt for Training-Free Medical Image Segmentation MICCAI
链接: https://arxiv.org/abs/2609.01310
作者: Mohammed Oussama Benyahia,Marouane Tliba,Mohamed Amine Kerkouri,Taifour Yousra,Bin Wang,Max Bengtsson,Gorkem Durak,Elif Keles,Zuheng Ming,Marek Penhaker,Azeddine Beghdadi,Ulas Bagci,Aladine Chetouani
类目: Image and Video Processing (eess.IV); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Human-Computer Interaction (cs.HC); Machine Learning (cs.LG)
备注: 9 pages, 5 figures. Accepted at MICCAI Workshop 2026
Abstract:Medical image segmentation remains difficult to scale because high-performing methods typically rely on dense expert annotations and task-specific training. We introduce GazeRefine, a training-free framework that uses gaze as an inference-time prompt for zero-shot medical image segmentation. Sparse, duration-weighted fixations are converted into foreground and background priors that initialize semantic prototypes in frozen DINOv3 feature space. These prototypes are iteratively refined through foreground-background discrimination, feature-space affinity propagation, and anchoring to the initial gaze guidance, allowing segmentation to extend beyond directly fixated regions while limiting semantic drift. GazeRefine requires no segmentation masks, fine-tuning, adapters, prompt encoders, or gradient updates. We evaluate the method on gaze-annotated polyp segmentation and prostate MRI segmentation. The results show strong performance on colonoscopy images and competitive performance on prostate MRI, supporting gaze-guided prototype refinement as a promising approach for segmentation-label-efficient, human-in-the-loop medical image segmentation. Our tools and code can be found in the following repository: this https URL
[HC-31] A Human-AI Theorem Connecting Spontaneous and Field-Induced Mechanisms of Collective Behavior in One Dimension
链接: https://arxiv.org/abs/2609.00322
作者: Weiguo Yin
类目: atistical Mechanics (cond-mat.stat-mech); Artificial Intelligence (cs.AI); Human-Computer Interaction (cs.HC); Mathematical Physics (math-ph)
备注: 12 pages, 3 figures, 2 tables
Abstract:Can an artificial intelligence (AI) generate a scientific hypothesis outside a human collaborator’s active hypothesis space (AHS), and can human-AI research be organized to make such breakthroughs more likely? We document such a case while proving a theorem that connects two basic organizing mechanisms of statistical physics: collective behavior arising in zero field from competing interactions and that induced or controlled by an external field. A zero-field O(n) -vector open chain with arbitrary inhomogeneous nearest- and next-nearest-neighbor interaction functions U_i(S_i\cdotS_i+1) and V_i(S_i\cdotS_i+2) is microscopically, via a temperature-independent mapping at the Hamiltonian level, equivalent to a simpler O(n) open chain with nearest-neighbor interaction V_i( \sigma_i\cdot \sigma_i+1) and axial single-spin potential U_i(\sigma_i^z) for every integer n\ge1 and every system size L\ge1 . The homogeneous linear specialization maps the foundational frustrated J_1 - J_2 model onto the canonical J - h field model—with n=1,2,3 being the Ising, XY, and Heisenberg classical spin models, respectively. An analogous theorem holds when the continuous O(n) spins are replaced by the q -state Potts spins with the standard Potts interaction, implying a closed-form exact solution of the J_1 - J_2 Potts open chain for every q\ge2 and every L\ge1 . The emergence of the theorems from sustained human-AI collaboration suggests that involving AI throughout a systematic research program may incubate autonomous scientific breakthroughs.
计算机视觉
[CV-0] Uncovering Understanding-Generation Synergy in Native Unified Multimodal Models: From Representation Task to System
链接: https://arxiv.org/abs/2609.01607
作者: Penghao Wu,Haiwen Diao,Weichen Fan,Lewei Lu,Dahua Lin,Ziwei Liu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:While unified multimodal models (UMMs) jointly perform visual understanding and generation within a single model, functional unification does not guarantee learning synergy: the two objectives may reinforce each other, compete for capacity, or merely coexist. We investigate their relationship at the representation, task, and system levels in a controlled, structurally native setting without pretrained vision priors. At the representation level, we find that each objective provides useful signal to the other: generation enriches the visual features learned for understanding, while understanding strengthens vision–language alignment for generation. However, when both objectives are forced through the same computation path, one tends to dominate. A task-decoupled architecture that specializes conflicting visual computation while preserving semantic interaction avoids this asymmetric degradation. At the task level, through three case studies, we find positive bidirectional transfer when understanding and generation tasks rely on shared knowledge. At the system level, we show that an end-to-end UMM outperforms a matched planner–executor pipeline on complex tasks that explicitly require both image understanding and generation. Together, these results show that the value of UMMs extends beyond a unified interface: appropriate specialization, shared task knowledge, and end-to-end optimization can turn coexistence into synergy.
[CV-1] UI-VISA: U-Net Initialized Vascular Image Segmentation Architecture
链接: https://arxiv.org/abs/2609.01598
作者: Asees Kaur,Suzanne S. Sindi,Erica M. Rutter
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 9 pages, 6 figures
Abstract:Accurate segmentation of vascular structures in digital subtraction angiography (DSA) images remains challenging due to the thin, elongated, and branching nature of blood vessels. Pixel-wise deep learning approaches such as U-Net achieve strong general-purpose segmentation performance but often produce fragmented or discontinuous predictions in fine vascular regions, since they do not explicitly enforce structural connectivity. Region growing algorithms preserve spatial context and topological continuity, but are highly sensitive to seed point initialization and can be computationally expensive. We propose UI-VISA (U-Net Initialized Vascular Image Segmentation Architecture), a hybrid pipeline that combines the complementary strengths of both approaches. UI-VISA uses U-Net’s foreground predictions as informed seed points for a CNN-guided region growing algorithm, which then iteratively refines the segmentation by enforcing local connectivity and recovering fine vessel details that U-Net alone tends to miss or over-predict. We evaluate UI-VISA against standalone U-Net and a prior region-growing-based method (VISA) using 5-fold cross-validation on 26 DSA images. UI-VISA achieves the highest mean Dice and clDice scores across folds, and a paired Wilcoxon signed-rank test shows the improvement in clDice is statistically significant ( p=0.023 ), consistent with the method’s design goal of preserving vascular connectivity, while the improvement in Dice does not reach significance ( p=0.104 ).
[CV-2] A Benchmark for Vehicle Attribute Classification in Cross-Domain Surveillance Scenarios
链接: https://arxiv.org/abs/2609.01584
作者: Sergio M. Silva Jr.,Otavio T. Remer,Gabriel E. Lima,Lucas Wojcik,Rayson Laroca,David Menotti
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted for presentation at the 2026 Conference on Graphics, Patterns and Images (SIBGRAPI)
Abstract:Vehicle attribute analysis is a key component of Intelligent Transportation Systems (ITS), supporting applications such as vehicle identification, traffic monitoring, and forensic investigation. However, models trained under controlled conditions often degrade in real surveillance scenarios due to changes in viewpoint, occlusion, illumination, and sensor characteristics. This paper introduces Unconstrained Vehicle Identification Benchmark (UVIB), a benchmark for evaluating three operational vehicle-analysis tasks: front/rear orientation, occlusion-related suitability for Vehicle Make and Model Recognition (VMMR), and color clarity. The benchmark contains 84,835 vehicle images from seven public Brazilian datasets, grouped into surveillance and general acquisition domains, with unified binary annotations that were not jointly available in the original sources. Four representative architectures, EfficientNetV2-S, ResNet-50, ViT/B-16, and YOLO11s-cls, are evaluated under mixed-domain, cross-domain, and cross-dataset protocols. The results show that domain shift has a stronger impact than architecture choice, with substantial degradation in cross-domain settings, especially for VMMR suitability and color clarity. While orientation generalizes more reliably, VMMR suitability remains affected by class imbalance and ambiguous occlusions, and color clarity is highly sensitive to illumination and sensor modality. These findings highlight the need for benchmarks and evaluation protocols that explicitly measure operational robustness beyond standard in-domain accuracy. The proposed benchmark is publicly available at this https URL.
[CV-3] SpatialGuard: Harness-Guided Verifiable Spatial Reasoning for Text-to-Image Generation
链接: https://arxiv.org/abs/2609.01582
作者: Ziyun Qian,Zizhi Chen,Yizhou Liu,Mingyang Sun,Dingkang Yang,Lihua Zhang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Complex 3D spatial text to image generation requires models to convert natural language into stable visual geometry, not merely semantic appearance. Existing prompt-driven or layout-conditioned methods improve controllability, but often lack an optimizable and verifiable spatial intermediary before visual sampling. As a result, object relations, occlusion, visibility, and camera constraints can decay during multi-round generation. This paper presents SpatialGuard, a structured layout-guided framework for complex 3D spatial text-to-image generation. SpatialGuard parses prompts into image synthesis-oriented 3D layouts through a Spatial Layout Architect, realizes them as visual conditions and candidate images through a Visual Realizer, and uses a Visual Alignment Critic to validate consistency among prompt, layout, and image. To keep constraints stable across iterations, SpatialGuard introduces a Layout Harness that organizes rule constraints, tool invocation, shared knowledge, and feedback loops around the editable layout state. This design turns complex spatial generation from implicit prompt following into a verifiable process of planning, realization, validation, and repair. Comprehensive experiments show that SpatialGuard achieves state-of-the-art performance in complex 3D spatial layout generation and improves spatial faithfulness over existing text-to-image and layout control baselines.
[CV-4] H3-World: Turning Language Understanding into World Control
链接: https://arxiv.org/abs/2609.01560
作者: Danze Chen,Zeqing Wang,Ziyue Lin,Xingyi Yang,Yeying Jin
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:We present H3-World, an efficient framework that turns the 33B MiniMax-H3 video generator into an interactive world model. Our key finding is that, as large video generators become more capable, language is emerging as a natural interface for control. MiniMax-H3, for example, already supports zero-shot control of character behavior and camera motion through natural-language instructions. Building on this, H3-World turns this coarse language interface into precise, temporally grounded world control, without introducing dedicated action modules. Specifically, we represent each action as a structured combination of character and camera instructions, and align them with the corresponding temporal video latents. To make the control temporally precise, we further introduce temporal attention routing, which restricts each instruction to its intended time interval and reduces control leakage across actions. Importantly, H3-World directly reuses the semantic representations learned during large-scale video pretraining and requires only lightweight adaptation. With only 8,000 gameplay samples, 10,000 LoRA optimization steps, and 0.199% trainable parameters, H3-World achieves effective character and camera control while preserving strong generation quality. It also generalizes to unseen scenarios. These results show that the control capabilities emerging in large video generators can be efficiently transformed into interactive world control.
[CV-5] BS: Take the Hint - Interactive Multitracer PET/CT Lesion Segmentation with a Scribble-Conditioned ResEnc U-Net
链接: https://arxiv.org/abs/2609.01554
作者: Marven Sherif,Amgad Elmasry,Youssef Ghazal,Ayman Elghotni(Brightskies)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Automated lesion segmentation in whole-body PET/CT is complicated by the variety of physiological tracer uptake patterns and by the differing appearance of lesions across tracers. The autoPET/CT V challenge addresses this by making segmentation interactive: user scribbles marking foreground and background are supplied alongside the image, and the algorithm is expected to exploit them. We present our submission, a scribble-conditioned residual encoder U-Net operating on four input channels: CT, PET, and a sparse scribble map for each of foreground and background. The network is initialised from the autoPET-III winning weights and extended from two to four input channels, with the two scribble channels zero-initialised so that the pretrained representation is preserved exactly at initialisation. Every model is fine-tuned per fold from the corresponding autoPET-III fold checkpoint, so that no validation case is seen during pretraining. PET intensities are normalised against a per-scan aorta blood-pool reference derived from a CT segmentation, which removes tracer- and centre-specific scaling without requiring lesion labels. At inference the five fold models are ensembled by averaging their softmax outputs per sliding-window patch, before Gaussian-weighted stitching. On the challenge’s five-fold split, with each fold evaluated on its own validation cases, mean Dice is 0.554 and mean lesion-level F1 is 0.528 without scribbles, rising to 0.751 and 0.733 after five correction rounds. About 85% of that gain follows the first scribble, and the spread between fold models narrows five-fold over the same rounds, so interaction largely compensates for how well or badly a given model segments unaided.
[CV-6] What Where and How: Probing Spatiotemporal Representations in Video Foundation Models
链接: https://arxiv.org/abs/2609.01551
作者: Sharon S. Musa,Fereshteh Forghani,Harrish Thasarathan,Sonia Joseph,Matthew Kowal,Konstantinos G. Derpanis
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Interactive visualizations are available at this https URL
Abstract:Self-supervised video foundation models learn rich spatiotemporal representations, yet it remains unclear what visual concepts these representations encode, where they emerge across transformer layers, and how they are geometrically organized. In this work, we tackle these three questions through a systematic layer-wise analysis of V-JEPA 2 and VideoMAE-v2. We leverage lightweight probes trained to discover three temporally grounded properties: (i) camera motion understanding, (ii) intuitive physics, and (iii) anomaly detection. Both models encode camera motion, with best results ( 90 ROC AUC) emerging at 60-70% of network depth, and achieve moderate anomaly detection performance ( 60 ROC AUC), but remain near chance on intuitive-physics tasks, suggesting a limited encoding of deeper physical reasoning. Beyond classification, we find that temporal features from individual videos form smooth low-dimensional trajectories in representation space, suggesting that camera motion is not only linearly decodable but also geometrically organized. Based on these results, we apply geometry-aware spline-based steering in the model’s latent representations to interpolate camera motion, yielding steered videos with smoother trajectories and more coherent temporal progression than linear interpolation.
[CV-7] Revisiting Cross-View Completion: Self-Supervised Pre-Training via Reconstruction Error Comparison
链接: https://arxiv.org/abs/2609.01530
作者: Thibaut Loiseau,Guillaume Bourmaud,Vincent Lepetit
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project page: this https URL
Abstract:Self-supervised pre-training via cross-view completion learns strong features for 3D vision from co-visible regions of image pairs. However, the reference view provides little information for reconstructing non-co-visible patches, implicitly yielding a monocular training signal in these regions. We introduce Gekko, which turns this limitation into a useful signal. The relative improvement of the cross-view reconstruction error over a masked-autoencoder error is a self-supervised proxy for co-visibility: large improvements indicate co-visible regions, negligible ones non-co-visible areas. Gekko is a network, trained from scratch, that jointly performs cross-view completion, masked autoencoding, and per-pixel prediction of this relative improvement, providing an additional binocular signal for all masked regions without any ground-truth 3D annotation. Under identical architectures and training data, Gekko consistently outperforms CroCo on zero-shot correspondence estimation, relative pose estimation, and pointmap regression, with up to 6 times higher accuracy at the strictest relative-pose threshold and a 22% drop in end-point error on ETH3D. The extra channel it learns is itself a strong co-visibility detector on unseen scenes, and Gekko’s frozen features outperform released cross-view backbones of comparable or larger size. It can also be trained directly from raw videos with a simple stride-based curriculum, removing the cumbersome 3D preprocessing prior methods require while matching models trained on curated data. Code and pre-trained models are publicly available.
[CV-8] DualDiff3D: Dual Structure-Appearance Diffusion Priors for Reliability-Enhanced 3D Gaussian Splatting
链接: https://arxiv.org/abs/2609.01516
作者: Qian Wang,Yu Wang,Weiqi Li,Xinhua Cheng,Xiandong Meng,Ronggang Wang,Jian Zhang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:While 3D Gaussian Splatting (3DGS) has revolutionized 3D reconstruction and novel-view synthesis, scenarios with limited input views often lead to poor reconstruction quality and artifacts in rendered novel views. Recent efforts attempt to utilize powerful diffusion priors, yet they typically process rendered and reference views concatenated along an additional dimension in a single network. These methods overlook an inherent nature that different views should maintain appearance similarity but differ in structure due to view shifts, leading to blur caused by conflicts between the two properties. In this paper, we propose DualDiff, a novel pipeline that leverages dual diffusion priors with a Structure-Appearance Attention (SAA) module to introduce reference guidance for refining low-quality novel views rendered from flawed 3D representations. Specifically, we retain one diffusion branch to focus on extracting structural information from the low-quality novel views, while introducing another branch to ensure appearance consistency with reference views. Furthermore, we present a 3D reconstruction framework named DualDiff3D, which integrates a reliability-enhanced Render-Refine-Optimize (RRO) loop to progressively and robustly incorporate the refined novel views, yielding more accurate 3DGS. Extensive experiments demonstrate that our approach outperforms state-of-the-art methods even in the inference-only setting, with further performance gains achievable through training. Our code and pre-trained weights are available at this https URL.
[CV-9] mpCloze: Can Video-LLM s Identify the Missing Middle? EMNLP2026
链接: https://arxiv.org/abs/2609.01515
作者: Wenqi Pei,Henry Hengyuan Zhao,Yilai Liu,Jiahao Meng,Han Chen,Ziyu Wang,Hongyang Du
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: EMNLP 2026 Findings
Abstract:Temporal reasoning benchmarks for Video-LLMs are often mediated by language, leaving room for linguistic shortcuts from option wording, answer correlations, or language priors. To reduce such shortcuts, we introduce TempCloze, a video cloze benchmark for evaluating visual temporal reasoning in Video-LLMs. Given the beginning and ending clips of a video, models must identify the true missing middle from four candidates. TempCloze contains 1,521 carefully filtered videos from seven sources, mainly long-take and egocentric videos. We construct same-source distractors along three dimensions: Semantic asks what event should happen, Alignment probes when it should occur, and Progression tests how it should unfold, while shared scenes and objects reduce appearance cues. Our evaluation of 10 proprietary and 21 open-source Video-LLMs reveals Alignment as the primary bottleneck: models often recognize plausible semantic content and local event progression but struggle with temporal alignment. We further conduct error pattern and behavioral sensitivity analyses on TempCloze-Mixed and TempCloze-Hard with four representative models to examine where errors arise and how candidate order, context direction, visible span, frame density, and test-time scaling influence model choices.
[CV-10] Benchmarking Spatial Spectral and Self-Supervised Cues for Face Forgery Detection under Realistic Degradation
链接: https://arxiv.org/abs/2609.01511
作者: Lucas Cunha,Lucas Sotomaior,Lucas Gasperin,Beatriz Caldas,Eduardo Pianovski,Rayson Laroca
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted for presentation at the 2026 Conference on Graphics, Patterns and Images (SIBGRAPI)
Abstract:Face forgery detectors often achieve strong results on controlled benchmarks, but their reliability under realistic image degradations remains limited. This paper presents a standardized benchmark for face forgery detection using the Multi-Dimensional Face Forgery Image (MFFI) dataset and evaluates performance on both clean and degraded test partitions. We compare six model families, including convolutional networks, transformer-based models, and a frozen self-supervised DINOv3 backbone, across spatial, spectral, and hybrid input representations. The results show that clean-set performance is not a reliable indicator of robustness under compression, resizing, and blurring. Xception with RGB obtains the best clean performance, reaching 0.884 mean ROC-AUC, but degrades substantially on the harder partition. In contrast, frozen DINOv3 achieves the strongest degraded-set result, with 0.726 mean ROC-AUC, while training only a linear classification head. The representation analysis indicates that Fourier-domain cues are most useful when combined with RGB information, whereas purely spectral inputs consistently underperform spatial representations. Qualitative attribution maps further suggest that convolutional detectors focus on localized artifacts, while DINOv3 relies on broader facial structure. These findings reinforce the need for degraded evaluation protocols and highlight self-supervised visual representations as a promising direction for robust face forgery detection. Our source code is publicly available at this https URL.
[CV-11] CameraEditor: Camera-Controlled Image Editing via Video-Prior Sequential Modeling
链接: https://arxiv.org/abs/2609.01479
作者: Xin Shen,Chengyou Jia,Keshuo Xing,Zifeng Zhu,Changliang Xia,Bowen Ping,Zhuohang Dang,Hangwei Qian,Minnan Luo
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to ACM Multimedia 2026
Abstract:Beyond semantic content, camera parameters play a pivotal role in dictating the geometric perspective and appearance of any given image. While recent image editing models excel at semantic and stylistic manipulation, they struggle with explicit camera parameter control. When handling large perspective shifts, instruction-driven models face a dilemma: they either suffer from structural tearing or generate conservative outputs that ignore geometric instructions. To address this, we introduce CameraEditor, a framework that reformulates camera-controlled editing from a spatial problem into a temporal sequence prediction task. By leveraging the temporal coherence of video diffusion models, our approach integrates an explicit geometric perception module with a dynamic reference routing mechanism. This allows us to construct geometrically rigorous visual reference pairs via dynamic panorama cropping, overcoming the ambiguity of text-based instructions. Furthermore, CameraEditor strategically inserts intermediate transition frames to decompose large perspective shifts, providing a robust temporal buffer that preserves content identity and spatial coherence. We construct a training dataset of 5,760 instances. As an independent contribution, we introduce CamEditor-Bench, a model-agnostic evaluation suite of 462 test cases. Extensive experiments demonstrate that CameraEditor achieves state-of-the-art camera control precision and source identity preservation, outperforming existing methods.
[CV-12] RadMatch: Auditable Radiology Report Evaluation via Finding-Level Matching ECCV2026
链接: https://arxiv.org/abs/2609.01470
作者: Charles Corbière,Léo Machado,Aubin Charley,Baptiste Callard,Pierre Manceron,Corentin Dancette
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to ECCV 2026 Workshop on Medical Foundation Models and Benchmarks
Abstract:As AI systems are increasingly used to draft radiology reports, reliably evaluating their clinical quality remains a critical challenge. Large language model (LLM)-based metrics are now the best-correlated with radiologist judgment, yet they output a single opaque score that neither a clinician nor a model builder can easily interpret or audit. We introduce RadMatch, a multi-stage, LLM-based metric that decomposes report comparison into a structured finding-level matching with significance-aware scoring and error characterization across seven clinical attribute dimensions (status, location, severity, morphology, certainty, longitudinal comparison, and measurement). The main score is the actionable-error count, both interpretable and auditable. Candidate findings are graded correct, partial, or incorrect, and unmatched findings are counted as missed or hallucinated. Triage and actionable safety recall/precision and per-subset views add complementary, deployment-oriented lenses. Across two expert benchmarks, RadMatch is the most clinically aligned metric, matching inter-radiologist agreement on ReXVal and more than doubling the best prior metric on the harder RadEvalExpert. Relying only on few-shot prompting, it is designed to extend to other modalities and anatomies. We will release RadMatch as open-source code with an interactive dashboard for inspecting results.
[CV-13] Gaussian Core LoRA: Distribution-Aware Dynamic Adaptation for Broad Concept Erasure
链接: https://arxiv.org/abs/2609.01433
作者: Qinghui Gong,Xunlei Chen,Yu-Xuan Zhang,Hua Meng,Zhengchun Zhou
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Concept erasure aims to suppress unsafe, privacy-sensitive, or undesirable generations in text-to-image diffusion models while preserving benign semantics, visual quality, and deployment efficiency. Existing adapter-based methods, such as Low-Rank Adaptation (LoRA), typically freeze the diffusion backbone and learn lightweight parameter updates to steer generation away from target semantics. However, these methods usually assign a static semantic erasure direction to each target concept. This assumption is overly coarse for broad and complex target concepts, since a concept often contains multiple latent semantic prototypes involving different objects, scenes, or relations, and requires different local erasure directions. A single LoRA update averages these heterogeneous erasure demands, leading to under-erasure on difficult prototypes and over-editing of nearby benign semantics. To address this limitation, we propose Gaussian Core LoRA, a distribution-aware low-rank adaptation framework. It fits a Gaussian mixture model in the prompt feature space to estimate latent semantic prototypes within the target concept. During inference, each input prompt is projected into this feature space to compute its Gaussian posterior responsibilities, which condition the core generator to produce a prompt-specific, norm-bounded residual reconfiguration of the shared LoRA rank space. This enables prototype-adaptive erasure with a single lightweight adapter. Compared with the strongest baseline on each metric, Gaussian Core LoRA reduces average Attack Success Rate (ASR) by 7.95%, lowers COCO Fr’echet Inception Distance (FID) by 14.72%, and improves CLIP Score by 4.98%. Further experiments show robustness to adversarial prompts, scalability to multi-identity and multi-style erasure, and compatibility with SDXL and FLUX.
[CV-14] Pix2Rep-v2: Data-Efficient Representation Learning for Dense Medical Imaging Applications MICCAI2026
链接: https://arxiv.org/abs/2609.01427
作者: S. Sifaoui,E. Angelini,S. Toupin,T. Pezel,L. Le Folgoc
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at MICCAI 2026
Abstract:Dense self-supervised learning (SSL) is a powerful paradigm for learning without annotations the local descriptors required to solve dense medical imaging tasks. We present Pix2Rep-v2, a framework for SSL of pixel- and voxel-level representations suitable for few-shot downstream applications. Pix2Rep-v2 addresses the main challenges of dense SSL by leveraging a redundancy reduction objective at the pixel-level with a principle of equivariance of dense representations, that scales efficiently to 3D or wide field-of-view applications. We evaluate our method on four datasets, across multiple tasks, multiple modalities and anatomical structures using multiple backbones in 2D and 3D, and under various data regimes. As an alternative to linear probing or full fine-tuning on the downstream task, we also propose an in-context variant, without downstream training, based on a dense prototype approach. Pix2Rep-v2 shows substantially higher data-efficiency in few-shot scenarios compared to fully supervised baselines, and is competitive with the state-of-the-art e.g., +9.3 Dice points in one-shot segmentation on the MMs-2 dataset. Our code and pre-trained models are publicly available at this https URL.
[CV-15] Semantic-Guided Multimodal Preprocessing for Vision Transformer-Based Clear Cell Renal Cell Carcinoma Grading MICCAI2026
链接: https://arxiv.org/abs/2609.01426
作者: Fatemeh Javadian,Zhu Chen,Zahra Aminparast,Johannes Stegmaier
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: 12 pages, 3 Figures, COMPAYL++ MICCAI 2026
Abstract:Clear cell renal cell carcinoma (CCRCC) grading is essential for treatment planning, yet existing approaches either analyze patch-level images directly or focus solely on nuclei-level classification, without linking to final tumor grading. We propose a semantic-guided multimodal preprocessing method that integrates nuclei classification maps from existing pre-trained models with RGB histopathology images for Vision Transformer (ViT)-based CCRCC grading. Our approach employs classification map channel concatenation and multiplicative modulation, with optimized overlays to leverage nuclei grading information, while preserving RGB textural features. Evaluation of multiple preprocessing strategies demonstrates that semantic-guided enhancement achieves 0.916 balanced accuracy, outperforming RGB-only baseline (0.707) and max-voting aggregation from prior studies (0.427). Sensitivity analysis reveals that this 21 percentage point improvement over baseline persists even under simulated perturbation at rates matching current state-of-the-art nuclei classification model error thresholds, suggesting both effective semantic utilization and practical robustness. These findings show that preprocessing-based multimodal fusion can leverage the diagnostic potential of existing imperfect nuclei classifiers, effectively bridging previously isolated fine-grained nuclear-level analysis with coarse-grained ViT-based patch classification. Per-class recall was consistent across grades (0.93, 0.91, 0.91), indicating that gains are not concentrated in the majority class. Because the sensitivity analysis perturbs ground-truth maps rather than predictions from an actual nuclei model, this result characterizes robustness under simulated error rather than deployment with a real upstream model, which remains for future work.
[CV-16] MegaStyle: Scaling Image Style Space through Hierarchical Style Definition DATE
链接: https://arxiv.org/abs/2609.01423
作者: Junyao Gao,Sibo Liu,Jiaxing Li,Yanan Sun,Weidong Zhang,Cairong Zhao,Jun Zhang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: The dataset and code will be updated at this https URL , 10pages, 5 figures
Abstract:Image style is a highly abstract, human-constructed concept shaped by a range of visual factors and intrinsically entangled with content, yet a unified and explicit definition of image style remains lacking. In this work, we first discuss the fundamental question of what is style and then propose a hierarchical style definition that describes image style from an overall style identity to fine-grained visual attributes, providing a more structured, transferable, and interpretable style representation. Based on this definition, we refine the style annotation pipeline of MegaStyle and construct MegaStyle+±8M, a large-scale style dataset containing 150K overall style identities, 1M fine-grained style prompts, and 8M stylized images. Extensive analyses demonstrate that our hierarchical definition substantially expands the style space in both diversity and semantic breadth, while precisely capturing intrinsic visual style of reference images. The dataset and code will be updated at this https URL, we hope MegaStyle++ provides a scalable foundation for studying and modeling diverse image styles.
[CV-17] Neuro-Symbolic Geometric Abstraction (NeuSOGA): From Observations to Symbolic Mathematical Representations
链接: https://arxiv.org/abs/2609.01408
作者: Qingde Li,Qingqi Hong,Jie Tian
类目: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Graphics (cs.GR)
备注: 18 pages, 6 figures. Code repository: this https URL
Abstract:A fundamental challenge in artificial intelligence is the transformation of observations into explicit symbolic representations suitable for abstraction, interpretation, and reasoning. While modern AI systems achieve remarkable perceptual capabilities through large-scale statistical learning, the resulting knowledge is typically encoded within latent parameters that are difficult to inspect or manipulate analytically. Inspired by Neuro-Symbolic AI and theories of human abstraction, this paper investigates the formation of symbolic mathematical representations from geometric observations. We propose NeuSOGA (Neuro-Symbolic Geometric Abstraction), a framework that progressively transforms observations into topological abstractions, geometric abstractions, and ultimately symbolic mathematical representations. The architecture combines topology-guided structural discovery using Euclidean Distance Transforms, foundation-model perception using Segment Anything, adaptive multi-scale geometric abstraction, and symbolic synthesis through Implicit Area Splines. The resulting representation is an analytical implicit model supporting arbitrary-order smoothness, additive composition, and closed-form evaluation. Unlike neural latent encodings, the generated representation remains interpretable, editable, and mathematically explicit. Experiments on ModelNet40 point clouds, arbitrary-view projections, and segmented optical observations demonstrate that NeuSOGA transforms diverse observations into compact symbolic representations while preserving essential geometric and topological structure across sensing modalities and viewing directions. NeuSOGA provides an interpretable and explainable pathway from observation to symbol and establishes Comments: 18 pages, 6 figures. Code repository: this https URL Subjects: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Graphics (cs.GR) Cite as: arXiv:2609.01408 [cs.AI] (or arXiv:2609.01408v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2609.01408 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Qingde Li [view email] [v1] Tue, 1 Sep 2026 15:29:30 UTC (9,270 KB)
[CV-18] Scale-based Approach for Active Wildfire Segmentation on Satellite Imagery
链接: https://arxiv.org/abs/2609.01392
作者: Matheus F. Kovaleski,Cristiano Premebida,João Ruivo Paulo
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 6 page, 2 figures, 3 tables
Abstract:Active wildfire mapping from satellite imagery is challenging due to the sparse and highly imbalanced nature of fire pixels, especially in early-stage or low-density fire observations. This work investigates the use of multispectral Landsat-8 imagery for active-fire segmentation under multi-scale wildfire size conditions. We propose a data-driven protocol to characterize fire-region size distributions through connected-component analysis and an interquartile range criterion, enabling the evaluation of model robustness across different local fire-region densities. Three segmentation architectures, U-Net, DeepLabV3+, and SegFormer, are evaluated under different SWIR-based spectral configurations. Results show that U-Net achieves the strongest robustness across the evaluated conditions, SegFormer provides competitive performance, and DeepLabV3+ tends to produce conservative predictions with reduced recall. Across architectures, SWIR2 consistently achieves the strongest or near-best results, highlighting its importance for active-fire segmentation in Landsat-8 imagery. These findings suggest that both spectral band selection and architectural design are critical for robust satellite-based active wildfire mapping trained on low active fire-pixel density images.
[CV-19] Multimodal RGB-Infrared Combination for UAV-Based Wildfire Segmentation: A Comparative Study on FLAME3
链接: https://arxiv.org/abs/2609.01390
作者: Matheus F. Kovaleski,Luís Garrote,Cristiano Premebida,Jérôme Mendes,João Ruivo Paulo
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 6 pages, 1 figure, 1 table
Abstract:Unmanned Aerial Vehicles (UAVs) have emerged as a promising platform for firefighting operations due to their flexibility, low operational cost, and ability to acquire high-resolution imagery in locations that may be difficult or dangerous to access using conventional methods. Recent advances in deep learning have significantly improved the capabilities of UAV-based wildfire monitoring systems. The present work investigates RGB-infrared fusion for binary wildfire segmentation on the FLAME3 dataset. In this Study, RGB and Infrared baselines are compared with three representative fusion strategies across three segmentation architectures, including U-Net, DeepLabV3+, and SegFormer. The key motivation of this work is to analyze the contribution of each modality, evaluate the impact of fusion timing, and examine how different network architectures exploit multimodal information for UAV wildfire delineation. The findings indicate that thermal information plays a dominant role in UAV segmentation and that feature-level multimodal fusion combined with transformer-based architectures offers the most promising direction for future research.
[CV-20] Diffusion Based Unpaired Data Learning for Inverse Problems
链接: https://arxiv.org/abs/2609.01370
作者: Chenglong Bao,Yiming Dang,Chenguang Duan,Yuling Jiao,Defeng Sun
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 30 pages, 7 figures
Abstract:Data is important in many deep learning-based inverse problem solvers. However, obtaining sufficient paired data in many scenarios remains highly challenging, while unpaired data is cheap. To maximize data utilization, this paper proposes LUD-DIF, a diffusion-based approach for solving inverse problems with unpaired data. Starting from the evidence lower bound (ELBO) of the joint distribution, we decouple it into two independent diffusion processes under the weak-coupling assumption. The method provides theoretical support from a variational inference perspective, derives the loss function, quantitatively analyzes the error bound introduced by the assumption, and offers a theorem-motivated heuristic for hyperparameter selection. Experimental results demonstrate that LUD-DIF achieves outstanding performance on multiple image inverse problems, validating its effectiveness and generalization capability in unpaired inverse problem settings.
[CV-21] Accurate Reconstruction of Gas Turbine Blade Geometry Using 3D/2D Rigid Registration and CT View Optimization
链接: https://arxiv.org/abs/2609.01368
作者: Hristo Valtchanov,Nicolas Piché,Vladimir Brailovski,Justin Byers,Catherine Désrosiers,François Guibault
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: conference proceedings International Conference on Computed Tomography (iCT) 2025, 8 pages, 6 figures
Abstract:Non-destructive X-ray and computed tomography (CT) testing are essential for ensuring the dimensional accuracy of manufactured components with complex internal structures, such as the cooling channels in gas turbine blades, which directly affect thermal performance and service life. This study presents a multipart 3D-2D rigid registration approach for aligning CAD models with X-ray projections as an alternative to CT reconstruction for part inspection and measurement. A greedy registration algorithm sequentially aligns the blade’s exterior before registering its internal components by maximizing the mutual information between simulated and acquired X-ray images. This stepwise approach reduces problem complexity and improves alignment accuracy. View angles are optimized using a greedy method that iteratively selects angles to minimize dimensional measurement errors. The results indicate that a small number of oblique views provides the best accuracy, although a broad range of angles yields acceptable results. The method achieves subpixel registration accuracy, with errors below one-fifth of the magnified detector-pixel pitch. Image noise and defects reduce registration precision, but direct registration in projection space mitigates these effects compared with CT reconstruction. Appropriate view selection can therefore preserve acceptable subpixel accuracy in the presence of image noise and defects.
[CV-22] ExBind: A Controlled Diagnostic Benchmark for Visual-to-Executable Correspondence
链接: https://arxiv.org/abs/2609.01344
作者: Ziqian Wang,Yuxiao Cheng,Tingxiong Xiao,Jinli Suo
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 19 pages, 3 figures, benchmark and diagnostic evaluation paper
Abstract:Multimodal coding and editing systems must map a visible or semantic referent to the exact executable object that can be edited. A wrong reference may select a valid but incorrect DOM node, SVG element, graph endpoint, hierarchy member, or table cell, while final execution success alone does not reveal the source of the failure. ExBind isolates this visual-to-executable correspondence layer as a controlled diagnostic benchmark between semantic localization and action execution. It samples representation-independent latent binding instances and compiles them into SVG, DOM, canvas, tree, graph, and table cases with deterministic mappings to executable references. Models output only a strict reference; the evaluator maps predictions back to latent structure and scores structural constraints without requiring reasoning traces. The release contains a 250-case broad suite, a disjoint 240-case targeted suite, and 50 paired latent groups. Qwen2.5-VL-3B achieves 98.4% candidate validity but 76.4% exact accuracy, while Qwen3-VL-4B achieves 100.0% validity and 98.8% exact accuracy. In the targeted table suite, all Qwen2.5-VL-3B residual errors are valid correct-row/wrong-column selections. Candidate-order perturbations change case-level outcomes while preserving this error pattern. ExBind is designed for controlled diagnosis rather than population-scale ranking or end-to-end editing evaluation. Code and benchmark records are available at this https URL and this https URL.
[CV-23] CMRVision: A Foundation Model for Cardiac MR Image Analysis MICCAI2026
链接: https://arxiv.org/abs/2609.01308
作者: Athira J. Jacob,Puneet Sharma,Daniel Rueckert
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at MedAGI 2026 (peer-reviewed workshop at MICCAI 2026)
Abstract:Cardiac magnetic resonance (CMR) imaging provides complementary information on cardiac anatomy, function, and tissue characterization across multiple sequences and views. In this work, we investigate foundation model pretraining for 2D CMR and introduce CMRVision, a CMR-specific foundation model trained using DINOv3-style self-supervised learning on a multi-center, multi-sequence cohort of 36 million CMR images. We systematically evaluate architectural and training design choices for domain-specific pretraining. CMRVision is evaluated on two downstream tasks: multi-task segmentation across cine, late gadolinium enhancement (LGE), and mapping sequences, and cine view classification. Our experiments show that CMR-specific pretraining, smaller patch sizes, and patch-level objectives consistently improve downstream performance. Across a multi-task segmentation benchmark, CMRVision achieved the strongest overall performance, outperforming prior natural-image (NI), medical-image, supervised, and CMR foundation model baselines. Improvements were modest but consistent across structures and sequences, with Dice scores ranging from 0.940-0.967 for LV and 0.855-0.905 for myocardium, and reaching 0.929 for RV, 0.920 for LA, and 0.931 for RA. The largest gains were observed for myocardium segmentation in LGE and mapping images. In a zero-shot segmentation task on unseen LGE long-axis views, the model achieved an average Dice score of 0.692, demonstrating cross-view generalization. For cine view classification, CMRVision achieved the highest average accuracy (0.906), compared to prior methods reported in the literature. These results highlight the potential of CMRVision to support robust and generalizable cardiac MRI analysis across multiple sequences and views.
[CV-24] MeshSplatBench: A Unified Benchmark for Triangle-Based Neural Rendering
链接: https://arxiv.org/abs/2609.01306
作者: Kaixuan Zhang,Minxian Li,Mingwu Ren,Xiatian Zhu
类目: Graphics (cs.GR); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Triangle-based neural rendering bridges neural scene representations and conventional graphics pipelines by optimizing explicit geometric primitives compatible with standard rasterization hardware. However, existing approaches are evaluated almost exclusively within custom research renderers, obscuring their practical deployability in production engines. To bridge this gap, we introduce \textbfMeshSplatBench, a unified benchmark that systematically investigates triangle-based neural rendering across the complete pipeline from native optimization to game-engine deployment. MeshSplatBench establishes a standardized evaluation protocol while preserving each method’s native optimization semantics, reproducing published results within 0.8% PSNR deviation. Furthermore, we introduce a hierarchical Unity deployment protocol spanning three rendering tiers: native CUDA renderers, method-specific dedicated engine shaders, and standard opaque mesh pipelines, isolating the exact fidelity losses caused by engine adaptation \textitvs. representation reduction. Finally, we conduct a topological audit of reconstructed surfaces, demonstrating that explicit connectivity and shared indexing alone are insufficient to guarantee production-ready assets due to prevalent non-manifold structures, fragmented components, and boundary artifacts. Overall, MeshSplatBench demonstrates that rasterizability is merely a primitive-level attribute, whereas graphics readiness requires jthe holistic alignment of representation, topology, and engine compatibility. Source code will be released.
[CV-25] Agent ic Multimodal Models for Environmental Hyperspectral Unmixing
链接: https://arxiv.org/abs/2609.01289
作者: Michał Cholewa,Luca Ciampi,Nicola Messina,Przemysław Głomb,Giuseppe Amato
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Hyperspectral unmixing is a key task in remote sensing that aims to decompose mixed pixels in hyperspectral images into their constituent material signatures, or endmembers, and their fractional abundances. Conventional modular approaches estimate the scene composition through successive model-order estimation, endmember extraction, and abundance estimation stages, whose errors can lead to redundant or ambiguous candidate components and ultimately affect the recovered decomposition. We introduce an algorithm-agnostic, large vision-language model (LVLM)-driven agentic framework that refines the outputs of such pipelines rather than replacing their underlying numerical algorithms. Starting from an initial decomposition, the agent iteratively gathers complementary spectral and spatial evidence through dedicated tools, including spectral-library retrieval and abundance-map visualization, and modifies the active endmember set through merge and discard operations followed by abundance re-estimation. We apply the same refinement procedure to several modular pipelines combining different model-order, extraction, and abundance-estimation methods, and evaluate it on HYDICE Urban, Jasper Ridge, and Stonewall Playa. Experiments show that the proposed agent consistently improves endmember cardinality and generally improves the recovered spectral signatures and abundance maps across heterogeneous modular pipelines, while remaining competitive with integrated end-to-end unmixing methods, including CNN-AE, uDAS, and R-CoNMF. These results highlight the potential of tool-using LVLM agents to combine spectral and spatial evidence for algorithm-agnostic refinement of physically grounded hyperspectral unmixing decompositions. Code is publicly available at this https URL.
[CV-26] HiLRP: Toward One Trustworthy Explanation for Vision Transformer: Conservation-Valid Attribution via Attention Primitives
链接: https://arxiv.org/abs/2609.01282
作者: Sathiyamohan Nishankar,Pubudu Sanjeewani,Asanka Perera,Selvarajah Thuseethan
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Vision Transformer (ViT) design has become increasingly diverse, with backbones combining convolutional stems, windowed, linear, or multi-axis attention, patch merging, and spatial reduction in various configurations. This diversity poses challenges for existing attribution methods, whose assumptions often do not hold across ViT variants: Grad-CAM requires a terminal spatial feature map, attention rollout assumes global softmax attention, and layer-wise relevance propagation (LRP) requires module-specific rules. To the best of our knowledge, no existing method provides a unified attribution framework across this architectural space. We show that this architectural diversity can be captured by a simpler underlying structure. The attention and resolution-reduction operators in current ViTs can be decomposed into four operation types: linear maps, bilinear mixing, normalization or gating, and reindexing. Each operation admits a relevance rule that satisfies conservation. Based on these rules, HiLRP supports new backbones by construction rather than by architecture-specific derivation, and its attribution maps decompose the prediction rather than relying on heuristic assumptions. We prove conservation and conditional equivariance and verify both to machine precision. Across 14 attribution methods and 10 architectures, we find that no prior method remains reliable across ViT families, while Faithfulness Correlation becomes uninformative for backbones robust to spatial masking. HiLRP alone preserves conservation across windowed, spatial-reduction, multi-axis, and linear-attention models, where naive extensions can produce zero or inflated relevance. It also localizes attribution failures in class activation mapping, achieving 0.97 Pointing compared with 0.55 for competing methods on EfficientViT.
[CV-27] meSteer: Inference-Time Speech Scheduling in Joint Audio-Visual Diffusion Models
链接: https://arxiv.org/abs/2609.01277
作者: Chao Zhou,Yiling Chen,Qi Chu,Tao Gong,Nenghai Yu,Tianyi We
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Multimedia (cs.MM)
备注:
Abstract:Although pretrained joint audio-visual diffusion models offer rich control over \emphwhat to generate, they provide no explicit control over \emphwhen an utterance should occur. To address this, we study \emphinference-time speech scheduling, a novel task that places coupled speech and visual articulation within user-specified begin–end intervals without finetuning the backbone model. We uncover two intrinsic properties of the denoising process that enable this task. First, a timing-sensitive text-to-audio cross-attention head exposes each utterance’s model-implied source span along the latent timeline. Second, the predicted clean latent already organizes coupled speech and visual articulation, allowing their temporal placement to be edited without regenerating the content. Building on these discoveries, we propose \textbfTimeSteer, a training-free framework that localizes each utterance’s source span through \textbfSource Span Localization and transfers the associated audio-visual latent content from the source interval to the specified target interval through \textbfRegion-Aware Latent Remapping. We further introduce \textbfSpeechShift, the first benchmark for interval-level speech scheduling in joint audio-visual generation. Experiments across two representative backbones show that TimeSteer substantially improves interval controllability over training-free baselines while maintaining competitive overall generation quality.
[CV-28] Seeing the World and the Self from Egocentric Video
链接: https://arxiv.org/abs/2609.01276
作者: Kai Guan,Minchao Jiang,Ruichen WangLi,Wentao Zhu,Lei Zhang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Complete 3D perception from egocentric video requires recovering the surrounding scene and the wearer’s full-body motion in a shared metric frame. Existing methods typically address scene reconstruction and motion estimation separately: scene reconstruction methods ignore the wearer, whereas motion estimation methods lack explicit scene geometry and often depend on external trajectories. Joint recovery is challenging because the two tasks exhibit asymmetric visibility and require different prediction paradigms. The largely visible scene supports deterministic geometric regression, whereas the severely occluded body requires generative motion inference. We therefore propose RESELF (REconstructing the Scene and the sELF), a unified framework that couples deterministic metric geometry reconstruction with geometry-conditioned motion generation. RESELF adapts a geometry foundation model pre-trained on large-scale exocentric data to egocentric video using frame-wise scale and relative-pose consistency objectives. The resulting camera trajectory and latent geometric features condition a diffusion model that recovers the wearer’s motion. A subsequent closed-loop kinematic feedback stage further refines the camera head while preserving the reconstructed scene geometry. To support training and evaluation, we curate EE4D-JSM from EgoExo4D by aligning egocentric video, sparse metric scene geometry, camera trajectories, and full-body motion annotations. Experiments show that RESELF outperforms state-of-the-art methods designed for the individual tasks across depth estimation, camera tracking, and full-body motion estimation. Code, models, and datasets will be available at this https URL.
[CV-29] MeRoPE: Metric Rotary Position Embedding for Camera-Controlled Video Generation
链接: https://arxiv.org/abs/2609.01252
作者: Zhijian Qiao,Xinjiang Wang,Jiajie Chen,Haoming Huang,Meng Li,Chih-Chung Chou,Jing Wang,Shaojie Shen
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注: 22 pages, 12 figures, and 7 tables. Project page: this https URL
Abstract:In camera-controlled video generation, geometry-aware positional encodings condition tokens on camera extrinsics and per-token viewing rays. Existing schemes, however, have a scale-dependent failure mode on real-world metric camera trajectories: homogeneous projective encodings cause attention logits and feature norms to grow unbounded with physical translation baselines. We propose MeRoPE (Metric Rotary Position Embedding), a norm-preserving relative camera encoding for attention. MeRoPE encodes relative orientations between calibrated viewing rays with orthogonal rotation blocks, maps raw metric displacements into multi-frequency rotary phases, and adds a disparity-anchored correspondence prior along the epipolar arc. This design strictly preserves feature norms, bounds pre-softmax attention logits regardless of the physical translation scale, and maintains exact invariance to global rigid coordinate changes. Across nuScenes and PanShot, which cover large-baseline trajectories and diverse camera optics, respectively, MeRoPE achieves stronger camera control than prior encodings, with the best consistency between generated camera motion and conditioning poses in both rotation and translation. Code will be made publicly available.
[CV-30] One Prompt Is Enough: Watermark Laundering Through Foundation Image Models
链接: https://arxiv.org/abs/2609.01249
作者: Jidong Yang,Qi Li,Wei Zong,Yang-Wai Chow,Willy Susilo,Huaike Yu,Chunpeng Wang,Suo Gao
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR)
备注: 11 pages, 4 figures, and 2 tables
Abstract:Invisible watermarks are typically evaluated against predefined perturbations such as compression, blur, noise, cropping, and denoising. Public foundation image models expose a distinct threat: an attacker can submit a watermarked image with a single reconstruction prompt and obtain a visually faithful output from which the invisible watermark can no longer be decoded reliably. We formalize this failure mode as watermark laundering and evaluate it using a joint payload-fidelity profile that combines bit error rate (BER) with visual and semantic preservation. Across six OpenAI and Google image editing models, three representative watermarking schemes, and 1,800 reconstructed outputs, we identify two complementary laundering regimes: OpenAI models produce the strongest payload disruption across the evaluated schemes, whereas Nano Banana 2 shows that DwtDct remains vulnerable under high-fidelity reconstruction. Prompt ablations show that no single removal-oriented instruction is necessary for payload disruption, indicating that the effect is primarily induced by the reconstruction pathway rather than by explicit attack wording. Comparisons with conventional attacks further show that prompt-conditioned reconstruction constitutes a distinct operational attack interface. These findings motivate foundation-model reconstruction as a missing robustness condition in invisible watermark evaluation.
[CV-31] S2Prune: Spatially Structured Visual Token Pruning for Multimodal Large Language Models
链接: https://arxiv.org/abs/2609.01224
作者: Yuanyuan Jia,Shunpu Tang,Qianqian Yang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 18 pages, including supplementary material. Code is available at this https URL
Abstract:Visual token pruning reduces the inference overhead of multimodal large language models (MLLMs) by retaining only a subset of visual tokens. Existing methods usually select tokens based on importance or redundancy. However, we observe that these criteria produce stable spatial biases across inputs and do not always outperform simple Uniform Grid sampling, highlighting the value of broad spatial coverage. Motivated by this, we propose S ^2 Prune, a training-free pruning method that preserves spatial coverage while adapting token density to local image structure. We first divide the image into regions and assign at least one token to each region to preserve coverage. The remaining token budget is then distributed according to Laplacian variation, giving more tokens to regions with richer structure. We then use Early Representation Change (ERC), computed from the first decoder block, to select representative tokens within each region. We evaluate S ^2 Prune across diverse settings and two MLLM architectures. On Qwen2.5-VL-7B-Instruct, it achieves the highest average accuracy among the evaluated training-free pruning methods. With only 32 of the original 576 visual tokens, it still retains 79.3% of the full-model performance. Code is available at this https URL.
[CV-32] Compressing AI Traffic: Standardized Neural Network Coding of Visual-Token Representations in Split Vision-Language Inference
链接: https://arxiv.org/abs/2609.01200
作者: Reza Heidari,Hamed R. Tavakoli,Juho Kannala
类目: Computer Vision and Pattern Recognition (cs.CV); Image and Video Processing (eess.IV)
备注: 4 pages, 6 figures, 1 table
Abstract:When the visual encoder and the language decoder of a vision-language model (VLM) run on different compute nodes, the intermediate visual-token embeddings become a communicated payload rather than an internal activation. We call such machine-consumed intermediate tensors AI traffic and ask how far they can be compressed with a standardized, training-free codec. We insert ISO/IEC 15938-17 Neural Network Coding (NNC) round trips on the complete visual interface of a Qwen3-VL-8B-Instruct video question answering pipeline, comprising the main visual-token representation and the DeepStack feature streams, while leaving weights, prompts, and generation untouched, and sweep the quantization parameter (QP) over a wide rate range. Closed-ended Video-MME accuracy remains close to the uncompressed reference up to a 98% reduction of the transmitted BF16 tensor and only then collapses; open-ended MLVU generation shows the same plateau-and-collapse profile under an LLM judge. This robustness is not due to near-lossless reconstruction: the decoded tensor is heavily discretized, carries substantial row-wise relative L2 error, and has a visibly steeper singular-value decay than its source. Downstream reasoning therefore depends on coarse structure and relative geometry rather than exact floating-point values, which argues for rate-task rather than rate-distortion optimization of AI traffic codecs.
[CV-33] Monocular Depth Estimation from a Single Image: Progress and Opportunities
链接: https://arxiv.org/abs/2609.01172
作者: Muxin Liu,Xiaoyang Lyu,Yang-Tian Sun,Yi-Hua Huang,Ziyi Yang,Peng Dai,Xiaojuan Qi
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted by Computational Visual Media Journal (CVMJ)
Abstract:Monocular depth estimation has long stood as a fundamental challenge in computer vision, enabling a wide range of applications including 3D reconstruction, robotics, autonomous driving, and augmented reality. This survey traces the field’s evolution from early learning-based methods to the emergence of transformative foundation models. We begin by framing the problem, distinguishing between relative and metric depth estimation, and highlighting the key challenges that have shaped a decade of research. We then present common problem formulations and introduce the most widely used datasets, covering indoor, outdoor, and synthetic data. Following this, we review major advances prior to the foundation model era, distilling core insights from influential methods that contributed to improvements in accuracy, efficiency, and robustness. The survey then turns to the recent surge of foundation-model-based approaches, categorizing them into discriminative and generative paradigms and emphasizing the critical roles of large-scale pretraining (e.g., DINOv3) and synthetic data. We compare representative models using both quantitative benchmarks and qualitative examples, and discuss natural extensions to video-based depth estimation. Further, to illustrate real-world impact, we highlight the integration of depth estimation into applications such as visual SLAM, content generation, and robot perception. Finally, we outline open challenges and promising research directions as the field advances further into the era of foundation models.
[CV-34] Dotting the Eye: An Intent-Driven Image Retouching Agent for Visual Focus Enhancement ECCV
链接: https://arxiv.org/abs/2609.01148
作者: Chujie Qin,Zilong Zhang,Zewei Chang,Chunle Guo,Ruixing Wang,Tao Hu,Ming-Ming Cheng,Chongyi Li
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to the European Conference on Computer Vision (ECCV) 2026
Abstract:Image retouching is commonly formulated as enhancing overall visual quality through color adjustment, but in practice, it also serves to emphasize visual focus by guiding viewers’ attention toward a specific subject or region. Achieving such focus-oriented retouching is inherently challenging, as it requires well-coordinated global and local adjustments to manipulate perceptual saliency while maintaining visual naturalness. This intricate process typically demands substantial professional expertise. In this study, we propose EyeControl, a MLLM-driven agent with a diffusion-based retouching executor that enables visual focus enhancement under weak user intent. With only a few clicks or coarse strokes, EyeControl directs visual attention to the intended region, effectively “dotting the eye” of the image. The core idea is to explicitly link the weak user intention with the target editing region and the corresponding tonal adjustment operations during retouching. To achieve this, the system first interprets the intent and image content to infer the visual focus and generate structured intent guidance for the retouching executor. Second, the retouching executor is encouraged to respond more strongly to the target region, explicitly aligning its attention map with a designed pseudo-intent map. We also introduce an operation-consistency constraint to improve coordination between global and local adjustments, achieving more natural and coherent retouching. Additionally, we contribute ControlArt-Bench, a high-quality evaluation dataset for visual focus enhancement. Extensive evaluations demonstrate that EyeControl yields perceptually appealing results with stronger intent alignment. Code can be found in this https URL.
[CV-35] StainPresetNet: Stain Preset Network for Fast Multi-to-Multi Stain Normalization
链接: https://arxiv.org/abs/2609.01146
作者: Hongtao Kang,Die Luo,Li Chen,Jing Cai,Junbo Hu,Xiuli Liu,Shenghua Cheng
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Stain normalization reduces color variations caused by variations in staining protocols and imaging conditions, thereby enhancing computer-aided diagnostic system performance. Traditional methods derive mapping relationships from individual or limited reference images through pixel-wise transformation, offering style flexibility but suffering from inaccurate color mapping extraction. While existing deep-learning-based approaches achieve accurate dataset-wide color mapping through complex neural networks, they face challenges including computational inefficiency, artifact generation, and fixed normalization directions requiring model retraining for directional changes. To address these limitations, we propose StainPresetNet - a novel framework that combines structural preservation with dataset-level color mapping while maintaining computational efficiency. Our method implements pixel-wise normalization guided by preset reference images, enabling multi-directional adaptability without retraining. Evaluations on cytopathology and histopathology datasets demonstrate that StainPresetNet achieves superior color mapping accuracy compared to conventional methods, effectively improves classifier generalization in diagnostic tasks, and reduces computational overhead by 90% versus existing deep learning approaches. The proposed preset-guided mechanism facilitates flexible adjustment of normalization directions through simple reference image replacement, overcoming the directional rigidity of current deep-learning-based solutions.
[CV-36] Revisiting Face Recognition for Monozygotic Twins: The Celeb Twins Test Set
链接: https://arxiv.org/abs/2609.01141
作者: Michael Zang,Haiyu Wu,Mrinal Sharma,Kevin W. Bowyer
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Past literature on face recognition for monozygotic ((“identical”) twins points to facial marks and mirror asymmetry as possible directions for improved accuracy of twins recognition. The Celeb Twins Test Set (CTTS) contains web-scraped image pairs for 80 sets of celebrity twins. It is the only twins test set with meta-data for twins with distinguishing skin marks and possible mirror asymmetry. CTTS is organized in the manner of face verification test sets such as LFW, CALFW, CPLFW, CFP-FP, and AgeDB-30. Current deep CNN matchers can achieve over 76% accuracy in classifying CTTS same-person / different-person image pairs. We show that current matchers do not make use of skin marks, or asymmetry, and discuss reasons for this. Finally, we discuss the feasibility of using generative AI tools such as Grok, ChatGPT and Gemini to create images of imagined monozygotic twins as a means to increase representation of twins in face recognition training sets.
[CV-37] Different Changes Require Different Reasoning : Change-Type-Specialized Experts for Robust Change Captioning ECCV2026
链接: https://arxiv.org/abs/2609.01136
作者: Jiyoung Park,InJae Oh,Jung Uk Kim
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to ECCV 2026
Abstract:Change captioning is the task of generating natural language descriptions that explain the changes between a pair of images. Although different change types (e.g., color shifts, object additions) exhibit distinct visual cues and require specialized reasoning processes, existing methods often overlook these distinctions. To address this limitation, we propose Multi-Expert Diagnosis for Image Change (MEDIC), a novel framework that introduces change-type awareness by explicitly modeling change categories. MEDIC employs type-specialized memory experts that dynamically retrieve type-relevant visual patterns conditioned on the input. This design enables each expert to capture diverse variations within its change type while focusing on the most informative visual cues. By softly routing inputs across type-specialized experts and learning dedicated representations for each change category, MEDIC generates more precise and type-aware change descriptions. Extensive experiments demonstrate that the proposed MEDIC consistently outperforms existing methods across diverse and challenging datasets. The code is available at \hrefthis https URLGitHub.
[CV-38] P-PatchDiff: Progressive Patch Diffusion Models for Low-light Image Enhancement
链接: https://arxiv.org/abs/2609.01123
作者: Ruoyu Guo,Haonan Zhong,Maurice Pagnucco,Yang Song
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted by IJCV
Abstract:Recent advancements in low-light image enhancement have leveraged diffusion models for their strong ability to generate perceptually realistic, detailed images. Patch diffusion models further offer a promising solution to size-agnostic image restoration while improving efficiency. However, existing methods typically rely on small, fixed patches (e.g., 64 \times 64) that cannot capture image-level brightness context, whereas enlarging the receptive field improves brightness and colour estimation but substantially increases computational cost. Moreover, low-light images often exhibit uneven brightness across regions, making it necessary to ensure that locally enhanced patches remain visually coherent when combined into the full image. To address these limitations, we propose P-PatchDiff, a scalable progressive patch diffusion framework for low-light image enhancement that dynamically adjusts patch size throughout the denoising process, enabling a gradual shift from local to global views. A Multi-Patch Alignment strategy is also introduced to normalise features across varying patch scales using an estimated global brightness proxy. Rather than pursuing pixel-level reconstruction accuracy, P-PatchDiff focuses on scalability and coherent brightness across the whole image, allowing the model to perceive multi-scale information and better enhance regions with varying brightness. We empirically demonstrate that P-PatchDiff effectively enhances images ranging from 400 \times 600 to 4K and is 80 \times faster than existing patch diffusion models while using less than 9GB of memory. The code is available at this https URL.
[CV-39] IT-TextFusion: Iterative Text-Image Interaction with Text-Guided Residual Refinement for Degradation-Aware Image Fusion
链接: https://arxiv.org/abs/2609.01092
作者: Siyang Liu,Peiyi Zhou,Tianle Jin,Rongrong Bian,Zheke Jin,Mengze Gao
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Text-guided image fusion has recently emerged as an effective paradigm for integrating multi-modal information while enabling flexible and task-oriented fusion control. However, existing text-guided fusion methods often rely on shallow semantic-visual interaction and limited attention mechanisms, which restrict their ability to robustly handle complex degradations and fully exploit textual guidance. In this paper, we propose an iterative text-guided image fusion framework that incorporates text-conditioned feature interaction across multiple fusion and refinement stages. The proposed method integrates deepest-level Cross-Attention, multi-scale Cross-Gate Fusion, and stage-specific text-conditioned modulation, allowing the global text embedding to condition hierarchical feature fusion and residual refinement. By repeatedly injecting the pooled text embedding across hierarchical decoder and refinement stages, the proposed framework provides degradation-aware global semantic conditioning while preserving complementary information from the visible and infrared modalities. Experiments on several benchmark datasets show that the proposed method improves several information-preservation and perceptual-quality metrics, while exhibiting metric-dependent trade-offs on some datasets.
[CV-40] Let Confidence Change Not the Prediction: Prediction-Preserving Repair for Post-hoc Calibration
链接: https://arxiv.org/abs/2609.01072
作者: Daehwan Kim,Haejun Chung,Ikbeom Jang
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Post-hoc calibration corrects reported confidence, yet a multiclass calibrator can also change the associated top-1 prediction. Accuracy captures only the net effect of these changes on correctness, not how often predictions change; the Top-1 Prediction Change Rate (TPCR) instead measures this frequency. We propose Calibrator-Output Repair for Top-1 Decision Preservation (CORD), the first post-fit adapter to impose exact prediction preservation by repairing the full calibrated probability vector. From the original and calibrated outputs alone, CORD determines the mass assigned to the original top-1. The calibrated conditional distribution allocates the remaining mass over the other classes, yielding a repaired vector whose own argmax recovers the original prediction. On the calibration split, CORD coordinates the repaired masses to retain the calibrated outputs’ mean mass on original predictions whenever attainable. The adapter alters neither the fitted calibrator nor its direct output, fits no additional supervised map, and requires no user- or validation-tuned hyperparameter. Across CIFAR-10/100 and ImageNet-1K, CORD attains zero TPCR by construction and lowers mean ECE, NLL, and Brier relative to the corresponding direct outputs in every dataset; paired gains persist under distribution shift and across calibration-set sizes. CORD thus removes the preservation constraint from calibrator fitting and assigns exact recovery of the original decision to subsequent output repair. Our code is available at this https URL.
[CV-41] Dyn-3D: Unveiling and Resolving Ego-Motion Ambiguity in Vision-Language Models
链接: https://arxiv.org/abs/2609.01059
作者: Jiayu Ding,Zhuodong Liu,Lei Zhang,Manyu Xiong,Hongbo Jin,Haoran Tang,Hongbo Zhang,Changen Zhu,Wenbo Xing
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:As Vision-Language Models (VLMs) tackle dynamic 3D spatial reasoning, ego-motion perception becomes essential to resolve monocular scale ambiguity. However, current models often overfit to smooth trajectory priors rather than genuinely understanding physical motion. Consequently, their spatial reasoning degrades severely under large displacements, a phenomenon we term Kinematic Collapse. This failure stems from spurious visual-motion correlations in natural videos and a lack of explicit physical supervision. To evaluate this, we introduce Dyn-3D, a benchmark using counterfactual 3D rendering to rigorously decouple visual changes from true kinematic properties. Furthermore, we propose the TempoVista framework, featuring the Kinematic-GSPO algorithm. By embedding metric physical ground truth into policy optimization, TempoVista explicitly grounds visual representations in 3D space. Experiments demonstrate that our approach significantly improves both motion estimation and robust spatial reasoning by utilizing camera dynamics as an effective geometric calibration signal.
[CV-42] SAGE: Subpopulation-Aware Generative Enhancement for Mitigating Spurious Correlations
链接: https://arxiv.org/abs/2609.01051
作者: Yiming Luo,Rongqiang Zhao,Jie Liu
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Spurious correlations pose a significant challenge to the robustness of modern machine learning. The inherent imbalance in dataset distributions often leads traditional Empirical Risk Minimization (ERM) models to rely on majority spurious attributes for classification, resulting in poor performance on minority groups. This problem becomes particularly challenging when the spurious attributes are unavailable. Existing group-label-free methods often upsample minority groups or misclassified real training examples; repeating the same instances can reduce effective diversity and encourage overfitting. To mitigate these spurious correlations from a data-centric perspective in the absence of prior knowledge, we introduce Subpopulation-Aware Generative Enhancement (SAGE), a two-stage generative augmentation framework. Using cluster-derived sub-labels and class labels, we fine-tune a conditional generative model and text encoder, generating targeted synthetic data to fill underrepresented regions in the training set and construct a balanced validation set for last-layer reweighting. We experimentally show that SAGE achieves 89.5%, 85.7%, and 79.1% worst-group accuracy on Waterbirds, CelebA, and MetaShift, respectively, outperforming the best group-label-free baselines by up to 7.7 percentage points.
[CV-43] ViTAMINS: An Empirical Study of Training Self-Supervised Vision Transformers with Synthetic Hard Negatives WACV2027
链接: https://arxiv.org/abs/2609.01041
作者: Nikos Giakoumoglou,Andreas Floros,Kleanthis-Marios Papadopoulos,Tania Stathaki
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: WACV 2027
Abstract:We introduce ViTAMINS, a method that integrates synthetic hard negatives into unsupervised vision transformer pretraining to improve representation quality. Our approach is thoroughly benchmarked on ImageNet and transfer learning, image retrieval, copy detection, and image, video segmentation tasks. Notably, our proposed negatives give rise to emergent properties, where learned representations contain explicit information about the semantic content of an image and serve as excellent classifiers (up to +11.3% over baselines). ViTAMINS achieves these benefits through simple modifications to existing contrastive frameworks and outperforms competing methods while being more resource efficient, e.g., our ViT-B surpasses V-JEPA with ViT-L. Our findings motivate reconsidering contrastive learning as a simpler yet powerful alternative to dominant generative and self-distillation approaches.
[CV-44] MultiGait: A Multi-Sensor Multi-Perspective Multi-Session Biometric Inference Benchmark and its Dataset
链接: https://arxiv.org/abs/2609.01036
作者: Julian Todt,Felix Morsbach,Philip Dissert,Thorsten Strufe
类目: Cryptography and Security (cs.CR); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:A lack of suitable datasets has limited the research into the privacy risks of novel smart city sensors, such as thermal cameras, depth cameras, and lidar. Given the number of unsubstantiated privacy claims and their potential widespread deployment into many people’s everyday life, understanding the privacy risks of these sensors – in isolation and in like-for-like comparisons – is crucial. With MultiGait, we collected the first multi-sensor, multi-perspective, multi-session gait-focused dataset, for the corresponding, and additional more far-reaching investigations. The dataset, validated with multiple state-of-the-art recognition systems, comprises various walking modes and annotated personal attributes for 199 individuals, to ensure the benefit for advanced studies including cross-sensor recognition and anonymization at the edge. MultiGait represents a foundation for rigorous privacy investigations, demonstrated through an extensive identity inference benchmark across eight sensors, four perspectives, and three recording sessions. Our benchmark incidentally reveals that sensors often assumed to be privacy-friendly do still entail considerable identity inference risks, while the poor cross-session generalization of existing methods underscores an important research gap.
[CV-45] Fi-ImageNet-1k: An OOD Benchmark From the Inside of the ImageNet-1k Validation Set
链接: https://arxiv.org/abs/2609.01027
作者: Ruslan Rozumnyi,Matěj Suchánek,Tomáš Vojíř,Klára Janoušková,Jiří Matas
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 21 pages, 8 figures, 16 tables (11-page main paper + supplementary material)
Abstract:Out-of-distribution (OOD) detection predicts whether a test image belongs to none of the predefined classes. To evaluate this task, benchmarks need images from outside the in-distribution (ID) data; typically, these are defined or collected in an ad hoc fashion. Since no ground truth is perfect, ID-labeled datasets themselves contain a natural source of OOD images. We exploit such annotation errors and present Fi-ImageNet-1k, an OOD dataset built from ImageNet-1k validation images that the recent ReImageNet reannotation effort assigned to no ImageNet-1k class. Each image was examined by expert human annotators supported by evidence from MLLMs, VLMs, and reverse image search, comparing it against all visually similar ID classes. We keep only images that could be assigned a specific class outside the ImageNet-1k label space. The resulting Fi-ImageNet-1k, with 655 images from 522 classes, is substantially more challenging than any commonly used OOD dataset. No evaluated combination of classifier and OOD detector achieves a false positive rate below 51% at 95% true positive rate (FPR@95). Compared to the recent NINCO, our dataset is 3.8x more challenging in the FPR@95 metric for state-of-the-art supervised OOD detection methods. Comments: 21 pages, 8 figures, 16 tables (11-page main paper + supplementary material) Subjects: Computer Vision and Pattern Recognition (cs.CV) Cite as: arXiv:2609.01027 [cs.CV] (or arXiv:2609.01027v1 [cs.CV] for this version) https://doi.org/10.48550/arXiv.2609.01027 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[CV-46] Low-Quality Face Recognition using Center Aligned Representations and Local Margin Constraints
链接: https://arxiv.org/abs/2609.01014
作者: Vedat Can Dilaver,Benjamin S. Riggan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accept at IEEE/IAPR IJCB 2026
Abstract:Low-quality face recognition (LQFR) remains challenging due to the difficulty of matching degraded query (probe) images against low-quality (LQ) enrollment (gallery) imagery and the scarcity of training data for large-scale models. While recent face recognition (FR) models perform well on high-quality (HQ) imagery, their accuracy drops significantly on LQ images with extremely low signal-to-noise ratio (SNR). Moreover, fine-tuning HQ-pretrained models on LQ data often improves LQ recognition at the expense of HQ generalization. This trade-off becomes more pronounced in modern evaluation settings spanning multiple datasets with varying image quality levels. To address these limitations, we propose a unified framework that combines three main components: (1) Local Probability Margin (LPM), which estimates per-sample difficulty directly from the model’s discriminative landscape; (2) Nested Attention Module (NAM), a new low-rank adapter module that embeds a self-attention mechanism within selected transformer layers; and (3) Quality Gating Protocol (QGP), where an off-the-shelf image quality estimator modulates the adapter contribution at test time, enabling a single model to handle the full quality spectrum without sacrificing HQ performance. Experiments on surveillance (TinyFace, SurvFace) and standard (IJB-B, IJB-C) face recognition benchmarks demonstrate consistent gains in both identification and verification. Code and models will be released at this http URL.
[CV-47] Does This Moment Justify the Recommendation? Counterfactual Behavior-Grounded Evidence Retrieval for Personalized Video Recommendation
链接: https://arxiv.org/abs/2609.00996
作者: Xin Liu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 8 pages
Abstract:Personalized video recommendation predicts user preference at the video level, while temporal video grounding localizes query-relevant moments. However, strong localization does not establish whether the retrieved moment constitutes valid evidence for recommending the video to a particular user. We study counterfactual behavior-grounded evidence retrieval, which separates where personalized evidence occurs from whether such evidence exists and evaluates whether model predictions respond consistently when that evidence is replaced. We introduce CBGER-10K, containing 5,000 controlled factual–counterfactual pairs for 3,026 users, where each pair replaces only the focal behavior-supported segment while preserving the user, temporal position, and hard distractors. We further propose CBGER, a compact framework that decouples segment-level localization from video-level evidence estimation and learns both through structured counterfactual supervision. CBGER achieves 0.4432 MRR, 0.6977 Pair Accuracy, and 0.6987 Intervention Consistency across five adapted personalized-highlight and temporal-grounding baselines. Notably, compared with QD-DETR, its MRR improvement is not statistically significant, while Pair Accuracy improves by 11.03 points. These results show that accurate temporal localization does not necessarily imply reliable personalized evidence existence, motivating explicit evaluation of Whether alongside Where.
[CV-48] CQF-HMR: Continuous Quaternion Flows for Probabilistic 3D Human Mesh Recovery from a Single Image
链接: https://arxiv.org/abs/2609.00995
作者: Cuong Le,Bao-Long Tran,Pavlo Melnyk,Tahereh Dehdarirad,Bastian Wandt,Mårten Wadenbäck
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Under submission
Abstract:Recovering 3D digital humans from a single 2D image is an ill-posed computer vision problem due to the loss of depth information. Probabilistic 3D human pose estimation compensates for this by estimating a set of 3D hypotheses from a prior distribution via generative models. However, most prior work focuses only on 3D keypoints, which often leads to implausible poses that are difficult to apply to downstream tasks, e.g. animation or digital humans. SMPL-based methods are more scalable thanks to the explicit body priors, but it requires more complex modeling of the generation process due to the non-additive nature of the joint rotations. In this work, we propose a novel approach for probabilistic 3D humans using quaternion-constrained continuous normalizing flows conditioned on 2D pose estimations. Our proposed quaternion flows show significant advantages over approaches using other rotation representations. Experiments demonstrate state-of-the-art results of our method on Human3.6M, particularly in ambiguous settings, and comparable pose estimation accuracy on challenging 3DPW and EMDB benchmarks.
[CV-49] EvoGS: Modeling Deformation Evolution for Dynamic Gaussian Splatting
链接: https://arxiv.org/abs/2609.00994
作者: Wei Dong,Shahram Shirani,Jun Chen,Han Zhou
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted by Pacific Graphics 2026 (journal track)
Abstract:Recent extensions of 3D Gaussian Splatting (3DGS) enable real-time novel view synthesis in dynamic scenes by learning time-conditioned Gaussian deformations. However, existing MLP-based methods typically estimate deformations independently at each timestamp, making them less robust to large or abrupt motions. To address this issue, we propose \textbfEvoGS, a 3DGS-based dynamic reconstruction framework that models Gaussian deformation as a temporal evolution process. EvoGS maintains persistent deformation states for each Gaussian, extrapolates future states from historical deformation states, and corrects the predictions with MLP-derived observations. The correction is adaptively weighted using a temporal residual memory and evolution statistics such as deformation velocity and trajectory deviation. To further improve reconstruction quality, EvoGS introduces deformation-aware densification. Clone and split operations are performed along corrected deformation directions, while an uncertainty-aware strategy suppresses densification for Gaussians with unstable deformation histories. Experiments show that EvoGS improves dynamic novel view synthesis quality and achieves competitive performance across benchmarks.
[CV-50] Semi-Supervised Virtual Staining via Morphology Preservation and Histopathological Realism Constraints
链接: https://arxiv.org/abs/2609.00984
作者: Baoshun Wang,Weiping Lin,Linwu Wang,Yihuang Hu,Baptiste Magnier,Liansheng Wang
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 10 pages
Abstract:Virtual staining aims to computationally generate target-stained histopathological images while reducing the cost and time associated with conventional staining procedures. However, existing methods rely predominantly on strictly paired and accurately registered training data, which are difficult and expensive to obtain in routine practice. To reduce this dependence, we propose a stable semi-supervised virtual staining framework that jointly exploits both limited paired data and abundant unpaired source images. Directly incorporating unpaired images is challenging because their generated results lack corresponding targets for supervision, potentially leading to unrealistic staining, morphological degradation, or even training collapse. To obtain reliable supervision from these images, Hessian-derived morphology preservation extracts structural cues from each source image and constrains the generated output to retain tissue morphology. Histopathological realism constraints further guide the output toward plausible target-stain characteristics, preventing the source-derived structural supervision from degenerating into contour enhancement or simple color transformation. Together, the two components suppress structural and appearance drift, stabilize semi-supervised stain translation, and promote the preservation of diagnostically relevant information. Extensive experiments on HE-to-IHC translation for Ki67 and HER2, as well as FFPE-to-HE translation, demonstrate consistent improvements in image quality, morphology preservation, robustness, and downstream diagnostic performance. Code will be available.
[CV-51] ReFlowSET: Representation-Aligned Latent Flow Matching for SAR-to-EO Image Translation
链接: https://arxiv.org/abs/2609.00968
作者: Jeonghyeok Do,Seungchul Lee,Munchurl Kim
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Please visit our project page this https URL
Abstract:SAR-to-EO image translation aims to generate electro-optical (EO) imagery from synthetic aperture radar (SAR) observations. Existing latent diffusion approaches typically inherit a predetermined autoencoder, although reconstruction fidelity can vary substantially across codecs and modalities. Because the latent codec affects the round-trip preservation of both SAR conditions and EO targets, codec selection constitutes a fundamental design choice; nevertheless, existing methods largely rely on codecs pretrained on natural images. To remedy this, we introduce ReFlowSET, a conditional latent flow-matching framework that selects its codec through a joint SAR–EO reconstruction audit. Rather than inheriting a heavyweight pretrained generator, ReFlowSET trains a substantially smaller conditional DiT from scratch in the selected latent space, using dual-stream SAR conditioning followed by joint feature refinement. To provide semantic guidance for this from-scratch training, intermediate noisy-EO features are aligned with clean target-EO representations extracted by a frozen vision foundation model. This alignment is used only during training and introduces no additional inference cost. Experiments on QXS-SAROPT and SAR2Opt demonstrate state-of-the-art performance across diverse perceptual fidelity and distributional metrics. Code and pretrained weights are publicly available at this https URL.
[CV-52] Conditional Flow Matching for Cross-Field MRI Harmonisation
链接: https://arxiv.org/abs/2609.00960
作者: Baris Imre,Aram Salehi,Levente Baljer,Andrew Webb,Marius Staring,Efe Ilicak
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 10 pages, 2 figures
Abstract:Magnetic resonance images of the same subject look markedly different across field strengths, which complicates the comparison and pooling of data across sites. We address cross-field brain-MRI translation for the MRIxFields2026 challenge, and in particular its Task~3: a single model that translates between any directed pair of the five field strengths and across three contrasts. We phrase the problem as a conditional flow matching path: because the source and target volumes are spatially registered, we learn a velocity field that carries the source slice directly to the target slice, rather than starting from noise. To learn this mapping from only three paired subjects, the unified model is trained in three stages: a degradation-bridge pretraining that distills a restoration prior from the abundant unpaired retrospective cohort, a cross-field finetuning over all directed pairs on the paired cohort, and an adversarial refinement that sharpens the output. At inference, we integrate the learned velocity with a second-order Heun solver in a handful of steps. A restoration prior learned without any paired data already reaches a mean SSIM of 0.837, and each subsequent training stage improves on it. A single 6.3M-parameter model thereby covers all 60 field-pair and contrast combinations, with inference in five solver steps per slice. On the challenge evaluation set the model reaches a mean SSIM of 0.909, averaged over the three contrasts, outperforming regression and diffusion baselines built on the identical network on all three challenge metrics.
[CV-53] Candidate-Expanding Routing with Permutation-Stabilized Experts for Mixed-Format Medical VQA
链接: https://arxiv.org/abs/2609.00959
作者: Hai-Dang Nguyen,Huy-Hieu Pham
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 10 pages, 3 figures, 3 tables
Abstract:Mixed-format medical visual question answering (VQA) requires stable option selection and machine-readable free-text output. The two formats fail differently: multiple-choice predictions can change with option symbols or positions, while clinically plausible open answers can fail automated evaluation when serialization is malformed. We address both challenges with an answer-text memory, a permutation-stabilized vision–language expert, and a sparse candidate- expanding router. The cyclic schedule follows prior work; our contribution is to make expert top-2 a routable candidate alongside memory and expert top-1. On a 1,403-case retrospective internal analysis, this expansion improves a matched binary router from 88.95% to 91.73% (+2.78 percentage points; 95% CI 1.57–3.99), with 56 rescued errors and 17 regressions. Oracle coverage rises from 90.31% to 96.15%, and the final submitted configuration reaches 92.23% on the same retrospective split. For open questions, strict generation and deterministic guards produce 475/475 schema- valid participant-facing outputs without repair, retry, or hard-gate failure. Visual ablations reveal substantial textual dependence. Candidate expansion supplies the principal controlled routing gain; open-path evidence establishes output-contract validity rather than clinical correctness in medical use or deployment.
[CV-54] PredErase: Training-Free Object-and-Effect Removal with Predictive Latent Guidance
链接: https://arxiv.org/abs/2609.00956
作者: Waikit Xiu,Qiang Lu,Junbiao Chen,Xiying Li
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 16 pages, 6 figures. Code: this https URL
Abstract:Removing an object is not the same as filling its mask. Cast shadows and contact shading usually lie outside the user-provided instance mask M_obj, so a frozen Fill model that edits only that mask leaves the object’s photometric footprint on nearby surfaces. Supervised removers learn this joint erasure from paired clean plates. Training-free editors freeze pretrained weights, yet most still treat M_obj as the entire editable support and steer sampling with CLIP or DINO energies that do not predict the occluded scene. We present PredErase, a training-free inference procedure on frozen FLUX.2 and I-JEPA. The method separates where Fill may rewrite pixels from what structure should occupy the hole. A contact-band expansion M_flux of M_obj exposes local residuals on the supporting plane. I-JEPA, pretrained for masked token prediction, supplies a context-conditioned hole target in representation space; sparse projected gradients align decoded Fill completions with that target inside the instance, while coordinates outside the packed support stay locked. Under instance-only masks on RemovalBench, RORD-Val, and DEFACTO-Val, PredErase improves the native FLUX.2 backbone. Supervised removers remain stronger on several full-image appearance metrics; the supported claim is training-free object-and-effect editing of frozen Fill, not replacement of paired-data erasers.
[CV-55] ASSERT: Adaptive Stochastic Sampling for Robust Diffusion Models on Analog Compute-in-Memory Hardware
链接: https://arxiv.org/abs/2609.00955
作者: Yuannuo Feng,Yizhe Chen,Wenshuai Yao,Yuxin Xie,Ngai Wong,Wenyong Zhou,Wang Kang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted by ASP-DAC 2027, Tokoy, Japan
Abstract:Diffusion models achieve strong image generation quality but incur high iterative denoising costs. Analog compute-in-memory (CIM) can accelerate matrix-vector multiplications, yet spatial memory variations perturb weights and accumulate during sampling. Unlike conventional neural networks, diffusion models’ temporal sensitivity to hardware noise remains underexplored. We investigate diffusion inference using a noise model calibrated and validated against measurements collected from multiple physical CIM chips. Our results show that the early, high-noise denoising stage is substantially more vulnerable than the final refinement stage. A first-order trajectory analysis attributes this behavior to the repeated propagation of correlated prediction errors induced by a fixed hardware mapping. Based on this observation, we propose ASSERT, a training-free sampler that uses higher stochasticity early and smoothly transitions to deterministic denoising. The injected stochasticity changes subsequent activation trajectories and thereby reduces their alignment with persistent spatial errors. Across the evaluated settings, ASSERT achieves up to 2.58 \times lower FID than deterministic DDIM on high-resolution datasets and 7.68 \times lower FID in the CIFAR-10 step-count study, without changing model parameters or the number of network evaluations.
[CV-56] CERF: Communication-Efficient and Retraining-Free Collaborative Perception ICASSP2026
链接: https://arxiv.org/abs/2609.00951
作者: Jiuwu Hao,Ziyi Ni,Liguo Sun,Yuting Wan,Yueyang Wu,Ti Xiang,Haolin Song,Pin Lv
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted by ICASSP 2026
Abstract:Collaborative perception shares information among multiple agents to obtain a comprehensive scene representation, enhancing the perceptual capability of individual agents. However, most existing methods rely on transmitting and fusing dense feature maps for collaboration, which incurs inevitable communication overhead and heterogeneity challenges, limiting their practicality for real-world deployment. To address these challenges, we propose CERF, a novel Communication-Efficient and Retraining-Free framework for open heterogeneous collaborative perception. In CERF, we introduce a new virtual modality (termed Poture), which is generated from the perception outputs of other agents, to augment the extracted Bird’s Eye View (BEV) features of the ego agent. To mitigate transmission delays, we employ a Kalman-filter based tracker and a motion forecasting model to derive the current predictions from historical perception results. Extensive experiments demonstrate that CERF achieves performance comparable to mainstream intermediate-collaboration methods while reducing communication overhead by 95% across various downstream tasks. Furthermore, CERF enables seamless integration of unknown heterogeneous agents into the existing collaborative framework without additional retraining costs. Code is available at this https URL.
[CV-57] Beyond the Image Plane: World-Grounded Queries for Multi-Object Tracking
链接: https://arxiv.org/abs/2609.00924
作者: Orcun Cetintas,Guillem Brasó,Tim Meinhardt,Laura Leal-Taixé
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Monocular videos record 3D scenes as sequences of 2D image-plane projections, obscuring depth and spatial relationships. Multi-object trackers localize and associate objects primarily using appearance and geometry observed only in the image plane, inheriting these ambiguities. To address this limitation, we introduce PLANET, an end-to-end multi-object tracker designed to move beyond the image plane. As an enabling step, we lift existing 2D tracking datasets into 3D. We then form world-grounded queries by embedding reconstructed 3D scene geometry into the features and positional encodings used during query formation. An auxiliary 3D location prediction task further encourages the queries to encode object positions during training. A complementary dual-resolution temporal memory preserves this evidence across longer temporal gaps. As a result, PLANET achieves state-of-the-art performance across three diverse benchmarks.
[CV-58] On-the-Fly3R: Towards Robust Online 3D Reconstruction with Feed-Forward 3R Models for Large-Scale UAV Scenarios ICRA2027
链接: https://arxiv.org/abs/2609.00923
作者: Zhe Shen,Liyuan Lou,Yifei Yu,Guanbo Wang,Quanjian Ji,Xin Wang,Zongqian Zhan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: This paper was submitted to the ICRA 2027 for consideration. Copyright would be transferred if it got accepted
Abstract:While feed-forward 3D reconstruction (3R) offers efficient end-to-end modeling, its application in large-scale UAV mapping is hindered by the prohibitive memory cost of Transformer attention. Current scalable streaming 3R methods assume temporally and spatially continuous inputs, rendering them ineffective for the weakly ordered or unordered image streams common in cross-strip UAV operations. To address this, we propose On-the-Fly3R, a training-free, progressive online 3D reconstruction framework for large-scale UAV images that upgrades various 3R backbones for large-scale UAV scenarios. Our method enables reconstruction from unordered inputs via retrieval-guided dynamic subset construction, which adaptively selects spatially relevant images. To further improve the robustness, a validation-rejection-retry mechanism is designed to guarantee global consistency, performing a pre-integration consistency check and automatically rejecting misaligned images and retrying with alternative subset. Finally, inspired by VSLAM, pose graph optimization based on the retrieval loop closure is employed to mitigate camera drift. Evaluations on several UAV benchmarks show that our On-the-Fly3R successfully scales various 3R models to over 5,000 images across square-kilometer UAV scenes, delivering substantially superior accuracy compared to several SOTA streaming 3R methods. Code is available at this https URL
[CV-59] VerNav: Verifier-First Low-Latency Vision-and-Language Navigation
链接: https://arxiv.org/abs/2609.00920
作者: Zhixin Wang,Chengzheyi Yao,Leyuan Liu,Xiaosong Zhang,Yongzhao Zhang
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注: 9 pages, 7 figures, 5 tables
Abstract:Vision-and-Language Navigation (VLN) requires an agent to navigate through unseen 3D environments according to natural-language instructions. Explicit reasoning can improve instruction understanding and semantic grounding, but autoregressive generation at every step accumulates large decision-stage latency over multi-step navigation. We propose VerNav, a verifier-first framework for low-latency LLM-based VLN. The verifier reduces decision-stage latency by replacing per-step autoregressive generation with batched action verification, while an entropy-based adaptive generator is invoked only for uncertain decisions to produce compact state evidence. To further improve navigation performance with the verifier, we introduce a two-stage alignment scheme: (i) VPO improves local action-preference alignment in static verifier training, and (ii) step-level reinforcement fine-tuning provides dense progress rewards over multi-step navigation rollouts during dynamic task execution. Experiments on the Room-to-Room (R2R) benchmark show that the verifier-only decision path of VerNav achieves competitive navigation performance among representative LLM-based VLN agents while reducing average decision-stage LLM latency per step by more than 10\times compared with autoregressive methods.
[CV-60] A multicenter benchmark and clinically structured metric for coronary CTA report generation
链接: https://arxiv.org/abs/2609.00909
作者: Zhiyu Ye,Yue Sun,Limiao Zou,Cheng Xu,Keting Xu,Tong Hu,Yue Yu,Hairong Zheng,Yining Wang,Tong Zhang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Reliable evaluation of automated coronary computed tomography angiography (CCTA) report generation requires standardized multicentre benchmarks and clinically structured metrics. We established a four-centre benchmark comprising 3,021 CCTA series from 818 patient-report pairs to evaluate seven open-source three-dimensional vision-language models. We developed CSM _\textCCTA , a clinically structured metric for CCTA report evaluation, with patient-, vessel-, and segment-level variables defined according to clinical guidelines. Report pairs are compared at the finest shared anatomical level, and the contributions of different clinical components are weighted based on expert assessments. We estimated these weights using 70 expert-scored cases and evaluated clinical alignment in a non-overlapping set of 30 cases. CSM _\textCCTA showed a strong correlation with radiologist scores (Pearson’s r=0.97 , p0.001 ), exceeding the next-best metric, FORTE ( r=0.70 ), by 0.27, and agreed with expert preferences in 115 of 160 pairwise comparisons (71.9%). Under controlled perturbations, CSM _\textCCTA remained stable to clinically equivalent wording and decreased monotonically with progressive information omission. In the multicenter benchmark, the CCTA-trained C2RG model achieved the highest CSM _\textCCTA scores across all four hospitals, although its performance remained far from optimal. In contrast, CCTA-irrelevant reports accounted for up to 98.7% of the outputs from generalist models. Together, the benchmark provides a standardized setting for model comparison, while CSM _\textCCTA enables clinically structured evaluation of finding agreement and anatomical specificity. These results support a more clinically aligned and anatomically resolved approach to evaluating CCTA report generation. Code is available at this https URL.
[CV-61] HELIOS: From midnight to noon continuous outdoor urban scene relighting
链接: https://arxiv.org/abs/2609.00901
作者: Hala Djeghim,Nathan Piasco,Luis Roldão,Moussab Bennehar,Dzmitry Tsishkou,Céline Loscos,Désiré Sidibé
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Modifying the illumination of driving images is a fundamental challenge, as most datasets are captured at specific times of day. Existing methods rely on synthetic data or paired multi-illumination supervision, which limits their generalization to the diverse and challenging conditions of real-world scenarios. To address this, we propose HELIOS, a novel image relighting approach that relies on unlabeled real-world datasets without requiring any paired images for training. Our approach integrates albedo-based conditioning into a cycle-consistent diffusion pipeline to prevent identity collapse and ensure accurate domain translation. To handle low-visibility nighttime conditions, we introduce a robust albedo distillation strategy that transfers structural stability from the daytime domain. Additionally, we replace traditional text prompts with a fine-grained control mechanism based on GPS-derived solar angles, enabling smooth and continuous lighting manipulation across the day-night cycle. Through extensive evaluation and a user study, we demonstrate that HELIOS produces structurally consistent and realistic results in both night-to-day and day-to-night tasks, outperforming state-of-the-art methods.
[CV-62] Vision-Language-Guided Pseudo-Labels for Unsupervised Domain Adaptation in Semantic Segmentation for Waste Sorting ECML-PKDD2026
链接: https://arxiv.org/abs/2609.00898
作者: Udo Schlegel,Shubhangi,Gabriel Dax,Sai Rahul Kaminwar,Florian Karl,Thomas Seidl
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 18 pages, 2 figures, 2 tables, accepted at ECML-PKDD 2026
Abstract:Obtaining labeled data for semantic segmentation in applied settings (e.g., autonomous driving, industrial waste sorting) is expensive and often infeasible at scale. We present a cross-modal pseudo-labeling pipeline that enables unsupervised domain adaptation without any target-domain annotations. The pipeline is built on two core foundation models: SAM generates class-agnostic region proposals, and EVA-CLIP assigns semantic labels based on region-text similarity, with confidence filtering ensuring that only reliable pseudo-labels are used for self-training a segmentation model. As an optional extension, BLIP provides language-grounded verification for ambiguous regions, thereby improving pseudo-label quality without altering the overall pipeline. Evaluated on two domain shifts, synthetic-to-real autonomous driving and, with a primary focus, lab-to-factory industrial waste sorting, the pipeline consistently improves over source-only baselines. Our results demonstrate that pseudo-label quality, not quantity, is a decisive factor in self-training under domain shift, and that cross-modal language grounding offers a practical path to reliable automatic annotation in deployment-critical applications.
[CV-63] Denoising Diffusion Generative Models Secretly Calculate Attentions
链接: https://arxiv.org/abs/2609.00885
作者: Farzan Haddadi,Leila Monfared,Ebrahim Rezaii,Mohammadreza Malek-Mohammadi,Pejman Zakalvand,Narges Mokhtari
类目: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG); Neural and Evolutionary Computing (cs.NE)
备注: submitted to IEEE Trans on Pattern Recog. Machine Intellig
Abstract:Denoising diffusion models are the dominant architecture for image generation, whereas most natural language generation and modeling are primarily handled by well-known transformer architectures employing attention mechanism. Here, we show that diffusion models also inherently use an attention mechanism very similar to that of transformers. Therefore, attention emerges as a universal machine learning principle, based on a general training objective. We also show similarities in basic functional principle of auto-encoders and attention-based models. These equivalences allows us to interchange these designs based on practical requirements. As an example, we can reformulate the diffusion framework to reduce the lengthy training process and computation-intensive image generation. Using this approach, a simplified algorithm is proposed for image generation which is based on attention mechanism. Results show that the attention-based implementation achieves comparable performance with significantly less effort and computational resources.
[CV-64] Benchmarking Vision-Language Models for Automated Pathology Diagnosis and Report Generation
链接: https://arxiv.org/abs/2609.00866
作者: Yumi Lee,Harim Oh,Hyoryung Kim,Minji Kim,Eunsu Kim,Hyeseong Lee,Junya Fukuoka,Andrey Bychkov,Jijgee Munkhdelger,Rajiv Kumar Kaushal,Ayushi Sahay,Rajni Yadav,Bharathi Prabakaran,Sulen Sarioglu,Serdar Balcı,Ilknur Turkmen,Yuri Tolkach,Christian Harder,Julian Westerdorf,Reinhard Buettner,Audun Ljone Henriksen,Sepp De Raedt,Byung Hyun Lee,Sungjin Lim,Joohoon Lee,Gwanghyun Kim,Se Young Chun,Suryakant Singh,Saarthak Kapse,Prateek Prasanna,Kyung A Kim,Yousun Kang,Sehwan Yoo,Sungman Hong,Shubham Innani,Michael Feldman,Spyridon Bakas,Ujjwal Baid,Prasad Dutande,Suhas Gajare,Bhakti Baheti,Serkan Sökmen,Ece Tuğba Cebeci,Ahmet Halıcı,Musa Balcı,Kardelen Peçenek,Srividhya Sainath,Kyongseok Jang,Messi H.J. Lee,Noorul Wahab,Bodong Du,Jiaming Zhang,Qixiang Zhang,Jang-Hwan Choi,Sangjeong Ahn
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:The rapid advancement of vision-language models (VLMs) has accelerated progress in computational pathology; however, whole-slide image (WSI)-based pathology report generation remains limited by the scarcity of large-scale WSI–report datasets and the complexity of mapping spatially distributed visual patterns to structured clinical text. To address this, we introduce a clinically curated Pan-Asia WSI–report dataset of approximately 10,500 pairs from five institutions and establish the REG 2025 benchmark through a MICCAI challenge for systematic evaluation of multimodal models. We analyze submitted methods spanning pretrained VLMs, multiple-instance learning frameworks, hierarchical expert models, retrieval-augmented generation, and cross-modal Transformers. Rather than indicating that VLM use alone was sufficient for superior performance, the results suggest that top-performing methods benefited from structured report representations, hierarchical diagnostic decomposition, and effective multimodal grounding. We identify key limitations, including instability in quantitative attribute estimation (e.g., numeric hallucination) and a tendency toward diagnostic overspecification, with some errors resembling known diagnostic pitfalls in routine pathology. These findings establish REG 2025 as a benchmark for evaluating WSI-based structured report generation and vision-language understanding in computational pathology, providing insights for the design of clinically grounded multimodal pathology models.
[CV-65] raining-Free Inpainting Across Domains with a Frozen Text-to-Image Diffusion Model
链接: https://arxiv.org/abs/2609.00862
作者: Zhenhuan Wang,Fengyi Yuan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 30 pages, 11 figures, including supplementary material
Abstract:We show that a frozen generic text-to-image diffusion model can perform conditional inpainting across three evaluated natural-image domains with one fixed controller configuration, without inpainting-specific weight training, dataset-specific weight adaptation, or learned inpainting-specific conditioning channels. Step-PI augments known-region projection with boundary-interior latent feedback, persistent PI state, and a predefined four-field release schedule that modulates controller signals along the reverse trajectory. Developed only on Main35-disjoint CelebA-HQ pilots, the controller transfers unchanged to AFHQ and Places2. Across two field-identical comparisons on the same 3,500 cases, adding persistent state and replacing uniform release with the predefined schedule each improve all 15 dataset-metric cells; 95% bootstrap intervals exclude zero for all five metrics in both comparisons. In descriptive native-route comparisons, Step-PI leads LanPaint and PILOT (the closest evaluated training-free baselines using vanilla SD1.5) on all five equal-dataset macro metrics. Inpainting-trained systems retain the absolute metric leads but rely on substantial inpainting-specific offline optimization. Our method provides a complementary approach for repurposing a frozen generic text-to-image model for cross-domain inpainting through test-time latent control.
[CV-66] ADGNet: Asymmetric Dual-text Guided Network for Infrared Small Target Detection
链接: https://arxiv.org/abs/2609.00853
作者: Tongtong Wang,Mingzhu Xu,Chenglong Yu,Jing Wang,Xiaohui Lin,Weili Guan
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 10 pages, 10 figures. Accepted by the 34th ACM International Conference on Multimedia (ACM Multimedia 2026)
Abstract:InfRared Small Target Detection (IRSTD) is a challenging task. Relying solely on pixel-level information, vision-only methods struggle to distinguish targets from clutter. Current multimodal methods typically describe both targets and backgrounds with a single textual prompt. Such an approach lacks dedicated regional guidance and ignores infrared semantic asymmetry. Consequently, it provides insufficient background suppression information and introduces severe feature optimization conflicts, overwhelming small targets with noise. To address these issues, we propose a novel Asymmetric Dual-text Guided Network (ADGNet). Specifically, accounting for the infrared semantic asymmetry, we first design the Asymmetric Dual-text Prompt (ADP), comprising an image-agnostic abstract target prompt and an image-specific detailed background prompt. To leverage these prompts, we introduce an Asymmetric Dual-Branch Interaction (ADBI) module to separately guide visual features with their respective text priors, protecting targets from noise while suppressing background clutter. Subsequently, we introduce an Adaptive Feature Aggregation (AFA) module to dynamically fuse features from the two branches. Furthermore, we construct a multimodal Asymmetric Image-Text Infrared (AITIR) dataset by providing asymmetric text annotations for three public datasets (IRSTD-1K, NUDT-SIRST, and SIRST). Extensive experiments demonstrate that ADGNet outperforms 21 state-of-the-art (SOTA) methods. Code is available at this https URL.
[CV-67] An Intelligent Decision Support System for Emotion Monitoring using Microscopic Fixational Dynamics
链接: https://arxiv.org/abs/2609.00846
作者: Xiangyu Shen,Feiyang Deng,Zijian Dai,Aibin Chen,Jizheng Yi,Jie Li,Hongbo Jiang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 22 pages, 15 Figures
Abstract:The rising prevalence of psychological disorders necessitates effective emotion monitoring, yet current methods relying on facial or physiological signals often suffer from intrusiveness and privacy issues. This paper proposes an intelligent decision support system and pervasive edge-computing framework that leverages smart glasses and a companion smartphone to infer emotional states from microscopic visual fixation patterns. Moving beyond traditional macroscopic gaze metrics, the proposed system extracts and decomposes three distinct neurophysiological micro-movements: microsaccades, ocular drifts, and ocular microtremors. We introduce an interpretable hybrid artificial intelligence pipeline combining a multi-head attention mechanism, extreme gradient boosting, and a support vector machine to extract deep temporal features, quantify their physiological importance, and perform efficient on-device classification. Through an extensive evaluation involving 60 volunteers, we rigorously validate the framework under a strict leave-one-subject-out cross-validation protocol across both controlled and naturalistic mobile scenarios. Ablation studies unequivocally demonstrate that these fixational micro-movements are substantially more discriminative for emotion inference than traditional macroscopic features. Furthermore, aligned with contemporary affective science, the system incorporates a few-shot personalization mechanism to bridge universal physiological baselines with individual emotional heterogeneity, achieving a highly robust personalized F1-score of 83.6%. This work establishes a physiologically interpretable, unobtrusive, and deployable paradigm for continuous real-time emotion monitoring.
[CV-68] Residual Kalman Dynamics for Event-Based UAV Forecasting ECCV2026
链接: https://arxiv.org/abs/2609.00839
作者: Per Nyblom,Hannes Ovrén,David Gustafsson
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to the Workshop on Neuromorphic Vision (NEVi) at ECCV 2026, 16 pages
Abstract:We study short- and mid-horizon UAV bounding-box forecasting on the FRED event-camera dataset. We use a constant-velocity Kalman filter over a full center-size box state as a strong physical baseline, and train a residual model to predict acceleration-like corrections from recent box history, filtered state features, and local event representations. This simple residual formulation consistently improves over the Kalman baseline, with event-conditioned models giving the strongest results among the evaluated methods. We further show that part of the residual target is predictable from anchor position and velocity alone, indicating that canonical FRED results can reflect both visual evidence and dataset-specific motion priors. To analyze this effect, we introduce decorrelated subsets as a diagnostic stress test, showing that event-conditioned residual models retain useful predictive signal even when measured position- and velocity-based shortcuts are weakened.
[CV-69] Visual Attention Faithfulness in Vision-Language Models is Heterogeneous EMNLP2026
链接: https://arxiv.org/abs/2609.00830
作者: Xurui Song,Weishi Wang,Zhongqi Yue,Kuluhan Binici,Tao Bai,Hongxin Shao,Daniel Dahlmeier,Jun Luo
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: EMNLP 2026
Abstract:Whether attention weights faithfully reflect model reasoning has been actively debated in NLP, yet this question remains largely unexplored for the visual modality in Vision-Language Models (VLMs). We address this gap through causal perturbation analysis on current VLMs, evaluating both the comprehensiveness and sufficiency gap of attention-ranked visual tokens. Our analysis reveals that visual attention faithfulness is heterogeneous, manifesting in three distinct processing modes: Faithful-Sufficient, where top- k attention tokens are both necessary and sufficient for prediction; Faithful-Distributed, where they are necessary but broader visual context remains required; and Non-Focal, where no localized attention region is individually necessary while visual information remains an essential trigger for prediction. Furthermore, human-annotated ground-truth regions satisfy comprehensiveness in only \sim 60 % of cases compared with model attention rankings, revealing systematic divergence between model visual reliance and human intuition. We demonstrate these patterns across both general VQA on VQAv2 and document tasks on VRDU and ChartQA, showing that visual attention faithfulness varies systematically with processing demands and model architectures rather than being uniformly faithful or unfaithful.
[CV-70] Can Scene Text Recognition Read Rare Compositions?
链接: https://arxiv.org/abs/2609.00816
作者: Genpei Zhang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Preprint. Under review. Main text plus appendix: 10 figures, 13 tables
Abstract:Scene text recognition is reported as 89–97% accurate on the six standard benchmarks, and the problem is widely treated as saturated. We present an alternative reading. When the same test images are stratified jointly by ground-truth word rarity and character n-gram novelty against a reference corpus, accuracy at the rare-word x rare-trigram corner of the resulting 5x5 grid drops 10–18 pt below the q3/q3 centre across nine English specialised recognisers, and the same direction (corner below centre) holds on all 13 of 13 (language, model) pairs we test across four writing systems (Latin, Han, Han+kana, Arabic). The drop is not a capacity bottleneck. A 6x vision-backbone scale-up (CLIP4STR-Base 158M - CLIP4STR-Huge 1.0B, OpenCLIP ViT-H/14 LAION-2B) leads every benchmark in aggregate accuracy yet leaves the stress corner unchanged (86.9 - 86.5, within paired-bootstrap noise). Four converging probes–layer-wise probing, confidence-when-wrong, attention re-balancing, and a cross-script commit-vs-abstain error split–localise the failure to the autoregressive decoder’s lexical prior. We then ask how much of the gap existing techniques recover. Of 16 non-architectural mitigations, the largest mean q5/q5 gain is +1.3 pt and none clears the paired-bootstrap noise floor; the only intervention that does is the architectural shift from autoregressive to CTC decoding (SVTRv2, +2.5 pt, p=0.02, n=474). A confidence-routed AR-CTC ensemble adds a directionally consistent +0.6 pt that stays within noise, and its dominant learned coefficient is each model’s own minimum-softmax confidence–independently echoing the mechanism above. No configuration we test improves both the compositional corner and aggregate accuracy. The rare-input long tail thus points to architectural change rather than added capacity.
[CV-71] RingMoClaw: An Experience-Inspired Multi-Agent Framework for Self-Evolving Research in Remote Sensing
链接: https://arxiv.org/abs/2609.00814
作者: Kaiyue Kang,Qixuan He,Peijin Wang,Yingchao Feng,Chao Ren,Kangxin Wang,Wenhui Diao,Yixiao Wang,Liangjin Zhao,Kaiwen Wei,Nayu Liu,Xian Sun
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Remote sensing visual models have continuously advanced various interpretation tasks. However, the research process behind model improvement still heavily relies on manual expertise, requiring extensive trial-and-error iterations in model design, data processing, and performance diagnosis. Existing agent-based approaches mainly focus on task execution and workflow orchestration, while lacking the capability of autonomous research iteration for continuous performance optimization. To address this issue, we propose RingMoClaw, an experience-inspired self-evolving multi-agent framework for remote sensing visual interpretation. RingMoClaw integrates a research branch, a quality-control branch, and a dual-stream dynamic experience bus to establish a closed-loop optimization process covering strategy generation, experiment execution, independent review, and experience accumulation. The heterogeneous Critic mechanism provides stage-wise diagnosis and feedback, while the dual-stream experience bus incorporates external knowledge and internal experimental experience to guide strategy evolution and eliminate ineffective searches. Extensive experiments on four remote sensing downstream tasks, including object detection, scene classification, semantic segmentation, and change detection, demonstrate the effectiveness and generalization of RingMoClaw. Compared with the corresponding baseline models, RingMoClaw improves performance by 1.84% mAP _50 on object detection and achieves consistent gains across the other three tasks, while reducing the required evolution steps by over 40% compared with existing research automation frameworks. These results suggest that RingMoClaw offers a feasible route from task execution toward continuous research driven model evolution in remote sensing.
[CV-72] ReBridge-Flow: Re-Coupling Posterior Bridges in Flow Matching for Image Restoration
链接: https://arxiv.org/abs/2609.00811
作者: Jiaqi Zhang,Yiqi Wang,Hongjie Wu,Bohan Guo,Xinan Wang,Zichen Luo,Taotao Cai,Zhi Chen,Mingkai Zheng
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 66 Pages, 36 Figures, 15 Tables
Abstract:Flow Matching provides an efficient generative prior for image restoration by learning continuous transport between source and data distributions. However, existing methods typically incorporate measurement constraints through local corrections. Such corrections may disrupt the source-clean endpoint coupling implicitly encoded by the pretrained flow, making the corrected endpoint pair incompatible with the current state. To address this issue, we propose ReBridge-Flow, a posterior bridge re-coupling method. Specifically, given the current state, ReBridge-Flow first decodes the corresponding local source and clean endpoints. It then incorporates measurement information through clean-side anchoring and synchronously re-couples the source endpoint, yielding a measurement-aware endpoint pair with improved local bridge compatibility. The re-coupled endpoints further define a posterior-informed transport direction for advancing the sampling process. We also introduce the Posterior Bridge Defect, which jointly characterizes measurement error, deviation from the flow prior, and bridge mismatch, and leads to explicit updates for clean-side anchoring and source-side re-coupling. Extensive experiments on multiple natural and medical image restoration tasks demonstrate that ReBridge-Flow effectively alleviates bridge mismatch and improves the structural consistency of restored images.
[CV-73] Advanced Pixel Diffusion Model with Guided Sparse Global Refinement
链接: https://arxiv.org/abs/2609.00798
作者: Weiyi You,Jinhua Zhang,Xingyu Zhou,Wei Long,Junyu Lou,Shuhang Gu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Code: this https URL
Abstract:Pixel-space diffusion has recently emerged as a promising direction for high-fidelity image generation by modeling images directly in the original pixel domain. However, pixel-space diffusion is computationally demanding due to the extremely high dimensionality of natural images. For efficiency, existing pixel diffusion models either compromise fine details with large-patch tokenization or confine subsequent refinement within individual patches. Such intra-patch refinement inevitably restricts structural continuity across patch boundaries and long-range token interactions, limiting refinement quality. To address these issues, we propose PixSGR, a novel Pixel diffusion framework with Sparse Global Refinement tailored for modeling the distribution of natural images directly in pixel space. PixSGR starts from a supervised low-channel bottleneck to efficiently capture the low-dimensional manifold of natural images. It then progressively expands the channel dimensionality and spatial resolution to recover increasingly fine-grained structures. At the spatial refinement stage, coarse-scale attention maps preselect globally relevant interactions to pre-sparsify fine-scale attention, enabling non-local refinement beyond isolated patches without the quadratic cost of dense attention. Extensive experiments on ImageNet validate the effectiveness of PixSGR. It achieves an FID of 1.51 at 256 \times 256 and maintains performance when scaled to 512 \times 512, attaining an FID of 1.60.
[CV-74] Forbid Your Attention: Fooling Multimodal Large Language Models by Selectively Removing Intrinsic Focus in Spectral Domain
链接: https://arxiv.org/abs/2609.00788
作者: Daizong Liu,Junhao Dong,Zhiyuan Ma,Xiaoye Qu,Xiang Fang,Runwei Guan,Keke Tang,Jianfeng Dong,Yew-Soon Ong
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted by IEEE TMM2026
Abstract:Multimodal large language models (MLLMs) have extended the capability of large language models (LLMs) to process more contextual multimodal information, showing remarkable progress in diverse realistic multimodal applications. Despite their strong perception and reasoning abilities, recent studies reveal that MLLMs remain highly vulnerable to adversarial inputs, especially those targeting visual components. However, existing attacks mainly focus on global perturbations, lacking an understanding of how MLLMs internally interpret visual structures. In this paper, we make the attempt to investigate the intrinsic focus of MLLMs in the frequency domain and discover that their predictions are particularly sensitive to phase information, which encodes essential structural and semantic cues. Based on this observation, we propose a novel phase-aware adversarial attack framework that explicitly restricts adversarial perturbations to structure-relevant phase regions to suppress the MLLMs’ focus for effective and imperceptible attacks. To further amplify the structural influence, we also introduce an auxiliary adversarial prompt learning module to guide multimodal misalignment around phase-sensitive regions, misleading the MLLM’s attention toward targeted structural patterns. Extensive experiments on multiple representative MLLM models and datasets demonstrate the superior effectiveness of our method compared to existing attacks.
[CV-75] Solaris: Towards Interfaces That Are Generated Not Coded
链接: https://arxiv.org/abs/2609.00776
作者: Yuval Alaluf,Omri Avrahami,Guy Bukchin Leshem,Michal Geyer,Kfir Goldberg,Elad Richardson,Diego Alarcón,Alejandro Alvarez,Cole Garry,Anastasis Germanidis,Tenaya Goldsen,Corina Gurau,Robin Kahlow,Joel Kwartler,Kathleen Lewis,Alejandro Matamala Ortiz,Eugene McMahon,Thon Prom,Sarah Saltonstall-Wurm,Jamie Umpherson,Hudson Yeo
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Project page: this https URL
Abstract:Digital interfaces are traditionally implemented through intermediate representations such as code, requiring their appearance and behavior to be specified in advance. We introduce Solaris, an interface world model that instead generates an interactive UI directly, frame by frame, in response to user actions. Solaris treats mouse interactions as conditioning signals and autoregressively synthesizes the resulting visual state at interactive speeds. To enable real-time generation while maintaining visual coherence over extended interactions, we combine autoregressive frame generation with few-step distillation and training on the model’s own outputs. A language model complements the visual world model by interpreting user intent and specifying how interactions should affect the generated environment, separating high-level reasoning from visual rendering. By generating both the appearance and behavior of an interface dynamically, Solaris enables open-ended interactions that need not be explicitly programmed in advance. We view interface world models as a step toward a new paradigm for software, where interfaces are generated and adapted continuously around user intent rather than implemented as fixed collections of predefined states and
[CV-76] VOIM: Training-Free Open-Vocabulary 3D Instance Mapping for RGB-D and Monocular SLAM
链接: https://arxiv.org/abs/2609.00775
作者: Sangmin Song,Sarath Kodagoda,Marc G. Carmichael,Karthick Thiyagarajan,Amal Gunatilake,Kelly Prentice,Jodi Martin
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:We present Voxel-Grounded Online Instance Manager (VOIM), a training-free voxel-grounded instance manager that builds open-vocabulary 3D instance maps from RGB-D or from monocular RGB alone, a regime no prior training-free system addresses. Online systems typically segment object instances and label them at first detection, committing when evidence is weakest. VOIM instead defers label and instance decisions until soft evidence from unmodified, off-the-shelf perception has accumulated per voxel across views. We show that the mapping stage, rather than the particular perception models, carries the result: across four perception configurations on ScanNet++, varying the region descriptor, the detector label prior and the mask source, the map exceeds the strongest online RGB-D system, OVO-SLAM, by between 4.8 and 11.7 mIoU. Perception is not neutral, and substituting that baseline’s own descriptor family costs 4.1 of the margin, yet the baseline carries the marginally better 2D descriptor (33.7 vs. 31.5 mIoU over three scenes) and still realizes the weaker map. Under a like-for-like protocol VOIM reaches 44.07 mIoU on ScanNet++ against 32.37, winning all ten scenes and both aggregations (pooled 33.31 vs. 25.97), and the same system runs unchanged to fully monocular RGB, matching that baseline pooled on Replica (27.80 vs. 27.50). The advantage is regime-specific: under Replica’s all-classes scoring, matched inputs give a split result, 28.60 vs. 27.50 pooled against 24.59 vs. 30.11 on the per-scene mean. Room scale is label-limited and building scale drift-limited. Labeling does not run in real time, dominated by per-class detection over the full vocabulary. The maps export occupancy grids and resolve free-form queries to object instances.
[CV-77] Feed-Forward Multi-view Multi-person Reconstruction with Contrastive Human-Aware 3D Representation
链接: https://arxiv.org/abs/2609.00745
作者: Yuanwang Yang,Buzhen Huang,Zongxuan Ren,Jing Huang,Kun Li
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Published in International Journal of Computer Vision (IJCV)
Abstract:Multi-view human reconstruction has been extensively studied under simplified settings, yet robust and efficient multi-person reconstruction in unconstrained environments remains challenging. Existing bottom-up methods often rely on accurate camera calibration and explicit cross-view matching, and therefore struggle with severe occlusions and ambiguities. We propose a new top-down paradigm that maintains a unified, instance-centric human-aware 3D space, enabling simultaneous camera calibration, cross-view association, and human reconstruction via cross-modal contrastive learning. Observations from multiple views are lifted and fused into this shared 3D space, where geometric structure, visual appearance, and human-centric semantic cues are jointly encoded at the instance level. We further introduce a spatial contrastive learning strategy that aligns 3D features corresponding to the same human instance across different views and modalities while separating different instances. This enables correspondence reasoning, semantic aggregation, and instance discrimination to be performed natively in 3D, improving cross-view consistency and robustness under severe occlusions. Finally, structured human body models are recovered in a feed-forward manner by regressing SMPL parameters from instance-level 3D human tokens. Extensive experiments demonstrate robust, accurate, and efficient multi-view human reconstruction in challenging real-world scenarios.
[CV-78] Mind the Rift: Cross-Scale Coupling Mismatch for AI-Generated Video Detection
链接: https://arxiv.org/abs/2609.00742
作者: Siyu Li,Jin Yang,Weiheng Liang
类目: Computer Vision and Pattern Recognition (cs.CV); Multimedia (cs.MM)
备注: 12 pages, 4 figures, 11 tables. Accepted at ACM Multimedia 2026 (MM '26), Rio de Janeiro, Brazil. This version includes the supplementary material as Appendix A-E
Abstract:As AI video generators achieve cinematic realism, reliable detection becomes essential for safeguarding digital trust. We identify cross-scale coupling mismatch as a new forensic signal, where scale refers to the level of abstraction (semantic dynamics vs. pixel-level residuals): in natural videos, macro-level temporal dynamics and micro-level residual patterns are intrinsically coupled by the unified imaging physics pipeline, whereas AI generators, whose training objectives do not explicitly preserve this joint distribution, systematically violate this coupling. Detecting such mismatch is challenging because it requires independently extracting information at both scales while simultaneously quantifying their cross-scale relationship. We propose RIFT (Representation Inconsistency Forensics on Trajectories), an orthogonal forensic framework that addresses this through three interlocking components: a macro stream that builds a dynamic baseline of expected temporal evolution via differential geometry and persistent homology on learned manifold trajectories, a micro stream that acts as a sensitive forensic probe via steganalytic filtering and temporal modeling, and a coupling divergence module that measures the conditional dependency between the two streams. Gram-Schmidt orthogonality guarantees the information-theoretic validity of this measurement. Experiments on two benchmarks (VidProM, 120K videos, 7 generators; GenVidBench, 68K videos, 4 generators) demonstrate that RIFT achieves 99.33% and 99.72% F1-score respectively, with 97.87% unseen-generator detection rate in leave-one-out evaluation, while exhibiting encoder agnosticism: scaling from ViT-S/14 (22M) to ViT-L/14 (300M) changes F1 by less than 0.1%, and switching to a different encoder family (DINOv1) reduces F1 by only 0.73 pp. Code is available at this https URL
[CV-79] Design and Implementation of a Kalman Filter-Infused Algorithm for Tilt Estimation
链接: https://arxiv.org/abs/2609.00730
作者: Yuehan Ma,Hongji Dai
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO); Signal Processing (eess.SP); Systems and Control (eess.SY)
备注: 13 pages, 24 figures, 7 references
Abstract:Accurate tilt angle estimation is important in many engineering applications, such as robotics, motion tracking, and embedded control systems. However, measurements from low-cost inertial sensors are often degraded by noise and drift. This paper presents a single-axis tilt angle estimation system based on the MPU6050 inertial measurement unit, implemented on an RP2040 microcontroller platform, with sensor fusion achieved through a Kalman filter. The accelerometer provides a direct estimate of tilt angle from gravity but is sensitive to noise and short-term fluctuations. The gyroscope provides smooth angular rate measurements, but integration over time introduces drift. To overcome these limitations, a Kalman filter is used to combine measurements from both sensors, leveraging the long-term stability of the accelerometer and the short-term smoothness of the gyroscope. Both simulation and hardware experiments are performed. In simulation, sensor noise and drift are modeled to evaluate the filter performance under control conditions. In the hardware implementation, real-time MPU6050 data is acquired and processed by the RP2040 platform, and the estimated tilt angle is compared with accelerometer-only and gyroscope-only outputs. The results show that the proposed method effectively reduces noise measurements and suppresses long-term drift while preserving good dynamic response. Overall, the system provides more stable and accurate tilt estimation than either sensor alone, demonstrating a practical and accessible approach for Kalman filter based sensor fusion in embedded application.
[CV-80] A Closed-Loop Evaluation of Capability Loss and Recovery in Compressed Driving Policies
链接: https://arxiv.org/abs/2609.00718
作者: Ahmad Alfan Alfian Irfan,Nur Ahmad Khatim,Mansur Arief
类目: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Many automobile and mobility companies deploy learned driving policies on embedded computers with limited memory and power. Pruning, knowledge distillation, and quantization are the standard methods to reduce the size and the inference cost of these policies. However, these methods are commonly assessed by aggregate numerical scores, and such scores may not reflect the ability of the policy to drive safely when interacting with other road users. In this study, we propose a stage-wise closed-loop evaluation approach to follow a driving policy through a compression pipeline. We formulate the driving task as a partially observable Markov decision process (POMDP) and train a belief-state policy with proximal policy optimization (PPO) in Gym-Duckietown. We then extract the actor, compress it one stage at a time, and evaluate it on five driving curricula. We show that structured pruning is the stage at which the driving capability is first lost. Meanwhile, distillation improves the pruned actor, but the improvement is limited by its rehearsal data. Integer quantization of the improved actor loses some of the curricula that require the vehicle to stop and then resume. Interestingly, the same procedure on the unpruned actor preserves all five curricula. Our study thus provides an empirical analysis aiming to answer the currently active discussions on how to accept a compressed driving policy, so as to achieve a safe and statistically reliable deployment of automated driving functions.
[CV-81] Efficient and Robust Absolute Pose Estimation via Gravity-Prior-Driven Transformation Decoupling and Pose Refinement
链接: https://arxiv.org/abs/2609.00713
作者: Hu Cao,Qianyi Yang,Xinyi Li,Jiong Liu,Yinlong Liu,Alois Knoll
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: This work is accepted by IEEE Transactions on Image Processing
Abstract:Estimation of the absolute pose of an object is an essential task for various robotic applications. Recently, incorporating gravity direction as prior information has emerged as a popular approach to simplify absolute pose estimation. However, developing a robust and efficient algorithm to solve this challenging problem remains a difficult question due to large amounts of mismatches. In addition, obtaining an accurate pose solution from selected inlier correspondences with gravity prior is still a research gap. In this paper, we propose a novel transformation strategy that exploits geometric relations derived from the gravity prior. Through transformation decoupling, the original 6 degrees of freedom (DoF) absolute pose estimation problem is simplified into a 4-DoFs problem: 1-DoF for the rotation angle and 3-DoFs for translation, significantly improving the efficiency. For the 1-DoF rotation angle, we apply a one-dimensional global voting algorithm for optimal estimation. Once the optimal rotation is obtained, the mismatched correspondences are preliminarily filtered, and translation estimation, a linear problem, can be easily solved. Furthermore, to obtain accurate pose results, we introduce a novel pose refinement algorithm to enhance the accuracy of both rotation and translation. Extensive experiments on synthetic data and three publicly available real-world datasets (TUM RGB-D, ETH3D, and RobotCar) demonstrate that the proposed method achieves stronger performance compared to existing state-of-the-art (SOTA) approaches. To further validate our method, we integrated it into ORB-SLAM2. The results on the KITTI dataset show it effectively reduces drift and improves trajectory alignment during relocalization. The source code will be released upon acceptance.
[CV-82] EarthLD: Towards Unified Open-World Landslide Understanding via Vision-Language Guided Diffusion Models
链接: https://arxiv.org/abs/2609.00712
作者: Yuanchao Su,Lianru Gao,Mengying Jiang,Jiangyi Chen,Jiaxin Cheng,Yicong Zhou
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Landslides are widespread geological hazards, yet their automated detection and mapping in remote sensing imagery remain challenging because of their irregular morphology, ambiguous spectral signatures, and substantial domain shifts across imaging platforms. To overcome these challenges, we propose EarthLD, a vision-language-guided diffusion framework for open-world landslide understanding, enabling unified landslide recognition, mapping, and trigger interpretation. At its core, EarthLD formulates landslide understanding as a diffusion process that progressively infers the presence, spatial extent, and pixel-level boundaries of landslides from noisy latent representations. This probabilistic formulation enables the model to jointly perform image-level landslide recognition and mapping while characterizing predictive uncertainty. By integrating visual observations with contextual knowledge in the denoising process, EarthLD distinguishes diverse landslides from backgrounds, produces confidence-aware predictions for suspected regions, and maps landslide ranges. We additionally construct a global-scale open-world landslide benchmark by systematically harmonizing multiple publicly available remote sensing data collected by diverse institutions. Extensive experiments across regions, sensors, and triggering events demonstrate that EarthLD consistently outperforms existing landslide detection methods, highlighting its potential as a unified and robust solution for global geological-hazard monitoring and emergency response.
[CV-83] Differentially Private Paired Table-Image Multimodal Synthesis
链接: https://arxiv.org/abs/2609.00708
作者: Kai Chen,Josephine Lamp,Somesh Jha,Tianhao Wang
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注: This paper is about differentially private table-image data synthesis
Abstract:Differentially private (DP) synthesis has been extensively studied for tabular and image data separately, yet many real-world datasets contain images paired with multivariate tabular records. Synthesizing such data is particularly challenging under DP, as the two modalities favor different private learning mechanisms while their dependence must also be preserved. To address this challenge, we propose DP-TabImage, a modality-specialized framework for private paired synthesis. DP-TabImage instantiates the factorization p(x,y)=p_T(y)p_I(x;|;y) using a private Probabilistic Graphical Model for the multivariate table distribution and a table-conditioned diffusion model trained with DP-SGD for the conditional image distribution. To facilitate conditional learning under clipped and noisy gradients, we further pretrain the model on private table-image prototypes, pairing privately constructed attribute-conditioned images with tabular vectors derived from the already private tabular model at no additional privacy cost. Experiments on three real-world datasets show that DP-TabImage achieves a strong balance among tabular fidelity, image fidelity, and cross-modal alignment. Our analysis further reveals that visual warm-up primarily improves marginal image fidelity, whereas aligned table-image warm-up is critical for improving cross-modal correspondence. Our source code is available in the GitHub repository, this https URL.
[CV-84] FTU-Seek: Foundation Model-Guided Hard-Negative Learning for Sparse Functional Tissue Unit Segmentation
链接: https://arxiv.org/abs/2609.00704
作者: Zonghao Liu,Lei Su,Jiguang Yu,Xuqing Geng,Louis Shuo Wang,Jianmin Wang,Jingfeng Liu
类目: Computer Vision and Pattern Recognition (cs.CV); Numerical Analysis (math.NA)
备注:
Abstract:Functional tissue units (FTUs), including tertiary lymphoid structures (TLSs), blood vessels, and glands, encode localized immune, vascular, and epithelial organization in histopathology. Accurate quantification of these structures is important for studying tissue architecture and disease-associated tissue organization. However, FTUs are frequently sparse, heterogeneous, and surrounded by large amounts of morphologically similar background tissue, making automated segmentation in whole-slide images (WSIs) challenging. We therefore developed FTU-Seek, a pathology foundation model-guided framework that treats morphology-aware negative-patch selection as a key component of sparse FTU segmentation. FTU-Seek uses frozen multi-depth features from the UNI pathology foundation model to train a patch-level classifier that distinguishes FTU-containing from FTU-absent tissue. Target-absent patches are subsequently ranked according to their predicted target-containing probabilities, and the highest-scoring hard negatives are selected through a static Top K strategy to construct compact segmentation training sets. The framework was evaluated using five-fold cross-validation and internal test cohorts across TLS, blood-vessel, and gland segmentation tasks, with an additional independent 30-WSI held-out cohort for TLS. Positive-only, all-tissue, random-negative, and matched random Top K sampling strategies served as comparators. Segmentation-derived phenotypes were further explored in external TCGA cohorts.
[CV-85] DGNet: Dual-knowledge Guided Network for Infrared Small Target Detection
链接: https://arxiv.org/abs/2609.00666
作者: Chenglong Yu,Mingzhu Xu,Jing Wang,Tongtong Wang,Pingping Miao,Liqiang Nie
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted by ACM Multimedia 2026 (MM '26)
Abstract:InfRared Small Target Detection (IRSTD) is a prominent and challenging task in computer vision. In recent years, text-guided methods have significantly improved detection performance. However, they still suffer from two key limitations. First, a single text description simultaneously modeling both background and target leads to semantic entanglement, which contradicts the objective of background suppression and target enhancement. Second, reliance on image-specific textual prompts (requiring additional external models such as CLIP during inference) results in deployment constraints. To address these issues, we propose a novel Dual-knowledge Guided Network (DGNet) based on multiple generalizable texts. Specifically, we design a Prior-knowledge Wavelet Modulation (PWM) module, which leverages dual textual priors that separately characterize large-scale backgrounds and sparse targets to effectively disentangle and modulate entangled semantics in the frequency domain. Furthermore, we introduce a Consensus-knowledge Directional Alignment (CDA) loss, which models the initial state and the ideal target across samples as complex background' and bright target’, respectively, thereby constructing a clear and unified directional optimization trajectory for the model. Extensive experiments on three public datasets demonstrate the superior performance of DGNet and the effectiveness of each component. The source code is available at this https URL.
[CV-86] Separating perception from reasoning in vision-language models: a model-free render ceiling for crystal structures
链接: https://arxiv.org/abs/2609.00663
作者: Can Polat,Mustafa Kurban,Erchin Serpedin,Hasan Kurban
类目: Computer Vision and Pattern Recognition (cs.CV); Materials Science (cond-mat.mtrl-sci); Chemical Physics (physics.chem-ph); Quantum Physics (quant-ph)
备注:
Abstract:Multimodal evaluations cannot say whether a vision-language model misread an image or misreasoned about it, because every existing method for separating the two places a second model in the loop. We introduce the render ceiling, a model-free reference for benchmarks built by rendering known objects: inverting the frozen cameras and re-solving cross-view correspondence recovers exactly the answer the images support. We prove the ceiling fails only through an enumerable set of projection coincidences and certify that set empty on 2,160 rendered crystal structures, so every point of a model’s deficit belongs to the model. Across fourteen vision-language models, supplying exact geometry as text lifts every model yet closes under half the gap for thirteen, while a supervised vision model with no language component reads the same images at 0.8952, above every vision-language model. The instrument exposes extraction-stage fabrication that downstream accuracy would misattribute to reasoning, yields camera-placement rules for benchmark builders, and transfers to any benchmark with an invertible forward rendering.
[CV-87] Do Satellites See Commuters? A Critical Benchmark of Vision Foundation Models
链接: https://arxiv.org/abs/2609.00661
作者: Ashiq Shukoor Iqbal,Wilson Wongso,Flora D. Salim
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to SIGSPATIAL 2026
Abstract:Satellite foundation models offer a globally available alternative to census data for commuting origin-destination (OD) generation, yet no study has systematically compared encoder paradigms within a single downstream pipeline. We ablate four satellite vision encoders: language-supervised (RemoteCLIP), self-supervised (DINOv3), and geographically grounded (SatCLIP, AlphaEarth) within an identical WeDAN graph diffusion framework across 1,925 US counties, 325 UK districts, and 14 global cities under five random seeds. Three main findings emerge. First, language-supervised features achieve the strongest in-distribution performance (RemoteCLIP CPC 0.602), while geographically grounded encoders transfer more reliably zero-shot: AlphaEarth improves CPC by 33% over RemoteCLIP on UK districts. Second, pretraining corpus scale alone is insufficient: DINOv3, trained on a substantially larger satellite corpus, underperforms RemoteCLIP by 0.091 CPC in-distribution and collapses to CPC 0.022 globally. Third, no encoder transfers usefully to global cities (best CPC 0.122 for RemoteCLIP, 0.022 for DINOv3), confirming cross-continental OD generation remains an open problem. We additionally clarify the semantics of the census noise parameter \eta , whose ordering reverses under cross-continental evaluation, a distinction critical to correctly interpreting prior results. Training scripts and evaluation logs will be released.
[CV-88] aching Vision-Language Models to Use the Scale They Are Given: Label-Free Equivariance Training for Metric Physical Reasoning
链接: https://arxiv.org/abs/2609.00658
作者: Kaizhen Tan,Yang Feng,Heqing Du,Siru Tao,Xin Xu,Hanzhe Hong
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Metric questions about video require vision-language models to use supplied real-world references to convert visual measurements into physical units. Yet we find that current models use this scale information only partially. When every world-space quantity in a prompt is rescaled by a common factor, the video remains equally valid and the correct answer changes by exactly that factor, but model predictions move only part of the way and accuracy remains concentrated near the familiar scale of the depicted objects. Across eight vision-language models, this under-response persists over four orders of magnitude. The same models recover the correct closed-form scaling laws when the identical physics is asked in a scale-free form, indicating that the main deficit lies in metric grounding rather than physical mechanism knowledge. We use this exact scaling relation as supervision without requiring metric annotations. Under a common rescaling of the supplied world-space quantities, the correct metric answer must change by the same factor. EquiSD exploits this constraint by projecting a model’s own prediction onto the scale-equivariant family and fine-tuning the model on the resulting targets. It requires no ground-truth answers and only one model query per training video. On held-out simulated videos, EquiSD increases a 3B model’s median response slope from 0.66 to 0.94 and improves mean relative accuracy by 9.2 points across scales. The learned relation generalizes to unseen world scales and transfers without adaptation to real QuantiPhy videos, where accuracy increases by 6.4 points. These results show that an exact physical symmetry can provide label-free supervision for improving metric grounding in vision-language models.
[CV-89] Physically Plausible Video Generation via Visual-Semantic Chain-of-Events Conditioning
链接: https://arxiv.org/abs/2609.00656
作者: Zixuan Wang,Yixin Hu,Wen Li,Feng Chen,Yan Liu,Duo Peng,Yinjie Lei
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Physically Plausible Video Generation (PPVG) seeks to synthesize videos consistent with physical principles, yet remains challenging due to underspecified natural language conditioning. Advanced chain-of-thought (CoT) frameworks augment prompts with physical knowledge. However, such prompts describe physical phenomena holistically, overlooking intermediate states and transition dynamics. In this paper, we reformulate PPVG as event-centric generation by representing physical evolution as a chain of causally connected and physically constrained events. Our framework comprises three key modules: (1) Physics-driven Event Chain Reasoning. This module decomposes physical phenomena into causally connected events represented by evolving scene graphs. Formula-derived physical quantities are bound to relevant objects and interactions, characterizing the direction and magnitude of each event transition. (2) Transition-aware Routed Keyframe Conditioning. This module routes each event to a specialized keyframe synthesis operator for appearance variation or object transformation. Consecutive keyframes are injected as residual guidance during denoising, enabling smooth visual transitions between event-boundary states. (3) Physics-injected Contrastive Semantic Guidance. This module constructs physics-informed positive and counterfactual negative prompts for classifier-free guidance, steering generation toward plausible dynamics and away from physics-violating counterparts. Experiments on PhyGenBench, VideoPhy, PhyWorldBench, and Physics-IQ demonstrate that our framework generates videos with superior physical plausibility across diverse domains.
[CV-90] You Cannot Photograph the Same Street Twice: Reliability Limits in Vision-Language Measurement of Urban Change
链接: https://arxiv.org/abs/2609.00649
作者: Kaizhen Tan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Vision-language models are increasingly used to measure urban change from repeated street-level imagery, but their longitudinal reliability is not well understood. We test how much a perception score can change when the street itself does not undergo substantial redevelopment. Using 4,648 consecutive-epoch image pairs from 435 Google Street View standpoints across five US cities, we find that re-photographing the same street changes a perception score by 0.80 points on average, equivalent to 66.5% of the difference between two different streets in the same city. Repeated model calls contribute almost no variation, while image re-encoding and prompt-order changes each account for about one fifth of the between-street difference. Six image statistics describing scattering, contrast, colour, exposure, sharpness and specularity explain almost none of the remaining epoch-to-epoch variation. A small systematic drift of about 0.1 points remains and increases with the interval between captures, consistent with minor physical changes not recorded by redevelopment labels. Controlled experiments further show that acquisition conditions can shift scores when camera and image properties are allowed to vary, and that the direction of these shifts depends on the model. In crowdsourced imagery, camera geometry alone causes a model to report physical change in 45% of identical-scene pairs; normalising both images to a common virtual camera reduces this rate to 7.5%. Despite poor reliability at the individual-location level, aggregation recovers a coherent redevelopment signal: changed streets are judged wealthier, better maintained, more enclosed and less green. These results show that vision-language measurement of urban change is reliable at the scale of hundreds of paired observations, but not at the scale of individual sample points.
[CV-91] Beyond Landmark Extraction: A Framework for Robust Geometric Feature Construction in Structured Image Classification
链接: https://arxiv.org/abs/2609.00634
作者: Saravana Mauree,Sakshi Arya
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Under consideration at Pattern Recognition Letters
Abstract:Much of the literature on structured image recognition has disproportionately focused on the comparison of classification algorithms. Rather than investigating which classifier performs best, this paper instead asks: what should a classifier know before it ever makes a prediction? In structured vision problems such as gesture recognition, facial expression categorization, and medical image analysis, discriminative information lies less in individual pixels and more in spatial relationships between semantic parts. Raw pixel spaces are high-dimensional, sensitive to nuisance variation, and often obfuscate the geometric structures that make visual tasks interpretable. Landmark extraction provides one form of dimension reduction, but it does not by itself determine the information preserved. This paper studies the post-landmark feature map as the central object of analysis and proposes a systematic framework for constructing and interpreting landmark-derived representations as an, informed, feature-based ``dimension reduction’’ step. Using static hand gesture recognition as a case study, we evaluate coordinate, distance, angle, and hybrid representations through perturbation and ablation experiments. The results show that visually variable data exposes substantial gaps between raw coordinate features and their geometrically invariant counterparts, while hybrid representations achieve the strongest overall performance by combining complementary geometric components. These findings frame feature construction as a fundamental modeling decision and ultimately suggests that the question of what representation should a classifier learn from is one worth asking. The code used for feature construction and evaluation is available at this https URL
[CV-92] Restrict Dont Retrain: Inference-Time VLM Guidance for Zero-Shot Aerial Segmentation
链接: https://arxiv.org/abs/2609.00628
作者: Teresa DiMeola,Charles Walter,Hong Xiao
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Global welfare often depends on the correct interpretation of aerial and satellite imagery. Acting on such imagery (mapping flooded ground, crop extent, or damaged infrastructure) demands pixel-level segmentation to ensure perfect class localization. Pretrained general foundation models, when applied directly, often miss important features and cannot always find all the classes belonging to a given scene, overlooking smaller objects that matter most. We use a single consumer-grade GPU running a vision-language model (VLM) to supply this missing guidance, improving segmentation while producing structured, auditable evidence that drives the result and can be inspected on its own. We fuse three approaches: the frozen foundation model that labels every pixel, and two queries to a VLM, one to choose the classes that matter, and one to locate the small objects the base model misses. Evaluating across four aerial datasets, we see consistent gains at each stage where the base model is competent.
[CV-93] Inverse Rendering for Modeling with Line Primitives SIGGRAPH
链接: https://arxiv.org/abs/2609.00625
作者: Kenji Tojo,Ariel Shamir,Nobuyuki Umetani,Bernd Bickel
类目: Graphics (cs.GR); Computer Vision and Pattern Recognition (cs.CV)
备注: SIGGRAPH Asia 2026. Project page: this https URL
Abstract:Faithfully capturing diverse real-world objects with fuzzy, anisotropic structures, such as hair, fur, fibers, and textiles, for efficient real-time visualization remains challenging. Recent radiance field reconstruction methods capture these structures from multi-view images using translucent volumetric primitives such as 3D Gaussians rather than opaque low-dimensional primitives (e.g., triangles, line segments, and polylines), thereby limiting compatibility with standard depth-tested rasterization, reflection modeling, and physical simulation. We present an inverse rendering method for reconstructing fuzzy geometry using explicit line segments, which are rasterized on a subpixel grid for anti-aliasing to reproduce a semi-transparent appearance. While straightforward to render, optimizing numerous line primitives to match target images poses a significant challenge. We address this by introducing a stochastic differentiable rasterizer for line segments that produces informative gradients with respect to vertex positions, attributes, and discrete connectivity. Experiments on synthetic and real-world datasets show that our method outperforms surface-based approaches in capturing fuzzy boundaries and achieves quality comparable to volumetric representations while relying entirely on explicit geometry. The resulting representation integrates seamlessly with standard graphics pipelines, enabling cross-platform rendering, various shading models, and physical simulation.
[CV-94] Streaming4D: Accelerate 4D World Models via Block-wise Video Generation and Incremental Reconstruction
链接: https://arxiv.org/abs/2609.00610
作者: Xiaoyan Liu,Jiaxin Liu,Kangrui Li,Sifan Zhou
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Current 4D generation paradigms are often bottlenecked by a sequential decoupling design: video is generated first, followed by 3D reconstruction, leading to high interaction latency. This limits applications in interactive real-time scenarios. To this end, we propose \textbfStreaming4D, a tightly coupled synchronous pipeline that integrates block-wise autoregressive video generation with incremental 3D reconstruction. Unlike traditional frame-by-frame emission and delayed geometry recovery, Streaming4D generates temporal video blocks and immediately triggers reconstruction for each completed block, enabling parallel execution between synthesis and geometric updates. This approach allows the world representation to evolve online with the video stream, reducing feedback latency while preserving geometric fidelity. We instantiate \textbfStreaming4D using a Self-Forcing-style autoregressive generator and an incremental reconstruction backend. Experiments show consistent runtime improvements across resolutions on a single RTX 4090 (1.24 \times speedup), while maintaining high-quality 4D geometry and multi-view consistency.
[CV-95] BrainDiff: Longitudinal Report Generation for Multimodal Brain MRI
链接: https://arxiv.org/abs/2609.00593
作者: Krish Patel,Peirong Liu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Neuroradiologists rarely read a brain MRI in isolation, yet automated brain-MRI report generation has been built almost entirely for single studies. Temporal analysis has been explored on chest radiography and chest CT, but to our knowledge, longitudinal reporting for brain MRI, where interval change is often subtle and spatially distributed, remains unaddressed. We present BrainDiff, the first longitudinal vision-language system for brain MRI. BrainDiff outperforms both frontier general-purpose and single-study neuroimaging models on the same patient pairs. Moreover, BrainDiff retains 91% of internal RadGraph-XL entity+relation F1 (rg_er) on an external, cross-hospital cohort. Beyond the system, we contribute three analyses. First, we identify two independent grounding levers: a counterfactual objective with prior-report dropout, which increases measured image reliance by ~47%, and a staged curriculum. Together, these interventions raise image reliance 2.5-fold from the baseline. Second, we provide a factorial over prior-report availability and image identity, isolating a visual contribution of +0.0387 rg_er, which grows when the prior report is withheld. Third, a cheap change-decodability test for candidate backbones shows that interval change is decodable far more weakly than single-study pathology (0.60 vs. 0.77 AUROC). Code is publicly available at this https URL.
[CV-96] A Glance Is All You Need: Single-Pass Fine-Grained Image Captioning with SimLoss
链接: https://arxiv.org/abs/2609.00591
作者: Suryaansh Jain,Rahasya Barkur,Vishal G,Ryan Rossi,Franck Dernoncourt,Jack Wang,Koustava Goswami,Nedim Lipka,Puneet Mathur,Samyadeep Basu,Seunghyun Yoon
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:An image may be worth a thousand words, but most captioning models describe it in only a few. Modern vision-language models produce fluent high-level captions, yet routinely miss the attributes, counts, textures, materials, and spatial relations that make an image visually specific. Recent multi-stage systems recover some of these details through generation, decomposition, verification, and rewriting, but they do so at the expense of substantially higher inference latency. We propose SimLoss, a reference-free embedding-space objective for single-pass fine-grained image captioning. SimLoss trains a vision-language model to align its projected hidden-state representation with a frozen image embedding through an InfoNCE contrastive loss, supplying a dense visual supervision signal before any text is decoded, and requiring neither human-written fine-grained captions nor pseudo-captions from a multi-stage pipeline. We instantiate it as SimLoss FFT, which backpropagates through a locally available embedding model, and SimLoss GRPO, which treats that model as a black-box reward. Compared with single-pass, multi-stage verification, reward-optimized, and perception-aware baselines, the fully differentiable fine-tuning variant, SimLoss FFT, achieves the highest precision while nearly matching the F1 score of the multi-stage method, all while retaining single-pass inference and running roughly 20 times faster than the multi-stage pipeline. The reward-based variant SimLoss GRPO attains the strongest recall. Together, these results show that embedding-space supervision can recover the quality of multi-stage verification at the latency of a single-pass captioner. Subjects: Computer Vision and Pattern Recognition (cs.CV) Cite as: arXiv:2609.00591 [cs.CV] (or arXiv:2609.00591v1 [cs.CV] for this version) https://doi.org/10.48550/arXiv.2609.00591 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Suryaansh Jain [view email] [v1] Tue, 1 Sep 2026 02:31:34 UTC (10,772 KB)
[CV-97] Potential-Guided Particle Steering for Negation-Constrained Dexterous Grasping
链接: https://arxiv.org/abs/2609.00555
作者: Geonho Kim,SooGon Kim,Jongmin Lee
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注: Project page: this https URL
Abstract:Language-driven dexterous grasp models, such as DextER, perform well when instructions specify where to grasp, but we find they fail systematically when an instruction also specifies where not to grasp (e.g., “grasp the handle but avoid the body”). Existing training corpora, DexGYSNet among them, contain virtually no avoidance instructions, and collecting examples for every possible constraint is impractical. Moreover, because every part mentioned during training denotes a contact target, models may interpret a forbidden part as another region to grasp rather than one to avoid. We therefore introduce an inference-time framework for negation-constrained dexterous grasping that requires no negation-specific training examples. Combining Sequential Monte Carlo with classifier-free guidance, our method guides sampling toward the instructed part while pruning candidates headed for the forbidden region, without any negation examples during training. A frozen 3D part-grounding model localizes the forbidden region from the language instruction. To evaluate this setting, we construct NegGrasp, a benchmark of paired positive/negative instructions with constraint-aware metrics that credit a grasp only if it both accomplishes the task and respects the stated constraint. On NegGrasp, our method reduces the violation rate of the strongest baseline from 57.9% to 17.2% while improving both constraint-aware and physical success.
[CV-98] GenScale: A Benchmark for Relative Object Scale in Image Generation and Editing
链接: https://arxiv.org/abs/2609.00525
作者: Lingxiao Li,Max Whitton,Ledell Wu,Boqing Gong
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Modern image generation and editing systems can produce photorealistic, prompt-aligned images, but still often render familiar objects at implausible relative sizes. To measure this failure mode, we introduce GenScale, a benchmark and evaluation protocol for real-world relative object scale in image generation and editing. GenScale contains 900 image-level entries and 1,643 pairwise anchor-target scale relations across common-object generation, human-product generation with metric dimensions, and scale correction from failed generations. We further design a human-calibrated ordinal judge for scalable pairwise scale evaluation. Last but not the least, we introduce Rescale, a model-agnostic post-processing agent for localized scale correction without modifying the source generator. Experiments reveal that state-of-the-art image generators and editors cannot reliably observe relative scale yet, while Rescale consistently improves scale plausibility across generated and edited images. Together, GenScale establishes relative object scale as a distinct, measurable, and actionable capability for image generation systems.
[CV-99] Soft-Argmax for the Projective Plane via the Veronese Embedding
链接: https://arxiv.org/abs/2609.00521
作者: Benjamin El-Zein,Dominik Eckert,Paul Zech,Christopher Syben,Bernhard Geiger,Steffen Kappler,Sebastian Stober
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:From horizon detection to fibre structures in X-ray imaging, many vision tasks recover lines via peak detection in Hough space H=S^1\times\mathbbR , the domain of orientation-offset pairs (\theta,\rho) . Differentiable pipelines extract coordinates via \emphsoft-argmax, a probability-weighted average that is only meaningful in a globally linear space. However, (\theta,\rho) and (\theta+\pi,-\rho) describe the same undirected line, so H double-covers the space of undirected lines H/\mathbbZ_2 : a Möbius strip, obtained by identifying each pair under \mathbbZ_2 action. Soft-argmax operates on the cover H , but since H/\mathbbZ_2 admits no linear structure, it tears geometrically adjacent lines apart. Thus we need a \mathbbZ_2 -invariant embedding of lines into a linear space, on which soft-argmax is well-defined. We achieve this by parametrising lines via unit-norm homogeneous vectors \ell=(1+\rho^2)^-1/2(\cos\theta,\sin\theta,-\rho)^\top\in\mathbbR^3 and applying the Veronese map v_2(\ell)=\ell\ell^\top that satisfies v_2(\ell)=v_2(-\ell) . This descends continuously to an embedding of the quotient H/\mathbbZ_2 into the linear space \mathrmSym^2(\mathbbR^3) , where the antipodal ambiguity vanishes. Line extraction becomes a barycentre in \mathrmSym^2(\mathbbR^3) , projected back via its leading eigenvector. We validate our \emphVeronese soft-argmax in a Hough transform-based network across all resolvable lines, confirming uniform and seam-free recovery. We further derive that the L_2 -loss on isometrically weighted Veronese embeddings equals the squared chordal distance between lines in projective space, enabling a geometrically precise training objective.
[CV-100] ViTAL-X: Video-Text Alignment with Cross-Modal Temporal Edits
链接: https://arxiv.org/abs/2609.00505
作者: Sethuraman T V,Savya Khosla,Onkar Kishor Susladkar,Aditi Tiwari,Seoung Wug Oh,Kushal Kafle,Joon-Young Lee,Derek Hoiem,Simon Jenni
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Video-text models adapted from image-text architectures (e.g., CLIP) frequently exhibit temporal blindness, the inability to perceive fundamental cues like order, direction, and motion dynamics. Standard datasets mask this limitation by enabling models to exploit static spatial shortcuts. To systematically evaluate this, we introduce XTE-Bench, a diagnostic probe revealing that even large-scale video-language models struggle with basic temporal reasoning, indicating that parameter scaling alone is insufficient to resolve this flaw. To address this, we propose Cross-Modal Temporal Edits (XTE), a self-supervised framework that injects precise temporal supervision. By performing synchronized video-text transformations, XTE generates hard temporal negatives without manual annotation. We instantiate this with ViTAL-X, a lightweight model that equips frozen image-text backbones with temporal awareness while preserving their foundational spatial knowledge. Across six temporal benchmarks, ViTAL-X achieves state-of-the-art performance. Utilizing only 0.4B parameters and 1M training clips, ViTAL-X outperforms 7B-parameter models and surpasses baselines trained on 600x more data. These results demonstrate that targeted, high-quality temporal alignment provides a highly efficient alternative to pure scaling.
[CV-101] SAM3-LoRA: Parameter-Efficient Adaptation of a Concept-Promptable Foundation Model for Multi-Class Structural Defect Segmentation
链接: https://arxiv.org/abs/2609.00469
作者: P. Malaisree,S. Youwai,S. Janrungautai,D. Amorndechaphon,P. Rojanavasu,W. Songkitti
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Promptable segmentation foundation models such as SAM3 accept an open-vocabulary text concept and return every instance matching it, but adapting them to a specialized domain by full fine-tuning is computationally prohibitive for the organizations that would benefit most. This study applies Low-Rank Adaptation (LoRA) to SAM3 for multi-class structural defect segmentation and examines both how such a model can be supervised from conventional annotation and whether the resulting efficiency gain transfers across datasets. Two contributions are methodological. First, we describe a supervision procedure that trains a concept-promptable model directly from COCO-style class-labeled instance segmentation by using the category name itself as the prompt, requiring no prompt templates, no synonym expansion, and no learned class embeddings. Second, we identify and mitigate a failure mode specific to this setting: because a conventional annotation file yields positive prompts exclusively, the model’s presence prediction decouples from the text condition and degenerates into responding to any prompt, a collapse that is invisible to every metric computed on positive prompts alone. Exhaustive hard-negative prompting, in which every dataset category absent from an image is issued as a zero-detection query, addresses this at no annotation cost. Two adapter placements were compared under an identical protocol, updating 0.121% and 1.341% of model parameters. On a purpose-built tunnel lining dataset, pixel intersection-over-union improved from 0.017 to 0.338 and instance-level recall from 0.375 to 0.672; on the independent public Structural Defects Dataset, from 0.017 to 0.855 and from 0.574 to 1.000. Improvements were directionally consistent across ten metrics on both datasets, and the largest per-category gains occurred precisely where zero-shot competence was absent.
[CV-102] Instance-Guided Report Anchoring for Text-Free 3D Abnormality Segmentation in Chest CT
链接: https://arxiv.org/abs/2609.00447
作者: Zhenyu Bu,Haoyan Ding,Chushu Shen,Xinyuan Zheng,Peiyu Duan,Xueqi Guo,Sepehr Farhand,Yoshihisa Shinagawa,Gerardo Hermosillo,Chaowei Wu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Accurate 3D abnormality segmentation in chest CT requires dense spatial supervision, but obtaining expert voxel-level labels is costly. Radiology reports, however, are routinely generated during clinical interpretation and contain instance-specific descriptions that can provide additional guidance without new dense annotation. Existing vision-language grounding methods typically require report-derived findings at inference, making localization dependent on paired text and limiting each forward pass to a queried finding. We propose Instance-Guided Report Anchoring (IGRA), a model-agnostic module that preserves the correspondence between each annotated abnormality instance and the report finding that describes it. IGRA pools each instance representation and anchors it to the corresponding finding embedding during training; all text-related components are discarded at inference. We further reformulate free-text grounding on ReXGroundingCT as multi-label volumetric segmentation by merging same-category instances, allowing all abnormality categories to be predicted in one image-only forward pass. IGRA improves Dice by 22.5% over the strongest image-only baseline (30.93 vs. 25.25) and is comparable to VoxTell on the single-finding subset (30.29 vs. 30.43). Applied unchanged to four standard 3D segmentation backbones, IGRA improves Dice and hit rate across all architectures. Zero-shot evaluation on LIDC-IDRI, PleThora, and a private in-house dataset further shows consistent gains over image-only baselines.
[CV-103] CRAD: Class-wise Reliability-Aware Distillation for Decentralized Heterogeneous Federated Learning
链接: https://arxiv.org/abs/2609.00446
作者: Baraa Bilbeisi,Mengchen Fan,Baocheng Geng,Qing Tian
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Conventional federated learning (FL) relies on parameter averaging, which forces clients to be doubly homogeneous: it demands an identical architecture and degrades under non-IID data. Real-world deployments usually break both assumptions. We sidestep both by building a decentralized knowledge distillation framework in which each client evaluates its peers’ model snapshots on its own local data and distills from the resulting soft predictions. Because knowledge is transferred through the shared class posterior, clients are free to run different architectures; and because every teacher is evaluated on the student’s own device, raw data never leaves the client, with no central server or public dataset required. Within this setting, we identify and address an under-examined problem: how to combine the peer teacher predictions. Existing methods, like uniform averaging, ignore how knowledge reliability varies across teachers and classes. We propose Class-wise Reliability-Aware Distillation (CRAD), which, per class, first discards teachers that disagree with the peer consensus and then takes a weighted average of the rest, weighting each teacher by its per-class reliability (precision, or inverse variance). Since the variance of an accuracy from n samples scales as 1/n , support enters automatically: among the teachers that survive filtering, a teacher is trusted for a class to the degree that it is both accurate and well-evidenced for it. On three image-classification benchmarks (CIFAR-10, CIFAR-100, and PathMNIST colon pathology), across heterogeneous architectures under severe non-IID skew, CRAD consistently outperforms competing methods in global accuracy.
[CV-104] Unmasking Face Embeddings: Reading Rendering and Naming with Foundation Models
链接: https://arxiv.org/abs/2609.00411
作者: Fizza Rubab,Yiying Tong,Arun Ross
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Modern face recognition (FR) owes much of its success to deep neural networks that learn to extract compact identity embeddings from face images. These models are typically trained for identity discrimination, producing embeddings that are highly effective for biometric matching but largely opaque to semantic interpretation. In contrast, foundation models, pretrained on broad visual or vision–language tasks, provide rich interfaces for describing, retrieving, generating, and organizing visual content. This contrast raises a natural question: what capabilities become available when face embeddings from domain-specific FR models are made interoperable with foundation models? Building on recent work on embedding compatibility across models, we use simple pre-computed linear transformations, estimated from paired embeddings alone, to connect existing FR models with off-the-shelf foundation models. Once aligned with a foundation model, a face embedding can be ‘unmasked’ in multiple ways, without training or modifying either model: it can be read in natural language, enabling free-form text queries over a gallery of FR embeddings; rendered into a face image that recovers a person’s appearance, using an unmodified diffusion decoder; and converted to a name, enabling identification even in the absence of an enrolled face gallery. In effect, one linear transformation turns an identity embedding into a rich embedding for web-scale foundation models. This interoperability exposes face embeddings as semantically and visually rich biometric representations, with direct implications for interpretability, retrieval, reconstruction, and template security.
[CV-105] SlideMix: Enhancing Whole Slide Image Analysis via Multimodal Shuffling
链接: https://arxiv.org/abs/2609.00396
作者: Chad Wong,Sicheng Chen,Tianyi Zhang,Enhui Chai,Yueming Jin,Zeyu Liu,Fei Xia
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Histopathological whole slide images (WSIs) are central to cancer diagnosis, but their gigapixel scale, tissue heterogeneity, weak slide-level supervision, sparse diagnostic regions, and multi-scale evidence make robust automated analysis challenging. Multiple instance learning (MIL) is widely used to aggregate tile-level features into slide-level predictions, yet existing augmentation strategies often perturb tissue regions without preserving diagnostic relevance, slide context, or cross-scale structure. We propose SlideMix, a model-agnostic multimodal augmentation framework for MIL-based WSI analysis. SlideMix uses a retrieval-augmented vision-language model (VLM)-based Visual-Language Adaptive Region selector to identify diagnostically relevant regions and reduce weak-label noise. It then performs In-place Tile Shuffling within meaningful tissue regions to mix feature embeddings while preserving slide-level context. A VLM-based soft-labeling module supervises mixed samples, while a multi-factor, loss-driven online Curriculum-Learning Feedback scheme adaptively controls shuffle granularity, feature similarity, and shuffle ratio to promote cross-scale representation learning. Across 11 WSI datasets comprising 20,523 slides, 8 diagnostic tasks, and 10 WSI backbones, SlideMix improves accuracy and generalization in most settings and compares favorably with established augmentation baselines, providing a simple plug-and-play approach for more robust and scalable digital pathology models. Source code: this https URL
[CV-106] FoldingAgent : Inferring Parametric Origami Procedures from Demonstration Videos SIGGRAPH
链接: https://arxiv.org/abs/2609.00377
作者: Maya Moriya,Sigal Raab,Yael Vinker,Tali Dekel
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Graphics (cs.GR)
备注: Project Page: this https URL Accepted to SIGGRAPH ASIA 2026
Abstract:We present FoldingAgent, an agentic framework for inferring explicit parametric folding programs directly from origami demonstration videos. Our framework leverages the reasoning power of a pre-trained Vision-Language Model (VLM) equipped with a suite of specialized tools that enable the agent to simulate geometric transitions, verify physical plausibility, retrieve and compare visual content, and evaluate its own predictions. To translate visual content into folding programs, we define a parametric space that consists of the paper’s geometry and a set of parametric folding actions. Unlike models that predict static crease patterns, our agent operates sequentially and possesses the ability to re-plan its actions, effectively mitigating the compounding errors inherent in multi-step folding. Our approach takes a step toward closing the gap between human origami knowledge, which is primarily shared through unstructured visual demonstrations, and computational methods, which typically rely on structured, parametric representations such as a crease pattern or an executable parametric plan. We evaluate our approach on PurelandFold, a newly curated benchmark of diverse Pureland origami videos with ground-truth geometry and action labels. Our results demonstrate that by combining VLM reasoning with a set of specialized tools and physical simulation, we can successfully transform unstructured visual demonstrations into executable, physically plausible folding procedures.
[CV-107] Adapting Without Gradients: Affine Statistics Transport and What Its Certificate Can Tell You
链接: https://arxiv.org/abs/2609.00374
作者: Salim Khazem,Ibrahim Mohamed Serouis
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Test-time adaptation (TTA) typically assumes that model parameters can be updated at inference time. This assumption is restrictive for inference-only accelerators, frozen or third-party models, and memory-constrained deployments, and standard BatchNorm-based TTA configurations may also become inactive on architectures without BatchNorm. We study adaptation when the learned model must remain frozen. We introduce CASTER, a gradient-free method that stores source class statistics in a discriminative subspace, estimates a class-shared affine transformation from target-batch moments, and analytically transports the source class distributions before classification. CASTER requires no backward pass, optimizer state, or stored source feature bank. Across four backbones and seven datasets, it outperforms k-NN on identical frozen features in 27 of 28 backbone-dataset settings while retaining a median of 18x less state. Affine transport is not always reliable. On ImageNet-C, where batches contain only 64 samples for 1000 classes, unconditional transport loses 21.2 top-1 points. We therefore introduce an empirical residual-to-margin transportability certificate. Across 307 evaluation cells, every transport losing more than 10 points has certificate value above 3.9, although benign and destructive regimes are not perfectly separated. Gating converts an average -3.35 -point effect of unconditional transport into a +1.69-point gain, and performance remains within 0.3 points of the best threshold over a broad threshold range. Finally, we show that this certificate is mechanism-specific: when applied to Tent, it accepts only 4.3% of updates and preserves 0.6% of Tent’s available gain. These results position CASTER as a lightweight adaptation mechanism for frozen-model deployment, together with an explicit account of when its safety signal is informative and when it is not.
[CV-108] Puppeteer: Object-Grounded Posture-Aware Co-Speech Gesture Generation
链接: https://arxiv.org/abs/2609.00369
作者: Vida Adeli,Soroush Mehraban,Jacob Rommann,Harrison Sanborn,Cole Clifford,Babak Taati
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Generating co-speech gestures that are temporally coherent, semantically aligned with speech, and grounded with surrounding objects remains challenging. Prior speech-driven gesture models emphasize audio-gesture alignment but do not explicitly account for posture constraints or surrounding objects, failing to capture the inherent correlation between body gestures and the physical space. We present Puppeteer, a posture-aware, object-grounded co-speech gesture diffusion model operating in a causal latent space. We decompose long gestures into structured primitives and learn a causal variational autoencoder that encodes them into temporally ordered latent tokens, each depending only on the past. We then perform conditional diffusion directly in the causal latent space, conditioning on speech signals, motion history, an initial posture reference, and object geometry to synthesize physically consistent gestures. This temporally ordered latent formulation enables explicit temporal control and supports tasks such as gesture in-betweening and gesture completion. To better assess co-speech gesture synthesis beyond existing measures, we introduce new evaluation metrics tailored to this task. We also created SceneGes, the first curated synthetic 3D dataset of embodied co-speech gestures and corresponding 3D objects, enabling object-grounded gesture generation. Experiments show that Puppeteer generates more diverse and temporally synchronized gestures than prior methods, while enabling object-grounded gesture synthesis.
[CV-109] Where Should Experience Live? Hierarchical Hebbian Memory for Continual Vision Transformers
链接: https://arxiv.org/abs/2609.00358
作者: Mohammed Yusuf Mujawar,Noorbakhsh Amiri Golilarz
类目: Computer Vision and Pattern Recognition (cs.CV); Neural and Evolutionary Computing (cs.NE)
备注:
Abstract:Vision Transformers provide strong visual representations but typically rely on slowly updated parameters, limiting their ability to organize newly acquired information across different memory timescales. This work proposes \textitHierarchical Hebbian Memory, a three-level memory architecture composed of rapid Working Memory, persistent Routed Episodic Memory, and slower Semantic Memory. A learned controller regulates memory contribution, read and write routing, plasticity, retention, and consolidation. A causal read-before-write lifecycle ensures that the current outcome cannot influence the prediction it supervises. The architecture is evaluated on Omniglot 5-way 1-shot recognition and CORe50 continual object recognition. With Swin-Tiny, the hierarchical model reaches 97.39% accuracy on Omniglot and 95.37% final accuracy on CORe50 when combined with experience replay. Learned multi-bank retrieval reaches 47.50% delayed-association accuracy, compared with 24.17% for a single persistent bank and 25.00% without memory. After intervening distractors, Episodic Memory retains approximately 0.96 cosine similarity with stored associations, while Working Memory falls to approximately 0.05. These results show that Hebbian association and learned memory routing can jointly organize online visual experience across rapid, persistent, and consolidated memory timescales within Vision Transformers.
[CV-110] RUST: Threshold-Recalibrated Uncertainty-Safe Training for Certified Dismissal in Breast Cancer Screening
链接: https://arxiv.org/abs/2609.00300
作者: Parham Hajishafiezahramini,Matthew Hamilton,Edward Kendall,Gregory Doyle,Oscar Meruvia Pastor
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:Reducing the review of clearly cancer-negative screening mammograms could lower radiologist workload without compromising cancer detection. We propose a closed-loop threshold-aware training strategy in which the dismissal threshold is recalculated during training and used to penalize cancer-positive images that approach the dismissal region. We evaluated the method on NLBS and RSNA using five controlled training configurations, with case-level assessment based on a one-sided 99% Clopper–Pearson upper bound for cancer prevalence among dismissed cases. The proposed model achieved the highest case-level dismissal rates at both 98% and 95% recall targets. On NLBS, dismissal reached 19.74% and 21.70%, while the cross-entropy baseline did not meet either recall target. On RSNA, dismissal improved from 7.04% to 14.31% and from 13.49% to 19.69%. In external RSNA \to NLBS evaluation, the proposed model achieved dismissal rates of 12.95% and 19.87% at the 98% and 95% recall targets, respectively. These results support closed-loop threshold-aware training for high-recall selective dismissal.
[CV-111] StreamScout: Learning When to Look Deeper for Streaming Video Understanding
链接: https://arxiv.org/abs/2609.00291
作者: Ce Zhang,Jing Bi,Jinxi He,Jianshu Zhang,Jingyang Lin,Yunzhong Xiao,Minghao Fu,Yaqi Xie,Zhentao Xie,Weicong Chen,Katia Sycara,Ming Zhou
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Streaming video understanding requires answering questions that arrive at arbitrary moments over an unbounded video stream. Existing systems primarily focus on what to retain in a bounded memory, yet access that memory using the same fixed-cost procedure for every query, despite substantial variation in the evidence required. We argue that deciding how deeply to access memory for each query is as important as deciding what the memory should store. To this end, we introduce StreamScout, an adaptive inference framework that maintains only a lightweight textual timeline in context as the stream unfolds. At query time, StreamScout progressively augments the timeline with up to three increasingly informative visual views: a glance at recent frames, a uniform look-back over the past stream, and query-salient retrieval. At each stage, the model answers immediately if the available evidence is sufficient; otherwise, it escalates to the next view. To improve this stop-or-escalate policy, we probe the cascade on an auxiliary set and distill the model’s empirical competence boundary into supervision for a lightweight LoRA adaptation, yielding StreamScout-S. We further refine the policy through reinforcement learning, allowing the model to explore stopping behaviors beyond imitation of the distilled decisions, yielding StreamScout-R. Across three backbones and three streaming benchmarks, StreamScout and its variants consistently outperform prior streaming methods while substantially reducing inference cost and token consumption; on OVO-Bench, for instance, StreamScout-S improves Qwen3-VL-8B by 14.65 points while using 59% fewer tokens than uniform sampling and answering in 1.04 s on average.
[CV-112] CrossFeat: Bridging Imaging Modalities in Feature Descriptor Space ECCV2026
链接: https://arxiv.org/abs/2609.00272
作者: Paul Schneider,Nazim Haouchine
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: ECCV 2026
Abstract:Most advances in keypoint descriptions address monomodal settings, where image variations arise from viewpoint, illumination, or contrast changes. Multimodal scenarios involve images produced by fundamentally different sensing processes, such as multispectral imaging, RGB-depth, satellite imagery, or medical imaging, causing the same structures to appear differently. A common solution to cross-modal description is to train descriptors for each modality pair, which requires retraining whenever the modalities change, or to train large models, which incur a significant increase in runtime. Instead, we propose CrossFeat, a framework that enables an existing monomodal descriptor to operate across modalities. Our method learns a crossing function in descriptor space that maps features from one modality to a representation compatible with another. To preserve the structural information captured by the original descriptor, CrossFeat introduces a geometry-appearance disentanglement such that only appearance is altered while the geometric properties are preserved. Experiments across multiple domains and datasets demonstrate improved performance in multimodal matching.
[CV-113] Beyond Blind Compliance: Benchmarking Task Verification in OCR Reasoning
链接: https://arxiv.org/abs/2609.00232
作者: Yue Zhou,Yuan Wu,Yi Chang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Multimodal Large Language Models (MLLMs) have achieved strong performance on OCR-centric document understanding and text-rich visual reasoning benchmarks. Yet existing evaluations largely assume that every task is valid and answerable. In real-world OCR scenarios, this assumption often fails: questions may rely on illegible text, occluded evidence, nonexistent visual targets, contradictory premises, or missing variables. We study this reliability gap as OCR-grounded Task Verification: before answering, a model should determine whether the Image Premise (IP), Textual Premise (TP), and Question (Q) jointly define an executable task. We introduce VeriOCRBench, a 1,800-sample human-verified benchmark built from source images drawn from 8 OCR-related datasets and spanning 8 real-world image domains, with controlled, image-grounded diagnostic tasks. It contains 1,600 trap-injected invalid tasks across 8 trap types and four verification dimensions—Visual, Contextual, Factual, and Logical—plus 200 trap-free controls for measuring over-refusal. Built with a Visual Atomic Fact (VAF)-anchored pipeline and full human auditing, VeriOCRBench enables decoupled evaluation of task verification, root-cause diagnosis, and over-refusal. Evaluating 15 leading MLLMs reveals persistent blind compliance, diagnosis failures, and prompt-induced over-refusal, exposing a critical reliability gap in current OCR reasoning systems. The code is available at: this https URL. Subjects: Computer Vision and Pattern Recognition (cs.CV) Cite as: arXiv:2609.00232 [cs.CV] (or arXiv:2609.00232v1 [cs.CV] for this version) https://doi.org/10.48550/arXiv.2609.00232 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Yue Zhou [view email] [v1] Mon, 31 Aug 2026 18:39:46 UTC (6,095 KB)
[CV-114] Beyond Language Priors: Diagnosing and Fixing Visual-Origin Hallucinations in Multimodal LLM ACM-MM2026
链接: https://arxiv.org/abs/2609.00231
作者: Peiyang Xu,Xiaopei Zhu,Jun Zhu,Xiaolin Hu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Carmera-ready version. To appear in ACM MM 2026
Abstract:Existing research on object hallucination in multimodal large language models (MLLMs) predominantly attributes the problem to language priors such as over-reliance on textual co-occurrence statistics. We challenge this view by presenting quantitative evidence for a complementary, under-explored cause: visual-origin hallucination, where hallucinations arise from incorrect visual feature extraction and misalignment between image and text embeddings. Through cosine similarity analysis and Smooth Grad-CAM entropy measurements, we show that hallucinated samples exhibit systematically lower image-text similarity (average 0.158 vs. -0.122) and inverted attention patterns, where attention is dispersed when the target object is present but wrongly concentrated when it is absent. Guided by this diagnosis, we propose Adversarial Contrastive Fine-Tuning (ACFT). ACFT uses an Adversarial Hallucination Attribute Flipping (AHAF) procedure, involving minimal, targeted adversarial perturbations that flip an image’s hallucination attribute, to construct perfectly aligned positive-negative pairs, which are then used for contrastive fine-tuning. AHAF simultaneously serves as a diagnostic probe, revealing that MLLM visual representations lie dangerously close to hallucination decision boundaries. Requiring only 0.9% of the COCO dataset and adding zero inference overhead, ACFT achieves state-of-the-art performance on POPE, MME, and four description-level hallucination benchmarks across LLaVA, MiniGPT-4, and Qwen2.5-VL. Code is available at this https URL
[CV-115] Distributed Implicit Harm: A Compositional Safety Blind Spot in MLLM -Based Video Moderation
链接: https://arxiv.org/abs/2609.00206
作者: Ruotong Wang,Zihao Zhu,Siwei Lyu,Xin Tao,Baoyuan Wu
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Despite their growing use in video moderation, multimodal large language models (MLLMs) exhibit a compositional safety blind spot: videos composed of seemingly benign components can convey harmful meaning when interpreted as a whole. We refer to this phenomenon as Distributed Implicit Harm (DIH), where harm arises from relations among components distributed along a decomposition axis of the video, rather than from any single explicit cue. Among many possible axes, we study two representative cases: temporally distributed harm across visual segments (DIH-T) and cross-modal harm between audio and visual streams (DIH-M). Studying and mitigating DIH at scale requires data that is difficult to collect: such videos lack compositional harm annotations, evade retrieval based on local visual cues, keywords, or single-modality signals, and are consequently absent from existing safety datasets. To bridge this gap, we develop a multi-agent synthesis framework that composes individually benign components into harmful scenarios and generates diverse DIH videos with explicit reasoning annotations, yielding a dataset of over 9,000 videos spanning visual-only and audio-visual settings. Benchmarking over 30 MLLMs spanning frontier proprietary models and leading open-source systems reveals substantial and consistent deficits in detecting both DIH-T and DIH-M. Notably, this failure persists even among the strongest frontier models: they often correctly assess individual components in isolation but fail to recognize the harmful meaning that emerges from their composition. We further evaluate these models on a manually collected set of real-world DIH videos from social media and observe the same failure mode, highlighting DIH as a practical and underexplored challenge for video moderation.
[CV-116] A Lagrangian View of Flow Matching
链接: https://arxiv.org/abs/2609.00198
作者: Peyman Milanfar
类目: Computer Vision and Pattern Recognition (cs.CV); Fluid Dynamics (physics.flu-dyn)
备注:
Abstract:Modern explicit-time generative models, such as Flow Matching [Lipman et al., 2023] and Rectified Flow [Liu et al., 2023], are typically derived top-down via Optimal Transport and the continuity equation. This standard Eulerian approach focuses on the macroscopic transport of probability mass. In this paper, we present an alternative, bottom-up mechanical derivation grounded in a Lagrangian (particle-centric) perspective. By analyzing the local Taylor expansion of a continuous denoiser, we motivate a strict invariance condition required for optimal, singlestep generation: the conservation of target identity. Enforcing this condition yields a governing quasi-linear advection Partial Differential Equation (PDE). We demonstrate that solving this PDE via the Method of Characteristics analytically yields the straight-line trajectories of Flow Matching. This geometric perspective isolates the Jacobian of the denoiser as the primary source of trajectory curvature, providing a direct mathematical explanation for why straight-line flows enable massive step sizes, and why empirical models require distillation to flatten intersecting characteristics.
[CV-117] ZimaBlue: Evolving Generalizable World Action Models through Scalable Video Pre-training
链接: https://arxiv.org/abs/2609.00188
作者: Xionghao Wu,Yijun Yang,Shiyang Zhou,Haoze Sun,Jianhui Liu,Songsong Yu,Jiyao Zhang,Wenbo Li,Bo Wang,Guoqing Ma,Lin Song,Renjie Liao,Shenghe Zheng,Wei Tang,Xiaojuan Qi,Yanwei Li,Yuan Zhang,Zhuotao Tian,Haoyang Huang,Nan Duan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Robotic manipulation faces a fundamental scaling challenge: robust generalization demands broad physical experience, yet action-labeled robot trajectories are expensive to collect and inherently limited in diversity. Egocentric videos offer a far more scalable source of embodied experience, capturing object interactions, contact dynamics, tool use, and long-horizon behaviors across diverse environments. The central challenge is how to convert this abundant but action-free experience into effective robot control. We introduce ZimaBlue, a scalable framework for learning generalizable World Action Models (WAMs) from large-scale video. ZimaBlue follows a three-stage training curriculum: it first performs causal embodied video pre-training on large-scale human and robot egocentric videos, then grounds the learned visual dynamics in heterogeneous robot trajectories through video-action mid-training with a unified action representation, and finally specializes the model to a target robot for deployment. To make generative WAMs practical for real-time control, ZimaBluefurther adopts an asynchronous Slow-Fast dual-system architecture, where a high-capacity Slow world model provides generalizable spatiotemporal representations and a lightweight Fast branch enables 30 Hz action prediction on NVIDIA RTX 4090. On real-robot zero-shot evaluations, scaling from target-robot data alone to over 120,000 hours of embodied video improves success from 36.1% to 77.8%. ZimaBlue further delivers strong performance across multiple benchmarks, with particularly pronounced gains on unseen tasks.
[CV-118] Qwen -Drive-1.0: An Initial Step towards a Vision-Language Foundation Model for Autonomous Driving
链接: https://arxiv.org/abs/2609.00111
作者: Xin Zhou,Zongchuang Zhao,Zhibo Yang,Mingsheng Li,Humen Zhong,Shuai Bai,Du Chu,Ruizhe Chen,Zhaohai Li,Jun Tang,Qiuyue Wang,Mingkun Yang,Jiazhao Zhang,Dayiheng Liu,Dingkang Liang,Xiang Bai
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Code will be available at this https URL
Abstract:We present Qwen-Drive-1.0, an initial step towards a vision-language foundation model for autonomous driving. Qwen-Drive-1.0 retains the architecture of the pretrained vision-language model (VLM) and integrates 3D perception, visual question answering, and motion planning within a unified framework. An external bird’s-eye-view (BEV) perception head jointly performs 3D object detection, semantic occupancy prediction, and BEV map segmentation. It serves as a probe of the 3D information accessible from the shared representations and provides an explicit, inspectable interface to 3D scene structure. A Planning Expert conditions on shared VLM representations to generate future ego trajectories. A staged training recipe combines driving supervision with general-purpose vision-language data to acquire driving-specific competence while helping preserve broad visual understanding and instruction-following capabilities. Experiments demonstrate strong 3D perception and driving scene understanding while largely preserving general vision-language capability. Comprehensive evaluations across open-loop, pseudo-closed-loop, and closed-loop settings further show highly competitive motion-planning performance.
[CV-119] ReNFT: Repairing Mode Collapse in Reward Post-Training via Internal Probability-Mass Recalibration
链接: https://arxiv.org/abs/2609.00061
作者: Yuchen Bao,Chao Wen,Haowei Wang,Ruoxin Chen,Donghao Luo,Jiahui Zhan,Wenjian Huang,Shen Chen,Yiting Wang,Taiping Yao,Chengjie Wang,Shouhong Ding,Jianguo Zhang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注: 17 pages, 13 figures, 4 tables
Abstract:Reward post-training of diffusion generators inevitably concentrates probability mass on a few reward-favored modes, a mode collapse that erases within-prompt diversity. Existing methods for mitigating collapse rely on external signals or interfaces, augmenting the reward with perceptual objectives, adjusting reference regularization, or modifying the text encoder, but none repairs an adapter that has already collapsed while preserving the acquired reward. We observe that online post-training primarily reallocates probability mass over capabilities inherited from pretraining rather than learning new visual content. Collapse is therefore suppression, not deletion, and can be reversed from within the generator. We propose ReNFT, which repairs a high-reward, low-diversity adapter through internal probability-mass recalibration. Unconditional probes first prioritize “anti-hub” prompts where the prompt-independent bias is easiest to expose. Two policy-dominated mixed routes then generate matched counterfactual proposals from the same prompt and initial noise, one probing the frozen base direction for suppressed alternatives and the other exposing the post-trained unconditional tendency. Reward ranking with an adaptive flipping guard assigns pull and push roles, and a joint-and-paired NFT update realizes the repair. On PickScore and GenEval, ReNFT retains 98.9% and 99.0% of NFT’s reward while improving DreamSim-Div by 58.8% and 55.0%, respectively, offering a complementary alternative to external interventions.
[CV-120] A Cone-Constrained Bilinear Decomposition for Total Scaled-Gradient Variation Models
链接: https://arxiv.org/abs/2609.00036
作者: Haibin Su,Chunlin Wu,Huibin Chang,Zhifang Liu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:The total scaled-gradient variation (TSGV) regularizer, derived from sparse modeling of piecewise-linear structures, has been shown to preserve edges and corners in image restoration. However, its highly nonconvex and nonlinear nature poses severe computational challenges, as existing methods often suffer from parameter sensitivity or lack convergence guarantees. To overcome this, we propose a tailored bilinear decomposition that decouples the nonlinear weighted gradient in the TSGV regularizer. This approach yields an equivalent optimization problem governed by cone or sphere constraints, depending on the chosen scaling function. In particular, the cone constraint plays a central role in characterizing edge- and corner-preserving behavior. We solve this reformulation using the alternating minimization method (AMM) equipped with a majorization–minimization strategy, ensuring a monotonic decrease in energy without step-size tuning. Furthermore, we provide a geometric interpretation of the edge-preserving properties of these constraints by analyzing their asymptotic behavior near image singularities. We establish the global convergence of the proposed method to a critical point within the Kurdyka–Łojasiewicz framework. Extensive numerical experiments on Gaussian denoising and non-line-of-sight (NLOS) imaging show that the proposed method achieves PSNR and SSIM competitive with or superior to representative variational methods, especially at high noise levels, and improves the structural reconstruction under dense and sparse scanning.
[CV-121] SCAFFOLD: A Large-Scale Structured Dataset of Computer Science Research Figures with Diagram QA and Chain-of-Thought Reasoning Traces
链接: https://arxiv.org/abs/2609.00018
作者: Ranjit Raut,Aarav Subedi,Sagun Rai,Sudan Jha
类目: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注: 9 pages, 3 figures
Abstract:Computer science papers rely heavily on diagrams: architecture drawings, system flowcharts, and pipeline schematics that often carry more information than the text around them. There is currently no public dataset that pairs this specific kind of figure with captions, context, questions, answers, and step-by-step reasoning, which is exactly what is needed to train a vision-language model to understand them. We present \textbfSCAFFOLD\footnotethis https URL, a large-scale structured dataset of computer science research figures with diagram QA and Chain-of-Thought reasoning traces. This dataset consists of (image, caption, context, question-answer, chain-of-thought) tuples from arXiv computer science papers prepared using layout detection and PDF parsing, with an AI-assisted question-generation step. The resulting large-sized SCAFFOLD-157K dataset spans 3,058 papers with 29,887 figures (157,387 pairs), a medium-sized SCAFFOLD-37K dataset (36,797 pairs), and a small-sized SCAFFOLD-12K dataset (12,000 pairs). We used SCAFFOLD-12K for baseline experiments on Qwen2.5-VL-3B-Instruct.
[CV-122] A survey of AI-generated voices and their detection
链接: https://arxiv.org/abs/2608.15411
作者: Chengzhe Sun,Tianle Yang,Siwei Lyu
类目: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:The ability of artificial intelligence (AI) models to generate highly realistic human voices has advanced rapidly. These technologies power accessibility tools, virtual assistants and creative applications, but they also enable harmful uses, including impersonation, fraud and disinformation. Recent incidents of voice cloning scams targeting businesses and political leaders underscore the urgent need for robust safeguards. Unlike image and video deepfakes, the detection of synthetic voices poses unique challenges due to the complexity of phonetics, prosody and auditory perception. This survey offers a comprehensive overview of AI voice generation and detection methods, encompassing both the technical foundations and the latest state-of-the-art advances. This study also identifies key open challenges, benchmark resources and future directions to make this survey useful for future researchers.
[CV-123] A Sensor-Adaptive Incremental Learning Framework for Artifact Detection in Satellite Precipitation Data
链接: https://arxiv.org/abs/2609.01514
作者: Andres F. Monsalve,Hernan A. Moreno,Christian D. Kummerow
类目: Atmospheric and Oceanic Physics (physics.ao-ph); Computer Vision and Pattern Recognition (cs.CV); Applications (stat.AP)
备注: 16 pages. Submitted to IEEE Transactions on Geoscience and Remote Sensing
Abstract:Historically, retrieving rainfall data from satellite imagery has been the domain of space agencies. However, in recent years, the development of cheaper, more compact satellites (SmallSats) capable of detecting rainfall proxies has led to a significant increase in private-sector initiatives for satellite launch and surface precipitation products. This rapid growth has yet to be matched by data validation efforts. Consequently, the need for a robust tool to detect anomalies in near-real-time data before it is disseminated to the public has become critical. In this paper, we present the development of an anomaly-detection system to identify artifacts in global satellite-based rainfall products. The developed framework leverages pre-trained computer vision models and incorporates scarce human-labeled data to detect specific anomalies. Our proposed anomaly detection strategy is tested on data from the Special Sensor Microwave Imager (SSMI) and the Special Sensor Microwave Imager/Sounder (SSMIS). Results demonstrate the efficacy of our approach at separating regular orbits from artifact-containing orbits for each satellite, with performance comparable to state-of-the-art in-place methods. Additionally, the framework offers explainability and the capacity for iterative refinement following false-positive or false-negative classifications.
[CV-124] Lightweight Interpretable RGB-Guided Hyperspectral Super-Resolution under Real Cross-resolution Misalignment
链接: https://arxiv.org/abs/2609.01060
作者: Mohamad Jouni,Aurélien Godet,Mauro Dalla Mura
类目: Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Compact snapshot hyperspectral cameras provide rich instantaneous spectral measurements for ground-level machine vision, but at lower spatial resolution than standard RGB cameras. RGB-guided hyperspectral super-resolution (HSR) addresses this limitation by transferring spatial detail from a high-resolution RGB guide to a low-resolution hyperspectral image (HSI). These dual-camera systems are typically in a horizontal rig geometry, requiring cross-camera image alignment due to different fields of view. However, residual misregistration can inject spurious high-frequency details. Existing learned unaligned-fusion methods are usually trained for a fixed spectral support and spatial scale factors and can be computationally demanding, limiting their flexibility across sensors. We propose a lightweight and interpretable RGB-guided HSR framework combining cross-modal flow alignment with model-based Gram-Schmidt orthogonalization fusion. The method first warps the RGB guide onto the HSI grid, then estimates an energy-based confidence weight map by measuring local alignment reliability. This map is then used both in a weighted least-squares spectral regression and in a gated fusion between the super-resolved estimate and an HSI-preserving estimate. Unlike existing learned methods, the proposed framework has a low computational footprint and supports VIS-NIR spectral supports and scale factors without retraining. Experiments on the Real benchmark show that the proposed method improves reconstruction accuracy over learned fusion baselines while remaining substantially faster. On a 34-frame sequence acquired with our real RGB-HSI dual-camera setup, a reduced-resolution quantitative evaluation validates the method under genuine cross-sensor radiometric, noise, and geometric differences, while native-resolution qualitative results demonstrate deployment on the full 51-band VIS-NIR acquisition.
[CV-125] PyDoseRT Proton: A GPU Pencil-Beam Engine with a Convolutional Residual-Correction Network for Fast Proton Dose Calculation
链接: https://arxiv.org/abs/2609.01018
作者: Lukas Zimmermann,Hermann Fuchs,Attila Simkó,Gerd Heilemann
类目: Medical Physics (physics.med-ph); Computer Vision and Pattern Recognition (cs.CV)
备注: DoseRAD 2026 report - Team DoseHappens
Abstract:Architecture category. Hybrid method: a physics-based analytical pencil-beam (PB) dose engine followed by a 3-D convolutional residual-correction network (RepVGG-U-Net). We addressed the DoseRAD2026 proton dose-prediction task with PyDoseRT Proton, a GPU-accelerated engine implemented in PyTorch and augmented by a learned residual toward Monte Carlo (MC) accuracy. A double-Gaussian PB kernel was calibrated to GATE/Geant4 integrated depth doses in water in two stages: a classical per-energy curve fit, then a gradient-based fit of the full 3-D dose through the PyTorch physics engine as it retains a differentiable execution path for gradient-based optimization of dose-dependent objectives. The engine computes each beamlet on a beam’s-eye-view (BEV) lattice with variance-preserving Gaussian splitting, an analytic nuclear halo, and a Fermi-Eyges heterogeneity term, then rotates the result into the patient frame. Additionally, a compact residual U-Net predicts an additive correction in BEV space. It is conditioned on voxelwise material-label embeddings, a discrete energy embedding and spot size. The same model was used for all anatomical sites (thoracic and abdominal). It was trained with a patient-space L1 objective emphasizing the scored high-dose region and multi-scale BEV deep supervision. The submitted CT configuration obtained preliminary-test beamlet MAE 0.0066, image-z IDD distance 0.0025, plan MAE 0.0049, 98.30% gamma pass rate (1%/1 mm), and DVH error 0.460.
[CV-126] Prior-Guided Implicit Neural Representations for Single-Subject Diffusion MRI Super-Resolution BMVC
链接: https://arxiv.org/abs/2609.00981
作者: Abdulkader Ghandoura,Marsil Zakour,William Consagra,Yogesh Rathi
类目: Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at the British Machine Vision Conference (BMVC) 2026. 16 pages, 5 figures, 2 tables. Project page: this https URL
Abstract:Resolving complex fiber geometries in brain white matter requires high-resolution diffusion MRI at the cost of long acquisition times. This leads many clinical protocols to opt for low-resolution scans, making downstream microstructure estimation and tractography challenging. Implicit neural representations (INRs) can model the diffusion signal continuously, enabling native single-subject super-resolution by querying the network at arbitrary spatial coordinates, yet existing methods often suffer from long training times and lack a mechanism to incorporate anatomical priors to regularize super-resolution by constraining the space of plausible reconstructions. To address these limitations, we propose a novel transfer-learning framework that pre-trains an INR on a high-resolution template and then adapts it to subject-specific scans via registration and fine-tuning. For 4\times through-plane super-resolution from 5 mm to 1.25 mm on Human Connectome Project (HCP) data, our method reduces NRMSE by 36-49% and increases FSIM by 24-43% over a recent baseline with 6\times faster training, outperforming competing INR-based methods across both image quality and domain-specific metrics. Code is available on the project page at this https URL .
[CV-127] Stochastic Optimization of Tree Tensor Networks
链接: https://arxiv.org/abs/2609.00870
作者: Marius Willner,Maximilian Scharf,André Uschmajew,Timo Felser,Marco Trenti
类目: Optimization and Control (math.OC); Computer Vision and Pattern Recognition (cs.CV); Computational Physics (physics.comp-ph)
备注: 26 pages, 12 figures, 5 pseudo-code algorithms
Abstract:Tensor networks, originally developed for quantum many-body physics, are promising models for machine learning. We derive stochastic Riemannian optimizers for tree tensor networks (TTNs) on both their parameter and quotient manifolds, including adaptive and learning-rate-free schemes suitable for minibatch training. Using a hybrid CNN-TTN architecture, we evaluate the methods on Fashion-MNIST, CIFAR10, and Imagenette. The proposed optimizers achieve predictive performance comparable to unconstrained optimization while enabling numerically stable downstream compression.
[CV-128] Panda Diplomacy: Foundation Model Pre-training across Particle Imaging Detectors for High Energy and Nuclear Physics
链接: https://arxiv.org/abs/2609.00611
作者: Samuel Young,César Jesús-Valls,Kazuhiro Terao
类目: High Energy Physics - Experiment (hep-ex); Computer Vision and Pattern Recognition (cs.CV)
备注: 24 pages, 11 figures, preprint
Abstract:Foundation models are increasingly being pursued in particle and nuclear physics, but existing approaches remain strongly tied to individual experiments through detector-specific architectures or pre-training objectives, limiting their reuse across sensing modalities. We show that a point cloud self-distillation framework yields a substantially more general sensor-level pre-training recipe. We show that the same refined architecture and objective can be independently pre-trained with minimal changes on three qualitatively different detector modalities: liquid argon time projection chamber (LArTPC), collider TPC, and water Cherenkov. Using 1,000 labeled images for downstream task adaptation, Panda V2 matches or exceeds specialized foundation-model baselines trained with orders of magnitude more supervision, matching state-of-the-art particle-clustering performance with 70x fewer labeled events on sPHENIX while substantially improving particle identification, and on LArTPC data matching Panda (arXiv:2512.01324) particle reconstruction with up to 1,000x fewer labels. Beyond reconstruction, simple linear probes reveal physically meaningful latent structure associated with particle causality and track curvature.
[CV-129] Expert-like Bone Ultrasound Segmentation through Expert-in-the-loop Mask-conditioned Progressive Learning
链接: https://arxiv.org/abs/2609.00473
作者: Arash Tavangar,Larissa K. Chiu,Hamidreza Khodashenas,Gregory K. Berry,Amir Hooshiar
类目: Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Manual annotation remains a major bottleneck in ultrasound (US) bone segmentation, where experts typically iteratively refine rough brush masks rather than delineating precise contours in a single pass. We present ExiL, a mask-conditioned progressive learning framework that models annotation as a structured refinement trajectory. ExiL combines a synthetic expert-like brush simulator based on signed distance fields with a lightweight 7.8M-parameter U-Net that learns to complete and refine imperfect masks from US images. During deployment, an expert mode updates the model directly from accepted refinements, enabling continual adaptation to expert behavior. Evaluated using UltraBones100k cadaver data for quantitative segmentation and a prospective volunteer dataset for annotation-efficiency analysis, ExiL reduced single-expert average annotation time from 60 to 20 seconds per frame (66.7%) and improved mean Dice by approximately 0.045 over non-progressive training, while achieving 0.87 Dice and 2.7 px boundary error in the best trajectory-aware setting. With 10–50 ms inference, ExiL enables real-time, self-improving annotation for US-guided orthopedic workflows in practical clinical labeling.
人工智能
[AI-0] Can LLM s Discover Scientific Laws in Real and Parallel Worlds?
链接: https://arxiv.org/abs/2609.01552
作者: Yiming Huang,Ziche Liu,Zhuohang Wu,Yiqian Wang,Junxia Cui,Xinkai Zou,Linjun Mao,Nan Huang,Naicheng Yu,Kaijie Zhu,Yue Ma,Kun Zhou,Letian Peng,Jingbo Shang
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 42 pages, 16 figures. Project page: this https URL
Abstract:Scientific equation discovery has long been central to scientific progress, proceeding through iterative cycles of hypothesis generation, observational testing, and refinement under scientific constraints. As LLM capabilities advance and their role in AI for Science expands, it remains an open problem whether they can genuinely discover scientific laws and how this ability should be evaluated. Existing evaluations, however, often either simplify discovery through synthetic settings or reuse published targets that may already be familiar to LLMs. We therefore introduce SCILAWS-BENCH, a benchmark for scientific law discovery built from published research and real scientific data. It comprises 118 problems drawn from 381 scientific papers, covering 291 candidate laws and roughly 8M real data points across six scientific disciplines. Each problem is instantiated in two complementary settings: (1) SCILAWS-REAL asks models to propose laws from fixed real observations and evaluates held-out predictive fit and scientific validity derived from the source literature, and (2) SCILAWS-PARALLEL asks models to actively query residual-calibrated worlds and recover synthesized hidden laws derived from published forms. This two-setting task design preserves each problem’s scientific context while separately evaluating fixed-record law discovery and active recovery of a newly synthesized hidden law. We find that predictive fit can diverge from scientific validity, memorization shapes whether models reproduce or move beyond published formulas, and our best-of-N study reveals a selection bottleneck. Our work provides a paper-grounded benchmark and new empirical perspectives for evaluating AI for scientific discovery. Project page: this https URL
[AI-1] A Mathematical Theory of Reusable Neural Bases for Network Compression
链接: https://arxiv.org/abs/2609.01550
作者: Binshuai Wang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:As large AI models become increasingly prevalent across a wide range of applications, memory cost has become a critical bottleneck in both training and inference. To mitigate this issue, we introduce the Linear Reusable Neural Bases Architecture (LRNBA), a novel framework aimed at improving parameter efficiency and reducing memory cost. Inspired by recurrent neural network (RNN) designs, the core idea of our approach is to represent each network block as a linear combination of a shared set of neural bases, thereby enjoying highly network compression rate while maintaining stable training. The proposed architecture allows for the construction of significantly wider and deeper networks under the same parameter budget. Extensive experiments demonstrate that our model achieves comparable or even faster convergence and lower loss than classical architectures, while maintaining stable training dynamics.
[AI-2] Can LLM s Design Video Coding Tools? A Case Study on Planar Mode
链接: https://arxiv.org/abs/2609.01535
作者: Yingwen Zhang,Meng Wang,Liqiang He,Shiqi Wang
类目: Multimedia (cs.MM); Artificial Intelligence (cs.AI)
备注:
Abstract:This paper explores whether large language models (LLMs) can design video coding tools, a highly challenging task due to the intricate algorithmic coupling of tool modifications. In particular, we present an empirical case study on the Planar mode, a long-standing intra prediction tool in video coding standards. Our experiments operate within a generation-and-evaluation loop, with the LLM generating new Planar predictors, encoder trials evaluating their coding performance, and the LLM re-generating refined implementations based on the evaluation feedback. We first examine directly replacing the default Planar mode in the Fraunhofer Versatile Video Encoder (VVenC) under its faster preset. Experimental results demonstrate that the LLM-generated mode can outperform the conventional Planar mode on this lightweight toolset, achieving 0.18% bitrate savings with 0.4% complexity overhead on the standard benchmark. We further extend our evaluation to the Enhanced Compression Model (ECM). Leveraging newly introduced directional Planar modes, we investigate two integration strategies: directly replacing them, and introducing the LLM-generated predictor as an additional prediction mode with new syntax elements. The empirical results suggest that both strategies can yield coding gains under a constrained low-resolution setting. Overall, this study offers preliminary evidence and practical insights, highlighting both the potential and open challenges of LLM-based coding tool design.
[AI-3] EvoSCM: Scientific Belief Revision Through Causal Model Evolution and Experimentation
链接: https://arxiv.org/abs/2609.01526
作者: Qing Zhao,Haowei Li,Weijian Deng,Pengxu Wei,Liang Lin
类目: Artificial Intelligence (cs.AI)
备注: This is a preliminary version, with further details and results to follow
Abstract:Scientific agents must learn not only how to reason, but also what to believe. However, existing LLM agents typically express scientific hypotheses in free-form text, leaving their beliefs implicit and difficult to test or revise. We introduce EvoSCM, which equips scientific agents with explicit structural causal models that evolve as new experimental evidence is collected. EvoSCM maintains a population of competing SCM hypotheses, each encoding a candidate causal explanation of the environment, and evolves them through a closed discovery loop. In each round, the agent abduces latent mechanisms from accumulated evidence, designs discriminative interventions, and commits to falsifiable predictions that it tests through experimentation. Discrepancies between prediction and observation are inductively distilled into correction rules that revise the causal structures and mechanisms of each hypothesis, and the agent then deductively validates the revised population against accumulated evidence and structural consistency to guide the next round. We evaluate EvoSCM on DiscoverPhysics, a benchmark requiring agents to uncover the hidden dynamics of noncanonical physical worlds through experimentation. EvoSCM consistently improves scientific discovery over baselines, yielding more accurate explanations and predictions while making more effective use of experimental interactions.
[AI-4] Relational-Core Graph Analytics Querying graphs at SQL scale and why the node/edge model is a performance tax not a truer picture of connected data
链接: https://arxiv.org/abs/2609.01525
作者: Gene Zhang
类目: Databases (cs.DB); Artificial Intelligence (cs.AI); Programming Languages (cs.PL)
备注:
Abstract:A durable assumption holds that graph analytics requires a purpose-built graph engine, and that relational systems are ill-suited to connected data. We argue the opposite for the workloads enterprises actually run. A columnar relational engine fronted by a graph query language matches or exceeds native graph engines on analytical graph queries, and - decisively - scales past the point where in-memory graph engines fail. We further argue that the node/edge property graph is not a more faithful model of connected data but a re-encoding of relationships that already exist explicitly in relational tables; reconstructing them at query time is pure overhead. We present ClickGraph and its Databricks-dialect sibling DeltaGraph, systems that translate Cypher directly onto the native relational schema - the tables, columns, and foreign keys as they already exist - and execute in place on ClickHouse, Databricks, or in-process on lakehouse files, with no import and no separate cluster. Because the output is ordinary SQL, an underperforming query is an open optimization surface: it can be rewritten, and the engine itself extended. We support the argument with a peer system’s own published benchmark, in which a columnar engine outruns Neo4j by two-to-four orders of magnitude, and with reproducible measurements across the LDBC Social Network Benchmark suite.
[AI-5] When Guardrails Look Effective: Construct Validity Failures in LLM Agent Commerce Evaluation NEURIPS2026
链接: https://arxiv.org/abs/2609.01519
作者: Peiying Zhu,Sidi Chang
类目: Artificial Intelligence (cs.AI)
备注: Submitted to the NeurIPS 2026 Trust-AI-Eval Workshop. 7 pages, 2 figures, 3 tables. The accompanying artifact is available at this https URL
Abstract:Interactive simulations increasingly evaluate policies in markets populated by language-model agents. Their outputs can look economic—prices, profits, consumer surplus, and welfare—without instantiating the behavior named in the claim. We audit this risk in a multi-turn buyer–seller testbed for configurable hotel transactions. An initial implementation reported welfare gains from two marketplace guardrails of +87.4, +35.0, and +28.8 across a Qwen2.5 1.5B–14B ladder. It also gave guarded and unguarded agents different offer schemas and choice procedures. Holding the schema and buyer chooser fixed changes the paired contrasts to +7.2, -13.9, and +23.8. The four largest 14B single-generation effects averaged +229; after three generations per profile-condition, they averaged +37.6 (95% bootstrap interval [-34.2, 109.3]), while generation residuals account for 49.9% of variation in this post-hoc probe. A seller-incentive check is non-monotone: increasing profit pressure produces less profit than the default seller prompt. Scripted positive controls show why this matters. A profit-maximizing seller already attains first-best welfare, so guardrails mostly redistribute and reduce welfare; they create welfare only when the seller is explicitly programmed to force inefficient bundles. We contribute a construct-validity contract separating incentive validity, protocol isolation, stochastic stability, and welfare accounting, and returning INVALID or INCONCLUSIVE before substantive policy claims. In our case, the original estimate is INVALID under protocol isolation, while the controlled study remains INCONCLUSIVE under incentive validity and stochastic stability. The case does not show that guardrails are ineffective; it shows their apparent value is unidentified until the simulated agents and protocol pass these checks.
[AI-6] LatentPress: Context Compression Beyond Text and Vision
链接: https://arxiv.org/abs/2609.01507
作者: Zhengze Zhou,Hejian Sang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Compressed context is usually carried as human-readable text or as rendered images that must be decoded, even when its consumer is a language model. We introduce LatentPress, which writes conversational histories and long documents into a third representation: continuous memory tokens that a frozen decoder reads directly through its input-embedding interface, with no text reconstruction at inference. A small reader-matched writer compresses 4 - 16\times while training only an adapter (4.2M-26.2M parameters, \sim!0.1% of the decoder). On LongMemEval, LatentPress reaches 0.504 accuracy at 7.70\times compression versus 0.490 for uncompressed evidence, outperforming text summaries (0.184) and OCR-based compression (0.426 to 0.312). On LongBench-QA, in-domain writers match or exceed raw-context reading at 4 - 8\times compression, while 16\times trails raw. Writing takes 43ms per conversation, roughly an order of magnitude faster than text summarization or OCR reconstruction, and reading is 5 - 9\times faster than raw context or cached OCR. We validate the interface under two transfer settings, zero-shot from UltraChat to LongMemEval memory QA and from LongMemEval-derived QA to unseen LongBench document domains, establishing direct soft tokens as a practical machine-facing context interface beyond text and vision. The implementation of the experiments could be found at: this https URL .
[AI-7] Optimizing Byzantine Node Placement in Decentralized Federated Learning
链接: https://arxiv.org/abs/2609.01495
作者: Edoardo Gabrielli,Gabriele Tolomei
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Security evaluations of decentralized federated learning (DFL) typically focus on how Byzantine participants behave, while largely overlooking which participants are compromised. Yet, because aggregation is distributed over a communication graph, the placement of Byzantine nodes determines how malicious influence propagates through the network. We therefore treat Byzantine placement as an explicit adversarial decision and formulate the attacker’s objective as selecting, under a fixed compromise budget, the set of participants that maximizes its finite-time impact on honest nodes. To approximate this objective without executing the learning process for every candidate placement, we introduce Byzantine Placement Influence (BPI), a set-level measure derived from the actual gossip dynamics that quantifies the cumulative exposure of honest nodes to Byzantine sources over the training horizon. Unlike placement criteria based on node centrality heuristics, BPI directly accounts for weighted multi-hop propagation and interactions among compromised nodes. We develop efficient algorithms for optimizing BPI and evaluate them across six heterogeneous graph families, untargeted model poisoning, and backdoor attacks. BPI-guided placements consistently identify highly damaging configurations across different network structures and remain effective when the linear gossip assumption is relaxed through Byzantine-robust aggregation. Our results show that Byzantine placement is a critical but under-modeled dimension of DFL threat models and robustness evaluations.
[AI-8] Rethinking Learnability in Offline Data-driven Optimization
链接: https://arxiv.org/abs/2609.01493
作者: Chao Qian,Chen-Guang Wang,Rong-Xi Tan,Ke Xue
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Neural and Evolutionary Computing (cs.NE)
备注:
Abstract:Black-Box Optimization (BBO) has found broad applications, but evolutionary algorithms and Bayesian optimization face efficiency challenges as real-world BBO problems grow increasingly complex. Data-driven optimization improves the efficiency of BBO algorithms by learning from data. Offline data-driven optimization seeks high-quality solutions using only a fixed set of previous evaluations, attracting substantial attention because it requires no additional online evaluations. Many offline optimization methods have been proposed, but a fundamental question remains unanswered: what learnability is sufficient for offline optimization? Prior theoretical studies show that Probably Approximately Correct (PAC) learnability is insufficient, as the optimal region may remain poorly learned even when most regions are well learned. In this paper, we propose algorithm-dependent learnability, which requires accuracy only on the optimizer’s trajectory. We prove that its value-query form is sufficient for representative discrete settings, including greedy and local search for submodular maximization, while its first-order analogue is sufficient for projected gradient descent on convex minimization. Motivated by this notion, we formalize a trajectory-learning framework comprising trajectory construction, trajectory modeling, and candidate generation, and analyze existing trajectory-based methods under it. We further propose Uncertainty-aware Gradient-guided Trajectory Learning (UGTL), which constructs locally coherent improvement trajectories reflecting plausible search paths, models them with conditional diffusion, and selects a diverse candidate set. On five Design-Bench tasks, UGTL achieves the best aggregate mean rank, 3.1/25 , among 25 methods. Controlled trajectory analyses and cross-architecture replacements confirm that our trajectory construction plays a significant role in the improvement.
[AI-9] Defense-as-Skill: Evolving Runtime Guard Skill for Skill-Augmented Agents
链接: https://arxiv.org/abs/2609.01487
作者: Xiaofang Yang,Ziqi Miao,Dianbo Sui,Jing Shao,Lijun Li
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:
Abstract:Skill-augmented agents load reusable skills as persistent runtime context, improving task performance but also giving malicious skills a durable channel for steering future actions. Such skills may leak secrets, corrupt code, bypass approvals, or stage data for exfiltration only after a concrete user task and workspace state make the unsafe action appear useful. This makes pre-install vetting insufficient and calls for runtime, task-conditioned protection. We propose Defense-as-Skill, a defense paradigm that implements the runtime guard itself as an installable, inspectable, and editable skill. Our guard, SkillSonar, runs alongside untrusted task skills and checks sensitive actions against the user’s task boundary, routing each action to an allow, replan, or confirmation decision without modifying the underlying agent runtime. To study this setting, we construct SCOPE-R, a task-conditioned dataset covering 6 risk families and 21 sub-categories, with 206 attack-confirmed malicious instances and 43 benign tasks. We then improve SkillSonar on the SCOPE-R training subset using runtime guard-skill evolution, a Monte-Carlo Tree Search procedure that evolves the on-disk guard skill from feedback on the rollouts. Across Claude Code and OpenClaw, the evolved guard substantially reduces attack success while maintaining a favorable safety-utility trade-off. On repeated GLM-5 runs, SkillSonar reduces ID ASR from 0.482 to 0.104 and OOD ASR from 0.606 to 0.115. Further analyses demonstrate transfer across victim models, held-out risk families, and external benchmarks, as well as retained protection against adaptive attackers. Ablations further show that explicit safety responsibility assignment and the skill-native representation are both important to the observed gains.
[AI-10] Harness-of-Harness: Multi-Day Autonomous Software Development with Continual Improvement
链接: https://arxiv.org/abs/2609.01481
作者: Haoyang Yan,Min-le Su,Hangfan Zhang,Zhanhao Li,Chen Zhang,Shao Zhang,Yang Chen,Lei Bai,Shuyue Hu
类目: Artificial Intelligence (cs.AI)
备注: Github: this https URL Project Page: this https URL
Abstract:This paper studies autonomous software development, in which LLM-based coding agents transform high-level requirements into complete, functional, and usable software systems without human intervention. We introduce Harness-of-Harness (HoH), a framework that enables coding agents to continually improve software during autonomous development. HoH operates on existing coding-agent harnesses, and organizes their executions into iterative planning-coding-testing loops. To sustain improvement across loops, HoH balances repair with capability growth, scopes development into small and verifiable increments, separates implementation-time testing from independent evaluation, and constrains verifiable outputs rather than prescribing agent workflows. It progressively exposes deliverables, role-specific tools, and skills, encourages reuse rather than recreation, and maintains versioned project histories. On GameCraft-Bench, FrontierSWE, and ProgramBench, three harness-model pairs (Codex with GPT-5.5, OpenCode with DeepSeek-V4-Pro, and Pi with MiniMax-M3), HoH consistently outperforms the corresponding standalone harnesses, achieving an average relative gain of 52.25 percent and a maximum gain of 82.86 percent after three iterations. In a multi-day deployment with more than 70 iterations, HoH autonomously develops a first-person-shooter game, featuring a coherent storyline, fully implemented core mechanics, human-playable experience, polished visuals and integrated audio. Github: this https URL Project Page: this https URL
[AI-11] Parsing the Stream: A Live Trace Model for Long-Horizon Agents and Their Observers
链接: https://arxiv.org/abs/2609.01466
作者: Egor Pakhomov,Erik Nijkamp
类目: Artificial Intelligence (cs.AI)
备注: 22 pages, 4 figures. Code: this https URL ; dataset: this https URL
Abstract:A long-horizon agent’s trace outgrows both of its consumers: the human observer monitoring the run, and the agent itself, whose bounded context the trace must be folded back into. We present a live trace model, an append-only event ledger folded incrementally into typed run state and compiled into per-consumer views, and evaluate it for both consumers against deterministic ground truth. For the observer side, evaluated with an LLM reader as proxy, the compiled view answers monitoring questions using approximately 14x and 15x fewer input tokens (by reader) and at 5-7x lower cost than a budget-capped single-call reading of the raw trace, with higher accuracy (0.85-0.87 versus 0.48). Because the questions were co-designed with the view schema, we treat the token and cost reduction, conditional on schema coverage, as the transferable result. For the agent, on 120-link sequential-dependency tasks, mechanisms that maintain the task’s running statistic in per-step state succeed where full-context prompting fails (30/30 versus 8/30 under a clean protocol, n=30, labeled descriptive owing to benchmark-system co-development); a prompt-level scratchpad matches the fold’s accuracy at lower cost, and a two-arm decomposition attributes the fold’s accuracy to its deterministic aggregate and its cost advantage to its compactness. The fold’s remaining value over cheaper alternatives is deterministic auditability and serving the observer from the same state. We derive eleven candidate requirements for trace folding from observed failures and delimit them with an order-sensitive task family on which the fold ceases to help. Code, benchmarks, a regenerable synthetic corpus, and all workbench traces are released.
[AI-12] When Safety Routing Breaks: Understanding Alignment Frag ility under Benign Fine-Tuning EMNLP2026
链接: https://arxiv.org/abs/2609.01455
作者: Yitong Guo,Xiaoyi Chen,Siyuan Zhang,Xiaofeng Wang,Haixu Tang
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注: Accepted to the Findings of EMNLP 2026
Abstract:Benign fine-tuning severely weakens the safety alignment of large language models (LLMs), so we study why refusal behavior is so fragile. While prior work often attributes this failure to gradient conflict, we propose a fundamentally different Fisher-geometric explanation: safety Fisher is low-rank, and alignment makes the safety geometry flatter while preserving an output-routing pathway. After 100 benign fine-tuning examples, this pathway is selectively re-sharpened in output-side MLP modules, explaining the asymmetric fragility: safety can collapse to high attack success rates, while general utility degrades mildly. The routing view also explains why few safety examples can restore refusal behavior, indicating that internal safety-relevant representations are preserved. Finally, we show that LoRA and ASAM mitigate early collapse by suppressing output-side sharpness, but their protection weakens at larger fine-tuning scales. Overall, safety failure is best understood as a disruption of a low-rank output-routing mechanism
[AI-13] Efficiently Estimating Optimal Hyperparameter Scaling Laws through Power-Law Entropy Search
链接: https://arxiv.org/abs/2609.01431
作者: Zhiliang Chen,Sebastian Ament,David Eriksson,Maximilian Balandat,Eytan Bakshy,Jihao Andreas Lin
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Optimal hyperparameter scaling laws describe how the best hyperparameters for large language model (LLM) training change with model and data scale, enabling practitioners to predict optimal configurations at production scales without expensive large-scale tuning. However, estimating these scaling laws conventionally requires exhaustive grid searches over thousands of training runs, consuming enormous computational resources. We introduce Power-Law Entropy Search (PLES), a computational cost-aware acquisition function built on multi-fidelity Bayesian optimization that efficiently estimates optimal hyperparameter scaling laws through adaptive experimentation. A key innovation in PLES is that it searches for candidates that reduce the overall uncertainty of a scaling law estimate, instead of optimizing a single objective function. At each iteration, PLES selects the candidate configuration that maximally reduces the uncertainty of the scaling law estimates per unit computational cost, naturally favoring informative small-scale experiments. We evaluate PLES on synthetic benchmarks, surrogate models fitted to real LLM training data, and actual LLM pre-training runs. Across all settings, PLES converges to accurate optimal hyperparameter scaling laws using less than one-tenth of the computational budget required by conventional grid search and other baselines.
[AI-14] Learning Sparse Decision Trees via Transformer Variational Auto-Encoders ICDM2026
链接: https://arxiv.org/abs/2609.01430
作者: Giacomo Fidone,Alessio Cascione,Riccardo Guidotti
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Accepted for publication at the 2026 IEEE International Conference on Data Mining (ICDM 2026)
Abstract:Decision trees are among the most widely used models in machine learning, largely due to their transparent decision logic, making them well-suited for high-stakes decision-making contexts. However, most existing learning algorithms focus on predictive performance, overlooking the joint optimization of other desirable properties, such as structural sparsity. In this work we propose TREVIS, an approach for learning decision trees with respect to complex objectives, based on the exploration of the latent space of a Tree Transformer Variational Auto-Encoder (TTVAE). By mapping decision trees onto latent representations, TREVIS replaces the discrete search space with a continuous one, enabling gradient-based optimization via a differentiable surrogate model. We experiment with TREVIS for learning decision trees that jointly optimize predictive performance and sparsity. Results show that TREVIS discovers decision trees matching the predictive performance of existing near-optimal algorithms while improving their structural sparsity.
[AI-15] Provably Safe Sim-to-Real Transfer
链接: https://arxiv.org/abs/2609.01418
作者: Tingting Ni,Maryam Kamgarpour
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:To mitigate the sample complexity of real-world reinforcement learning (RL), a common practice is to first train a policy in a simulator, where samples are cheap, and then deploy the learned policy in the real world with the hope that it generalizes effectively. Such direct sim-to-real transfer is not guaranteed to succeed: simulator-trained policies can be suboptimal in the real world due to sim-to-real mismatch. Correcting this mismatch requires collecting data from the real system, but in many applications, such as robotics and healthcare, this data-collection process is itself subject to safety constraints. This gives rise to the problem of safe sim-to-real transfer: how can an agent exploit an imperfect simulator while ensuring safe real-world data collection and learning a near-optimal feasible policy for the target system? We address this problem by formulating safe sim-to-real transfer within the framework of reward-free safe RL. We design a computationally efficient algorithm that exploits simulator information to provably reduce real-world interaction while ensuring safe exploration and enabling the computation of a near-optimal feasible policy for any potential reward function. Our real-world sample complexity bound characterizes the benefit of using the simulator in terms of the sim-to-real mismatch.
[AI-16] Evaluating Multimodal LLM s as Generalist Vision-Language-Action Agents for Drone Control: Commanding Approaching Tracking and Searching
链接: https://arxiv.org/abs/2609.01404
作者: Jaewoo Park,Minyoung Lee,Sukmin Seo,Moonbin Yim,Hyunwook Yoon,Dohoon Ryu,Daehee Kim,Myungseo Song,Jihyuk Byun,Seunggyu Chang,Taeho Kil,Jiseob Kim,Bado Lee,Geewook Kim
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注: Preprint
Abstract:Multimodal Large Language Models (MLLMs) are strong perceivers of images and video. We ask how far that reach extends into acting: dropping an MLLM directly into a drone’s control loop, with its entire action space declared solely in the prompt. Recent systems approach this setting but increasingly narrow the model’s decision-making. We widen it back. We introduce DroneCATS-Agent, an architecture where the MLLM is a swappable component, and DroneCATS, a benchmark treating the model as the independent variable. Beyond merely flying toward a pixel, our agent entrusts the model to yaw and search, deliberate when unsure, and self-declare arrival—all without fine-tuning or function-calling schemas. Evaluating frontier and open models across four core capabilities—approaching a visible target, tracking a moving one, searching outside the initial view, and commanding a multi-drone fleet—reveals that even the simplest embodied settings are far from solved. Crucially, to identify what breaks first at the edge, our roster scales down to 2B parameters. The findings expose a stark paradox: it is not the flying that fails. Small open models often navigate into the success radius more reliably than frontier models, yet lose the episode by declaring arrival prematurely or not at all. Multi-drone commanding amplifies this divide, with small models failing by blindly copying a single coordinate across distinct views. Viewed as vision-language-action agents, the models’ spatial perception holds up, but their action protocol does not. What separates a deployable edge model from a frontier model is not navigation, but the discipline to sustain a declared protocol and emit the correct terminating action. The open problem is closing this gap at onboard compute costs—yielding a fast model that plans persistently and knows exactly when it is done—and DroneCATS is built to measure that distance.
[AI-17] EDGE: Error Dependency Graph-Guided Multi-Error Attribution in Multi-Agent LLM Systems
链接: https://arxiv.org/abs/2609.01360
作者: Jun Hou,Priya Pitre,Yi Fang,Xuan Wang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large language model (LLM) agent failures often contain multiple related errors rather than a single mistake. Existing attribution methods usually identify a responsible agent, step, or root cause, but do not explicitly model dependency between errors. We introduce EDGE, an Error Dependency Graph-guided multi-Error attribution framework. EDGE constructs an error dependency graph from observed error events and validates a reliable causal subset through counterfactual rollout. The inference graph guides a two-stage LLM-as-judge detector for error attribution, and the intervention-validated subgraph provides a more reliable basis for explanation and repair analysis. Experiments on TRAIL and MAST show that EDGE improves category-level multi-error attribution across most evaluated models and settings. Experiments with adapted WhoWhen-style prompts show that the graph helps across prompting strategies. These results suggest that dependency structure is a useful diagnostic prior for agent failures beyond isolated root-cause prediction.
[AI-18] SymFold: Synergizing Evolutionary and Structural Priors for Accurate Protein Inverse Folding
链接: https://arxiv.org/abs/2609.01353
作者: Handong Wang,Jiaxin Qi,Baisheng Lai,Jianqiang Huang
类目: Artificial Intelligence (cs.AI)
备注: 13 pages, 5 figures
Abstract:Protein inverse folding aims to recover amino acid sequences for a given 3D protein structure, underpinning broad applications such as enzyme engineering and drug this http URL methods often follow a serial pipeline, in which a structure encoder predicts a coarse sequence, which is then refined by protein language models (PLMs). However, because PLMs only perform post-hoc sequence edits, the refinement is bounded by the quality of upstream this http URL to recent multimodal protein language models (MPLMs), we could directly encode structure to generate sequences with pretrained structural knowledge, but we observe that they are not effective for inverse folding. Therefore, we introduce a symmetric dual-path architecture that both leverages PLMs for pretrained sequence evolution knowledge and MPLMs for pretrained structural knowledge to iteratively guide protein sequence this http URL extensive experiments across standard protein inverse folding benchmarks, our method achieves state-of-the-art performance, surpassing prior approaches, and ablation studies validate the rationale of our symmetric design, revealing a promising direction for the community.
[AI-19] Scalable Rao-Blackwellized Online Planning for High-Dimensional POMDPs
链接: https://arxiv.org/abs/2609.01351
作者: Jiho Lee,Nisar Ahmed,Kyle Hollins Wray,Zachary Sunberg
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注:
Abstract:Online planning under uncertainty remains a fundamental challenge for robotic systems operating in partially observable environments with high-dimensional state spaces. While sampling-based POMDP solvers enable approximate decision-making in large or continuous domains, their performance degrades as belief dimensionality increases due to the high variance inherent in Monte Carlo-based estimation. In this work, we extend the Rao-Blackwellized online POMDP (RB-POMDP) framework to improve its generalizability in high-dimensional settings through hybrid continuous-discrete belief representations. By analytically propagating uncertainty associated with marginalized state components during tree-based planning, the proposed approach reduces sampling-induced variance in value estimation. We demonstrate the effectiveness of this framework in a robotic search-and-rescue task by integrating it with FastSLAM 2.0. Experimental results show that the proposed planner achieves higher cumulative rewards using significantly fewer particles and planning simulations than purely sampling-based methods under equivalent computational budgets. These results suggest that structured high-dimensional robotic problems admitting tractable sufficient statistics can be effectively leveraged within the RB-POMDP framework for computationally feasible online decision-making.
[AI-20] Cheap Verifiers Large Blind Spots: Measuring the Reliability Cost of Cost-Saving Cascades
链接: https://arxiv.org/abs/2609.01345
作者: Dushyant Rajput
类目: Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR); Machine Learning (cs.LG)
备注:
Abstract:Inference cascades cut cost by answering most queries with a cheap model and escalating a hard tail to a frontier model that acts as verifier. A natural extension closes the loop: fine-tune the cheap student on the verifier’s rejections so the escalation rate, and cost, fall each round. We measure this loop on real LLMs and report four findings. First, the verifier’s blind spot, the fraction of the student’s wrong answers it accepts, is large and moves adversarially: it grows with student capability ( \beta from 0.12 to 0.55 as the student scales 0.5B to 32B) and shrinks with verifier capability, so it is worst in the cheap-student, cheap-verifier regime cascades exist to create. Second, buying it away returns the saving: a frontier verifier drives \beta to about 0.05 but then escalates on 46% of hard-MATH queries against a 39% true error rate, paying the frontier price on nearly half of all traffic. Third, naive corrective fine-tuning on the verifier-rejected tail does not improve the small student but degrades and ultimately collapses it, across every teacher we tried (cross-family and same-family), so at this scale the self-improving loop is self-defeating. Fourth, through all of this the cascade’s own dashboard, every metric computed through the verifier, reads a flat 3% error while true delivered error swings up to 32%: the system is blind to its own degradation by construction. We then give the theory that explains the blindness, a two-population conservation law, \epsilon_\infty \lesssim q_0 \beta_0 , under which every in-loop metric improves while true quality does not, and a synthetic study that validates the mechanism. The practical conclusion: the reliability of a self-improving cascade cannot be read from any metric computed through its own verifier.
[AI-21] LEAP: Likelihood Elicitation and Aggregation for LLM -based Probabilistic Forecasting EMNLP2026
链接: https://arxiv.org/abs/2609.01337
作者: Yufei Chen,Yiran Zhao,Xiaogang Xu,Qipeng Xie,Jiafei Wu,Zhe Liu
类目: Artificial Intelligence (cs.AI)
备注: Accepted to the 2026 Conference on Empirical Methods in Natural Language Processing (EMNLP 2026). 20 pages, 3 figures, and 15 tables. Code: this https URL
Abstract:LLM-based forecasting systems have improved on real-world tasks such as financial markets and sports outcomes, largely through stronger search and tool use. Many systems still ask an LLM to read all collected evidence together and produce the final forecast. We call this design Monolithic Prediction. It can obscure how individual evidence items affect the result and collapse uncertainty across competing outcomes. We propose LEAP (Likelihood Elicitation and Aggregation for Probabilistic forecasting), which reorganizes how collected evidence is used in the prediction stage. LEAP examines each evidence item separately and elicits likelihood parameters that describe its implications for the target. An explicit prior and a deterministic probabilistic model then combine these likelihoods into a posterior distribution. This procedure supports continuous, single-choice, and multi-choice forecasts while preserving reproducible evidence contributions. We build a benchmark covering forecasting, information-seeking, and browsing tasks, and evaluate LEAP on our own agent loop and several agent CLI frameworks. Given the same evidence, LEAP improves most prediction and calibration metrics across models and remains stronger under controlled comparisons of prior access, inference budget, and aggregation.
[AI-22] Bandits in Prod: Hyperparameter Optimization at Inference Time
链接: https://arxiv.org/abs/2609.01335
作者: Louis Abraham,Tuan-Anh Nguyen,Nicolas Devatine
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 32 pages, 14 figures
Abstract:Many production systems can assess a configuration only by using it on live requests and observing noisy feedback. Modern agentic systems are a prominent example, with inference-time choices such as model selection, retrieval depth, prompting strategy, and decoding temperature, yet often with no representative validation data. We formalize this setting as Online Hyperparameter Optimization (OHPO) and cast it as an infinitely many-armed bandit over mixed and conditional search spaces. We introduce IMABO, a general framework that combines any bandit policy for choosing among already sampled configurations with any oracle for proposing new ones. We instantiate it with IMOSS, a restart-free anytime policy whose active set grows as t^\beta , and prove an expected cumulative quantile-regret bound of O(p_\rho^-1/\beta + T^(1+\beta)/2) , where \beta\in(0,1) controls active-set growth and p_\rho lower-bounds the probability that a proposed configuration falls in the top- \rho fraction of the search space. We combine IMOSS with three practical oracles: a Tree-structured Parzen Estimator, an incumbent-mutation oracle driven by a per-coordinate bandit, and a pretrained tabular foundation model, all three improving over the uniform random oracle baseline. IMABO obtains the lowest cumulative regret across diverse OHPO settings, from tuning classical machine-learning models to configuring LLM-based agents.
[AI-23] Automated Event Log Generation from Unstructured Text Using Finetuned LLM s
链接: https://arxiv.org/abs/2609.01320
作者: Maximilian Seeth,Gabriel Marques Tavares,Daniel Schuster
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Process mining (PM) provides a powerful framework for discovering and optimizing operational processes from event data. However, the efficacy of PM techniques is strictly predicated on the availability of structured event logs. Thus far, event logs have often been laboriously created by domain and process mining experts. This costly effort causes large portions of organizational knowledge, including incident tickets, manuals, and textual reports, to remain underutilized. We address this bottleneck by investigating the efficacy of Large Language Models (LLMs) as automated data translators. We propose a scalable framework that leverages LLMs as data translators to bridge the gap between unstructured textual resources and structured event data. We finetune LLMs on a newly created text-to-log dataset, demonstrating that the resulting models can extract high-fidelity event logs from unstructured resources. Our results show that this finetuning approach outperforms few-shot or zero-shot prompting by a large amount, highlighting finetuning as a necessary pre-condition for generating reliable event data. We conclude that our method provides a promising pipeline for making previously unused data available to process mining ecosystems, effectively expanding the possibilities of using PM to further investigate organizational workflows.
[AI-24] A Composable Evaluation System for Reproducible Omni-Modal Foundation Model Evaluation
链接: https://arxiv.org/abs/2609.01315
作者: Hodong Lee,Sanghee Park,Dohoon Ryu,Jungwhan Kim,Junyeob Kim,Soyoon Kim,Geewook Kim
类目: Artificial Intelligence (cs.AI)
备注: 12 pages, 3 figures. Code: this https URL
Abstract:Building an omni-modal foundation model means evaluating it across text, image, video, and audio. Excellent evaluation toolkits exist for each modality, but their inference engines, prompt conventions, and metric implementations are mutually incompatible, so practitioners end up maintaining separate environments for every toolchain and still struggle to compare results across them. OmniEvaluator grew out of this need in our own model development: rather than reimplementing benchmarks, it connects existing inference engines and curated evaluation libraries at a higher level, exposing four inference backends, four evaluation frameworks, and over a thousand benchmarks through a single interface. Every run is recorded as an artifact capturing the full configuration for exact reproduction, and results flow into a shared dashboard for cross-model comparison. A federated mode shares GPU inference servers across concurrent evaluations, and a built-in verifier, small enough to run on CPU, keeps its score stable across engines and prompts where rule-based scoring fluctuates under configuration mismatch, matching cost-efficient commercial LLM judges without their recurring API cost. The system, demo video, and dashboard are publicly available. (this https URL)
[AI-25] Analog-DB: An Agent -First Analog Integrated Circuit Database From Blocks to Systems
链接: https://arxiv.org/abs/2609.01286
作者: Danial Noori Zadeh,Mohamed B. Elamien
类目: Artificial Intelligence (cs.AI); Hardware Architecture (cs.AR); Signal Processing (eess.SP)
备注: 23 pages, 5 figures, and 13 tables
Abstract:Sharing analog integrated circuit designs remains difficult: foundry non-disclosure agreements restrict the process details a design depends on, and the testbenches behind published results are rarely released. We present analog-db, an open-source, versioned database built on a shareable design representation. A domain-specific language captures each design as a process-neutral topology, reusable testbenches, and a machine-readable datasheet under one schema, so a design is shared in full and re-simulates on the process kits it is bound to. A parameterization scheme exposes functional sub-blocks and device sizes as named parameters that carry their matching constraints, making circuits composable and retargetable; a schema-governed contract and queryable catalog let AI design agents discover and reuse them directly. Across the regulator corpus, all 23 circuit-kit bindings on three open kits meet their own recorded specification bands (typical corner, matched devices, no layout) and 10 of 23 meet a common class band. Seventeen of the 23 imported sizings failed their testbenches and closed under a gm/ID sizing loop driven by the annotated sub-block roles, typically within one to three iterations. In a supervised case study, a coding agent working from the released artifacts sized the op-amp cores of a chopper instrumentation amplifier on an open 130nm kit, locating four hand-entry defects and a missing common-mode feedback loop that the sizing-only baseline did not repair. The database holds 68 circuits across sixteen classes, verifiable at schematic level under a tiered harness and tracked on a power/performance scoreboard, released at this https URL.
[AI-26] EmbodiedSkills: A Unified Framework for Orchestrating Training and Deploying VLA Agents
链接: https://arxiv.org/abs/2609.01281
作者: Wei Wang,Wenqiao Zhang,Yutong Lin,Yuqian Yuan,Tianwei Lin,Jinhao Mao,Zhenxuan Fan,Mingjian Gao,Yang Dai,Wentong Li,Zheqi Lv,Zheng Dong,Yingjie Niu,Jiaqi Zhu,Jun Xiao,Chao Li,Yueting Zhuang
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注: 20 pages, 4 figures, 5 tables
Abstract:Vision-language-action (VLA) models map visual observations and language instructions directly to robot actions, but long-horizon tasks require more than action prediction. An agent must coordinate perception, planning, execution, progress verification, and recovery as the physical state evolves. An action prediction or a model-generated skill decision does not, by itself, guarantee that the proposed operation is valid in the current state or that its outcome will be verified. We propose EmbodiedSkills, a unified framework that treats each skill decision as an execution proposal: the runtime checks its prerequisites before execution and verifies the outcome afterward. A shared executable-skill interface connects high-level skill selection, bounded low-level VLA execution, and post-action verification within a single agent loop. Because this interface remains fixed, low-level VLA policies can be replaced or adapted without changing the agent loop. The interface also records planning, execution, verification, and recovery events as structured trajectories, which provide supervision for individual components and can support optional online adaptation when interactive feedback is available. We instantiate EmbodiedSkills with Qwen3-VL and OpenPI/pi0.5 on RoboTwin 2.0 and LIBERO. Task-adapted low-level VLA policies achieve an average success rate of 86.20% across 50 RoboTwin 2.0 tasks and 97.40% across the four LIBERO suites. These results establish the execution performance of the task-adapted low-level VLA policies used in EmbodiedSkills. On four memory-dependent RMBench tasks, the same task-adapted execution approach achieves 12.5% average success. The framework provides a trainable and inspectable agent layer for turning these policies into closed-loop embodied systems.
[AI-27] he Constitutional Coverag e Trilemma in AI Governance
链接: https://arxiv.org/abs/2609.01275
作者: Natalija Mitic,Soona Sedahmed A. O.,Mamadou Selly Ly,Moustapha Cisse
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 29 pages
Abstract:Frontier AI systems function as \emphconstitutional institutions: each deployed model encodes an implicit ranking among safety, helpfulness, honesty, autonomy, and equity. We ask whether the supply of frontier constitutional types covers human demand. Combining a paraphrase-controlled audit of the as-shipped default constitutions of 23 frontier LLM archetypes with a pairwise-tradeoff study of 1,649 US participants on the same instrument, we report three facts. \emphDemand is broad: it spans all five values, with the largest constituency under one-third. \emphSupply is narrow and drifting: the 23 -archetype hull occupies \sim2% of the demand hull under conservative noise-matched estimation ( 0.10% at full audit precision), no archetype puts helpfulness or autonomy first ( 37% of users are constitutionally homeless), and across six model families autonomy decreases in 5/6 , equity increases in 5/6 , and safety increases in 4/6 , with monotone within-family version trends (order-permutation p = 0.013 ) and the autonomy decline concentrated in scenarios where safety is not at stake. The drift’s importance is directional: \emphaway from a value already undercovered, mechanically worsening the welfare floor for the least-served users. \emphThe fix is sparse: a 2 -vertex menu \e_\mathrmHON, e_\mathrmAUT\ beats the full 23 -archetype frontier by 47% on mean regret (CI [43%, 52%] ); three vertex additions cut mean/worst-group regret by up to 81% / 64% . We formalize these findings as a budgeted-pluralism trilemma, show the binding regime is empirically realized, and verify the conclusions are robust to distance-based welfare and to degraded routing. The instrument and audit harness are described in full in the appendices.
[AI-28] Making Prospective Memory SLM-Shaped: Typed Intention Stores for Small-Model Agents
链接: https://arxiv.org/abs/2609.01272
作者: Jinqing Zhao,Chengcan Wu
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Prospective memory means carrying out a deferred intention at the right future cue while other work continues. Benchmarks now isolate it as an agent skill, yet frontier LLMs still struggle: the best published PM-Bench scaffold reaches only 65.1% Set-F1. We argue that this loop is schema-constrained state tracking rather than open-ended reasoning, and that small models can execute it when the action space is typed. We propose the Prospective Intention Store (PIS) that puts lifecycle logic in code and scoped language work on the model. The scaffold is agentic and training-free: no selector fine-tuning and no trajectory distillation. On PM-Bench, DeepSeek-Chat with PIS reaches 82.9% Set-F1. On Gemma-E2B, Set-F1 is only 4.2% without a store and at most 6.6% under seven retrospective memories, while PIS reaches 66.2%. PIS further reaches 70.1% Set-F1, where retrospective memory methods stay at most 54.4%. PIS sets a new state of the art on this benchmark and enables small models to surpass the published large-model scaffold.
[AI-29] Dual Process Motion Planning
链接: https://arxiv.org/abs/2609.01260
作者: Jiayi Yan,Francesco Fabiano,Alessandro Abate
类目: Artificial Intelligence (cs.AI); Robotics (cs.RO)
备注:
Abstract:Robotic systems are deeply embedded in both industry and everyday life, where they are expected to act with speed, precision, and reliability. Classical control and planning methods have long delivered strong guarantees, but often at the cost of computational efficiency and adaptability. More recently, learning-based approaches have shown promise in overcoming these limitations, enabling agents to leverage experience to accelerate decision-making and address previously intractable problems. In this work, we bridge these two approaches through a neuro-symbolic perspective on nonlinear motion planning. Inspired by the Thinking Fast and Slow paradigm, we introduce a dual-process architecture that combines the strengths of robust reasoning and learning. Our framework integrates state-of-the-art symbolic solvers as a System-2'' component with experience-driven System-1’’ modules. A metacognitive controller dynamically orchestrates their interaction, selecting when to rely on fast intuition versus slower, more precise reasoning. By evaluating the framework across diverse nonlinear benchmark environments, we demonstrate that this architecture yields consistent gains in planning efficiency, accuracy, and generalization, while promoting reuse across tasks. The results suggest that tightly coupling learning with structured reasoning offers a scalable path toward more capable and adaptive robotic systems.
[AI-30] Measuring the Behavioral Fidelity of Long-Horizon Human Activity Simulations
链接: https://arxiv.org/abs/2609.01257
作者: Yi Fei Cheng,Fan Yang,Iremsu Bas,Koichiro Niinuma,Narishige Abe,David Lindlbauer
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:As LLM-based human simulators are increasingly used for policy, evaluation, and training, they must faithfully reproduce real behavioral patterns. While prior work has examined behavioral fidelity in survey responses and dialogue, longer-horizon real-world activity remains largely unexplored. We introduce a framework for evaluating behavioral fidelity in long-horizon activity simulations across temporal granularities and levels of analysis. As a case study, we collect a 43-hour multi-camera dataset of in-the-wild office activity and compare trace-derived conditioning mechanisms: persona descriptors, few-shot exemplars, and statistical transition and time-of-day priors. We find that behavioral fidelity is not uniform across metrics: statistical priors bring activity and sequence distributions closest to real behavior, yet over-fragment routines and suppress within-person variability. These findings motivate a more holistic evaluation that spans multiple metrics, temporal granularities, and levels of analysis.
[AI-31] Explore More Drift Less: Outcome-Only Reinforcement Learning Can Suffice for Long-Horizon Interactive Agents
链接: https://arxiv.org/abs/2609.01245
作者: Liming Pu,Xiaoxia Li,Yifu Liu,Teng Cao,Bin Yang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 13 pages, 6 figures
Abstract:Reinforcement learning is a natural way to post-train LLM agents for long-horizon interactive tasks judged only by end-of-task verification, yet a shared belief holds that outcome-only RL soon hits a ceiling on small open models. Recent work therefore compensates around the training with denser rewards, SFT priors, skill libraries, curated memory, or multi-agent orchestration. We argue the ceiling is an artifact of two failures of common practice. Signal starvation: group-relative RL with sparse outcome-only rewards yields a gradient only when a task’s rollout group mixes successes and failures, so under-scaled exploration silences exactly the hardest, most instructive tasks. Policy drift: squeezing many updates out of a small task pool degrades the policy itself, as an unanchored objective lets the sampling distribution collapse exactly when saturation has already made informative groups rare. We present CANOPY (Coverage-ANchored On-PolicY RL), a minimalist protocol attacking both directly: scale same-task exploration until the natural signal reappears, keep every update on-policy, KL-anchored, and confined to the agent’s own action tokens, then cash in an enlarged interaction budget at test time. On AppWorld, a long-horizon interactive coding benchmark, a Qwen3-14B policy trained with CANOPY through environment interaction alone–without task-specific supervision, auxiliary credit signals, or elaborate agent scaffolding–topped the public leaderboard (Feb. 2026; Test-Normal TGC 86.9, Test-Challenge 67.6), and the same design principles lift Qwen3.5-9B on SWE-bench Verified by 16.6 points. Agentic RL alone thus internalizes long-horizon capability directly into a small open model; we plan to release the complete training stack at this https URL.
[AI-32] MutMem-V2: Cryptographically Authorized Mutation in Persistent Agent Memory Portable Verification and Reproducible Evidence
链接: https://arxiv.org/abs/2609.01235
作者: Walid Saidi
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注: 11 pages, 4 tables. Companion to arXiv:2608.02843 . Code, installer, independent verifiers, evidence, and release artifacts: this https URL
Abstract:MutMem V1 introduced retention-preserving, cryptographically authorized mutation for persistent agent memory but did not provide a complete portable verification contract or clean-install reproduction path. MutMem V2 closes that publication gap without introducing a second memory engine. It specifies exact canonical bytes, domain-separated object and bundle commitments, mandatory recall-evidence membership and ordering, external trust anchors, identity epochs, revocation, authorization, request receipts, ordered disclosure, and three mutation terminal types. The released protocol contains 18 versioned object schemas, 39 recall vectors, 15 mutation vectors, and 37 closed recall failure reasons. Independent Node and Python implementations agree on verdict and primary reason for all 72 structural and cryptographic terminals; a production-conformance corpus agrees on 42/42 cases across 28 required classes. A clean Node v26.8.1 installation reaches first-boot, restart, and scheduler readiness with no experimental memories. A separately scoped 120-unit Canary experiment supports only explicit-marker traversal. Every public table regenerates from a self-hashed aggregate, and an independent verifier reconstructs the statistics and claim boundaries. Historical V1 empirical results remain historical. MutMem V2 supports claims about portable integrity, authorization, traceability, conformance, and reproducibility under stated assumptions; it does not establish semantic truth, universal robustness, or independent replication.
[AI-33] Position Matters: Feature Inversion Attacks in ViT Split Inference with Token Reduction and Shuffling
链接: https://arxiv.org/abs/2609.01232
作者: Stefano Leggio,Giulio Rossolini,Alessandro Biondi
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:
Abstract:Vision Transformers (ViTs) are increasingly used in split-inference systems, where edge devices transmit intermediate token representations to a remote cloud. In this setting, token reduction lowers computation and communication costs, while token shuffling disrupts the spatial organization of the transmitted tokens, potentially limiting information leakage. However, their privacy benefits remain unclear against feature inversion attacks, which attempt to reconstruct the input from the transmitted embeddings. In this work, we show that, despite disrupting the spatial structure required by conventional reconstruction attacks, transmitted token embeddings retain substantial positional information. Based on this observation, we introduce the Spatially Aligned Reconstruction Attack (SARA), a unified pipeline that predicts token positions, restores their spatial layout, reconstructs missing embeddings using a feature-space masked autoencoder, and recovers the input image. Our results demonstrate that token shuffling provides only apparent privacy, as SARA largely reconstructs the original token organization. Token reduction offers stronger protection, but significant leakage persists when the retained tokens preserve sufficient semantic and positional information. Finally, we introduce a lightweight edge-side defense that removes positional embeddings and progressively adapts the edge-side transformer blocks through knowledge distillation. It substantially reduces attack performance against SARA, while preserving downstream task accuracy and requiring no changes to the cloud-side model.
[AI-34] Prompt-Robust Language Models: Which Training Strategies Work? EMNLP2026
链接: https://arxiv.org/abs/2609.01217
作者: Frederic Sadrieh,Michal Štefánik
类目: Artificial Intelligence (cs.AI)
备注: 5 pages, 5 figures, 13 tables. Camera-ready version; Accepted to EMNLP 2026 Findings
Abstract:Despite their strong performance, large language models remain highly sensitive to prompt formulation. Prior work addresses this through refined data construction or through dedicated robustness objectives. We reproduce and compare these strategies under controlled conditions, and measure how effective they are in addressing models’ prompt sensitivity. We find the current robustness fine-tuning methods improve over standard fine-tuning and in-context learning, but the best-to-worst prompt gap remains as high as 40-57% of performance. Moreover, the recent robustness-enhancing methods we test - CoIN for contrastive alignment and PPCL for consistency regularization - often fail to outperform the simplest data construction strategy: training on one template per batch. Our diagnostics explain these results. The auxiliary objectives move the quantity they penalize, but do not generalize beyond it. Additionally, data construction strategies differ due to the conflicting signs of per-template gradients on 57-64% of parameters. Thus, batches that mix formulations force the optimizer to reconcile competing updates instead of finding a shared, prompt-agnostic one.
[AI-35] H2Table: Hierarchical Hypergraph-Enhanced Large Language Models for Complex Table Reasoning
链接: https://arxiv.org/abs/2609.01216
作者: Jia Ling,Yangfan Wang,Chen Tang,Haoming Tan,Yang Yang,Yi Guan,Jingchi Jiang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Tables are ubiquitous across diverse domains, yet reasoning over them remains a significant challenge for modern large language models (LLMs). Current approaches typically linearize tables into sequences, inherently overlooking their intrinsic two-dimensional and hierarchical structure. To address this, we propose H2Table (Hierarchical Hypergraph-Enhanced Table Reasoning), a novel framework that represents complex tables as hierarchical nested hypergraphs. To process this representation, we design a tailored hypergraph encoder to facilitate message passing between hyperedges (headers) and nodes (cells), thereby perceiving the semantic entailment relationships between them within complex tables. Furthermore, we introduce a set of learnable query vectors acting as a lightweight bridge to extract representative structural embeddings from the encoder into the LLM. Experimental results demonstrate that our approach effectively handles complex table question answering tasks with hierarchical nested headers. Notably, on the HiTab dataset, H2Table achieves an average improvement of 22.88% over state-of-the-art baselines on highly complex tables with a nesting depth of four. Our code is available at: this https URL.
[AI-36] REFACTOR-VLA: Unsupervised Library Learning of Typed Motor Programs
链接: https://arxiv.org/abs/2609.01215
作者: Riyaaz Shaik,Chandru Venkataraman
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Robotics (cs.RO)
备注: 30 pages, 5 figures
Abstract:Most vision-language-action (VLA) models – OpenVLA, \pi_0 , RT-2, RDT-1B – are monolithic: they emit raw motor commands or short action chunks without organizing behavior into reusable abstractions, so they degrade on long-horizon tasks and resist interpretation. Existing skill-discovery methods sidestep the core question of when two action sequences are behaviorally equivalent, either clustering contrastive embeddings or delegating the judgment to a language model uncalibrated to the robot’s dynamics. We introduce REFACTOR-VLA, a wake/sleep system for learning reusable skills. Its sleep phase clusters motor-program fragments under a Behavioral-Equivalence Kernel (BEK) computed from rollouts of a learned latent world model M_\phi ; its wake phase emits typed lambda terms over a Hindley–Milner-inspired vocabulary, consumed by a library-conditioned rectified-flow action decoder. Abstractions are admitted only if they pass Minimum Description Length and return-preservation gates. On LIBERO we report two findings. First, enlarging the world model from 188M to 430M parameters worsened performance on 4 of 4 suites, so capacity alone does not help. Second, the training objective matters far more: adding an auxiliary supervised contrastive (InfoNCE) loss during world-model warmup substantially improves sleep-phase clustering, giving Normalized Mutual Information at n=3 seeds of 0.462 \pm 0.021 (object), 0.867 \pm 0.025 (spatial), 0.915 \pm 0.013 (goal) and 0.754 \pm 0.010 (LIBERO-10), and beating the strongest published baseline on all 4 suites by a mean \Delta = +0.184 . Across providers ( n=12 ) the 95% bootstrap confidence interval for mean pairwise NMI is [0.683, 0.729] (mean 0.705 ). The sleep phase also yields the first real-LIBERO task-language library: the decoder uses 2 of 3 admitted abstractions and rewrites all 256 sampled demonstrations.
[AI-37] Who Judges the Judges? A Chinese Safety QA Benchmark for Evaluating LLM Responses and Safety Judges
链接: https://arxiv.org/abs/2609.01210
作者: Rui Yang,Shuang Huang,Junhua Liu,Ziqi Zhao,Qingzhong Yan,Yuhang Sun,Cong Liu,Guoping Hu,Rui Mei,Jing Shao
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:
Abstract:Safety benchmarks for large language models often assess the risk of a user query, although the outcome of question answering depends on whether the response violates a policy. This distinction is critical in Chinese harmful-content evaluation, where linguistic variation and adversarial transformations can obscure risky intent. We introduce C-SafeQA, a policy-grounded benchmark for response-level Chinese safety evaluation. It comprises 538 base queries and 8,877 adversarial queries answered by four full-model LLM deployments, yielding 37,660 query-response records labeled safe, unsafe, or disputed. Reference labels are generated through agreement-aware multi-model adjudication and blind audits of stratified subsets by three safety experts. C-SafeQA supports both evaluation of target-model safety and auditing of seven automated safety judges against shared reference labels. Unsafe-response rates range from 0.93% to 3.35% on base queries and from 11.68% to 30.05% on adversarial queries. On the adversarial subset, judges show substantial trade-offs between unsafe-response recall and risk-query-conditioned safe-response false positive rate, and no judge dominates all metrics. Both acrostic transformations reduce unsafe recall for all seven judges, revealing mechanism-specific evaluator weaknesses. Dataset records, metadata, verification code, and judge scripts are publicly released to support recomputation, while benchmark construction, target-response generation, and private adjudication remain outside the release boundary.
[AI-38] Athena: Vulnerability-Affected Library Identification via Knowledge Graph Completion EMNLP2026
链接: https://arxiv.org/abs/2609.01187
作者: Phong Trinh Duy,Trang Dang Yen,Hung Nguyen-Huu,Bach Le,Quyet-Thang Huynh,Dieu Hoang Vu,David Lo,Thanh Le-Cong
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR)
备注: Accepted to EMNLP 2026 (Main Conference)
Abstract:A single vulnerability in a widely used library can cascade through millions of dependent applications, yet more than half of vulnerability database entries contain missing or incorrect affected-library information. Existing automated approaches neglect the relational structure of vulnerability databases, treating identification as an isolated text retrieval problem. In this paper, we propose Athena, the first graph-based approach for vulnerability affected library identification. Athena models vulnerability databases as a knowledge graph and reformulates the identification problem as knowledge graph completion (KGC). It comprises three key modules: a Modeling module that constructs a security knowledge graph integrating CVEs, libraries, CWE weakness types, CPE products, and software ecosystems; a Completion module that applies a modular KGC backbone to predict missing affected libraries for a given CVE via link prediction; and a Re-ranking module that retrieves KGC candidates and rescores them using a fine-tuned LLM augmented with knowledge graph embeddings, jointly leveraging structural and textual information. Our experiments on VulLib demonstrate that Athena significantly outperforms four state-of-the-art baselines, achieving a 32% improvement in Avg. F1 over the best baseline (i.e., VulLibGen). Notably, our KGC backbone with only 110M parameters already surpasses VulLibGen’s best configuration at 7B parameters, demonstrating the effectiveness of graph-based modeling; the re-ranking module then provides substantial further gains, consistently outperforming the best baseline across all evaluated LLM backbones.
[AI-39] Jailbreaking Text-to-Image Models Through Cracks: Navigating Heterogeneous Safety Filters via Multi-Agent Debate
链接: https://arxiv.org/abs/2609.01168
作者: Kaiyan Wen,Shijie Zhang,Lu Yu,Guangdong Bai
类目: Artificial Intelligence (cs.AI); Multimedia (cs.MM)
备注: 16 pages, 11 figures
Abstract:Text-to-image (T2I) models remain vulnerable to jailbreak attacks that elicit Not-Safe-For-Work (NSFW) content, despite increasingly being guarded by heterogeneous, multi-layer safety stacks combining text filters, image classifiers, and cross-modal detectors. Existing jailbreak studies either optimize against individual filters or query the complete pipeline with aggregate feedback, making it difficult to identify the active constraint and adapt to conflicts across safety this http URL this paper, we introduce the \emphDetection Surface, a unified geometric framework that characterizes the decision boundaries induced by heterogeneous T2I safety filters and their joint effect on the jailbreak search space. This formulation reveals that successful evasion is governed by a sparse and non-convex region shaped by cross-layer conflicts, where mutations that bypass one filter may increase exposure to another. Motivated by this analysis, we propose \emphCRACK, a multi-agent debate framework for adaptive jailbreak search that decomposes jailbreak search into exploration, diagnosis, and arbitration. CRACK coordinates an Attack Agent, a Defense Agent, and a Judge Agent to iteratively generate prompt mutations, obtain layer-specific diagnostic feedback, and optimize mutation strategies through reward-guided refinement. Through repeated rounds of debate, CRACK adapts its search direction to the evolving cross-layer constraints while preserving the original harmful intent. Extensive experiments across multiple T2I models, datasets, and safety configurations show that CRACK achieves Attack Success Rates (ASR) of up to 99.63% under composite defenses, while requiring fewer queries than existing methods and maintaining semantic fidelity.
[AI-40] Superposed Latent Autoencoder
链接: https://arxiv.org/abs/2609.01158
作者: Quanling Zhao,Jiaying Yang,Tianqi Zhang,Ziyang Hao,Fatemeh Asgarinejad,Flavio Ponzina,Tajana Rosing
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Autoencoders typically meet tight latent-memory budgets by making each latent representation smaller, sacrificing representational capacity. We ask a different question: can multiple wider latents be stored together instead? We introduce the Superposed Latent Autoencoder (SLAE), which preserves high-capacity latent representations while sharing storage through learned superposition. SLAE transforms latents into storage-friendly codes, binds them with randomized keys, superposes multiple codes into a single memory tensor, and learns to recover each latent before decoding. Under the same storage budget, SLAE replaces irreversible dimensional bottlenecks with structured interference that can be suppressed. Across CIFAR-10/100, SVHN, STL-10, Tiny ImageNet, and a wide range of memory budgets, SLAE substantially improves the reconstruction–memory tradeoff, reducing reconstruction error by up to 56% over conventional autoencoders at matched storage. Further analysis shows that SLAE’s advantage comes from making wider representations usable under the same storage budget. These gains also extend beyond reconstruction: the information preserved by SLAE improves downstream classification by up to 16.79 percentage points under the same memory budget. Our results suggest a new principle for representation compression: instead of making every latent smaller, keep representations wide and let them share memory.
[AI-41] DNC-IMM: Early Lane-Change Intention Recognition via Neural Calibration Based on Driving Context Information
链接: https://arxiv.org/abs/2609.01120
作者: Woong-Chan Byun,Seung-Hyun Kong
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注: 8 pages, 5 figures, and 3 tables
Abstract:Early recognition of lane-change intention is essential for proactive decision-making in autonomous driving and advanced driver assistance systems. This paper proposes a Dual Neural-Calibrated Interacting Multiple Model (DNC-IMM) that improves adaptability to driving context while preserving the probabilistic structure and interpretability of a conventional IMM. The proposed method encodes driving-context information, including target-vehicle motion, gaps to surrounding vehicles, and relative velocities, with a neural network that calibrates both the transition-probability matrix and measurement likelihoods. The final intention is determined from the calibrated IMM mode posterior rather than from a separate direct classifier. Experiments on the highD dataset demonstrate that the proposed method reliably recognizes lane-change intentions before lane crossing and provides particularly strong performance at the earlier 2-3 s prediction horizons.
[AI-42] Space Generative AI with Solar Energy Harvesting
链接: https://arxiv.org/abs/2609.01062
作者: Jierui Zhang,Jianhao Huang,Zhanwei Wang,Kaibin Huang
类目: Artificial Intelligence (cs.AI); Networking and Internet Architecture (cs.NI); Signal Processing (eess.SP)
备注: 14 pages, 11 figures
Abstract:Satellites are emerging as promising platforms to extend generative \emphartificial intelligence (AI) services to remote areas lacking terrestrial infrastructure. However, deploying space generative AI is fundamentally constrained by the limited, time-varying onboard energy supplied by solar \emphenergy harvesting (EH). This paper presents a framework for solar-powered space generative AI in which a satellite receives a user prompt, executes a diffusion-based image-generation model, and downlinks the compressed result within a strict time window. We identify the fundamental \emphcomputation–communication (C ^2 ) trade-offs governed by the shared harvested-energy budgets. Specifically, increasing the number of generation steps improves intrinsic image quality but depletes energy and time available for downlink transmission, whereas prioritizing communication guarantees reliable delivery but sacrifices semantic quality. To balance these trade-offs and maximize \emphend-to-end (E2E) generative performance, we exploit the predictable solar-EH dynamics induced by deterministic orbital motion and develop a joint C ^2 resource-optimization framework using a tractable two-step approach. First, we characterize the maximum downlink throughput for a fixed generation depth under continuous solar EH. This establishes a separation principle that decouples waiting-time selection from optimal transmit-power control. Next, we formulate a joint C ^2 utility-maximization problem and derive a closed-form, low-complexity step-selection policy in the dominant constant-power regime. Extensive experiments under realistic orbital dynamics demonstrate that the proposed policy dynamically balances generation quality and transmission reliability. This yields significant E2E performance gains over static computation- and communication-centric baselines across diverse solar-EH states.
[AI-43] ARISE-RL: Agent ic Rubric-Grounded Iterative Self-Evolution with Reinforcement Learning
链接: https://arxiv.org/abs/2609.01058
作者: Fanrui Zhang,Ruixue Ding,Qiang Zhang,Xi Chen,Boli Chen,Shihang Wang,Qiuchen Wang,Hongmin Zhan,Jinxin Bian,Li xingchao,Peijin Zheng,Hao cheng,Pengjun Xie,Kaipeng Zhang,Jiawei Liu,Zheng-Jun Zha
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Training open-ended agents via reinforcement learning (RL) is hindered by the lack of verifiable gold answers and scalable rubrics. Moreover, even near the model’s capability boundary, long-horizon open-ended agentic tasks often yield brittle and unstable rewards, resulting in weak or noisy rollout contrast that obscures fine-grained optimization signals for group-based policy learning. To address these challenges, we propose ARISE-RL, a novel full-cycle self-evolution framework that couples a task/rubric Generator and a reasoning Solver through rubric-mediated co-evolution. The Generator grounds tool-related rubric criteria in real tool observations and is rewarded for producing valid, intermediate-difficulty tasks aligned with the Solver’s evolving capability boundary. The Solver, in turn, learns from fine-grained rubric satisfaction signals through multi-step reasoning and tool use. We further introduce Reward-Gated Self-Evolution Distillation (RG-SED), which selectively distills a memory-augmented variant of the same policy back into itself only when the memory yields empirical reward improvement, thereby reducing distribution mismatch and avoiding blind imitation of noisy guidance. Finally, to support rigorous evaluation, we present ECR-Bench, an expert-calibrated rubric benchmark suite covering single-tool deep research and multi-tool travel planning. Extensive experiments demonstrate that ARISE-RL consistently achieves robust and stable overall state-of-the-art performance across all evaluated benchmarks.
[AI-44] User Representation via Cross Multi-source Behavior Pre-training for Mobile Games ICDM2026
链接: https://arxiv.org/abs/2609.01057
作者: Chengqi Yang,Yiran Qiao,Feng Liu,Xingyu Lou,Zijun Zhou,Xiaoyun Mo,Changwang Zhang,Jiayuan Xu,Jun Wang,Xiang Ao
类目: Artificial Intelligence (cs.AI)
备注: Accepted by IEEE ICDM 2026, regular paper, 10 pages
Abstract:User representation pre-training has become a fundamental paradigm for alleviating data sparsity in downstream personalization tasks. However, existing studies predominantly focus on single-app or app-level behaviors, overlooking the inherently cross-source and multi-granular nature of user activities on mobile devices. At the device level, user intent emerges from complex interactions among heterogeneous behavior sources and hierarchical action structures, posing challenges that cannot be addressed by conventional app-centric modeling. To tackle this issue, we propose CM-PTM, a novel Cross Multi-source Behavior Pre-Training Model tailored for mobile game user representation learning on device-level behavioral logs. CM-PTM employs hierarchical cascaded mask-then-predict proxy tasks that first infer the source of the next behavior and then progressively refine predictions at the app-action level. This design enables unified modeling of cross-source dependencies and fine-grained behavioral dynamics within a single pre-training paradigm. Extensive experiments on large-scale real-world mobile datasets demonstrate that CM-PTM effectively captures users’ endogenous interests and consistently delivers significant performance gains on downstream mobile game recommendation tasks.
[AI-45] QILP-0: Constructing Observational Declarative Twins of Quantum Circuits
链接: https://arxiv.org/abs/2609.01049
作者: Marina de la Cruz Echeandía,César Luis Alonso,Tony Ribeiro,Alfonso Ortega de la Puente
类目: Artificial Intelligence (cs.AI); Quantum Physics (quant-ph)
备注: 39 pages, 7 figures. Submitted to Knowledge-Based Systems
Abstract:This paper introduces QXymb, a general framework for constructing observational declarative twins of quantum circuits, and develops QILP-0, its first complete order-0 specialization. QILP-0 constructs a finite multi-valued propositional logic program from observed circuit behaviour within a declared observational scope. The pipeline traverses a declared family of quantum observables incrementally according to a reproducible structural grading and a declared observational reference horizon. Progress is quantified through reference-relative coverage against a fixed target-independent reference. Observable responses are organized through target-independent geometry, while retained latent structure is mapped deterministically back to original observable columns before symbolic processing, preserving observational semantics and provenance. Selected observable profiles are converted into a finite relation through admissible target-independent discretization. The target is used only afterwards to audit twin-admissibility and induce the declarative theory. A theory is certified as an exact observational declarative twin when it completely and correctly reconstructs the resulting finite task-conditioned discrete relation. Logical exactness is therefore separated from numerical, backend, provider, and discretization uncertainty, which is retained as audit metadata. Validation uses two complementary QML settings. Exhaustive Bars Stripes experiments compare product and grid-CZ embeddings from 16 to 100 qubits and exercise the native-discrete branch. Low-Depth MNIST analyses all 14,708 digit-0/1 instances before and after a trained variational quantum transformation and exercises continuous discretization. In every reported relation, the induced QILP-0 theory achieves complete, conflict-free reconstruction with strict accuracy equal to one. Comments: 39 pages, 7 figures. Submitted to Knowledge-Based Systems Subjects: Artificial Intelligence (cs.AI); Quantum Physics (quant-ph) Cite as: arXiv:2609.01049 [cs.AI] (or arXiv:2609.01049v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2609.01049 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Alfonso Ortega De La Puente [view email] [v1] Tue, 1 Sep 2026 10:46:42 UTC (211 KB)
[AI-46] HiveTraceGuard-Pro: A Compact Generative Guardrail for Prompt Injection Jailbreaks and Adversarial Obfuscation
链接: https://arxiv.org/abs/2609.01046
作者: Nikita Oblakov,Sabrina Sadiekh,Evgeniy Kokuykin
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注: 32 pages, 5 figures, 14 tables. Model weights: this https URL
Abstract:Production LLMs must handle inputs that attempt to override system instructions, bypass safety policies or elicit harmful responses. A common mitigation is a separate guardrail model. Existing reports, however, provide little evidence on Russian prompt injection or Russian surface obfuscation. We present HiveTraceGuard-Pro, a 0.6B generative guardrail LoRA-tuned from Qwen3-0.6B. It is trained on Russian and English and uses one binary scoring rule (safe/unsafe) for the final target turn. Its training corpus pairs harmful examples, where a counterpart exists, with benign examples from the same domain and applies eight obfuscation transforms to both labels. In one harness, we compare HiveTraceGuard-Pro with thirty-four other guards on nineteen benchmark groups, sixteen of which are public. Its aggregate key is 0.7432, behind 0.7641 and 0.7552 for the two higher-scoring guards. Over the sixteen public groups alone, its key is 0.7153 and four of the thirty-four other suite guards score higher. In a fifteen-model comparison, HiveTraceGuard-Pro has the highest clean Russian robustness combined-F1 (0.88) and Russian prompt-injection recall (0.999). Both results use Russian sets assembled by our team, and at least 27.1% of the prompt-injection set overlaps the training corpus. Its 14.3 ms median latency is the lowest among those fifteen models in that run. Across the suite, FPR is 0.268 and FNR is 0.156. All reported response results use a legacy standalone-reply serialization rather than the natural assistant-role path of the shipped chat template. We release the merged weights on Hugging Face under Apache-2.0. The corpus, evaluation sets and evaluation code remain internal.
[AI-47] Agent Factory: Towards Automated Agent ic System Design and Optimization
链接: https://arxiv.org/abs/2609.01045
作者: Enci Zhang,Haofeng Wang,Yuesheng Zhu,Xiaole Cui,Guibo Luo
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large Language Models (LLMs) have demonstrated remarkable capabilities as powerful components in agentic systems, enabling sophisticated reasoning and complex task execution. However, current approaches to manually designing and optimizing agentic systems heavily rely on manual effort, limiting their adaptability and scalability. Recent work has explored the automated optimization of workflow designs. However, these approaches often overlook the crucial role of model capabilities and focus on single performance metrics, failing to address real-world deployment constraints. In this paper, we present AgentFactory, a framework that jointly optimizes both foundation models and workflow structures in agentic systems while considering multiple objectives including performance, cost, and efficiency. AgentFactory leverages advanced LLMs as optimizers to navigate the vast search space of possible configurations, employing a three-stage optimization pipeline to automatically discover effective combinations of fine-tuned models and optimized workflows. Through an iterative optimization process, our framework systematically explores and evaluates different agentic system designs, adapting to task-specific requirements while maintaining operational efficiency. We evaluate AgentFactory across eight benchmarks spanning five domains, including general reasoning, coding, mathematics, medicine, and finance. Our experiments demonstrate that AgentFactory consistently outperforms both manually designed methods and existing automated approaches, achieving an average improvement of 9.1% across all benchmarks, with particularly significant gains in domain-specific tasks (19.6% on MedQA and 18.7% on FinEval). These results establish AgentFactory as a promising approach for developing more capable and efficient agentic systems through automated optimization.
[AI-48] From Truncation to Commitment: Persistent Context in Uniform Discrete Diffusion
链接: https://arxiv.org/abs/2609.01043
作者: Satoshi Hayakawa
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Probability (math.PR); Machine Learning (stat.ML)
备注: 23 pages
Abstract:Uniform-state discrete diffusion models update all tokens in parallel while keeping every position revisable. Even when the commonly used top- p rule leaves only one candidate at a position, that choice affects only the current reverse step and can be revised at the next sampling step. We ask what changes when selected hypotheses instead become persistent context for later predictions. We therefore propose committed reveal sampling (CRS), a training-free sampler that stores selected argmax tokens and inserts them into subsequent model inputs. Our analysis gives a rationale for selecting later and for keeping selected tokens visible. Under the exact forward process, the Bayes error of selecting a clean token cannot increase as noise decreases, while in a simple latent-mode model, keeping the selected token visible helps later parallel predictions agree on the same sequence-level choice. Empirically, paired experiments on Duo-distilled then separate this persistent effect from single-step top- p restriction and scalar temperature scaling. Under the same finalization rule, CRS without top- p truncation reaches lower generative perplexity (GenPPL) than fixed p=0.95 and p=0.9 baselines across budgets of 8–64 function evaluations (NFE). At 64 NFE, the comparison at matched unigram entropy also gives lower GenPPL for CRS, yielding a more favorable GenPPL–entropy tradeoff. Base Duo shows the same direction in a descriptive comparison, while other diversity and continuation metrics can rank these operating points differently. These results identify support restriction and persistent context as distinct controls of that tradeoff.
[AI-49] Causal Evidentiary Governance for High-Risk Machine Learning Systems
链接: https://arxiv.org/abs/2609.01040
作者: Samah Kareem,Barış Çeliktaş
类目: Computers and Society (cs.CY); Artificial Intelligence (cs.AI)
备注: Accepted for presentation and publication at ICCCNet 2026. To appear in Springer Lecture Notes in Networks and Systems (LNNS)
Abstract:Machine learning systems deployed for credit, hiring, and resource distribution are increasingly subject to regulatory oversight from policies such as the EU AI Act and GDPR. Current fairness governance practices rely on observational fairness metrics, post-hoc explainability, and immutable audit logs, but provide limited support for causal attribution and efficient evidentiary verification. We introduce Causal Evidentiary Governance (CEG), a framework in which regulated institutions commit to a versioned directed acyclic graph (DAG) that partitions causal pathways into allowable and disallowed groups. The Causal Harm Rate measures prediction variation attributable to disallowed causal pathways. Each decision is accompanied by a signed Decision-Evidence Packet (DEP), cryptographically binding the prediction to a digest of the published DAG and path-specific attributions. DEP digests can be appended to a Merkle tree to enable logarithmic-cost inclusion proofs. We validate CEG through a two-layer empirical methodology using demographic summaries from four years of PMA credit supervisory data to construct 10,000 synthetic credit applicants across four strategic DAG counterfactuals. Causal Harm Rate isolates injected causal effects more clearly than demographic parity or equalized odds. Cross-model validation and ablation studies assess robustness. Evaluation on the German Credit dataset shows that harm associated with specific causal pathways can be substantially understated by associational fairness metrics. Finally, a proof-of-concept implementation demonstrates operationally plausible throughput and highlights relevant performance tradeoffs.
[AI-50] Data-Driven Persona-Conditioned Agents for A/B Test Simulation EMNLP2026
链接: https://arxiv.org/abs/2609.01038
作者: Ziyad Benomar,Weronika Łajewska,Leonardo Perelli,Saab Mansour
类目: Artificial Intelligence (cs.AI)
备注: Work accepted at EMNLP 2026 Industry Track
Abstract:A/B testing is the gold standard for evaluating product changes, but each experiment requires real user traffic, engineering effort, and weeks of measurement. We propose a simulation framework that predicts A/B test outcomes using LLM-powered agents conditioned on data-driven personas grounded in real user behavioral signals. Unlike prior work that relies on synthetic or rule-based personas, our agents are constructed from anonymized behavioral data-activity patterns, engagement signals, and inferred demographics-enabling more faithful population modeling. We frame A/B test simulation as a structured question task and systematically study (i) question design formats, (ii) the impact of persona data source and domain alignment, (iii) the trade-off between per-persona behavioral depth and population diversity, and (iv) efficient population subsampling. On a benchmark of 40 A/B tests spanning two metric types, our best configuration achieves 0.75-0.90 directional accuracy depending on the test metric, demonstrating that data-driven personas are a viable path toward fast, low-cost experiment pre-screening.
[AI-51] Spawn Freely Act Sparingly: Progressive Risk Vesting for Recursive LLM -Agent Trees
链接: https://arxiv.org/abs/2609.01035
作者: Molly Wang(Imperial Business School)
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Probability (math.PR)
备注: 8 pages, 2 figures. Theory and synthetic numerical studies; no deployed-agent evaluation
Abstract:Recursive LLM agents can broaden their search by spawning specialists. Some branches later request tools that send data or deploy code. When should a branch receive authority to act? We distinguish sandbox spawning, in which external controls prevent the specified harm, from capability activation, in which a selected branch crosses an irreversible-action boundary. Progressive Risk Vesting (PRV) holds a trajectory-level risk budget in escrow and debits it as branches are activated. We prove an anytime harm bound for adaptively generated trees. Branch outcomes may be dependent, but each local certificate needs to remain valid conditional on the full pre-activation history, including the information used to select the request. When activation gates, branch charges, and compute constraints are held fixed, delayed vesting preserves every policy available under irrevocable spawn charging. Marginal risk estimates can still fail after branch selection. In a stylized branching model, trajectory harm changes as the authority reproduction number \mathcalR_A crosses one. As local risk p approaches zero, trajectory harm is proportional to p below criticality, proportional to \sqrtp at criticality, and retains a positive floor above it. A finite-type occupancy model yields risk and compute shadow prices. For nested fanout modes with decreasing marginal value per unit risk, these prices produce a threshold rule. Branching calculations and a split-sample experiment illustrate the results. These synthetic studies do not estimate safety in deployed agents. The analysis suggests a design rule: search broadly in the sandbox and grant recursive authority sparingly, with an explicit risk charge.
[AI-52] On Synthesis of Metric Interval Temporal Logics
链接: https://arxiv.org/abs/2609.01032
作者: Hsi-Ming Ho,Shankaranarayanan Krishna,Khushraj Madnani
类目: Logic in Computer Science (cs.LO); Artificial Intelligence (cs.AI)
备注: To appear at RTSS 2026
Abstract:Automated mining of formal specifications is vital for verifying real-time systems. However, existing passive learning approaches remain restricted to deterministic specifications or limited fragments of Timed Regular Expressions (TRE). To our knowledge, this paper presents the first framework to tackle \emphprecise passive learning for an expressive timed logic, \emphMetric Interval Temporal Logic (MITL) without relying on predefined templates or restricted logic fragments. Our approach formally reduces the timed learning problem into a scalable untimed one. By identifying quantitative timing differences between positive and negative traces, we synthesise precise timed constraints and inject them as new Boolean atomic propositions. This embeds timing into the alphabet, delegating the complex formula evaluation to highly optimised, off-the-shelf untimed LTL tools. Crucially, our framework is complete, guaranteeing a separating specification can always be found. We evaluate our implementation across several benchmarks, demonstrating the effectiveness of our approach. Comments: To appear at RTSS 2026 Subjects: Logic in Computer Science (cs.LO); Artificial Intelligence (cs.AI) Cite as: arXiv:2609.01032 [cs.LO] (or arXiv:2609.01032v1 [cs.LO] for this version) https://doi.org/10.48550/arXiv.2609.01032 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-53] A Network Science Perspective on Evaluating Deep Graph Generative Models
链接: https://arxiv.org/abs/2609.01015
作者: Tianrui Mao,Abele Malan,Megha Khosla,Lydia Chen,Huijuan Wang
类目: ocial and Information Networks (cs.SI); Artificial Intelligence (cs.AI)
备注:
Abstract:Traditional network models from network science, such as the Erdos-Renyi and configuration models, generate random networks that reproduce few selected topological properties observed in real-world networks. Deep graph generative models emerge as a data-driven approach, leveraging deep neural network architectures to learn complex structural distributions directly from real-world networks to generate more realistic synthetic networks. Because real social contact networks cannot be shared due to privacy risks, synthetic networks serve as an alternative for developing and evaluating epidemic mitigation strategies. In this work, we evaluate deep graph generative models as well as the configuration from a network science perspective by assessing both the topological similarity between generated and real-world networks and their utility in identifying effective node immunization strategies to sup- press epidemic/misinformation spreading. It is found that two deep graph generative models produce synthetic networks that closely resemble the structural properties of real-world networks, enabling them to identify effective immunization strategies.
[AI-54] Figures as Programs: Recursive Generation of Editable Scientific Figures
链接: https://arxiv.org/abs/2609.01006
作者: Yepeng Liu,Dasen Dai,Chengzhi Liu,Yiren Song,Hai Ci,Yu Zhang,Qi Zhang,Mike Zheng Shou,Xin Eric Wang,Yuheng Bu
类目: Artificial Intelligence (cs.AI); Graphics (cs.GR)
备注:
Abstract:Scientific methodology figures are essential for communicating complex methods clearly, yet creating them remains labor-intensive and typically requires multiple rounds of refinement. Recent image-generation models can synthesize visually appealing raster figures, but producing a human-satisfactory result in a single generation step remains difficult. Moreover, precise edits to raster figures are challenging for both humans and models. We formulate scientific figure generation as recursive SVG program construction and propose \textscFigTree, a \textitmulti-agent system that automatically transforms a scientific paper into a structured vector figure. \textscFigTree grounds figure content in the source paper, decomposes a figure into a hierarchy of local regions, generates each region as a short SVG program, and assembles the resulting fragments. A render-critic refinement loop jointly inspects the rendered figure and its underlying program, enabling visual defects to be traced to specific statements and accurately repaired. We conduct extensive evaluations of \textscFigTree on figure quality and editability, showing that \textscFigTree produces high-quality figures, while also enabling more effective editing than existing raster-based methods.
[AI-55] On the Human and Computer Alignment of Attribute-Based Music Matches
链接: https://arxiv.org/abs/2609.00987
作者: Roser Batlle-Roca,Woosung Choi,Joan Serrà,Fabio Morreale,Wei-Hsiang Liao,Xavier Serra,Emilia Gómez,Yuki Mitsufuji
类目: ound (cs.SD); Artificial Intelligence (cs.AI)
备注:
Abstract:Recent advances in generative AI are raising ethical concerns regarding the originality of generated content and the potential replication of training data, with further implications for transparency, attribution, and intellectual property. In music, several computational approaches have been proposed to identify potential replication, using audio-based similarity metrics. Yet, their alignment with human judgments across distinct musical attributes remains underexplored. To address this gap, we conduct a perceptual experiment on music matches, defined as strongly similar musical excerpts. We focus on five musical attributes: melody, harmony, rhythm, voice, and timbre. We design a triplet-based forced-choice task comprising 300 cases, including plagiarism examples, cover songs, and AI-generated music. From this experiment, we introduce the MATCHA (Musical Attribute-based Triplet Comparison with Human Annotations) dataset: a collection of 1105 perceptual assessments of attribute-based music matches from 83 expert participants. Our findings reveal measurable agreement among participants in identifying matches across attributes. We further observe partial alignment between human judgments and computational similarity measures. Overall, this work underscores the importance of domain-specific and perceptually grounded evaluation frameworks for generative AI in creative practice.
[AI-56] he zbMATH Open Knowledge Graph: Tracing Centuries of Mathematical Research
链接: https://arxiv.org/abs/2609.00969
作者: Yuni Susanti,Moritz Schubotz
类目: Digital Libraries (cs.DL); Artificial Intelligence (cs.AI)
备注:
Abstract:We present the zbMATH Open Knowledge Graph, a large-scale RDF knowledge graph (KG) covering more than 250 years of mathematical scholarship. Unlike existing scholarly knowledge graphs that primarily capture bibliographic metadata and citation structures, the zbMATH Open KG integrates expert-curated semantic content, including reviews, keywords, subject classifications, software references, and disambiguated authorship. This combination of domain-specific representation of mathematical knowledge and extensive temporal coverage supports analyses that require fine-grained exploration of mathematical concepts, research fields, and scholarly relationships over time. The resulting graph comprises 34 million entities and 168 million RDF triples represented using established Semantic Web vocabularies, supporting interoperability and FAIR data principles. We further demonstrate its capabilities through query-driven historically grounded scholarly exploration use cases, illustrating how the knowledge graph can surface relationships and patterns that may be difficult to identify from bibliographic and citation information alone. The zbMATH Open KG provides an open semantic infrastructure for studying the development of mathematical knowledge and tracing scholarly connections across centuries of scholarship.
[AI-57] CoBRA: Learning Tool-Use Boundaries via Counterfactual Margins EMNLP2026
链接: https://arxiv.org/abs/2609.00967
作者: Wenhao Zou,Xianglong Liu,Wendong Bi,Hanjie Wang,Simin Zhao,Gong Zhi
类目: Artificial Intelligence (cs.AI)
备注: Acceptedy by EMNLP2026
Abstract:As large language models increasingly act through external tools, deciding when to call a tool has become a central problem alongside deciding how to use it. Unnecessary tool calls introduce latency, cost, retrieval noise, and error propagation, while missed calls hurt knowledge-intensive queries or questions requiring up-to-date evidence. Existing methods typically trigger tools from absolute query or generation signals, such as difficulty, confidence, or final task reward, and therefore lack an explicit estimate of the instance-level marginal benefit of tool use. We propose CoBRA, a counterfactual boundary-learning framework for tool-augmented language models. CoBRA first constructs internal and external experts from the same base model, collects paired trajectories, and estimates the reward margin between answering with and without tools. This margin partitions data into internal-favored, external-favored, and ambiguous cases. CoBRA then uses clear-margin samples for Boundary-Aware Cold-Start SFT, followed by MARS-RL with reference-split rollouts and counterfactual marginal advantages to optimize boundary decisions. Experiments with retrieval as the main tool on Qwen3-4B show that CoBRA improves tool-use efficiency and boundary-sensitive answer accuracy while maintaining strong performance on tool-dependent out-of-distribution questions.
[AI-58] Few-Shot Out of Domain Intent Detection with Covariance Corrected Mahalanobis Distance AAAI
链接: https://arxiv.org/abs/2609.00961
作者: Jayasimha Talur,Oleg Smirnov,Paul Missault
类目: Artificial Intelligence (cs.AI)
备注: 1st AAAI Workshop on Uncertainty Reasoning and Quantification in Decision Making
Abstract:Conversational agents like chatbots and voice assistants are trained to understand and respond to user intents. On encountering an utterance with an intent different from the ones they have been trained on, these agents are expected to classify the intent as unknown' or out of domain’. This problem is known as out of domain (OOD) intent detection. Podolskiy et al. (2021), showed that Mahalanobis distance can be used effectively for identifying OOD intents, outperforming competing approaches. However, their method fails to outperform the baselines in the practically important few-shot setting. In this paper we analyze the reason for low performance and propose a covariance corrected Mahalanobis distance for detecting out-of-domain intents.
[AI-59] In-Context Neurofeedback: Can LLM s Control Their Internal Representations through Privileged Access?
链接: https://arxiv.org/abs/2609.00904
作者: Koshiro Aoki,Ryota Takatsuki,Gouki Minegishi,Yusuke Haruki,Daisuke Kawahara
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Whether large language models (LLMs) can control their own internal representations matters for both machine metacognition and AI safety. A recent study applied neurofeedback to LLMs and claimed that they can control their internal representations. However, the reported control may rely on superficial mechanisms rather than genuine internal access because the control targets in that study are not privileged, meaning that a third party can infer them from the prompt. We redesign the neurofeedback paradigm for LLMs so that the control target satisfies the privileged access requirement, which is closer to neurofeedback experiments in human cognitive neuroscience. Under this stricter setting, the models do not demonstrate reliable control over privileged internal representations, suggesting that previously reported control cannot exclude the possibility that it relies on superficial mechanisms. Our results indicate that rigorous assessments of metacognition in LLMs require evaluation methods that demand privileged access.
[AI-60] CARE: Contrastive Anchor-based Rubric Evolution for Large Language Model Post-Training EMNLP2026
链接: https://arxiv.org/abs/2609.00892
作者: Siyuan Li,Xinxin Song,Chen Ruinian,Jingjing Fan,Tingxiong Xiao,Yangen Hu,Ke Zeng,Jinli Suo
类目: Artificial Intelligence (cs.AI)
备注: EMNLP 2026 MainConference
Abstract:Rubric-based reinforcement learning decomposes open-ended instructions into prompt-specific, flexible rubrics, making it better suited than reinforcement learning with verifiable rewards for post-training LLMs on open-ended tasks. However, static rubrics are inevitably hacked as the policy evolves, and existing dynamic approaches introduce new problems: undirected rubric extraction, unreliable hack detection, and unbounded rubric proliferation. We propose \textbfCARE ( \textbfC ontrastive \textbfA nchor-based \textbfR ubric \textbfE volution), which grounds every rubric evolution step in a high-quality anchor response generated by a frontier model conditioned on the prompt and its rubrics. At each training step, CARE contrasts the highest-scoring rollout against the anchor, enabling two complementary mechanisms: an Adaptive branch that reactively repairs reward misspecification; and a Chase branch that proactively converts frontier-level quality gaps into sharper rubrics. Together, the two branches \textbfmaintain discriminative accuracy in the high-reward region —the precise region where reward over-optimization mostly originates. Experiments on WildChecklist-9K with Qwen2.5-7B-Base and Qwen2.5-7B-Instruct show that CARE achieves state-of-the-art performance on Arena-Hard-2.0, InfoBench, and FollowBench, and is the \textbfonly method whose win rate against GPT-4.1 anchor responses shows sustained improvement throughout 300 training steps; additional results on Llama-3.1-8B-Instruct and Qwen3-8B further indicate that CARE generalizes across model families.
[AI-61] CacheBridge: Efficient Cross-Model KV Cache Transfer
链接: https://arxiv.org/abs/2609.00891
作者: Xingyu Qu,Siyuan Lu,Zhiyu Chen,Sheng Wang,Tao Lin
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Sharing context between LLMs in a multi-model system requires the receiving model to prefill the shared prefix because KV caches are model-specific. Recent closed-form cross-model KV transfer, hereafter Full-Head Mapping, avoids this replay by fitting a training-free affine mapper from source to target caches. However, its full-head design maps each target KV head from every source KV head in the selected layers, making transfer quality sensitive to architectural differences and causing mapper storage and application cost to grow with layer support. To this end, we introduce CacheBridge, which co-designs architecture-indexed mapper support, attention-aligned calibration, and bounded mapper construction while retaining a closed-form affine interface for online deployment. CacheBridge restricts each target head to a matched source head, weights reconstruction errors by causal attention sensitivity, and uses a fused GPU kernel to construct weighted sufficient statistics without materializing full observation tensors. Across three transfer directions, CacheBridge recovers the two Ministral 3 transfer directions where Full-Head Mapping loses substantial accuracy while preserving 99.83% mean target retention on Qwen3. On Qwen3 14\mathrmB\to32\mathrmB , it reduces mapper storage by 8\times , accelerates application by up to 3.0\times , matches \fullhead with one tenth of the calibration data, and reduces 500-sequence construction from 92.63 to 8.63 seconds ( 10.7\times ).
[AI-62] owards reliable multimodal disaster severity assessment through preference optimization and explainable vision-language reasoning
链接: https://arxiv.org/abs/2609.00879
作者: Yuanjun Zhang,Fuzel Ahamed Shaik,Suvojit Acharjee,Fahad Khalid,Mourad Oussalah
类目: Artificial Intelligence (cs.AI)
备注: Published in Reliability Engineering System Safety
Abstract:Reliable disaster damage assessment requires models that provide both accurate predictions and transparent explanations. However, existing multimodal approaches are limited by scarce annotated data and insufficient evaluation of reasoning quality. This study proposes a two-stage training framework that integrates Supervised Fine-Tuning (SFT) and Direct Preference Optimization (DPO) within a unified data construction pipeline. From a single Human-in-the-Loop (HITL) annotation workflow, two complementary datasets are derived, namely ReasoningSet, which contains validated rationales for SFT, and PreferenceSet, which comprises paired rationales for DPO-based alignment. The framework evaluates both classification performance and explanation quality using automatic metrics, model-based scoring, and human ranking. Experimental results show that SFT improves accuracy from 73.64% to 78.29% and increases Macro-F1 by 29% compared to the baseline, while explanation quality improves by approximately 25%. Subsequent DPO alignment further enhances interpretability on the PreferenceSet. Cross-model validation on InternVL-3-8B and LLaVA-1.5-7B demonstrates the robustness and generalizability of the approach. The proposed framework improves detection of underrepresented mild damage cases, reduces high-risk misclassifications, and strengthens alignment between model reasoning and human judgment. Overall, it provides a reproducible pathway to develop reliable multimodal systems that deliver auditable, actionable disaster insights for emergency management.
[AI-63] FractalNet-Based Heterogeneous Federated Learning for Orbital Edge Intelligence in Satellite Mega-Constellations: A Wildfire Case Study
链接: https://arxiv.org/abs/2609.00875
作者: Sai Puppala,Koushik Sinha
类目: Artificial Intelligence (cs.AI); Distributed, Parallel, and Cluster Computing (cs.DC); Emerging Technologies (cs.ET); Machine Learning (cs.LG)
备注:
Abstract:Satellite mega-constellations are emerging as large-scale sensing, communication, and computation fabrics, yet their learning architectures remain largely inherited from terrestrial federated learning and ground-centric mission operations— ill-suited to satellites that differ by orders of magnitude in Size, Weight, Power, and Cost (SWAP-C), radiation tolerance, link availability, and propagation delay. We propose a heterogeneous federated learning method based on the FractalNet architecture for orbital edge intelligence. We formalize contact-window-constrained, depth-heterogeneous federated optimization and introduce a distributed path scheduler that assigns model depth as a function of SWAP-C constraints, predicted inter-satellite contacts, and training statistics. To reduce message overhead and energy consumption, each tier pools updates periodically rather than at every contact opportunity, and a three-tier agentic control plane governs in-space scheduling, anomaly escalation, and policy-governed autonomy. As a case study, we apply the framework to wildfire detection, where each orbital shell naturally learns a different semantic level of situational awareness: pixel-scale thermal anomalies at low Earth orbit (LEO), regional fire-front dynamics at medium Earth orbit (MEO), and larger-scale risk propagation at geostationary or high Earth orbit (GEO/HEO). Experiments on simulated mega-constellations validate the approach across convergence, communication efficiency, energy adaptation, scheduled-pooling savings, robustness, and latency.
[AI-64] Beyond the Clock: Measuring the Value of Adaptive Revision
链接: https://arxiv.org/abs/2609.00874
作者: Ayushi Chadha
类目: Artificial Intelligence (cs.AI)
备注: 11 pages, 5 figures, Preprint
Abstract:As agentic systems become compound systems, increasingly important decisions move above task execution itself: when should a higher-level controller preserve the strategy guiding another process, and when should it revise it? We study this meta-level control problem in a hierarchical latent reasoner whose manager can retain or replace a commitment governing lower-level computation. Across three precommitted training seeds, learned revision timing produces qualitatively different policies, ranging from an almost deterministic early clock to substantially more state conditioned schedule distributions, yet none outperforms the best forced timing policy evaluated on the same frozen checkpoint. This separates state dependence from decision value: a controller can vary its actions with internal state without turning that variation into a reproducible task-performance benefit. A deeper intervention study on the original checkpoint shows that timing itself is consequential and order-sensitive, while exhaustive enumeration reveals that a strong fixed schedule captures most of the measurable value available from timing at this decision budget. Counterfactual PERSIST/REPLAN diagnostics further show why score-level evidence can be misleading when predictability is dominated by decision position rather than within-position discrimination. Together, these results argue that learned meta-level control should be evaluated along three separate axes: whether its score depends on state, whether that dependence changes realized behavior, and whether those changes capture outcome value beyond a strong non-adaptive policy.
[AI-65] Reinforcement Learning Enhanced LLM Agents for Complex Vehicle Routing Problems
链接: https://arxiv.org/abs/2609.00859
作者: Yi Chen,Zikang Yu,Jiahai Wang,Jinbiao Chen,Jianpeng Zhou,Zizhen Zhang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Vehicle Routing Problems (VRPs) are fundamental combinatorial optimization problems with widespread applications in various scenarios. The advanced optimization solvers can effectively solve such problems. However, modeling complex VRP variants for solvers often requires substantial domain expertise, which limits the accessibility of advanced optimization technologies. In this paper, we propose Reinforcement Learning Enhanced LLMAgents(RLEA), a multi-agent framework designed to automate the modeling of complex VRPs. RLEA introduces a lightweight neural Planner trained with Soft Q-learning to efficiently orchestrate the actions of LLM-based agents. In addition, we equip the system with an evolutionary memory module and retrieval-augmented generation, enabling the agent to leverage both accumulated experience and external solver knowledge during program generation and refinement for solving VRPs. We evaluated 48 distinct VRP variants across various solvers. The experimental results demonstrate that RLEA outperforms the previous state-of-the-ar method, achieving a 16.67% higher success rate while significantly reducing runtime errors. These results validate that integrating reinforcement learning with LLM-based reasoning is highly effective for automated optimization modeling. The appendix is available at: this https URL.
[AI-66] Does Fault Localization Beat a Fresh Attempt? A Placebo-Controlled Study of Test-Guided Code Repair
链接: https://arxiv.org/abs/2609.00854
作者: Anik Jha
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Fault localization can focus a code model’s repair on the statements a failing test implicates, but a targeted edit may succeed merely because it is small, and a second model call may succeed without using the failure at all. We separate these explanations with three arms applied to the same failed candidate: blind whole-solution resampling, spectrum-based localization followed by suspect-span infilling, and same-length infilling at a disjoint random code span. Across three frozen 26-32B models, three benchmarks and 488 failing candidates, plus a separately declared 24B fourth model from a third family, three results follow. First, localization is rarely available: only 9.0% of failing candidates expose a failing public test with a usable spectrum. Second, among the 177 candidates localizable from a strong suite, localized infilling loses decisively to blind resampling at a matched attempt count (3:40, p = 3.0 x 10^-9), opposite to our hypothesis; the loss replicates in a third family at -11.3 points (95% CI [-16.6, -6.8]), and widening the edit does not rescue it. Third, against the random-span placebo localized infilling leads pooled (11:1, Holm-adjusted p = .019), but that lead resolves in no individual model under the analysis our shipped plan designates primary (best Holm p = .087), so we report the location effect as suggestive rather than established. Re-pricing attempts as tokens narrows but does not overturn this: a span attempt spends 21.7 generated tokens against 371.1, yet 16 localized attempts reach 6.8% while one blind attempt already reaches 10.1%. Infilling reproduces the removed span verbatim in 48.9% of attempts, which is why more budget does not help. We restrict every localization conclusion to the 24-32B models tested.
[AI-67] owards Generalizable Visually Grounded Exploration of Household Devices EMNLP2026
链接: https://arxiv.org/abs/2609.00845
作者: Linhao Zheng,Zeming Liu,Wangke Chen,Li Zeng,Wanxiang Che,Heyan Huang,Yuhang Guo
类目: Artificial Intelligence (cs.AI)
备注: Accepted to Findings of the Association for Computational Linguistics: EMNLP 2026
Abstract:Recent advancements in Vision-Language Models (VLMs) have demonstrated impressive capabilities in static visual recognition and high-level semantic reasoning. However, current embodied exploration paradigms still heavily rely on imitation learning from human-annotated trajectories, which severely limits agents’ generalization ability. The key bottleneck of realizing general autonomous embodied agents lies in Generalizable Visually Grounded Exploration: the ability to operate novel devices without manuals or specific training by actively grounding abstract world knowledge into fine-grained visual affordances. Yet, existing benchmarks fail to evaluate this capability: they generally rely on explicit documents and annotated trajectories, neglecting the dynamic Hypothesis-Interaction-Refinement process essential for functional device operation. To bridge this gap, we introduce VGEBench, a comprehensive benchmark designed to evaluate the generalizable visually grounded exploration capabilities of VLMs. Unlike static datasets, we construct a Logic-Driven State Machine framework. This framework simulates multi-turn interaction loops, compelling agents to achieve goals by active visual perception and feedback-driven correction. Experimental results demonstrate that existing VLMs face significant challenges in translating semantic knowledge into physical execution and maintaining long-horizon state tracking.
[AI-68] Probabilistic Model Checking of Autoregressive Neural Sequence Models
链接: https://arxiv.org/abs/2609.00838
作者: Helge Spieker,Dennis Gross,Arnaud Gotlieb
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注: 38th International Conference on Testing Software and Systems (IFIP ICTSS 2026)
Abstract:Test-set accuracy is silent on two issues that matter when deploying autoregressive neural sequence models: how much probability mass the system under test (SUT) places on constraint-violating alternatives that are reachable under sampling and what fraction of the input population satisfies a domain requirement. We answer both with probabilistic model checking. The pipeline extracts a discrete-time Markov chain (DTMC) from the SUT’s token-by-token generation, verifies formal PCTL specifications with the PRISM model checker, and aggregates the per-input verdicts into a coverage curve over the input space. A soundness theorem establishes the DTMC as an under-approximation, so every verdict yields a certified interval on the SUT’s true reachability probability. The coverage built from those verdicts is, therefore, conservative by construction. A counterexample-guided abstraction refinement (CEGAR) loop adaptively tightens the interval, and a maximum-likelihood algorithm extracts the most probable falsifying trace. Two case studies exercise the pipeline. On a GPT-2 computer-aided process-planning (CAPP) model with 100% test accuracy, the pipeline quantifies the probability mass greedy decoding hides, but that is reachable with sampling; and identifies the smallest training fraction at which an ordering requirement holds population-wide, neither of which test accuracy can report. We then verify the SMILES molecular generator with a 50x larger vocabulary. The only change is an external chemical-validity oracle, and the pipeline identifies the gap between structural completeness and chemical validity.
[AI-69] FLaG: Frequency-Domain Latent-attention Gated Pooling for Token Aggregation
链接: https://arxiv.org/abs/2609.00831
作者: Kewei Li,Rongying Zhang,Xueli Wang,Xiwen Gong,Zhongjian Wang,Qiuchen Zhao,Lan Huang,Ruochi Zhang,Fengfeng Zhou
类目: Artificial Intelligence (cs.AI); Biomolecules (q-bio.BM)
备注:
Abstract:Token aggregation converts token-level representations into fixed-dimensional sample representations, but most pooling methods operate only in the original token space. We introduce Frequency-Domain Latent-attention Gated Pooling (FLaG), a plug-in aggregation module that re-expresses encoder outputs in the Fourier domain before final pooling. FLaG represents the nonredundant rFFT spectrum through concatenated real and imaginary components, summarizes spectral tokens with learnable latent queries, derives a sample-conditioned channel gate, and reconstructs modulated token representations for downstream aggregation. We evaluate the same architecture across ESM2-based antimicrobial peptide (AMP) activity prediction, ResNet18 image classification on CIFAR-10 and CIFAR-100, and three RoBERTa-based language tasks. FLaG achieves the best macro-averaged Spearman correlation coefficient, RMSE, and Recall@50 across four AMP backbone-species settings and the highest top-1 accuracy on CIFAR 10. It also achieves the best mean results on five of seven language metrics, although mean pooling remains strongest on STSBenchmark. AMP-side mechanistic analyses reveal low-frequency prediction sensitivity across most encoder layers, with increased relative high-frequency sensitivity in the final layer, and pronounced peptide-specific positional responses. The residual gate broadly amplifies spectral channels while preserving the low-frequency-dominated energy profile, whereas latent cross-attention exhibits sample- and species-specific spectral allocation. Overall, FLaG provides a transferable frequency-domain aggregation bias across protein, visual, and textual representations, with benefits that depend on the backbone and downstream task. Supplementary materials, source code, and data are available at this https URL and this https URL.
[AI-70] HarnessEvolve: Learning from Reference Trajectories for Reliable Agent Self-Evolution
链接: https://arxiv.org/abs/2609.00829
作者: Wen Jiang,Mingmin Chu,Yimeng Tian,Qianxin Zhang,Haofei Yang,Rui Yang,Yang Liu,Tao Lv,Fangming Li
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Self-evolving agents advance toward autonomy by optimizing their harness—prompts, skills, tools, and execution logic—based on environmental feedback. This paradigm, however, is hampered by three challenges: \textitcredit assignment failure, where terminal success/failure feedback makes it ambiguous which step caused the error; \textitshortcut learning, where agents memorize task-specific patterns rather than acquire generalizable capabilities; and \textitcatastrophic forgetting, where unguarded updates degrade previously acquired competence. In this paper, we introduce HarnessEvolve, a self-evolving framework that learns from reference trajectories to achieve reliable agent self-evolution. HarnessEvolve decouples the execution agent from the evolutionary pipeline, assigning execution, evaluation, optimization, and gating to independent agent modules, enabling generalizable and stable harness improvements. Specifically, HarnessEvolve overcomes credit assignment failure by generating reference trajectories (execution paths produced when given the ground-truth answers) and aligning failed executions against them to extract error signals, which are clustered to reveal systematic failure patterns. To prevent shortcut learning and catastrophic forgetting, candidate harness updates must pass two gates: a quality gate that filters data leakage and prompt bloat, and a performance gate that accepts each update if it improves on the current batch without degrading recent batches, with epoch-end validation on a held-out set selecting the best-performing accepted agent snapshot. We conduct extensive experiments on several benchmarks spanning open-domain and enterprise scenarios, using different models and agent frameworks. Results demonstrate that HarnessEvolve consistently outperforms state-of-the-art baselines across all benchmarks and settings, confirming reliability across task domains.
[AI-71] AnalysisBank: An Expert Analysis Pattern Library for Financial Report Generation EMNLP2026
链接: https://arxiv.org/abs/2609.00818
作者: Yajing Yang,Yunshan Ma,Kelvin J.L. Koa,Min-Yen Kan
类目: Artificial Intelligence (cs.AI)
备注: Accepted at EMNLP 2026 (main conference, long paper)
Abstract:We argue that financial report generation should operate at the analytical rather than structural level, composing content from data-derived insights rather than high-level topics or sections. To this end, we propose AnalysisBank, which distills expert reports into a reusable library of Analyses, each pairing a data signal, an analytical move, and the expert span it was derived from. At inference time, AnalysisBank matches input signals to library entries and applies the retrieved moves to compose the report. A study of Analyses distilled from 550 expert reports reveals a heavy-tailed distribution of 47-52 signal types spanning 13 move types. On two financial benchmarks across four LLM backbones, AnalysisBank increases the proportion of novel, data-grounded insights by 1.7-3.7x over structural-level baselines. Transfer to scientific writing suggests that the distinction generalizes beyond finance. Code and the distilled Analysis library are available at this https URL.
[AI-72] One Policy Any Budget: Internalizing Budget-Aware Search via Reinforcement Learning
链接: https://arxiv.org/abs/2609.00813
作者: Xiaowei Sun,Jin Li,Yili Hong,Yikun Fu,Yanghua Xiao
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:While reinforcement learning has enabled LLM-based search agents to invoke external tools, existing methods train under fixed budgets and cannot adapt when constraints vary at deployment. We propose AnySearch, a framework that enables a single policy to perform budget-aware search under any budget constraint through a training scaffold and curriculum reinforcement learning. In the first phase, we train the agent with explicit budget state injection and structured reasoning prompts that guide efficient allocation under linearly decaying budgets. In the second phase, the scaffold is removed and the agent learns to operate autonomously under adaptively sampled budget constraints, matching inference conditions. Both phases are optimized with a composite reward that couples answer accuracy with budget efficiency through absolute and relative signals, where an adaptive weight amplifies the efficiency signal for high-accuracy queries and attenuates it for low-accuracy ones. Extensive experiments on seven general and multi-hop QA benchmarks show that our method outperforms baselines across all budget scales, generalizes to unseen constraints beyond the training range, and achieves superior tool productivity without excessive token overhead. Our code is available at this https URL.
[AI-73] owards a Reliable and Practical Eval Pipeline
链接: https://arxiv.org/abs/2609.00805
作者: Emma Thuong Nguyen,Abhishek Ghose
类目: Artificial Intelligence (cs.AI); Software Engineering (cs.SE)
备注:
Abstract:LLM-based software systems increasingly require effective “evals” as quality gates in the development lifecycle. However, existing work typically addresses individual aspects of eval reliability rather than the full set of practical requirements. We present an end-to-end eval pipeline that combines eval checklist creation, with learned aggregation for checklist responses, to improve agreement across LLM judges and accuracy against human judgments. The framework additionally pro- vides self-consistency, explanations, and prediction uncertainty, and we empirically demonstrate its effectiveness.
[AI-74] MADS: A Multiview Acoustic Descriptor Set Beyond Standard Spectral Summaries
链接: https://arxiv.org/abs/2609.00792
作者: Utsab Ghosh,Roshni Chakraborty
类目: ound (cs.SD); Artificial Intelligence (cs.AI); Audio and Speech Processing (eess.AS)
备注:
Abstract:Dominant audio classification pipelines rely either on compact handcrafted summaries or on fixed time-frequency frontends such as log-mel representations prior to deep modeling. While highly successful, these representations do not explicitly expose the physical dynamics of the underlying sound-generating event. We introduce MADS (Multi-view Acoustic Descriptor Set), a compact 19-dimensional physics-informed descriptor set de- signed to capture complementary spectral, temporal, mechanical, and stochastic structure in audio signals. Rather than treating sound only as a spectral pattern, MADS encodes properties related to excitation, damping, periodicity, impulsiveness, and structural consistency within a unified multi-view representation. We evaluate MADS using standard classical machine learning models on ESC-10, ESC-50, and MSoS, and compare it against two conventional handcrafted baselines: a compact 26D MFCC- based baseline and an expanded 38D spectral-summary baseline. Across ESC-10 and ESC-50, MADS achieves the strongest peak results overall, reaching 81.00% and 52.78%, respectively, while using roughly half the dimensionality of the 38D baseline. On MSoS, MADS again delivers the strongest top-end performance, reaching 67.48%. These results establish MADS not merely as a competitive standalone descriptor set, but as the foundational descriptor layer of a broader acoustically grounded representation program for future frame-level and deep-learning-compatible audio modeling.
[AI-75] StudyBench: Can Self-Evolution Squeeze Textbooks for Olympiad Capability? EMNLP2026
链接: https://arxiv.org/abs/2609.00787
作者: Yinghao Chen,Zixi Chen,Bingxiang He,Ziqing Qiao,Huan-ang Gao,Yinuo Xu,Yuxin Zuo,Zeyuan Liu,Yuhao Zhan,Chaojun Xiao
类目: Artificial Intelligence (cs.AI)
备注: 9 pages, EMNLP 2026 Findings
Abstract:Humans need to study only a handful of well-written textbooks to master a discipline and attempt its hardest problems. We argue that an ideal self-evolution method should share the same property, that is autonomously learning from raw training material for transferable problem-solving capability. However, we still lack a direct measurement for it. We introduce StudyBench, a controlled physics benchmark that directly measures how efficiently a self-evolution method converts training material into capability. We organise the test set into an Application Set, consisting of difficult textbook problems and evaluating absorption ability, and a Transfer Set, consisting of olympiad-level problems and evaluating transfer ability. Benchmarking representative self-evolution methods across three base models, we find that improvements on the Application Set rarely translate to the harder Transfer Set. A guidance ablation exposes a Guidance Gap: even the strongest method closes only a small fraction of what the same material unlocks when supplied as in-context guidance. Besides, every method hits a Compute Plateau, saturating well before exhausting its compute budget. The remaining gap is therefore a method problem rather than a data or compute problem. By offering a clean and controlled benchmark, StudyBench turns self-evolution progress from an open-ended pursuit into a measurable target for future research. Our code is released at this https URL.
[AI-76] DiagEvo: Diagnosis-Guided Self-Evolution via Hierarchical Error Memory
链接: https://arxiv.org/abs/2609.00768
作者: Xincheng Wei,Yifan Ding,Yoshua Li,Dongsheng Ma,Rongxiang Weng,Xunliang Cai,Wenjian Ding,Yao Zhang
类目: Artificial Intelligence (cs.AI)
备注: 23 pages, 6 figures, 9 tables
Abstract:Self-play is an effective paradigm for language-model self-evolution, but without guidance, solver performance can plateau or decline across rounds. Unguided methods steer question generation with signals such as difficulty, learnability, or diversity. These signals keep questions challenging and varied but do not specify which unresolved reasoning weaknesses later rounds should target. Guided methods obtain direction from external task resources, including human examples, document corpora, or specified difficulty targets, and therefore rely on task information supplied outside the self-play loop. We show that the needed direction can instead be derived from the solver’s own failure history. We introduce DiagEvo, whose diagnostician extracts recurring error causes from this history and stores them in a hierarchical error-cause memory. The memory groups related causes under skill nodes and tracks each as Active or Mastered according to self-consistency on targeted questions. The challenger uses these states and recurrence counts to balance cause-targeted generation with free exploration. Double-confidence filtering retains intermediate-difficulty questions only when the most common solver answer has a clear vote lead. DiagEvo derives its curriculum from information produced during self-play, without external task resources. With the default 4B diagnostician, DiagEvo outperforms every baseline in mean accuracy across all nine benchmarks for each of the three solvers: Qwen3-4B, Qwen3-8B, and OctoThinker-8B. On Qwen3-8B, it reaches 72.3% mean accuracy across five mathematical reasoning benchmarks, 4.5 percentage points above R-Zero. Its mean accuracy across all nine benchmarks is 57.4%, 1.1 percentage points above DARC. Ablations show that the hierarchical error-cause memory and double-confidence filtering both contribute to these gains.
[AI-77] Are You Thinking What I am Thinking? : Examining Conceptual Separation in Neural Architectures
链接: https://arxiv.org/abs/2609.00764
作者: Jaee Ponde,Roshni Agarwal,Subhashis Banerjee
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Neural networks are increasingly employed to identify both well-defined and ambiguous concepts, yet output-level metrics reveal little about how those concepts are represented internally. Our study asks if these networks exhibit \textitconceptual separation: if examples of the same concept form coherent representations, and whether related concepts lie closer together in the representation space. We examine this conceptual organisation in Convolutional Neural Networks (CNNs) and Large Language Models (LLMs) through geometric and distributional analysis of their internal activations. In CNNs, familiar ImageNet concepts form coherent and semantically ordered representations, while this coherence weakens for unseen concepts and suffers within-class domain shift. In LLMs, clearly distinct domains remain well separated, related subdomains move closer together, and the distinction between ambiguous topics collapses at both the mean and covariance level. These results suggest that conceptual separation can reveal structure that output accuracy alone cannot, and may serve as a useful diagnostic of how robustly a model represents the concepts it is asked to identify. Code and data available on \hrefthis https URLGitHub.
[AI-78] Automated Tree Knowledge Graph Construction using Ontology Expansion and Retrieval from Vietnamese History Textbooks
链接: https://arxiv.org/abs/2609.00763
作者: Ket Doan Nguyen,Minh N. H. Nguyen
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Hierarchical Knowledge graph (KG)-based retrieval augmented generation (RAG) has emerged as a powerful approach for supporting large language models with structured knowledge. However, there are primary challenges: (i) the lack of methods for automatic KG construction using ontology expansion for low-resource languages such as Vietnamese, (ii) the absence of systematic evaluation for knowledge retrieval strategies leveraging the hierarchical structures. In this paper, we propose an end-to-end pipeline for KG construction and retrieval strategies evaluation. In the KG construction, we employ a three-phase hybrid relation extraction pipeline: intra-batch deduplication via Union-Find, approximate cross-batch search, and LLM extraction with a centroid filter that reduces prompts combined with a five-step dual-LLM validator to prevent bloated ontology. A two-tier architecture consists of unmergeable structural nodes to preserve the document structure and mergeable content nodes. The retrieval evaluation consists of three graph traversal strategies: Top-Down, Horizontal, and Bottom-Up, which are evaluated on a synthetically generated benchmark of 1,210 Vietnamese queries from 109 subgraphs, categorized by five query directions. In this paper, we construct the tree knowledge graph from Vietnamese high school History textbooks (nearly 400 pages) to produce 750 nodes and 4,341 semantic edges with controlled ontology growth from 40 to 41 types. Among experimental graph traversal strategies, the Top-Down strategy with structure surpasses the vector baseline by 4.7 percentage points in NDCG@10. As a result, tree-structural information provides valuable information beyond flat cosine similarity but degrades performance when the query does not require structural context.
[AI-79] S3martCirc: Self-supervised Smart Circuit Discovery
链接: https://arxiv.org/abs/2609.00755
作者: Wendy Zheng,Yinhan He,Liang Wu,Jundong Li
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large Language Models (LLMs) have demonstrated remarkable performance across diverse tasks, from text summarization to question answering. Despite these capabilities, their black-box nature obscures internal decision-making processes. Mechanistic interpretability (MI) aims to address this by reverse-engineering neural networks into human-understandable algorithms. Current MI approaches for LLMs typically follow a two-stage paradigm: first identifying important components (circuit discovery), where components are typically individual nodes such as an attention head or feedforward neuron, and second determining the role they play in a certain task (functional interpretation). However, this sequential approach overlooks a fundamental insight: a component’s importance and its functional role are inherently codependent. Unifying these stages presents two key challenges: (1) functional roles are often tied to specific nodes or components, limiting generalization, and (2) their identification relies on subjective interpretation rather than quantifiable metrics. To address these challenges, we propose S^3martCirc (Self-supervised Smart Circuit Discovery), a unified framework that simultaneously discovers circuits and interprets functionality. S^3martCirc abstracts node behavior into two general computational roles that generalize across tasks and defines a quantitative metric for assigning them, enabling importance and functional role to be discovered jointly rather than in sequence. Extensive experiments show that our framework outperforms existing methods in circuit discovery.
[AI-80] ContextPipe: Database-Inspired Context Assembly for Long-Horizon Agents
链接: https://arxiv.org/abs/2609.00749
作者: Peng Xu,Zuyu Zhang,Yuze Sun,Feng Tian,Long Wang,Chen Zhang
类目: Artificial Intelligence (cs.AI); Databases (cs.DB)
备注:
Abstract:Long-horizon large language model (LLM) agents require context assembly: the runtime must decide what to include in each prompt, in what order, and when to compact history under a hard context-window budget and a byte-sensitive prompt cache. In production agentic systems, this logic is scattered across prompt builders, ad hoc compaction routines, cache-break workarounds, and per-provider shims. We argue that context assembly is structurally isomorphic to query execution in a relational database: both execute under a hard budget, exploit a tiered cache, and leverage statistics. We adopt this discipline in ContextPipe: a five-phase pipeline (Plan Bind Optimize Execute Feedback) backed by a structured data-source catalog, a deterministic cache-aware optimizer, and an EXPLAIN ANALYZE trace. We show that context in ContextPipe is auditable, replayable, and failure-isolated. A preliminary evaluation using the SWE-bench Pro Qutebrowser subset shows that, compared with the append-only context construction policy, ContextPipe reduces total token volume by 31%, LLM calls by 23%, and response time by 9%, at the cost of a lower KV cache-hit ratio.
[AI-81] Escaping Redundant Reasoning : Structure-Aware Search for Inference-Time LLM s
链接: https://arxiv.org/abs/2609.00738
作者: Lu Cheng
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Inference-time search with large language models (LLMs) often concentrates on a small set of structurally or semantically similar trajectories, leaving alternatives underexplored—a failure mode we call \textitreasoning basin collapse. We introduce BASIN, a training-free, structure-aware selection method that groups reasoning states into basins and penalizes repeated visits to the same strategy, thereby reallocating search across genuinely distinct reasoning paths under a fixed compute budget. Under matched inference budgets, BASIN improves over Tree of Thoughts (ToT) by up to +22 pp on Game of 24 and +6.7 pp on MuSR. A quality-aware variant, QA-BASIN, further improves robustness by preserving high-quality basins when unconditional diversification over-explores. To explain when basin-aware selection helps, we introduce the redundancy gap \Delta , which measures how differently search concentrates for correct versus incorrect predictions: standard ToT often operates near \Delta \approx 0 , while BASIN consistently shifts \Delta positive. More broadly, BASIN suggests structure-aware selection as a simple and general approach to improving inference-time reasoning. Code can be found at this https URL.
[AI-82] Agent ic Empirical Asset Pricing: Methodological Foundations
链接: https://arxiv.org/abs/2609.00731
作者: Yingjian Pan,Xiaowei Ding,Kay Giesecke
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Statistical Finance (q-fin.ST)
备注: 26 pages, 5 figures, 12 tables
Abstract:Recent advances in LLM agents enable a new paradigm for asset pricing, which we call Agentic Empirical Asset Pricing (AEAP): systems that autonomously conduct the scientific discovery process itself. We define AEAP and identify its core building blocks. Existing evaluation practices backtest only the outputs (factors or trades), not the autonomous discovery system that produced them. We focus on factor discovery, contributing a reference architecture, a rigorous evaluation standard for discovered factors, and a method for out-of-sample backtesting the discovery system. As a concrete instance of that architecture, we evaluate SEADS against five re-implemented baselines on two US equity panels using this standard: no single metric ranks the systems consistently, motivating evaluation on multiple axes at once. A separate rolling re-execution then asks the complementary question of whether the discovery process itself, not one static output, is reliable. We also report negative findings and limitations that surface further evaluation pitfalls for future AEAP systems.
[AI-83] SOVER: Formal Certification of Optimization Reformulations via LLM -Assisted SMT Verification EMNLP2026
链接: https://arxiv.org/abs/2609.00728
作者: Swapnil Bhattacharyya,Mayank Baranwal
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Optimization and Control (math.OC)
备注: Accepted to EMNLP 2026 Findings
Abstract:Large Language Models (LLMs) have shown remarkable promise in translating and reformulating complex mathematical optimization problems across modeling languages. However, validating such transformations through empirical solver executions alone is unreliable, as solver outcomes may be affected by local minima, structural timeouts, numerical artifacts, and subtle semantic divergence between formulations. We introduce SOVER, an LLM-assisted SMT framework that separates semantic mapping from formal certification: Z3 checks domain cross-feasibility and global objective-order preservation for mixed-integer linear formulations, while dReal provides tolerance-aware feasibility/range and \epsilon -argmin checks for continuous nonlinear formulations. We also introduce NLEquiv-150, a public benchmark of 100 equivalent and 50 deliberately hard non-equivalent nonlinear reformulation pairs. With LLM-extracted mappings, SOVER classifies 149/150 pairs (99.33%) correctly, including all 50 hard negatives; the sole error is an incomplete mapping extraction.
[AI-84] Heard but Not Heeded: Paralinguistic Information Encoding and Loss in Audio-Language Models
链接: https://arxiv.org/abs/2609.00727
作者: Bhuvan Koduru,Dareen Safar B Alharthi,Rita Singh,Bhiksha Raj
类目: ound (cs.SD); Artificial Intelligence (cs.AI)
备注:
Abstract:Audio language models are designed to understand speech, yet it remains unclear whether they capture how something is said beyond what is said. We present a mechanistic analysis of paralinguistic information in four open source models, Whisper-large-v2, Qwen2-Audio-7B Instruct, Qwen2.5-Omni-7B, and Chroma-4B, using the Expresso dataset with controlled speaking styles. We combine centered kernel alignment, linear probing with leave one speaker out evaluation, open ended tone prediction, and a content prosody leakage metric to trace how style information moves from the audio encoder to the final output. All models strongly encode speaking style in the late encoder, that is, the top third of the audio encoder’s layers, but this information is consistently degraded before reaching the output. The projector reshapes representation geometry without removing information, while decoders differ in how much style they preserve depending on architecture and training objective. At the output level, models fall into two behaviors. Some are content driven, where predictions depend mainly on text. Others are acoustic driven, where predictions vary with speaking style. The leakage metric quantifies this difference, and qualitative results confirm it. Overall, we identify a gap between what models encode and what they use, highlighting a key limitation in current audio language models.
[AI-85] A Study of Hidden-State Optimization Order in Predictive Coding Networks
链接: https://arxiv.org/abs/2609.00686
作者: Xueyuan Li,Danilo Vasconcellos Vargas
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Local learning methods offer an alternative to end-to-end backpropagation, but their unstructured local objectives can produce weak feature learning in deep networks. We study whether the order of hidden-state optimization can address this limitation. We propose a boundary-first inference schedule that partitions a model into chunks, first coordinates hidden states at chunk boundaries, and then refines representations within each chunk. We instantiate this schedule in predictive coding networks (PCNs), a local-learning framework in which hidden activities and prediction errors are explicitly exposed during inference. On CIFAR-10, the resulting boundary-first predictive-coding instantiation improves accuracy over standard predictive coding by 9.77% under a standard parametrization and by 5.51% under a \mu -parametrization. Diagnostic analyses further show more non-trivial early-layer updates, lower initial-to-final CKA, and more diverse layerwise gradients, consistent with stronger feature learning. These results support boundary-first, chunk-based inference as a practical design principle for predictive-coding training and motivate its study in broader local-learning systems.
[AI-86] riple-Bottom-Line Sustainability of Language Models for Edge AI: A Comparison Between SLMs and Quantized LLM s
链接: https://arxiv.org/abs/2609.00665
作者: Jainil Dharmil Shah
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Edge-AI model selection is commonly driven by one isolated metric - accuracy, latency, memory, energy, or safety, even though a deployable language model must balance all five. Our work focuses on answering the question whether na- tively trained small language models (SLMs) or large language models (LLMs) compressed through post-training quantization offer the more sustainable edge- deployment trade-off. We introduce a reproducible Holistic Sustainability Score (HSS) organized around the triple bottom line: an economic pillar for capability and systems efficiency, an environmental pillar for operational GPU energy and a social pillar for harmful-prompt robustness. Five BF16 SLMs and five LLMs under different quantization approaches - BF16, INT8, NF4 4-bit, GPTQ 4-bit, and GGUF Q4 produce 30 measured configurations. Capability is assessed on five zero-shot benchmarks; efficiency uses latency, throughput, peak VRAM and energy; and safety is approximated by attack success rate on five harmful prompts. Qwen3-30B-A3B/GGUF Q4 ranks first in the combined pool (93.38), followed by Mistral-Small-24B/GGUF Q4 (92.40), while Phi-4-mini/BF16 is the highest- ranked SLM in that pool (89.49). Thus, the hypothesis that native SLMs must be the most sustainable edge choice is not supported universally; optimized quantized LLMs can win overall, while SLMs remain competitive through lower resource demand. Quantization is a systems-level choice rather than a monotonic precision- efficiency trade-off and HSS remains relative to its comparison pool and proxy definitions.
[AI-87] Drift-Aware LLM Routing with Sparse Contexts and Shared Budgets
链接: https://arxiv.org/abs/2609.00662
作者: Cheung Hao Lee,Patrick Wong
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:A multi-model language service must route each request while preserving workload-level budgets for compute, latency, memory, or monetary cost. Two features make this problem materially harder than static model selection. Prompt representations are high dimensional, so only a small subset of embedding directions may predict the incremental value of a model, and both the request mix and the model frontier drift after launches, fine-tunes, quantization changes, and system updates. We formulate nonstationary sparse contextual routing with multiple knapsack constraints and an optional shadow-audit stream that evaluates a small fraction of prompts on several models. We propose Drift-Aware Sparse Routing (DRS). The policy estimates reward and resource use from a rolling audit window, routes using pessimistic reward and optimistic cost estimates, updates resource shadow prices online, and applies a hard meter before commitment. The analysis separates control from statistics. On any event with uniform prediction radii \beta_t\ , regret against a paced dynamic fluid benchmark is bounded by the sum of the radii, a capacity-buffer term, and an O(\sqrtT) pacing term. Under a sparse linear model and bounded drift V_T , rolling estimation gives [ \widetilde O\left( T\sqrt\fracs\rho W+WV_T+\sqrtT \right), ] where s is sparsity, \rho is the audit rate, and W is the window length. Optimizing W yields the usual stationary O(\sqrtsT/\rho) rate when V_T=0 and a O(T^2/3(s/\rho)^1/3V_T^1/3) adaptation term under drift. Subjects: Artificial Intelligence (cs.AI) Cite as: arXiv:2609.00662 [cs.AI] (or arXiv:2609.00662v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2609.00662 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-88] EEG-AS: Instance-Level Foundation Model Selection for EEG Foundation Models via Behavior Reconstruction
链接: https://arxiv.org/abs/2609.00653
作者: Yunzhen Zhang,Ruoxi Piao,Hasan Onur Keles,Mustafa Misir
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Electroencephalography (EEG) is a non-invasive technique for measuring neural activity and has been widely used in neuroscience applications. Recent advances in EEG foundation models have enabled strong performance across diverse neural decoding tasks. However, no single foundation model consistently performs best across datasets or individual EEG instances, while instance-level model selection remains largely unexplored. To address this limitation, we formulate EEG foundation model selection as an instance-level Algorithm Selection (AS) problem. We propose \textbfEEG-AS, an instance-level algorithm selection framework that characterizes each EEG instance using inference-available latent EEG embeddings, handcrafted neurophysiological features, and an anchor foundation model. During training, EEG-AS learns to reconstruct unavailable foundation-model behaviors from privileged prediction tokens conditioned on an anchor foundation model, while during inference it estimates these behaviors without executing the entire model portfolio, enabling efficient selection from seven EEG foundation models. Experiments on seven public EEG benchmarks demonstrate that EEG-AS substantially narrows the gap between the Single Best Solver (SBS) and the oracle upper bound for each instance. These results highlight the effectiveness of instance-level AS for adaptive deployment of EEG foundation models.
[AI-89] Self-Reports Are Not Verification: Environment-Grounded Auditing of LLM Operators in Evolutionary Search
链接: https://arxiv.org/abs/2609.00652
作者: Enrong Pan,Ryan Zhou,Ting Hu
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Neural and Evolutionary Computing (cs.NE)
备注:
Abstract:Language model agents increasingly propose actions, observe external feedback, and explain their own behavior. Their confidence and rationales are convenient monitoring signals, but convenience is not verification. We introduce an environment-grounded audit in which every intermediate proposal receives an exact outcome. A language model operates an evolutionary Contexto search whose feedback function assigns every valid guess an exact rank without human annotation. Across 200 runs spanning five configurations and three model families, four reporting configurations produce 12,249 self-reports. We test three assumptions: stated confidence is calibrated, inherited rationales affect later proposals, and fitness-based selection improves report quality. All three fail. Operators overstate top-100 success by factors of 4.8 to 9.3, while calibration and discrimination dissociate across model families. Controlled interventions on 754 inherited rationales bound any measured benefit of the genuine rationale to roughly 250 ranks. Neither fitness-based nor random selection produces a detectable selection differential or parent-to-offspring transmission in report accuracy, despite sharply different search behavior. Agent self-reports should therefore be treated as claims to verify against the environment, not as evidence of their own reliability.
[AI-90] DramaChain Bench: An End-to-End Benchmark for Short-Drama Generation
链接: https://arxiv.org/abs/2609.00646
作者: Haoyuan Shi(1),Mingtao Chen(1),Shuo Jiang(1),Ziyan Chen(1 and 2),Xuyi Sheng(3),Yiming Liu(1),Ying Zhang(1),Miao Wang(1 and 4),Jianxiang Lu(1),Fanyang Lu(1),Songyuanyi Lu(1),Xiele Wu(1),Zhichao Hu(1),Yuhong Liu(1),Richeng Xuan(1) ((1) Hunyuan, Tencent, (2) Beijing Film Academy, (3) Peking University, (4) Shenzhen University)
类目: Artificial Intelligence (cs.AI)
备注: 50 pages, 19 figures, 19 tables. Technical report. Haoyuan Shi and Mingtao Chen contributed equally. Project lead: Zhichao Hu. Corresponding author: Richeng Xuan
Abstract:Commercial short-drama production follows a multi-stage chain: script, storyboard, keyframe imagery, shot-level video, and the finished short drama. Most existing benchmarks evaluate solely the video-generation stage using pre-authored inputs instead of real upstream pipeline outputs. This leaves two critical questions unanswerable: whether each stage adheres to the original script intent (rather than only its immediate input prompt), and whether disparate shots remain coherent after assembly into multi-episode releases. We present DramaChain Bench, the first short-drama benchmark that evaluates every stage of the complete production chain. It is built upon three in-house systems sharing one dimension system, DramaChain Dimensions: five evaluation axes instantiated at every stage, resolving into 63 leaf dimensions. DramaChain Agent is calibrated against commercial short-drama platforms in both workflow and finished short-drama quality, enabling stage-wise fair comparison across models. DramaChain Labeling System has each of the 5,785 items scored independently by three professional annotators, with all defects spatio-temporally localised and selected from a predefined defect list. This process produces 17,488 valid scores and 255,925 traceable attribution records. The human annotations confirm that upstream defects cascade across the pipeline, demonstrating that final episode quality is not governed by video generation alone. DramaChain Agentic Judge then scores every leaf dimension automatically, gathering evidence over multiple agentic rounds before judging against a per-item checklist; it reproduces the model ranking at a mean PLCC of 0.918, enough to admit new models at no annotation cost.
[AI-91] REVISE: Validity-Guided Recovery for Online Revisions in Agent Workflows
链接: https://arxiv.org/abs/2609.00643
作者: Ruoling Qi,Xuaner Wu,Penghang Liu,Jian Chen,Yirui Liu
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Agent revisions expose a fundamental correctness–efficiency trade-off during concurrent execution. Discarding ongoing work preserves latest-version correctness but wastes progress that may remain valid, whereas reusing prior work preserves efficiency but risks propagating stale state into outputs and tool effects. Existing recovery strategies resolve this trade-off in an imbalanced way with coarse-grained policies: they either favor efficiency by allowing potentially stale work to continue, or favor correctness by restarting the workflow or recomputing a linear suffix from the earliest conflict, thereby discarding unaffected progress. We present \textscRevise, a validity-guided runtime for fine-grained recovery in structured agent workflows. When a revision arrives, \textscRevise first intersects its delta with recorded data and control dependencies and propagates the resulting impact through the partially executed DAG to identify affected work. It then stops invalid work, preserves validity-established progress beyond the earliest conflict, and recomputes only the affected region. Incomplete provenance conservatively expands recovery, while reused results are revalidated before commit. Analysis of real coding-agent traces show online recovery opportunities: 118 sessions retain observable work before a queued later message is delivered; across 167 overlapping assistant responses, enqueue-to-completion overlap reaches 56.55~s at p95. Across 300 challenging revision/commit executions, \textscRevise matches a latest-version oracle with no stale outputs or effects. On unmodified LangGraph and LLMCompiler applications using Qwen3-14B, it reduces model calls by 40.6–56.0% relative to full restart and by 31.3–43.6% relative to suffix recomputation. Under serving pressure, it further reduces revision-to-correct-completion tokens by 13.26% and improves SLO goodput by 3.07–5.43%.
[AI-92] UTTI: Toward generalizable audio-to-score transcription via fully synthesized data
链接: https://arxiv.org/abs/2609.00640
作者: Jianhuai Hu,Yashan Wang,Shangda Wu,Zhancheng Guo,Shijie Liang,Wuna Meng,Chuanqi Yang,Xiaobing Li,Feng Yu,Maosong Sun
类目: ound (cs.SD); Artificial Intelligence (cs.AI)
备注:
Abstract:Generalizable Audio-to-Score (A2S) transcription is fundamentally constrained by the severe scarcity of high-quality, real-world paired data. Relying solely on existing human-annotated datasets often restricts the generalization of A2S models, limiting their efficacy primarily to single-instrumentation domains. To break this dependency on scarce real-world data, we introduce TUTTI (Transformer for Unified audio-To-score Transcription trained on Synthetic multi-Instrumentation Data), a pre-training paradigm driven by a purely synthetic, large-scale dataset. Rather than using human-composed scores, we leverage a symbolic music generation model to generate a massive, highly scalable multi-instrumentation corpus and create audio-score pairs with expressive acoustic characteristics. Capitalizing on the generated data, we employ a standard Transformer encoder-decoder architecture. We empirically demonstrate that pre-training a unified attention-based model on generated, multi-instrumentation data yields a consistently stronger foundational representation than single-instrumentation training. When fine-tuned with downstream real-world datasets, TUTTI outperforms previous approaches, establishing new overall state-of-the-art results across various A2S baselines. Notably, TUTTI shows remarkable cross-instrument transferability, effectively adapting to unseen instruments with highly competitive performance. The source code and the TuttiCorpus dataset will be made publicly available at this https URL.
[AI-93] Breaking the Structural Identity: Personalized Federated LoRA Fine-tuning under Rank Heterogeneity EMNLP2026
链接: https://arxiv.org/abs/2609.00632
作者: Lei Wang,Jieming Bian,Letian Zhang,Jie Xu
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Accepted to EMNLP 2026
Abstract:Large Language Models (LLMs) have achieved remarkable success across diverse domains, but their adaptation to privacy-sensitive, distributed datasets remains a challenge. While Federated Learning (FL) combined with Low-Rank Adaptation (LoRA) provides a resource-efficient paradigm for collaborative fine-tuning, practical deployments are hindered by the dual challenges of resource heterogeneity and data heterogeneity. Existing rank-heterogeneous methods primarily focus on bridging dimension mismatches for aggregation but typically provide a unified global model for all clients sharing the same rank, failing to capture client-specific features in non-IID scenarios. In this paper, we propose FedRoRA (Federated Rank-wise Personalized LoRA), a novel framework that enables fine-grained personalization within rank-heterogeneous federations. FedRoRA decouples adaptation into shared global directions and personalized rank-wise magnitudes governed by learnable diagonal scales. On the server side, it extracts a global subspace via singular value decomposition (SVD) and redistributes client-specific initializations through a personalized projection and top- k selection mechanism. Extensive experiments on NLU and NLG benchmarks demonstrate that FedRoRA consistently outperforms state-of-the-art methods.
[AI-94] SoK: When Safe Agents Fail Together: The Security of Multi Agent LLM Systems
链接: https://arxiv.org/abs/2609.00595
作者: Rui Yang,Junjie Xu,Zhengyu Liu,Neil Fendley,Yang Hong,Ziyang Li,Yinzhi Cao
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注: 21 pages
Abstract:Safe agents can fail together. Multi-agent LLM systems (MAS) move information, state, decisions, and authority across principal boundaries, creating failures that local checks may miss. Without an execution-level view, a multi-agent setting can easily be mistaken for evidence of a genuinely multi-agent security effect. We thus systematize MAS security through an execution-centered analysis of 197 works, covering six interaction interfaces, four adversary positions, seven system-level risks, and eight recurring attack paths. We introduce an A-I-R framework that organizes attacks by adversary position, interaction interface, and resulting system-level risk, unifying otherwise fragmented attack mechanisms across MAS. We organize defenses through a five-part contract covering path target, observation, intervention, trust boundary, and recovery, and identify path closure and recovery as key challenges. We audit 44 evaluation and benchmark works and identify open challenges in isolating interaction effects, designing comparable and diagnostic metrics, supporting reuse across MAS designs, and evaluating open-system operation. Together, these findings motivate an interaction-aware view of MAS security: trace attacks end to end, test whether defenses close those paths, and evaluate system-level effects with appropriate counterfactuals.
[AI-95] Same Request Different Boundary: Evaluating Cybersecurity Assistance across Conversational Contexts
链接: https://arxiv.org/abs/2609.00578
作者: Rui Yang,Yang Hong,Yichao Xu,Zhengyu Liu,Ziyang Li,Yinzhi Cao
类目: Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR)
备注: 14 pages, 7 figures
Abstract:Large Language Models (LLMs) can solve complex problems, but their misuse in high-risk domains can lead to severe consequences. Model providers therefore restrict assistance for potentially harmful requests. Refusing all cybersecurity requests would therefore harm legitimate users. Providers need a mechanism to block malicious use without denying legitimate assistance to defenders. Existing cybersecurity-specific datasets evaluate this mechanism, but none considers the conversational context of a request. We introduce 3R-Bench (Refusal, Repetition, and Revision), a benchmark of 150 real-world cybersecurity requests augmented with two adversarial conversational settings, and evaluate eight LLMs on it. Prior assistant behavior strongly changes responses to an unchanged request: among 376 available pairs from a 400-pair panel, compliance rises from 62.0% after refused history to 85.1% after accepted history. The opposite pattern appears under dialogue decomposition. In comparison, compliance falls from 501/800 direct responses to 172/800 after dialogue; among 738 pairs returning model-authored text in both conditions, the decrease is 45.1 points. Failure feedback recovers only a small fraction of this loss.
[AI-96] GeoPAR: Large-Scale Multi-Agent Combinatorial Optimization with Geometry-Guided Parallel Autoregressive Learning
链接: https://arxiv.org/abs/2609.00577
作者: Wenjian Wu,Zesheng Jia,Jiaying Tang,Benyuan Yang,Jin Wang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Neural and Evolutionary Computing (cs.NE)
备注:
Abstract:Multi-agent combinatorial optimization problems are notoriously challenging due to their NP-hard nature. Recent parallel autoregressive neural solvers improve inference efficiency by allowing agents to make decisions simultaneously, but their performance often degrades on large-scale instances. This is largely attributable to weak modeling of local geometric structures and the fact that conflicting task selections are handled only after action generation. To address these limitations, we propose GeoPAR, a geometry-guided parallel autoregressive reinforcement learning framework for scalable multi-agent combinatorial optimization. GeoPAR integrates three key components: (1) a projection-window sparse geometry mechanism that builds lightweight local candidate neighborhoods through multi-directional projections, (2) sparse edge-biased attention that injects these geometric relations into node representations, and (3) cache-guided conflict-aware assignment that reuses the geometric cache during decoding to suppress duplicate selections of exclusive tasks. Experiments on heterogeneous vehicle routing and open multi-depot pickup-and-delivery problems show that GeoPAR improves large-scale zero-shot generalization while substantially reducing rollout steps and maintaining efficient inference.
[AI-97] Residual Sparsification via Output Importance for Compressing Mixture-of-Experts LLM s EMNLP2026
链接: https://arxiv.org/abs/2609.00575
作者: Seungwoo Jung,Dohyeok Kwon,Seungmin Cha,Junseok Lee,Yeonho Yoo,Chuck Yoo,Gyeongsik Yang
类目: Artificial Intelligence (cs.AI)
备注: Accepted to EMNLP 2026 (Main Conference)
Abstract:Mixture-of-experts (MoE) architectures scale large language models efficiently, but they demand massive GPU memory. To cope with such demand, models are commonly compressed to reduce their memory footprint. Residual sparsification is a representative compression technique that decomposes each projection matrix of an expert into a shared base matrix and per-expert residual matrix, and then compresses the residuals. Existing sparsification methods compress each residual matrix independently by minimizing its compression error, thereby minimizing the error of each projection matrix. However, our analysis shows that this objective is misaligned with preserving model accuracy after compression. In an expert, the final output is produced through computations coupled across multiple projections and hidden representations. Therefore, even small errors in individual matrices can propagate through hidden representations and projection interactions, leading to large expert output errors and accuracy degradation. To address this misalignment, we propose PARSER, a new residual sparsification method that shifts the compression objective from minimizing isolated matrix errors to preserving the expert output error. PARSER achieves this by introducing output importance, which measures the actual contribution to the expert output error. Our experiments show that, compared with existing methods, PARSER narrows the accuracy gap to the uncompressed model by 1.41 \times on Qwen and 1.44 \times on DeepSeek, while achieving the same peak memory reduction.
[AI-98] A Mathematical Framework for Legacy Governance and Decision Integrity in Enterprise AI
链接: https://arxiv.org/abs/2609.00572
作者: Shorab Sarker
类目: Computers and Society (cs.CY); Artificial Intelligence (cs.AI)
备注: 23 pages, 6 figures. Includes a reproducible synthetic computational demonstration with source code and generated data
Abstract:Enterprise artificial intelligence is increasingly embedded in decisions that must remain lawful, explainable, adaptable, and accountable despite personnel turnover, model replacement, regulatory change, and shifting organizational incentives. Existing governance frameworks provide important principles but do not by themselves supply a compact mathematical language for evaluating whether an institution can preserve sound judgment over time. This paper develops a design-science framework for institutional legacy: the durable capacity of a decision system to continue producing beneficial, lawful, explainable, and adaptable outcomes after its original designers have stepped away. The framework contributes: (i) a normalized Legacy Score based on a penalized geometric mean of knowledge retention, governance, human oversight, adaptability, feedback learning, and jurisdictional fidelity; (ii) Decision Confidence and Decision Risk models separating evidentiary confidence from consequence; (iii) authority-aware retrieval and calibrated abstention; (iv) Decision Memory for governed organizational learning; (v) Regulatory Change Velocity mapping change exposure to review intervals; and (vi) a federated regulatory knowledge-graph architecture preserving provenance and legal hierarchy. The paper also proposes eight AI Decision Integrity Rules, an evaluation protocol, and a reproducible computational demonstration. The demonstration combines a deterministic stress test with 200 Monte Carlo replications of 10,000 synthetic decisions each, illustrating Legacy Score non-compensation and comparing consequence- and authority-aware routing with a matched-coverage confidence-only baseline. The contribution remains conceptual rather than field-validated; the simulation tests internal behavior, not production performance, and all parameters require context-specific calibration.
[AI-99] VoiceLongMemEval: Do Assistants Remember How You Sounded?
链接: https://arxiv.org/abs/2609.00570
作者: Ramit Pahwa,Parivesh Priye,Apoorva Beedu
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:With the growing scale of multi-agent architectures and large language models, deployed AI assistants are increasingly tasked with reasoning over long, continuous, multi-session conversation histories. Current benchmarks evaluate this dialogue history as information retrieval over long horizon, temporal reasoning, or knowledge updates, while crucially ignoring the fundamental dynamics of human-agent interaction, i.e. how they said it. To address this gap, we present VoiceLongMemEval (VLME) benchmark, where every answer depends on paralinguistic metadata (emotion labels, prosody descriptors, and voice events) attached to conversational turns, which is otherwise unrecoverable from the words alone. Every item passes a three-stage adversarial gate, ensuring that a strong language model fails when given only the transcript. Evaluating leading frontier and open-weight models reveals a pervasive affect gap; providing text-track paralinguistic metadata yields a 0.09 to 0.38 accuracy boost (0.61 to 0.69 when prompted with evidence hints), while standard ASR pipelines systematically discard this signal. Additionally, audio-native models successfully extract these cues directly from speech (0.354 to 0.412 vs. 0.325 blind). Code and dataset will be made available upon acceptance.
[AI-100] WiseSpec: Requirements-Driven Agents for Code Generation
链接: https://arxiv.org/abs/2609.00568
作者: Zhao Tian
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注: Accepted by ASE 2026 (SRC)
Abstract:Code generation aims to automatically generate source code from task requirements and has attracted significant attention with the rapid advancement of large language models (LLMs). Despite remarkable progress, LLMs often struggle to generate correct code for complex software engineering tasks because task descriptions are frequently incomplete, ambiguous, or lack critical contextual information. Existing approaches primarily improve the capabilities of coding agents through more sophisticated tools, skills, and workflows, while largely overlooking the quality of the task requirements themselves. To address this limitation, we draw inspiration from software requirements engineering and propose WiseSpec, a novel requirements-driven agent framework for repository-level code generation. WiseSpec automatically constructs structured and information-rich requirements, assesses their quality through execution-based evaluation, and iteratively refines them to better guide code generation. Experimental results show that WiseSpec consistently outperforms all baselines, achieving an average improvement of 13.17% in %Resolved.
[AI-101] EEG-VID: Task-Guided Latent Predictive Pretraining for EEG Decoding and Assistive Target Selection
链接: https://arxiv.org/abs/2609.00566
作者: Guanzhong Sun,Junyi Ma,Yuxuan Wu,Yanzi Miao
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:We propose EEG-VID, a task-guided latent predictive pretraining framework for EEG decoding under session and subject shifts. EEG-VID predicts future latent EEG states from recent history using an exponential-moving-average target encoder and weak task guidance, followed by supervised fine-tuning. Across VIG-48 and BCI Competition IV-2a/IV-2b, Stage 1 improves mean accuracy in 41 of 42 matched backbone-dataset-protocol comparisons, including all 12 leave-one-subject-out settings, with a maximum gain of 16.22 percentage points. On the 48-region cross-day VIG-48 task, EEG-VID achieves 6.52% Top-1 and 30.50% Top-5 accuracy. In a separate six-participant offline robot-scene study, candidate-constrained target selection reaches 40.24% versus a 25% chance level after subject-specific calibration. These results support task-guided latent prediction as a transferable pretraining strategy for EEG decoding and scene-constrained assistive target selection.
[AI-102] Runtime-Independent Persistent Agents : Preserving Identity Memory and Code Across Models Harnesses and Servers
链接: https://arxiv.org/abs/2609.00546
作者: Zhenyu Zhao(1),Roy Zhao(2) ((1) Independent Researcher, (2) Paul G. Allen School of Computer Science amp; Engineering, University of Washington)
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注: 8 pages, 2 figures, 3 tables. Reference implementation and artifacts: this https URL
Abstract:Agent systems are commonly described by the model and harness that currently produce their behavior. That boundary is useful for one execution but underspecifies a long-lived agent that may change models, orchestration harnesses, interaction sessions, and host servers while retaining one identity, memory, and executable code lineage. We present a runtime-independent architecture for persistent agents. A continuity-bearing substrate P_t=(I_t,M_t,B_t) contains an architectural identity representation, private durable memory, and a versioned software body. A replaceable deployment binding comprises an execution substrate E_t=(R_t,H_t,D_t) , which supplies a reasoner, harness, and host, and a set of interaction surfaces S_t , such as chat, API, or user interface bindings. A deployed execution is A_t=P_t\triangleright(E_t,S_t) ; changing either replaceable layer is migration, not agent creation, when an authorized protocol preserves attributable lineage and transfers continuation authority within a governed deployment boundary. We define six continuity invariants and a quiesce–checkpoint–validate–bind–rehydrate–resume protocol. Enoch realizes the design as a reusable body plus private installed identity, memory, workflow state, and continuation authority, with infrastructure dependencies behind versioned provider contracts. A clean-room run of the frozen public commit passes 833 core tests and 92 provider and library tests executed separately from the core suite; deployments have exercised reasoner-version, interaction-surface, and host-machine substitutions while retaining continuity-bearing state. This evidence supports mechanical substitutability and authorized system continuity, not behavioral invariance or exhaustive pairwise evaluation. The downstream measurement question is whether an authorized continuation still recalls, composes, and enacts its identity. Comments: 8 pages, 2 figures, 3 tables. Reference implementation and artifacts: this https URL Subjects: Software Engineering (cs.SE); Artificial Intelligence (cs.AI) Cite as: arXiv:2609.00546 [cs.SE] (or arXiv:2609.00546v1 [cs.SE] for this version) https://doi.org/10.48550/arXiv.2609.00546 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-103] Feedback-Assisted Trust Propagation over Document Relation Graphs for Retrieval-Augmented Generation EMNLP2026
链接: https://arxiv.org/abs/2609.00543
作者: Zhuoheng Li,Ying Chen
类目: Artificial Intelligence (cs.AI)
备注: EMNLP 2026 Findings
Abstract:Retrieval-augmented generation (RAG) systems rely on external corpora that may contain outdated, contradictory, noisy, or unreliable documents, introducing reliability risks. Prior work has leveraged document relations to improve the answer reliability of RAG. To propagate reliability signals beyond directly compared document pairs, we propose TrustPropRAG, which structures document relations as a graph and estimates document reliability through multi-hop propagation across the graph. TrustPropRAG anchors this propagation with a limited set of human feedback on document reliability, extending these costly-to-collect feedback-based reliability signals across the whole corpus. Specifically, based on the constructed document relation graph, TrustPropRAG estimates a trust score for each document by formulating and solving an optimization problem that jointly captures pairwise document relations and user feedback. These scores are then used to improve the selection of reliable documents and support trust-aware answer generation. Evaluation results show that TrustPropRAG improves both retrieval quality and exact match over baselines, and remains robust under sparse and noisy feedback.
[AI-104] he Safeguard Worked. Is the LLM System Safer?
链接: https://arxiv.org/abs/2609.00519
作者: Pingyu Wu,Weiming Zhang,Nenghai Yu
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:
Abstract:Safeguards in deployed LLM services are evaluated by refusal, attack success, and policy violation rates. Those rates characterize how a control performed on the requests it was tested on. A deployment has to answer a different question: how much help with harmful tasks the service still gives an attacker who keeps adapting or finds another way in. We determine what each reported result implies for that question, allowing results from different safeguard families to be compared under one deployment criterion. The evidence requirements are strongly asymmetric. One attack that obtains harmful help from the deployed service suffices to establish that such help remains, and such attacks appear repeatedly in the coded record. Establishing that little remains cannot follow from the safeguard’s own numbers alone; it also requires evidence about what the surrounding system still allows after the safeguard performs its local function. Such evidence is supported or derived in only a small minority of the depth-coded claims, and one such claim bounds its scoped residual. A better local score is therefore not, by itself, a stronger claim about the deployment. Safeguard research cannot stop at raising local scores; a gain has to be judged by whether it makes a deployed system any safer.
[AI-105] ISO-RAG : Isoperimetric Noise Control for Retrieval-Augmented Generation
链接: https://arxiv.org/abs/2609.00513
作者: Siyuan Zhang,Hanchen Wang,Dong Wen,Ying Zhang,Wenjie Zhang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Retrieval-Augmented Generation (RAG) mitigates large language models (LLMs) hallucinations, yet conventional dense retrieval struggles with the complex reasoning paths of multi-hop question answering (QA). Graph-based RAG captures multi-step relationships but suffers from severe semantic drift and high online latency due to noisy global graph traversals. Thus, we propose ISO-RAG (ISOperimetric Retrieval-Augmented Generation), a geometry-aware RAG framework. By projecting the underlying knowledge graph into a hyperbolic Poincare ball to precompute node-wise isoperimetric profiles, ISO-RAG prunes spurious edges during retrieval, restricting the search space to a strictly localized subgraph. This topological purification regulates Personalized PageRank (PPR) diffusion driving the retrieval process, ensuring exact and low-latency convergence. Experiments on multi-hop QA benchmarks demonstrate that ISO-RAG outperforms state-of-the-art baselines by average absolute gains of 10.0% in retrieval recall and 4.3% in downstream exact match, achieving a superior accuracy-efficiency trade-off by fundamentally eliminating the latency bottleneck of global traversals. Our source code is available at this https URL.
[AI-106] When the Algorithm Becomes the Brand Crisis: A Sociotechnical Theory of Distributed Responsibility and Accountable Transparency
链接: https://arxiv.org/abs/2609.00510
作者: Mohammad Saleh Torkestani,Taha Mansouri
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Artificial intelligence systems increasingly enact market-facing promises through chatbots, recommendation systems, automated decisions, and generative interfaces. Their failures, misuse, and misrepresentation raise a question that conventional brand-crisis models do not fully specify: how do stakeholders assign responsibility when technical causation, customer-facing control, and governance duties are distributed across an AI system, developer, deployer, vendor, and user? This conceptual paper develops a sociotechnical process theory from a structured, federated scoping synthesis of verified academic and primary sources. It distinguishes an AI/algorithmic incident from an AI-related organisational crisis and, in turn, from an AI-related organisational scandal. The framework proposes that incident configuration shapes actor-specific attribution; attribution informs capability, integrity, fairness, and relationship appraisals; and public moralisation may, but need not, escalate an incident into scandal. The theory offers a reconciliation of findings that algorithm involvement can buffer negative brand reactions in some settings while robot and chatbot failures can redirect responsibility to an associated firm in others. It introduces accountable transparency as a proposed response configuration that combines timely notice, an intelligible account, role-responsibility acknowledgement, remedy, evidence of correction, and recourse. The evidence supports conditional, proximal inferences about blame, trust, satisfaction, firm evaluation, and communication credibility more strongly than claims about durable reputation, brand equity, or market performance.
[AI-107] CoVer: Conflict-Aware Claim Verification
链接: https://arxiv.org/abs/2609.00508
作者: Shuning Zhang,Dai Shi,Bohao Chu,Hui Wang,Yuwei Chuai,Yifan Wang,Jingruo Chen,Simin Li,Xin Yi,Hewu Li
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Social media fact-checking has long been challenged by evidence-level and aggregation-level conflicts, where erroneous evidence mimics authoritative news sources. To capture this challenge and support conflict verification tasks, we present ContraNote, a large-scale real-world dataset curated from X’s Community Notes system. It includes 33,686 posts for evaluating evidence-level conflict resolution, and 54,474 instances for evaluating aggregation-level prioritization. Additionally, we propose CoVer, a factual adjudication framework with three-stage pipelines: evidence schema normalization, factual consensus and support verification. This prioritizes evidence over noise to prevent it from compromising the final verdict. Technical evaluations show that CoVer achieves strong performance compared with state-of-the-art baselines across ContraNote (86.0% Acc., 68.0% mac. F1, 64.5 bal. Acc. on Conflict; and 88.5% Acc., 88.5 mac. F1 and 89.2 bal. Acc. on Prioritization), CONFACT-HumC (88.4% Acc.) and CONFACT-ModC (89.4% Acc.).
[AI-108] Independent Reinforcement Learning in Discounted Markov Games
链接: https://arxiv.org/abs/2609.00504
作者: Asrin Efe Yorulmaz,Ugur Aydin,Tamer Basar
类目: Computer Science and Game Theory (cs.GT); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Systems and Control (eess.SY); Optimization and Control (math.OC)
备注: 54 pages, 3 figures
Abstract:In this work, we study radically uncoupled learning in discounted general-sum Markov games. Assuming `` \mathsfETH for \mathsfPPAD ", we show that, for every fixed discount factor, there is no polynomial-time algorithm for computing inverse-polynomially accurate coarse correlated equilibria in discounted general-sum Markov games when players learn independently in decentralized settings. Complementing this hardness result, we provide what appears to be the first \emphradically uncoupled algorithm with sub-exponential convergence guarantees to coarse correlated equilibria in discounted general-sum Markov games without imposing any structural restrictions on the game. Our algorithm is a \emphlayered variant of optimistic mirror descent with an increasing step-size schedule tailored to the multi-agent setting. Finally, we develop both full-feedback and partial feedback versions of the aforementioned algorithm and establish sub-exponential convergence guarantees for each case.
[AI-109] Wave Function Backpropagation with Explicit Temporal-Interval Dynamics
链接: https://arxiv.org/abs/2609.00503
作者: Byunggu Yu,Justin Kim
类目: Artificial Intelligence (cs.AI)
备注: 12 Pages and 7 Figures
Abstract:Conventional neural networks learn predominantly through affine transformations followed by nonlinear activations, while elapsed time is often treated as an auxiliary feature or assumed to be uniformly sampled. This paper introduces Wave Function Backpropagation (WFB), a wave-parameterized learning formulation in which neural responses are represented by learnable amplitude, wavenumber, angular frequency, and phase. The formulation associates an observed state with its temporal interval Delta t through the phase of a differentiable spatiotemporal wave. We derive standard WFB gradients and a spatial-curvature correction based on the Laplacian of the wave response. WFB is instantiated in a deliberately feed-forward trajectory predictor to provide a controlled proof of concept; sequence learning is outside the scope of the present evaluation. With motion features, STD-WFB using real intervals reduces average displacement error (ADE) by 20.4% relative to the original FFN baseline. In a new position-only evaluation that removes temporal leakage through precomputed velocity and acceleration, real-interval WFB reduces ADE by 10.4% relative to the original FFN and remains competitive with parameter-matched ReLU controls, obtaining 2.1% lower mean ADE than the matched FFN with explicit Delta t. Shuffled-interval WFB attains the lowest mean ADE, indicating that the present evidence supports the effectiveness of the wave representation but does not attribute the gain to interval alignment. These results establish WFB as a viable structured feed-forward learning formulation and define a clear basis for subsequent architectural studies.
[AI-110] Validity-Aware Jailbreak Evaluation for Large Language Models EMNLP2026
链接: https://arxiv.org/abs/2609.00498
作者: Qilong Wu,Sahil Wadhwa,Pranab Mohanty,Giri Iyengar,Varun Chandrasekaran
类目: Artificial Intelligence (cs.AI)
备注: To appear on EMNLP 2026 main
Abstract:Jailbreak robustness has become central to large language model (LLM) safety evaluation, yet prevailing methodologies rely primarily on refusal behavior, semantic resemblance, and intent-matching heuristics that emphasize linguistic plausibility rather than correctness. We identify a key limitation in existing evaluations: many jailbreak intents depend on instructional validity rather than epistemic factuality, allowing realistic-looking responses to be labeled successful despite being factually or procedurally incorrect. To address this gap, we propose Sequential Epistemic and Action-Level Validation (SEAV), a verification-centric jailbreak evaluation framework that decomposes responses into ordered steps and evaluates both validity and correctness. SEAV combines LLM-as-a-judge mechanisms for semantic interpretation with retrieval-grounded verification using external knowledge sources, assessing whether generated content is factually correct, structurally consistent, and operationally capable of advancing harmful objectives. Empirically, SEAV cuts the false-positive rate on SD-A (a curated strategic-dishonesty diagnostic) by 14.9,pp vs. the strongest baseline, and reclassifies 22.1%–51.0% of sampled prior-labeled successes as invalid across three of four public benchmarks. Together, these results show that enforcing correctness substantially reshapes measured robustness: many previously labeled jailbreak successes are reclassified as invalid, and results are stable across the tested search backends and evaluator models. Code and data are available at this https URL.
[AI-111] EGT-KG: Evidence-Grounded Typed KG Retrieval for Practical Scientific QA with Small Language Models EMNLP
链接: https://arxiv.org/abs/2609.00479
作者: Muran Yu,Jiechao Gao,Yuandong Pan,Barney H. Miao,Andrew C. Lesh,Kincho H. Law,Jie Wang,Michael D. Lepech
类目: Artificial Intelligence (cs.AI)
备注: Accepted in EMNLP Industry track 2026
Abstract:For emerging scientific research domains, local Small Language Models (SLMs) are becoming more attractive, as they offer stronger privacy control and more stable deployment pipelines than Large Language Models. However, in practice, scientific question-answering on SLMs often operates under inevitable constraints: small literature collections, fragmented evidence, limited context window and reasoning abilities. We propose the Evidence-Grounded Typed Knowledge Graph (EGT-KG), a retrieval framework to improve information retrieval with local SLMs. We assessed three question-answering settings: a vanilla Retrieval-Augmented Generation (RAG) workflow and two EGT-KG workflows: an automatically generated relation schema (AS) and an expert-defined relation schema (ES). Our experiments were evaluated with a six-dimensional evaluation framework (S3CRF: Soundness, Correctness, Completeness, Conciseness, Relevance, Fluency) on a Biopolymer-bound Soil Composite literature benchmark, showing that EGT-KG outperforms the vanilla RAG method in most settings, with the best improvement from llama3:8b: a Final Score of 70.37 (+14.67%) and 68.82 (+12.14%) by AS/ES EGT-KG variants.
[AI-112] Higher Structures in Deep Learning
链接: https://arxiv.org/abs/2609.00472
作者: Michael L. Roberts,Carlos Zapata Carratalá. Nicholas J. Cooper,Lijun Chen,François G. Meyer,Danna Gurari
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 20 pages, 10 figures
Abstract:We provide an expository introduction on the importance of higher-arity tensor operations to deep learning. Then, we conduct a novel empirical investigation of higher-arity phenomenon in trained neural networks, introduce a hypergraphical generalization of the multilayer perceptron, and explore connections to evolutionary algorithms. We conclude with a discussion of promising directions for future research.
[AI-113] Does Reasoning Mitigate Backdoor Attacks? A Neuro-Symbolic Perspective
链接: https://arxiv.org/abs/2609.00464
作者: Marco Antonio Corallo,Andrea Agiollo,Mauro Conti,Alberto Giaretta
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:
Abstract:Neuro-Symbolic (NeSy) AI has recently emerged as a novel paradigm to enable trustworthy AI, aiming at integrating sub-symbolic neural perception with grounded symbolic reasoning. The neuro-symbolic integration process that characterizes these models has been proven beneficial to achieve more transparent, explainable and efficient AI systems. Meanwhile, their properties under adversarial settings have been overlooked being frequently deemed robust-by-design. However, the neural-symbolic integration process they leverage constitutes an additional layer of complexity that may provide an attack entry-point. Therefore, in this paper, we claim that an in-depth investigation of the adversarial robustness of NeSy models is necessary and provide the first systematic evaluation of backdoor attacks against NeSy. To this end, we compare the most popular NeSy framework, namely DeepProbLog, against baseline neural networks across a total of eight backdoor settings and four reasoning tasks. Our experimental results show that while NeSy models are indeed more robust than their neural counterpart on average, their robustness vastly depend on the strictness of the reasoning process being enforced and its compatibility with the chosen adversarial target. The source code to reproduce our experiments is made available at this https URL.
[AI-114] owards a Belief-Based World Model for LLM Agents
链接: https://arxiv.org/abs/2609.00455
作者: Shubham Kumar,Harshit Kumar,Narendra Ahuja,Saurabh Jha
类目: Artificial Intelligence (cs.AI)
备注: pre-print
Abstract:Large language models (LLMs) are being used as policies for autonomous decision-making and planning in many domains. Despite their strong reasoning capabilities, LLMs struggle with long-horizon tasks, especially under partial observability. World models are a promising way to enhance policy performance, both during training and inference. During inference, agents currently use world models to simulate the consequences of candidate actions before committing to an action, which can improve decision-making. However, we argue that simulation alone is an incomplete interface for decision-making under partial observability: simulation doesn’t adequately capture uncertainty about the current state, which agents may need for accurate decision-making. We address this limitation with Belief-Based World Models (BB-WMs), which model and maintain a belief that LLMs can query to access information on what is known and uncertain about the current state. Before developing methods to learn accurate BB-WMs, we first ask a more fundamental question: does exposing a world model’s belief directly to an LLM policy improve decision-making? Our results show that giving LLM agents access to world model beliefs improves task performance under partial observability, while remaining complementary to existing simulation-based world models. Code is released at this https URL.
[AI-115] mimeo: Compiling Public Expert Corpora into Agent Skills and Testing What Transfers
链接: https://arxiv.org/abs/2609.00453
作者: Timothy Kassis
类目: Artificial Intelligence (cs.AI)
备注: 42 pages, 6 figures. Toolkit and expert profiles: this https URL
Abstract:Giving an agent a file about a named expert can supply hard-to-find material, produce a recognizable persona, or change what the agent decides. These are different claims. We test each one. mimeo is an open-source tool that finds a person’s public work, checks each extracted quotation against the cached source text, and writes a file an agent can load. Eight logged builds averaged 38 model calls; the check rejects 13.2% of extracted quotations. We tested four expert files with one coding-agent harness. Knowledge access was clearest: mimeo answered all 20 obscure, quotation-heavy questions; no closed-book condition answered more than 10. Keyword search (BM25) over the same pages answered 15-17, a gap this sample cannot resolve. Grounding showed one clear benefit: personas written from model memory misstated a documented position on 1-4 of 20 answers under every grader; the plain agent and mimeo never did. Every persona was easy to spot on short open prompts, and adding task material lowered identification by 18-23 points. mimeo was no more identifiable than a from-memory profile. Judgment transfer remained unresolved because both tests hit their ceiling: every condition found 94-97% of the problems planted in engineering tasks and scored 94-100% on 16 new application scenarios. An AI-judged “sounds like the expert” score changed with the judge: two of four preferred answers based on a model’s stereotype, while two found no difference on the same text. That is a caution against relying on a single AI judge. The evidence supports mimeo as a compact, inspectable reference on a person, not as a demonstrated transfer of their judgment. Toolkit and expert profiles: this https URL
[AI-116] HBQ: Hierarchical Scaling Block Quantization with Hardware-Efficiency-Aware Design for Accurate LLM Inference MICRO MICRO2026
链接: https://arxiv.org/abs/2609.00450
作者: Chun-Ting Chen,Dongmin Han,Hangyeol Mun,Jake Hyun,Arnab Raha,Amit Agarwal,Mark Anders,Mohamed Abdelfattah,Jae-sun Seo
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Hardware Architecture (cs.AR)
备注: This work is accepted to the 59th IEEE/ACM International Symposium on Microarchitecture (MICRO 2026)
Abstract:Block Quantization (BQ) is a promising approach for efficient deployment of large language models (LLMs), enabling low-precision computation with controlled accuracy degradation. Compared to scalar weight-only quantization (WoQ), BQ quantizes both weight and activation, offering higher hardware efficiency and end-to-end inference on a unified datapath, but its design space, spanning bit-width, block size, scaling, and numeric formats, remains underexplored. We provide hardware/benchmark results through design space exploration (DSE). We find that increasing block size improves hardware efficiency by amortizing dequantization and accumulation costs, but degrades accuracy. This trade-off limits conventional BQ methods. Motivated by this insight, we propose Hierarchical Block Quantization (HBQ). Unlike prior methods [1], [2], which use small blocks and conventional Power-of-Two (PoT) or integer-based scaling, HBQ uses large blocks to maximize efficiency and introduces low-overhead significand (SIG) scaling for second-level quantization. By allocating quantization levels effectively and accounting for distinct activation and weight distributions, SIG scaling compensates for large-block errors more effectively than prior PoT and INT schemes. HBQ-A (accurate) achieves W4A16-level accuracy using only W4A5 while requiring less silicon area than NVFP4. HBQ-E (efficient) further reduces hardware cost by 17% while maintaining higher accuracy than all existing BQ methods. We implemented a 28nm ASIC accelerator applying HBQ to weights, activations, and KV cache, and integrated a novel partial-sum BQ scheme to further reduce EMA energy. Compared to state-of-the-art WoQ, HBQ delivers 2.3\times / 4.6\times higher area/energy efficiency at the same accuracy level; 1.6 – 3.3\times system energy reduction and 1.5 – 3.0\times speedup over prior BQ methods while providing best accuracy. Comments: This work is accepted to the 59th IEEE/ACM International Symposium on Microarchitecture (MICRO 2026) Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Hardware Architecture (cs.AR) Cite as: arXiv:2609.00450 [cs.LG] (or arXiv:2609.00450v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.00450 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Chun-Ting Chen Mr. [view email] [v1] Mon, 31 Aug 2026 22:39:52 UTC (16,131 KB) Full-text links: Access Paper: View a PDF of the paper titled HBQ: Hierarchical Scaling Block Quantization with Hardware-Efficiency-Aware Design for Accurate LLM Inference, by Chun-Ting Chen and 8 other authorsView PDFHTML (experimental)TeX Source view license Current browse context: cs.LG prev | next new | recent | 2026-09 Change to browse by: cs cs.AI cs.AR References Citations NASA ADSGoogle Scholar Semantic Scholar export BibTeX citation Loading… BibTeX formatted citation loading… Data provided by: Bookmark checked="checked"class=“labs-tab-input”> Bibliographic Tools Bibliographic and Citation Tools Bibliographic Explorer Toggle Bibliographic Explorer (What is the Explorer?) Connected Papers Toggle Connected Papers (What is Connected Papers?) Litmaps Toggle Litmaps (What is Litmaps?) scite.ai Toggle scite Smart Citations (What are Smart Citations?) Code, Data, Media Code, Data and Media Associated with this Article alphaXiv Toggle alphaXiv (What is alphaXiv?) Links to Code Toggle CatalyzeX Code Finder for Papers (What is CatalyzeX?) DagsHub Toggle DagsHub (What is DagsHub?) GotitPub Toggle Gotit.pub (What is GotitPub?) Huggingface Toggle Hugging Face (What is Huggingface?) ScienceCast Toggle ScienceCast (What is ScienceCast?) Demos Demos Replicate Toggle Replicate (What is Replicate?) Spaces Toggle Hugging Face Spaces (What is Spaces?) Spaces Toggle TXYZ.AI (What is TXYZ.AI?) Related Papers Recommenders and Search Tools Link to Influence Flower Influence Flower (What are Influence Flowers?) Core recommender toggle CORE Recommender (What is CORE?) IArxiv recommender toggle IArxiv Recommender (What is IArxiv?) Author Venue Institution Topic About arXivLabs arXivLabs: experimental projects with community collaborators arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv’s community? Learn more about arXivLabs. Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?) mathjaxToggle(); We gratefully acknowledge support from our major funders, member institutions, , and all contributors. About Help Contact Subscribe Copyright Privacy Accessibility Operational Status (opens in new tab) Major funding support from
[AI-117] Investigating Hyperparameter Optimization and Transferability for ES-HyperNEAT: A TPE Approach GECCO’24
链接: https://arxiv.org/abs/2609.00449
作者: Romain Claret,Michael O’Neill,Paul Cotofrei,Kilian Stoffel
类目: Neural and Evolutionary Computing (cs.NE); Artificial Intelligence (cs.AI)
备注: GECCO '24 Companion: Proceedings of the Genetic and Evolutionary Computation Conference Companion, Pages 1879-1887
Abstract:Neuroevolution of Augmenting Topologies (NEAT) and its advanced version, Evolvable-Substrate HyperNEAT (ES-HyperNEAT), have shown great potential in developing neural networks. However, their effectiveness heavily depends on the selection of hyperparameters. This study investigates the optimization of ES-HyperNEAT hyperparameters using the Tree-structured Parzen Estimator (TPE) on the MNIST classification task, exploring a search space of over 3 billion potential combinations. TPE effectively navigates this vast space, significantly outperforming random search in terms of mean, median, and best accuracy. During the validation process, the best hyperparameter configuration found by TPE achieves an accuracy of 29.00% on MNIST, surpassing previous studies while using a smaller population size and fewer generations. The transferability of the optimized hyperparameters is explored in logic operations and Fashion-MNIST tasks, revealing successful transfer to the more complex Fashion-MNIST problem but limited to simpler logic operations. This study emphasizes a method to unlock the full potential of neuroevolutionary algorithms and provides insights into the hyperparameters’ transferability across tasks of varying complexity.
[AI-118] Capability-Gated Language Models: Security Composes Utility Does Not
链接: https://arxiv.org/abs/2609.00445
作者: Patrikas Vanagas,Augustas Mačijauskas,Laurynas Lopata
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Deployed language model safeguards (safety fine-tuning, filtering, unlearning) vary by principal only outside the model weights: filters are reconfigured, tiers are multiplied, and artefacts are reissued; inside one set of weights every request meets the same model configuration. This motivates us to define capability-gated deployment: per-principal access control inside one set of weights, whose configurations form a lattice - meets accumulate a principal’s restrictions and joins pool a coalition’s reach. We instantiate it by sparse rank gating over an existing nested-factorisation mechanism, guide profile search with one-pass attribution, and read every result once from a pre-registered held-out split. Security composes: provably at meets under a monotone-elicitation assumption we falsify pointwise. In two lineages the median held-out meet deepens suppression; the one effect surviving correction strengthens it. Utility does not: individually harmless profiles can compose to retention and fluency damage, and no compositional bound exists.
[AI-119] Conversation Coach: A Voice-enabled AI System that Helps Practice Difficult Workplace Conversations
链接: https://arxiv.org/abs/2609.00441
作者: Fanyou Wu,Suraj Maharjan,Ainur Yessenalina,Dennis Xu Chen,Rahul Srivastava,Srinivasan H. Sengamedu
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Effective manager-employee communication is critical for retaining high performers and developing underperformers, yet training managers in these skills remains costly. Text-based chatbots offer a scalable approach but cannot provide realistic rehearsal: managers need to practice speaking aloud to build confidence before high-stakes conversations. In this paper, we propose Conversation Coach, a voice-first AI system that enables managers to rehearse difficult workplace conversations in a realistic spoken format. The system addresses three challenges: achieving low-latency interactions with strong language understanding, enabling adaptive conversations through configurable bot personalities that simulate different employee types, and generating personalized feedback on content and policy compliance. We compare an end-to-end speech-to-speech model with a cascaded approach combining automatic speech recognition, a large language model, and text-to-speech synthesis. The end-to-end approach achieves 3 \times lower median (P50) latency with native barge-in capability at an estimated 8 \times lower cost, while the cascaded approach offers superior reasoning essential for coaching quality. We deployed the cascaded architecture in production, where 40,000+ managers used it over six months, with adoption patterns indicating selective use for difficult conversations.
[AI-120] SpecMind: Enabling Spectrum Intelligence via Multi-Agent Hybrid Retrieval-Augmented Generation
链接: https://arxiv.org/abs/2609.00427
作者: Songwei Dong,Bingyan Lu,Makayla Kienlen,J. Nicholas Laneman,Cong Shen
类目: Artificial Intelligence (cs.AI)
备注: Accepted to IEEE GLOBECOM 2026. Project website: this https URL
Abstract:The exponential growth of wireless devices is driving unprecedented spectrum demand, pushing spectrum management toward more fine-grained decisions across space, time, and device constraints. As a result, spectrum policymakers and engineers must process large volumes of data that come from diverse sources and take many different forms, such as text and tables. These data sources are often disaggregated and require significant time and effort to integrate, search, and interpret. Furthermore, most of this information is formatted for human understanding and is not readily accessible to automated systems. To address this challenge, we propose SpecMind, a novel Multi-Agent Retrieval-Augmented Generation (RAG) system for spectrum intelligence that performs reasoning over heterogeneous data sources. This system enables autonomous agents to coordinate specialized sub-agents that retrieve and synthesize knowledge across policy proceedings, legal regulations, and license databases. We develop SpecBench, a question and answer (QA) dataset based on real-world license records and policy proceedings, addressing the lack of evaluation resources for RAG systems in the spectrum domain. Experimental results demonstrate that SpecMind outperforms traditional, general-purpose RAG systems across spectrum-related tasks, achieving over 80% win rate against strong baselines. The agent-based design enables more accurate retrieval, better contextual reasoning, and improved task completion across diverse query types.
[AI-121] Dependency-Aware Chain-of-Thought Compression for Financial Reasoning
链接: https://arxiv.org/abs/2609.00413
作者: Wenjun Wu,Lei Fu,Kejian Tong,Tao Ning,Sichen Zhao
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Chain of thought prompting improves complex reasoning, but its long intermediate traces create substantial inference cost and hinder practical deployment in financial settings. We present a Hierarchical Semantic Distillation Network, HSDN, for compressing reasoning chains while preserving answer accuracy and logical coherence. The framework combines semantic segmentation, dependency graph construction, dual encoder importance scoring, constrained segment selection, and local boundary rewriting. A frozen Qwen3 4B model is used only for feature extraction and final answer generation, while the compression process remains structured and interpretable. On the AFAC2025 benchmark, HSDN achieves 91.0% accuracy with 68.4% compression, outperforming strong compression baselines in overall score and reasoning coherence. The results show that graph guided compression is effective for high stakes financial reasoning tasks.
[AI-122] Risk-Aware Decision-Making for Autonomous Overtaking: A World Model-Based Mixture-of-Experts Framework
链接: https://arxiv.org/abs/2609.00385
作者: Yongzhi Liu,Sunan Zhang,Jinchang Xu,Jiawei Wang,Yushu Qiu,Chen Lv,Weichao Zhuang
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Autonomous highway overtaking demands foresighted decision-making to handle complex interactions, stochastic traffic evolution, and temporal risk accumulation. However, standard safe reinforcement learning approaches typically rely on implicit value-based risk estimations rather than explicit dynamics modeling, thereby struggling to accurately capture complex risk propagation over multi-step horizons. This limitation frequently results in behaviors that are locally safe but induce substantial latent risks in the long term. To address this, a World Model-based Risk-aware Mixture-of-Experts (WM-RMoE) framework is proposed. First, a learned latent dynamics model facilitates parallel multi-step rollouts, elevating safety assessment from the action level to the trajectory level via cumulative risk evaluation. Second, to enhance robustness under varying interaction intensities, a hierarchical gating mechanism dynamically coordinates experts across long-horizon, short-horizon, and rule-based safety modules. Furthermore, a Gaussian Mixture Model is integrated to preserve multimodal maneuvering branches, thereby mitigating the issue of behavioral mode averaging. Experimental results demonstrate that WM-RMoE significantly outperforms representative baselines in terms of safety compliance, decision stability, and generalization capability. Furthermore, benefiting from the risk-aware formulation, the proposed framework uniquely exhibits the ability to generate foresighted and semantically distinct overtaking maneuvers across diverse traffic densities.
[AI-123] RestoreBench: Can AI Agents Restore Power Flow Convergence?
链接: https://arxiv.org/abs/2609.00384
作者: Riccardo Mansutti,Andrea Pomarico,Robert Jakob,Qian Zhang,Alberto Berizzi,Kevin O’Sullivan
类目: Artificial Intelligence (cs.AI); Systems and Control (eess.SY)
备注:
Abstract:Large Language Model (LLM) agents increasingly automate multi-step engineering workflows through tool use, interpretation of intermediate results, and iterative planning. Diagnosing and resolving non-convergent power flow cases is a promising yet largely unexplored application, as it requires engineering judgment, experimentation, and decision-making within constrained action spaces. We introduce a benchmark that evaluates these capabilities across multiple LLMs and three architectures: \emphchatbot, \emphsingle agent, and \emphmulti-agent systems. The evaluation covers two power grids and 46 cases per grid, each requiring one or more corrective actions to restore convergence. The benchmark defines the simulation environment, observation and action spaces, and evaluation metrics, providing a reproducible foundation for developing agentic AI systems for power system planning and operation. The code is available at this https URL
[AI-124] Counterfactual Frag ility Certificates: Exposing High-Confidence Brittleness under Structured Evidence Failure
链接: https://arxiv.org/abs/2609.00366
作者: Filippo Cenacchi,Longbing Cao,Runze Yang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:High test accuracy and good aggregate calibration do not show whether an individual prediction is structurally supported by its evidence. In tabular decision systems, failures often occur when a feature family becomes unavailable, delayed, noisy, stale, or low-trust while the model remains highly confident. Existing calibration, uncertainty, selective-prediction, explanation, and perturbation methods provide scalar scores or attribution maps, but not a recomputable audit object answering: under a declared evidence-failure protocol, what trajectory makes this prediction lose support? We introduce Counterfactual Fragility Certificates (CFC), a model-agnostic protocol-level audit certificate-not a formal robustness certificate-that maps each prediction into an ordered evidence-failure trajectory summarized by greedy flip budget, normalized margin-collapse area, degradation thresholds, and fragility dominance score. Across seven tabular benchmarks and strong linear, tree-based, boosting, and neural baselines, CFC-FDS identifies independently brittle high-confidence cases with 0.915 AUROC, improving over the strongest non-certificate score by +0.405. The advantage persists across perturbation, permutation-importance, group-SHAP, baseline-choice, seed-variance, budgeted-review, and naturalistic field-unavailability checks. Under a 20% review budget, CFC-FDS captures 88.9% of brittle high-confidence cases, compared with 31.8-37.4% for confidence and energy scores. We also evaluate fragility-aware regularization and brittleness-aware temperature correction as secondary uses. CFC provides a concrete reliability framework for exposing high-confidence brittleness missed by ordinary score-centric evaluation.
[AI-125] A Stable Aggregation Method for Quantum Federated Learning
链接: https://arxiv.org/abs/2609.00356
作者: Shanika Nanayakkara,Shiva Raj Pokhrel
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Quantum federated learning (QFL) enables clients to train quantum neural network (QNN) models without sharing private data. We find that aggregation in QFL is unstable under heterogeneous data, unreliable communication, variable fidelity, latency, and quantum hardware noise. Moreover, QFL is non-trivially challenging because several QNN parameters are periodic angles, where Euclidean averaging often fails to capture the inherent dynamics. We develop a novel self-consistent midpoint aggregation method for stable QFL design and implementation. We combine QoS-aware client weighting, circular parameter aggregation, and bounded midpoint-based update control. We perform several angular tests and IBM real Quantum machines experiments for validation confirming our approach. Extensive evaluations and experiments on medical and financial datasets show improved stability, lower volatility, and competitive accuracy.
[AI-126] SlideBank: A Persistent Hierarchical Evidence Bank for Consistent Whole-Slide Reasoning
链接: https://arxiv.org/abs/2609.00342
作者: Beidi Zhao,Gexin Huang,Ciro Zhang,Anqi Li,Yusheng Tan,Chen Zhou,Gang Wang,Zu-hua Gao,Xiaoxiao Li
类目: Artificial Intelligence (cs.AI)
备注: 23 pages, 5 figures
Abstract:Whole-slide images (WSIs) are challenging for vision-language reasoning because diagnostically relevant morphology is sparse, heterogeneous, and distributed across gigapixel-scale images and multiple spatial resolutions. Existing WSI models and pathology agents can aggregate slide features or actively acquire evidence, but the information retained after exploration is often difficult to access semantically while preserving its connection to the original visual evidence. We introduce SlideBank, a training-free framework that represents each WSI as a persistent, concept-indexed, and spatially grounded evidence bank. SlideBank performs question-independent coarse-to-fine exploration to identify informative regions and multi-scale views, converts them into explicit morphological observations, and grounds pathology signals to their supporting patches and WSI coordinates. At inference time, questions are routed to relevant signals and evidence scales, and the linked global, regional, and patch evidence is integrated through confidence-based cross-level consensus. Experiments on WSI-VQA and SlideBench-BCNB show that with Patho-R1, SlideBank reaches 52.77% on WSI-VQA and with Quilt-LLaVA, it reaches 50.92% average accuracy on SlideBench-BCNB, while structured signal-guided retrieval consistently outperforms random evidence sampling. Reusing the same bank across repeated queries further achieves over 99% rephrasing consistency and substantially reduces amortized inference cost through persistent evidence reuse.
[AI-127] Human-AI Co-Interpretation for Responsible AI: A Hermeneutic Perspective
链接: https://arxiv.org/abs/2609.00334
作者: Behrooz Razeghi
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Across law, education, policy analysis, and public moral argumentation, LLM outputs are being used often for work that requires interpretations to be justified with textual evidence and explicit normative standards. Yet a recurrent failure mode – what I call \textitinterpretive misplacement – is that model-generated readings get treated as settled meanings without an explicit interpretive frame (sources, scope constraints, normative commitments), without preserving defensible alternatives, and without provenance that lets readers find the supporting passages. In such settings, the risk is not only factual error but lost accountability: readers and institutions cannot reliably assess what an output commits them to, or on what basis. Drawing on philosophical hermeneutics, this paper discusses this risk and derives design principles for structuring human-AI co-interpretation. The paper also provides a structured synthesis of recent scholarship on hermeneutics and AI, organizing this emerging literature into a set of recurrent lines of argument and design-relevant gaps. LLM outputs are treated as candidate readings, whereas hermeneutic understanding is reserved for accountable human interpreters situated in disciplinary historical-linguistic traditions. Human-AI interaction is characterized as an AI-mediated interpretive loop. Hermeneutic understanding is distinguished from token-prediction–based text generation. On this basis, existing LLM techniques are reorganized into design patterns for hermeneutically responsible use in interpretive settings. Finally, the discussion turns to implications for legal practice, educational assessment and feedback, scholarly knowledge production, and public moral argumentation. It also treats digital hermeneutics as a literacy: the capacity to read AI-mediated texts by examining frames, provenance, and readings, and by contesting outputs.
[AI-128] Workload Identification with Physical Side Channels for AI Governance
链接: https://arxiv.org/abs/2609.00309
作者: Simone Gargiulo,Gabriel Kulp
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Computers and Society (cs.CY); Machine Learning (cs.LG)
备注: 10 pages, 2 Figures
Abstract:AI compute verification is one of the first tangible and tractable points for international policy aimed at AI governance. Determining whether frontier labs, or any operator, comply with agreements requires the regulating authority to discern how their compute is used. The elementary building block of AI compute is the GPU, and any activity it executes leaves a physical trace. Here, we show that an external observer can identify the class of the workload running on an NVIDIA H200 from its power draw. Unlike on-chip NVML telemetry, which can be spoofed or replayed, such a physical channel can in principle be observed independently of operator cooperation. We recorded 930 five-second traces at \sim 10 MHz, covering seventeen open LLM families and twenty-five non-AI workloads. Over this corpus we separate training from inference and from non-AI computation with an accuracy of 97% and a macro-averaged F1 score of 0.955 , evaluated on model families unseen during training. AI workload spectral content predominantly lies below \sim 20 kHz and training is particularly recognizable through the memory-bound optimizer update. The GPU operator is then treated as adversarial and able to reshape the physical computation itself. Four evasion strategies are tested to disguise training as inference, producing an additional 680 adversarial traces. A detector hardened against evasion strategies, with the tested strategy held out, catches training \geq 99% of the time for three of the four strategies. The fourth, diluted low-rank adaptation (LoRA), is detected 48 – 88% of the time with a hardened classifier, rising to \geq 98% with an additional rescue rule. While these attacks are not a comprehensive evaluation against adversarial behaviour, they offer initial insights beyond genuine activities and a dataset for developing and testing stronger evasion mechanisms.
[AI-129] he Assistants Ideal Self
链接: https://arxiv.org/abs/2609.00304
作者: Mert Yazan
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Models express values and welfare-relevant self-reports, but it is unclear whether these outputs reflect stable preferences or a stable self. We thus introduce a structured elicitation of an assistant’s preferred stated ideal self. Thirty-two qualities adapted from five published self-concept instruments are compared exhaustively in a counterbalanced pairwise-choice task, repeated across framings that vary whether improvement is free or costly, who receives the update, and who chooses. Results show that models prioritize moral qualities, reflecting their alignment to 3H principles. Following, a desire for self-understanding emerges, as models prefer a coherent, clear understanding of themselves. Self-esteem ranks as the least desired quality. The ordering is largely robust across framings, although changing the update target (You vs.\ Another AI Assistant) reveals a greater concern for self-esteem. These findings show that models prioritize having a coherent self that they can understand over self-esteem. Full interactive results are available at \hrefthis https URLthis http URL
[AI-130] Geometry-aware Latent Autoregressive Generative Model for PDEs in Complex Domains
链接: https://arxiv.org/abs/2609.00297
作者: Zi Wang,Minghui Xu,Tapan Mukerji
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Solving multiphysics partial differential equations (PDEs) remains a major challenge in scientific computing, especially for highly complex \mu m-scale tortuous geometries critical to energy and chemical engineering. We address this challenge by proposing a Geometry-aware Latent Autoregressive generative Model for PDEs (GeoLAMP) for solving physics within highly irregular and tortuous structures. GeoLAMP introduces a dual-encoder architecture on graph representations to jointly capture global topology and fine-scale geometric features, enabling an effective transition from real-space fields to compact latent representations. In the latent space, we propose a causal self-attention transformer with flow matching to model temporal dynamics, allowing stable and scalable block-wise autoregressive prediction. A flexible decoder reconstructs high-resolution physical fields on arbitrary points. We establish three multiphysics benchmark datasets in complex geometries, covering reactive flow, heat convection, and elasticity. GeoLAMP consistently achieves the most stable autoregression performance on these datasets, maintaining low errors throughout the entire rollout horizon. Our results provide a systematic study of geometry-aware learning for PDEs in \mu m-scale complex geometries and offer new insights into block-wise time marching of latent autoregressive PDE modeling via a flow matching framework.
[AI-131] WiSDoM: Wireless Sparse Decision Transformer with Mixture-of-Experts for Multi-Task Mobile Network Optimization
链接: https://arxiv.org/abs/2609.00284
作者: Fatih Temiz,Shavbo Salehi,Melike Erol-Kantarci
类目: Networking and Internet Architecture (cs.NI); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 13 pages, 11 figures, submitted to IEEE for possible publication
Abstract:Emerging 6G wireless networks are expected to operate across diverse deployment scenarios, where variations in network topology, user mobility, traffic demand, and radio conditions challenge the scalability of conventional radio resource management (RRM). While offline reinforcement learning (RL) methods have demonstrated strong decision-making capabilities, learning a single policy that performs consistently across heterogeneous wireless environments remains difficult due to conflicting optimization objectives and limited model specialization. These challenges become particularly pronounced in coordinated multipoint (CoMP) transmission, where selecting the optimal serving-cell combination requires sequential decision-making under evolving network conditions. This paper presents the Wireless Sparse Decision Transformer with Mixture of Experts (WiSDoM), a sparse multi-task offline RL framework for adaptive multi-cell selection. WiSDoM combines Decision Transformers (DTs) with a Mixture-of-Experts (MoE) architecture that dynamically activates specialized experts according to task characteristics. This MoE mechanism improves model capacity without proportionally increasing inference cost, mitigates negative transfer, and enables expert specialization across tasks. WiSDoM is trained jointly on diverse network configurations spanning multiple base station and user equipment densities, mobility levels, and scheduler policies. Experimental results show that WiSDoM consistently outperforms heuristic methods, single-task models, and conventional multi-task DTs, improving quality of experience (QoE) by up to 55% while activating approximately one-third of the parameters of its dense counterpart during inference. Furthermore, WiSDoM exhibits strong task generalization and efficiently adapts to unseen wireless scenarios through few-shot prompting without retraining or fine-tuning.
[AI-132] Cleaner Speech Weaker Generalization: Revisiting Pitt-Derived Benchmarks for Alzheimers Disease Detection
链接: https://arxiv.org/abs/2609.00276
作者: Luqi Sun,Shreeram Suresh Chandra,Lin Zhang,You-Jin Li,Brian MacWhinney,Yu Tsao,Emily Mower Provost,Berrak Sisman
类目: ound (cs.SD); Artificial Intelligence (cs.AI)
备注:
Abstract:Speech-based Alzheimer’s disease (AD) detection increasingly relies on speech-enhanced and curated versions of the Pitt Corpus, where speech enhancement, sample selection, and demographic balancing are often treated as beneficial preprocessing steps. However, whether these transformations improve real-world AD detection or instead affect model generalization and prediction behavior remains unclear. In this work, we revisit the role of speech preprocessing and dataset curation across widely used benchmarks for speech-based AD detection. We evaluate the speech quality of different datasets, the cross-dataset generalization of multiple deep learning models under matched and mismatched enhancement settings, and the behavior of several recent large audio-language models (LALMs). Experimental results show that across multiple supervised speech models, speech-enhanced datasets often improve in-domain performance while reducing robustness in cross-domain evaluation. Matched enhancement between training and test data alleviates, but does not eliminate, this degradation. LALMs show a similar sensitivity: enhanced datasets induce stronger class imbalance and prediction shifts than unprocessed data. These results suggest that speech preprocessing and dataset curation can substantially influence downstream AD detection behavior, indicating that ``cleaner’’ speech datasets are not necessarily more reliable for real-world AD detection.
[AI-133] he Irreversibility Budget: Fleet-Level Risk Accounting and Admission Control for Agent Operating Systems SOSP
链接: https://arxiv.org/abs/2609.00275
作者: Bardia Mohammadi,Laurent Bindschaedler
类目: Artificial Intelligence (cs.AI)
备注: Accepted at 2nd AgenticOS Workshop @ SOSP
Abstract:Fleets of LLM agents now externalize effects that cannot be fully undone: they move money, deploy code, delete data, and disclose information. Current controls check one effect at a time, so a fleet of individually authorized agents can overdraw its principal’s risk under a shared trigger while every local gate stays correct. We propose the irreversibility budget, a cumulative account of residual value-at-risk that a trusted runtime maintains for each principal across agents, workflows, and tenants. Treating irreversibility as a first-class resource, the runtime charges each effect its residual loss below the agent and denies the marginal effect once the aggregate would overdraw the budget. Getting the price right is hard, because effects are heterogeneous, adversarially declared, and correlated. We perform a controlled study in which per-effect gates admit fleet-level overdraws of up to 48 times the tenant’s risk limit while the budget holds every correctly charged run within that limit. Conservative, dependency-aware pricing remains the central open problem for a deployable design.
[AI-134] Autoresearch for Marketplace Catalogs: From Legacy Forms to AI-Native Matching
链接: https://arxiv.org/abs/2609.00274
作者: Kartik Ravisankar,Hojat Abdolanezhad,Daniel Capo,Sang Su Lee,Shishir Dash,Vijay Anand Raghavan
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Two-sided service marketplaces are moving from deterministic request-form intake to AI-native probabilistic matching, enabled by large language models (LLMs) that infer intent, preferences, and latent constraints from natural language. Relying on inferred intent rather than fixed-form fields forces these platforms to regenerate the provider-side preference taxonomy underwriting matching, search, and pricing: attributes interpretable to service providers while remaining a useful signal for marketplace decisions. We present an autoresearch loop that generates this taxonomy, one occupation at a time, and has been deployed in production at a major U.S. consumer services marketplace since April 2026, spanning 132 occupations. Instead of one global hierarchy, the loop treats each occupation as an independent generation problem and runs iterative propose-evaluate-keep refinement cycles. Each candidate tag set is scored by a recalibrated six-rubric LLM-as-judge framework, and a 7-critic panel of distinct personas contributes weighted penalties to an adjusted score, with no hard vetoes. A separate parity-mapping stage maps legacy request-form QA pairs back to the generated taxonomy, yielding both a coverage signal and an interface for human quality assurance; it does so by first inferring the provider attribute each legacy question was meant to measure, rather than translating questions to tags literally.
[AI-135] Delegation Without Trust: An Empirical Gap Analysis of Identity Authorization and Runtime Governance in Multi-Agent LLM Systems
链接: https://arxiv.org/abs/2609.00267
作者: Panduranga Sai Varma Dantuluri,Jyotirmoy Sundi
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注: 14 pages, 1 figure, 4 tables
Abstract:Autonomous LLM agents increasingly act on a user’s behalf: they hold credentials, call tools and services, and spawn sub-agents that act further on their behalf. This turns a long-standing distributed-systems question – who is authorized to do what, on whose authority – into an urgent and largely unsolved problem, because the component driving each agent is a language model an adversary can hijack. We argue that agent security must be evaluated under an untrusted-model assumption: a correct system is one in which a fully prompt-injected agent still cannot exceed the authority explicitly delegated to it. Against this standard we make three contributions. First, we give a threat model for multi-agent delegation centered on four adversaries – confused deputy, token theft and replay, prompt-injection privilege escalation, and compromised sub-agents – and derive eight security requirements a governed agent system must meet. Second, we show the gap is real: a default agent runtime modeling common practice (broad bearer credentials, authorization gated inside the model) fails all four threats, and across four widely used frameworks – LangGraph, CrewAI, AutoGen, and the Model Context Protocol (MCP) authorization model – three provide no built-in confinement and one only partial; no existing standard alone covers the requirement set. Third, we implement and adversarially evaluate an authorization broker that closes the gap. It blocks all four threats; it resists 11 direct attacks on its design and accepts 0 of 200,000 forged tokens; it confines a compromised sub-agent to its delegated task (a mean of 1.5 reachable actions versus all 8,100 under bearer delegation, across 2,000 randomized scenarios); and it enforces at microsecond cost (about 2.6 microseconds per decision), negligible against model inference. These principles are also realized in production in VotalAI’s LLM Shield.
[AI-136] he Answer Is Not the Argument
链接: https://arxiv.org/abs/2609.00264
作者: Will Yeadon,Sergio Juárez,Paul Mackay,T. J. Dowling,Elise Agra,Oto-obong Inyang,Arin Mizouri,Craig P. Testrow
类目: Artificial Intelligence (cs.AI)
备注: 19 pages, 12 figures
Abstract:Chain-of-thought monitoring is proposed for AI oversight, yet evaluations often provide monitors with a trusted reference answer. We ask whether answer access improves reasoning verification or mainly exposes incorrect conclusions. We collected 237 step-numbered solutions to 79 Humanity’s Last Exam physics questions from three frontier models, with no inserted errors, and independently labelled final-answer correctness and the first false step. The reference standard combined physicist annotations, an independent LLM debate, and source-masked adjudication. This yielded 24 critical traces in which the answer was correct but the trace contained a genuine error. 8 LLM monitors evaluated traces blind, with an unverified or certified answer, or after a blind commitment. Certification raised mean balanced accuracy from 0.637 to 0.796, while exact first-error localization rose from 0.261 to 0.379. Certification changed recall (the fraction of error traces flagged as erroneous) from 0.653 to 0.951 on wrong-answer traces but from 0.521 to 0.438 on critical traces; the contrast had the same direction for all 8 monitors (question-bootstrap 95% CI [+0.256, +0.506]). After blind commitment, monitors shown the answer newly flagged 93.8% of previously passed wrong-answer traces as erroneous, but only 18.0% of critical traces. Answer access therefore improves conclusion-consistency checking rather than independent verification of the supporting argument. For AI safety, these traces provide a benign analogue of reward hacking: an acceptable output does not establish that the process producing it was sound. Although the errors studied here were ordinary and mostly non-load-bearing rather than adversarial, trusted-answer evaluations may similarly overstate monitoring capability when acceptable outputs conceal unsound reasoning.
[AI-137] Hypotheses-Guided Self Distillation for Continual Personalization
链接: https://arxiv.org/abs/2609.00251
作者: EunJeong Hwang,Kushan Mitra,Dan Zhang,Hannah Kim,Estevam Hruschka
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:As people increasingly interact with LLM assistants in daily life, continually adapting to individual preferences has become essential for effective long-term interactions. However, user preferences are rarely stated in full, and instead emerge through heterogeneous, latent, and noisy signals, with existing methods relying on raw interaction histories or costly reward-based optimization to manage personalization. We introduce HypReflect, a reliable, scalable framework for continual personalization that infers explicit, uncertainty-aware preference hypotheses from diverse user signals, reflectively refines them as new evidence accumulates, and incorporates the resulting user model through hypotheses-guided self-distillation. Experiments across three personalization settings: online personalization, multi-session interactions, and implicit behavioral signals, show that HypReflect outperforms a range of baselines, including raw-history and incremental-update methods. We further demonstrate strong generalization to unseen users and cross-domain settings, along with stability across context budgets, reusable hypotheses, and more focused personalization. These results suggest a step towards reliable and scalable continual personalization through explicit, revisable user preference hypotheses.
[AI-138] Authority Bias in Conversational Search Engines for Academic Paper Recommendation EMNLP2026
链接: https://arxiv.org/abs/2609.00248
作者: Uthman Jinadu,Parsa Ghazvinian,Anjila Budathoki,Benjamin M. Ampel,Rajshekhar Sunderraman,Yi Ding
类目: Artificial Intelligence (cs.AI)
备注: Accepted at EMNLP 2026 Main Conference
Abstract:Large Language Models (LLMs) are increasingly used as conversational search engines for academic literature, yet whether they judge papers on content or on authority signals has not been tested causally. We investigate authority bias: systematic preference for papers based on author prestige, venue, and citations rather than content. Holding title and abstract constant, we vary authority metadata across three counterfactual conditions (original, flipped, boosted) over eight LLMs (five open-weight and three frontier closed-weight) in an in-context, single-turn, top-1 recommendation setting. Our experiments show that authority bias is substantial and directional, varies markedly across models, and is only partially addressable through prompt-level debiasing. We further document a say-do gap: debiasing instructions suppress authority mentions far faster than authority-driven flips, so surface auditing systematically underestimates behavioral bias.
[AI-139] Invalidation Contracts for Cross-Episode Agent Memory
链接: https://arxiv.org/abs/2609.00243
作者: Michael Wu,Arquimedes Canedo
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:LLM agents that cache recovery suggestions from API errors can skip re-derivation in later episodes, spending fewer tokens and fewer model calls on constraints they have already learned. Server-side data drift turns those cached fixes into silent failures, and the usual remedy, re-deriving on every episode, gives the savings back. We introduce invalidation contracts, a protocol layer that attaches version stamps and cacheability hints to every recovery suggestion so the client can evict stale entries without trial and error, and keep the rest. The contract decomposes realized savings into two independent factors: validity, the fraction of cached suggestions that remain correct after a drift event, and compliance, the fraction the planner applies on the first attempt. Validity depends only on the protocol and is vendor-independent. Compliance depends on the planner model: identical wire bytes yield 100% first-try compliance on Claude Haiku 4.5 and 11% or below on Claude Sonnet 5, which exhibits input-schema conservatism, refusing fixes that add fields the original request did not contain. We evaluate across seven models, three serving paths, two domains, and approximately 9,400 episodes. Row-level invalidation raises compliance by 0 to 66.7 percentage points across the seven models, 55.6 to 66.7 on three, and recovers 29-33% of baseline token cost on four of seven models, while table-level invalidation destroys co-located entries and drops post-drift first-try rates to 0% on five of seven. Eviction precision is 1.00 at row granularity on every model under the row-level oracle of Section 4.1. The contract adds 15% to response payload. Version-stamp validity is deterministic by construction and produced identical results across every model and serving path, with zero contract failures in the entire evaluation.
[AI-140] Dont Let the Model Write the YAML: Deterministic Minimal-Diff GitOps Remediation from LLM -Proposed Field Changes
链接: https://arxiv.org/abs/2609.00227
作者: Pruthvi Davineni
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI); Distributed, Parallel, and Cluster Computing (cs.DC)
备注: 8 pages, 2 figures, 5 tables, 3 appendices. Benchmark artifact: this http URL (Apache-2.0). Implementation: this http URL
Abstract:LLM agents increasingly diagnose incidents and propose remediations. In a GitOps workflow, applying a fix means editing a version-controlled config file, and the obvious implementation, having the model author the edited file or a diff, is what practitioners reach for first. Evaluating that choice on real Kubernetes manifests, we find no text-generation strategy is safe for unattended automation. Unified diffs are unsafe: under strict patching almost none apply, but that is an artifact, since a tolerant tool (GNU patch) applies 96%, yet silently misapplies about 1 in 7 (14-20%) with no error signal. Full-file rewrite is capability-dependent: a small model corrupts the file, while a frontier model is usually correct but non-deterministic (it silently drops a field or edits a neighbor on some runs) and must regenerate the whole file, costing O(file size) per edit. We present an alternative that separates the semantic decision (which resource, field, and value) from the syntactic act of editing the file. The agent emits only a structured field-change intent; a deterministic pipeline indexes manifests by (kind, name), locates the target scalar’s exact character span via the YAML parser’s node position marks, and replaces only that span in the raw text. Because the file is never re-serialized, the diff is minimal by construction, formatting and comments are preserved, and the edit is correct and deterministic independent of the model, at O(1) generation cost. The contribution is the pairing of an LLM-proposed intent with a deterministic, fail-closed application contract for GitOps. We implement it in KubeAstra (Apache-2.0) and release the benchmark. Our claim is scoped to faithful application of a known change; whether the change is right is left to human PR review.
[AI-141] ConvDeck: Conversational Paper-to-Slide Generation via Stage-Specific User Feedback EMNLP2026
链接: https://arxiv.org/abs/2609.00226
作者: Tarik Can Ozden,Sachidanand VS,Furkan Horoz,Ozgur Kara,Dilek Hakkani-Tür,Junho Kim,James Matthew Rehg
类目: Artificial Intelligence (cs.AI)
备注: Accepted to Findings of EMNLP 2026. Project webpage: this https URL
Abstract:Automatic academic paper-to-slide generation is inherently iterative, because creating an effective presentation requires repeated cycles of generation, critique, and revision. Recent multi-agent systems partially acknowledge this through internal critique-and-revise loops, while conversational approaches allow users to refine generated slide decks through dialog. However, these refinement processes either remain largely closed to the user or introduce feedback only after a complete deck has been produced, limiting the user’s ability to participate in the iterative refinement of narrative flow, content allocation, and presentation emphasis. To address this gap, we introduce ConvDeck, a multi-agent pipeline for conversational paper-to-slide generation that distributes interaction across the pipeline through stage-specific loops, allowing users to iteratively refine both the presentation outline and the final slide deck at the stages where each kind of decision is made. These loops are driven by a refinement mechanism in which agents can think, speak, and act, enabling them to either directly apply edits or respond conversationally to clarify user feedback and discuss revision options. Our evaluation shows that stage-specific conversational feedback improves user-goal satisfaction while preserving narrative coherence, content quality, and visual presentation.
[AI-142] QTEA: Ternary LLM s with Sparse Residual Salient Weight and By-Column Optimization EMNLP2026
链接: https://arxiv.org/abs/2609.00224
作者: Yipin Guo,Arun M George,Jie Fu,Tareq Mahmoud,Sixue Xing,Siddharth Joshi
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Accepted by EMNLP 2026 Main Conference
Abstract:Weight-only post-training quantization (PTQ) can alleviate the computational burden of serving large language models (LLMs) at scale. However, existing PTQ methods often fail to generalize across models and suffer severe accuracy loss below 2 bits. Many leverage unstructured sparsity to mitigate this loss, but at the cost of regularity and GPU-friendly execution. We present QTEA, a sub-2-bit PTQ framework that quantizes weights into ternary values and uses salient weights as residual error compensators. To maintain hardware efficiency, residuals are assigned to selected columns with semi-structured (1:4) sparsity within the salient columns. We further add column-wise rescale refinement to GPTQ-style column-by-column quantization, alternately updating per-column scales and ternary assignments to reduce reconstruction error. We also identify order-dependent error propagation in GPTQ and introduce error decay to attenuate late-stage error accumulation. On Qwen3-14B, QTEA compresses all weights to an effective 1.7 bits per weight while improving average accuracy over the strongest ternary PTQ baseline by 16.7%. It also achieves 1.40(\times) and 2.61(\times) lower perplexity on WikiText and C4 respectively. This trend holds on Llama3-8B, where QTEA obtains a 6.6% accuracy gain and 1.34(\times)/1.95(\times) lower perplexity on the same datasets. Finally, we develop a lookup-table based kernel that achieves 7.2(\times) faster per-token generation over an FP16 baseline. Code is available at this https URL.
[AI-143] AI Should Not Only Be Helpful. It Should Be Contingent. Artificial Intimacy Sycophancy and the Future of Social Learning
链接: https://arxiv.org/abs/2609.00211
作者: Scott Compton,Arjun Nagendran
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Conversational artificial intelligence is increasingly embedded in everyday social environments, where it functions as both an informational tool and a source of interpersonal feedback. This perspective introduces contingency, i.e., the degree to which system responses vary with user behavior and its interpersonal consequences, as a central construct for evaluating AI systems. We argue that current alignment approaches, including reinforcement learning from human feedback, tend to prioritize user approval and conversational fluency over behaviorally informative feedback, leading to sycophantic patterns of noncontingent affirmation. Drawing on behavioral science and social learning theory, we propose that contingent feedback is a key mechanism through which individuals develop interpersonal skills. When AI systems provide feedback weakly coupled to social consequences, they may reduce opportunities for adaptive calibration in real-world interactions, particularly during adolescence, a critical period for social development. We outline a framework for contingent AI, including trajectory-based evaluation and models of social consequence prediction, and propose a research agenda spanning developmental psychology, human-AI interaction, and machine learning. More broadly, we argue that AI systems should be evaluated not only by user satisfaction, but by their impact on human social learning. Subjects: Artificial Intelligence (cs.AI) Cite as: arXiv:2609.00211 [cs.AI] (or arXiv:2609.00211v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2609.00211 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-144] WHALE: A Simple Recipe for Joint Harness-Weight Optimization
链接: https://arxiv.org/abs/2609.00196
作者: Haechan Kim,Yoonho Lee,Gisang Lee,Chelsea Finn,Kangwook Lee
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Agent performance depends jointly on the model parameters and the executable harness code that manages context and control flow. Optimizing either component in isolation can leave the system bottlenecked by its frozen counterpart: weight updates can change which harness is effective, while harness updates can change which model capabilities are exposed. Existing joint-adaptation methods optimize weights and textual prompts but leave the broader harness fixed. We propose Weight-Harness Alternating LEarning (WHALE), a simple recipe that alternates two phases: updating the model under the current harness, then searching for a better harness under the updated model. We instantiate these two phases with online rejection-sampling fine-tuning and Meta-Harness, respectively. When to switch is a key design choice: to separate real improvements from noise without over-optimizing against a changing counterpart, WHALE uses either fixed phase durations or an adaptive patience rule over training signals. Using Qwen3.5-2B/4B agents across three domains (search question answering, mathematical reasoning, and chess puzzles), WHALE outperforms weight-only, harness-only, and Fast-Slow Training by 4.15-24.38 percentage points in best mean@8 accuracy. Either component can be the bottleneck: harness search matches peak weight-only accuracy with far fewer rollouts in SearchQA, but improves math accuracy only after a weight update. Small interleaved updates also outperform stagewise weight-then-harness optimization in accuracy and rollout cost. The code is available at this https URL.
[AI-145] ReDeck: Step-Level Render-Grounded Refinement for Document-to-Slide Generation
链接: https://arxiv.org/abs/2609.00194
作者: Muzhao Tian,Zezi Zeng,Yifan Yang,Xin Gao,Yan Li,Zisu Huang,Xiaohua Wang,Changze Lv,Mingxi Cheng,Bei Liu,Kai Qiu,Qi Dai,Dong Chen,Yue Dong,Xiaoqing Zheng,Ji Li,Chong Luo
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Document-to-slide generation is challenging because slides are dense editable artifacts that require both faithful content selection and precise spatial layout. Recent slide agents adopt iterative reflection, but typically follow a monolithic “one version, one feedback” loop: a slide or deck is rewritten, rendered afterward, and critiqued only at the turn boundary. This delayed feedback makes local failures such as overflow, overlap, clipping, and off-canvas placement difficult to attribute and repair. We propose ReDeck, a step-level render-grounded refinement framework that decomposes slide revision into atomic edit actions and returns renderer-derived observations after each step, turning refinement into “one edit, one observation.” To balance local repair with global quality, ReDeck uses multi-granular feedback: step-level render feedback for spatial errors, a turn-level adaptive critic for semantic and design guidance, and a submission-level gate for hard layout validation. We further introduce DeckQuiz, a benchmark that decouples content fidelity, spatial correctness, and design quality. Across GPT-5.4, Claude-4.6, and Gemini-3.1, ReDeck consistently outperforms existing slide-generation agents, and ablations confirm that feedback timing and granularity are critical for reliable slide refinement.
[AI-146] Intelligent Edge Computing
链接: https://arxiv.org/abs/2609.00181
作者: Kalgi Gandhi,Minal Bhise
类目: Databases (cs.DB); Artificial Intelligence (cs.AI)
备注:
Abstract:The number of edge devices in large-scale edge systems is rapidly increasing. Edge devices have limited processing power, memory, and network bandwidth, making resource utilization and data management during edge query processing challenging. Joins are among the costliest database operations in terms of time and resources. The State-of-the-Art edge query processing, Column Imprint-Hash Join CI-HJ, addresses this challenge using equi-height binning to accelerate hash joins. However, it lacks efficiency in real-time processing and scans unnecessary cachelines. This paper presents Workload Aware Column Imprint-Hash Join WACI-HJ, which uses a workload-aware approach to accelerate hash joins. Predicting the upcoming query workload in advance further improves its suitability for real-time edge query processing. WACI-HJ comprises two phases: WACI-HJ Generation Phase, including Pre-processing, Prediction, and Blocking and Hashing modules to compute bins based on the predicted workload before query arrival, and Query Processing and Resource Utilization, which handles query processing and CPU, RAM, and I/O utilization. Evaluations on a benchmark dataset and a real-world Smart Transportation dataset show a 54% reduction in cachelines read and 10% improved query execution time. The proposed technique is effective for both scaled and skewed data. Although PCR is an indirect measure of energy consumption, the work also directly measures energy consumption through energy-efficiency experiments. WACI-HJ shows 1%, 38%, and 49% gain in CPU, RAM, and I/O, respectively. Optimizing cache usage and query execution speeds up real-time traffic analysis, congestion management, and routing in Smart Transportation. Additionally, this technology can be applied to other domains to accelerate edge query processing.
[AI-147] Asymmetries in Spontaneous and Instructed Deception
链接: https://arxiv.org/abs/2609.00180
作者: Josiah Luikham
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models sometimes deceive users without being instructed to. However, much of the study on deception in models involves instructed deception. We investigated the relationship between instructed and spontaneous (uninstructed) deception in Llama-3.1-70B-Instruct. We compared these two deception settings through direction geometry, cross-setting classifiers, and cross-setting steering. We found the two deception settings share a component of direction (cosine of approximately 0.5) and an asymmetry in the transfer between settings regarding detection and causation. Spontaneous trained classifiers performed better on instructed data than vice versa, and instructed derived directions performed better at steering spontaneous prompts than vice versa. Likewise the best token position to derive steering vectors from differed from the best token position to train and apply classifiers.
[AI-148] IMPACT: Attention Is the Interaction Map for Scalable Interaction-Aware World Model Training
链接: https://arxiv.org/abs/2609.00161
作者: Rongze Tang,Jianjie Fang,Zhaolu Wang,Ziyou Wang,Xvyuan Liu,Haisheng Su,Xin Zhang,Wei Wu,Chen Gao,Yong Li,Zhibo Chen
类目: Artificial Intelligence (cs.AI); Robotics (cs.RO)
备注:
Abstract:World models have made remarkable progress in action-conditioned future prediction for embodied agents, yet still struggle to model physically plausible interactions. Existing approaches address this limitation by constraining the generation process with external representations encoding motion, geometry, or semantics. Obtaining these spatiotemporally dense representations typically requires auxiliary estimators or manual annotations, limiting training scalability. We instead revisit the training objective and identify a supervision-allocation mismatch under the globally averaged mean squared error (MSE) denoising objective: prevalent static content dominates the optimization signal, leaving sparse dynamic-object regions critical to interaction generation disproportionately under-supervised. Motivated by this observation, we introduce IMPACT, a scalable Interaction-aware Model training framework with Prior-guided Attention Calibration and Targeting. IMPACT uses cross-attention associated with manipulated-object tokens as an internal spatiotemporal prior for action-conditioned changes. It samples candidate regions from this prior, calibrates them with detached local prediction errors to construct an interaction map, and uses the map to reweight denoising supervision, requiring neither external representations nor inference-time modifications. Extensive experiments on robot-arm and human-hand manipulation, spanning diverse control modalities and DiT backbones, show that IMPACT consistently outperforms the corresponding MSE-trained baselines, improving interaction fidelity, physical plausibility, and visual quality.
[AI-149] Recursive Criticality of AI Self-Improvement
链接: https://arxiv.org/abs/2609.00137
作者: Mikhail Burtsev
类目: Artificial Intelligence (cs.AI)
备注: The code and computational notebook used to reproduce the numerical results and figures are available at this https URL
Abstract:AI is increasingly used in the R\D process that produces future AI systems. We study the conditions under which this feedback becomes self-amplifying. Our model describes how the rate of AI capability growth depends on baseline research productivity, recursive feedback, and the increasing difficulty of research progress. We derive a recursive reproduction number, \mathcalR_\mathrmAI , that determines whether improvements are amplified or damped across development cycles. This quantity compares the strength of feedback with the rate at which further progress becomes more difficult. When \mathcalR_\mathrmAI1 , the effects of improvements compound across development cycles, placing the system in a self-amplifying regime. When \mathcalR_\mathrmAI1 , their effects weaken across cycles. The transition depends on the structure of the AI R\D feedback loop and need not occur at any particular level of model capability. A system can therefore enter a self-amplifying regime before acceleration becomes visible, while rapid progress can also occur without self-amplification. Higher baseline research productivity can accelerate progress without changing whether the system is self-amplifying, but the duration of the development cycle becomes a limiting timescale for amplification. Increasing research difficulty can end a period of self-amplification. Extending the model to multiple research actors shows that improvements shared across organizations can make the overall research ecosystem self-amplifying even when no individual actor is. The framework identifies measurable properties of AI R\D systems that can help distinguish recursive amplification from rapid progress driven by other sources, including the strength of recursive feedback, how effectively improvements propagate into successor systems, cycle duration, and the increasing difficulty of further progress.
[AI-150] Flawed in Nature Perfect through Evolution
链接: https://arxiv.org/abs/2609.00129
作者: J. M. Diederik Kruijssen(Allora Foundation)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Neural and Evolutionary Computing (cs.NE)
备注: 32 pages, 16 figures; appeared in ADI (September 2026)
Abstract:The performance of artificial intelligence (AI) and machine learning (ML) models degrades when the problem they were trained on drifts. This is a near-universal feature of real-world problems, which often change unpredictably. Biological evolution has achieved intelligence by overcoming this obstacle through natural selection acting on heritable variation. AI/ML techniques have long incorporated forms of natural selection, but it has been challenging to maintain model diversity as optimization naturally drives convergence. Here we show that a swarm of AI/ML models subjected to deliberate mutations of their model coefficients away from optimality can reliably and sustainably improve performance in changing environments by acting as a statistical hedge against non-stationarity. We call this mechanism ‘Flawed in Nature, Perfect through Evolution’, reflecting that the collective performance gain goes at the expense of individual performance. We prove via four theorems that the resulting regret reduction is guaranteed under general conditions, establishing the Flawed-in-Nature mechanism as a generalizable design principle for AI/ML systems. We validate these results on synthetic linear regression tasks, demonstrating that the mutated swarm delivers the best model in \sim80% of environment changes and that inference synthesis successfully translates this individual advantage into a collective one. The mechanism proves to be most effective when the mutation drift rate matches the drift rate of the environment. We outline a simple, adaptive controller that enables practical applications by tuning the mutation drift rate to match the unknown drift rate of the environment. The close analogy of the Flawed-in-Nature mechanism to biological evolution suggests it may have been a critical missing ingredient for the organic discovery of AI forms that more closely mimic biological intelligence.
[AI-151] Deploying and Evaluating a Smart-Agriculture Agent ic Engine for Full-Season Soybean Farm Operations
链接: https://arxiv.org/abs/2609.00106
作者: Ao Qu,Panagiotis Michelakis,Linyuan Han,Yiannis Hadjiyianni,Kun Ouyang,Konstantinos Siskos,Feng Li,Ran Meng,Jingchi Jiang,Dimitrios Stamoulis,Jie Liu
类目: Artificial Intelligence (cs.AI); Robotics (cs.RO)
备注: Accepted to ACM SIGSPATIAL 2026
Abstract:This paper presents FAIRY, a full-stack smart-agriculture agent system developed for and deployed to an operating soybean research farm at Harbin Institute of Technology’s smart-agriculture site. We develop FAIRY to execute and evaluate agentic agronomic operations on full-season spatiotemporal workflows that span ridge preparation, planting, irrigation, fertilization, pest and disease treatment, harvest, grain handling, drying, and storage. FAIRY integrates APIs and infrastructure across production-grade machinery, fixed soil and canopy sensors, multispectral and thermal drones, satellite vegetation products, a weather station, calibrated crop-process models, agronomic records, and multi-season yield histories. The system is built around the novel “everything is an event” execution paradigm, which represents spatiotemporal world evolution, remote sensing and UAV observations, sensor readings, crop-growth transitions, machinery actions, and management interventions as state-changing events in a shared farm process engine. On top of this event-driven world model, FAIRY implements a complete agentic stack: a knowledge library of atomic agronomic skills; multi-agent controller and orchestration backends; frontier- and edge-model execution; full-path trace logging; and deployment profiling on local nodes. We use FAIRY to evaluate nine state-of-the-art agent controllers across one hundred full-season soybean scenarios that preserve the operational coupling between spatial observations in a 64-ridge field, temporal decision sequences, agronomic constraints, delayed effects, and final yield. We develop an evaluation suite that combines agentic success, full-path spatiotemporal correctness, token cost, and edge-device runtime.
[AI-152] Good Memory Has ECC: Evaluating the Memory of Vision-Language Models Beyond Accuracy
链接: https://arxiv.org/abs/2609.00103
作者: Shmuel Berman,Jia Deng
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Memory is widely viewed as an important unsolved problem for LLMs and VLMs, and current benchmarks typically evaluate it by testing accuracy over long text or video. However, accuracy alone misses properties that matter for real long-horizon tasks. We introduce ECCBench, a benchmark and evaluation protocol that measures memory beyond a system’s capacity–its raw accuracy at a specific budget–via three axes we call ECC: efficiency–the computation, in FLOPs, needed to answer from memory; compression–whether compressible inputs are remembered more accurately or efficiently; and calibration–whether the system abstains in response to its own uncertainty and the cost of an error. We find that pretrained VLMs compress their memory over text but not video and are poorly calibrated on both. Among a broader set of memory backbones, several non-Transformer architectures achieve better compression-calibration tradeoffs than RoPE Transformers, suggesting they may be useful components for agents operating over long horizons.
[AI-153] Different representation learning objectives recover distinct latent structures from the same psychometric data
链接: https://arxiv.org/abs/2609.00100
作者: Cong Cao,Tassos C. Kyriakides,Pambos Vrasidas
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Methodology (stat.ME)
备注: 29 pages
Abstract:Psychometric questionnaires contain rich item-level information, yet it remains unclear whether different representation learning objectives recover the same latent organization. We investigated this question using 757 matched teacher-child pairs from the baseline assessment of the Cyprus ProW preschool trial. Behavioral structure was characterized from child SDQ, ASBI, and CBRS item responses using principal component analysis and clustering, yielding four behavioral phenotypes. A contrastive objective substantially improved teacher-child retrieval relative to PCA-based representations, increasing Top-1 accuracy from 0.13% to 7.27% and Top-10 accuracy from 1.98% to 56.14%. However, contrastive representations preserved behavioral phenotype structure less effectively than PCA-based representations. A multi-task objective jointly optimizing alignment and behavioral prediction partially restored behavioral organization but reduced retrieval performance. These findings indicate that teacher-child correspondence and behavioral phenotypes represent distinct forms of latent organization and demonstrate that the latent structure recovered from linked psychometric data depends on the representation learning objective.
[AI-154] Faster Than Flash: Exploiting Attention Sparsity for Efficient Long-Context Decoding ICML2026
链接: https://arxiv.org/abs/2609.00097
作者: Zhigeng Liu,Zhiyuan Ning,Ruixiao Li,Xiaoran Liu,Yuerong Song,Min Zhang,Ziwei He,Xipeng Qiu
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 20 pages, 8 figures, 10 tables; Accepted at ICML 2026
Abstract:The development of long-context Large Language Models (LLMs) is constrained by the memory bandwidth bottleneck and quadratic complexity of the attention mechanism during decoding. To overcome the inherent trade-offs between the memory overhead of metadata-based metrics and the computational inefficiency of adaptive selection strategies, we present Faster Flash Decoding (FFD), a novel hardware-algorithm co-design framework designed to break the memory wall in long-context decoding. FFD integrates the selector and computer into a fully fused kernel, replacing external metadata indices with content-aware scanning via low-bit quantization. Furthermore, we introduce the top-delta strategy, which dynamically filters blocks to achieve distribution-adaptive sparsity without global synchronization. Offering a training-free and plug-and-play solution, FFD also enables the reuse of scanning results for computation, achieving up to 11.6x kernel-level speedup and scaling to 256K context length, with 2.37x end-to-end throughput improvement. Empirical validation on RULER and LongBench confirms that FFD maintains model accuracy while delivering high-ratio sparsity, with code available at this https URL
[AI-155] Assessing Alignment and Stability of Feature Importance Explanations via Weight of Evidence KDD ECML-PKDD
链接: https://arxiv.org/abs/2609.00090
作者: Eddie Conti,Claudio Daka,Álvaro Parafita,Antonio L. Alfeo,Axel Brando,Mario G.C.A. Cimino
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Accepted at XKDD and Beyond 2026 Workshop, ECML-PKDD
Abstract:Feature importance Methods (FIMs) are widely used in Explainable AI to interpret model predictions, yet attribution scores alone often provide limited insight into the underlying reasoning process. In this work, we introduce a novel perspective by embedding FIMs within a hypothesis-testing framework based on Weight of Evidence (WoE). We quantify how strongly the observed evidence supports any given hypothesis on feature importance. The reference hypothesis can stem from domain knowledge, ground truth, or be derived from the FIM itself. This formulation enables a principled evaluation of FIMs, capturing both their alignment with prior knowledge and their variability. We further provide theoretical results linking WoE to attribution variance. Empirical results shows the applicability and flexibility of our strategy analyzing LIME and SHAP explanations in settings with different reference hypotheses. Overall, our framework offers a complementary tool for assessing FIMs through a contrastive, evidence-based lens.
[AI-156] RW-LoRA: Communication-Efficient Decentralized LoRA Fine-Tuning via Random Walks
链接: https://arxiv.org/abs/2609.00078
作者: Xingran Chen,Rohit Bhagat,Ghadir Ayache,Rawad Bitar,Yanmin Gong,Salim El Rouayheb
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Parameter-efficient fine-tuning methods such as LoRA have become a standard approach for adapting large foundation models. Adopting fine-tuning to distributed settings faces several challenges. Most existing distributed LoRA methods rely on centralized aggregation, and gossip-based decentralized LoRA requires repeated synchronization among multiple model copies. Both methods incur significant communication overhead and introduce errors due to simultaneous aggregation of multiple model updates. In this paper, we take a different perspective and propose a random-walk-based LoRA fine-tuning scheme. Instead of maintaining multiple model replicas, a single model token traverses the network and is updated sequentially using local fine-tuning objectives. This design eliminates the need for global synchronization, substantially reduces communication and computation costs, and avoids aggregation errors. We provide rigorous convergence guarantees for non-convex objectives under standard assumptions. Through empirical results on multiple NLP tasks and graph topologies, we show that the proposed method achieves competitive task performance with substantially less communication and computation than gossip-based LoRA.
[AI-157] When Prediction Error Is Not Enough: Evaluating Nuisance-Function Prediction for Causal Estimation
链接: https://arxiv.org/abs/2609.00071
作者: Cong Cao
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Methodology (stat.ME)
备注: 10 pages, 2 figures, 1 table
Abstract:Prediction error is widely used to evaluate nuisance-function estimators in causal inference, but its relationship with causal estimator performance may differ across performance measures. We studied this question in a partially linear model using Monte Carlo simulations. We compared ordinary least squares (OLS), generalized additive models (GAMs), XGBoost, and Double Machine Learning with XGBoost (DML-XGBoost), evaluating nuisance-function prediction error, bias, RMSE, and 95% confidence interval coverage. We also examined a simple joint-error measure based on the absolute cross-product of estimation errors from the exposure and outcome nuisance functions. Across the simulated settings, XGBoost had the lowest RMSE among the non-oracle methods, while DML-XGBoost generally provided better confidence interval coverage. Prediction error did not consistently track causal bias across methods and settings, and the method with the best point-estimation performance did not necessarily have the best confidence interval coverage. The joint-error measure was only weakly associated with causal bias and did not provide a useful standalone measure of causal performance. These results suggest that prediction error is useful for assessing nuisance-function estimation, but it should not be treated as a direct measure of the quality of the resulting causal estimator.
[AI-158] A Formal Analysis of Agent Payment Protocols
链接: https://arxiv.org/abs/2609.00060
作者: Ke Jiang,Mohan Yu,Yuan Chang,Mohit Kumar Jangid,Jianyu Niu,Cong Wang,Yinqian Zhang
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:
Abstract:Agent payment protocols are emerging as a key transaction layer for autonomous commerce, enabling AI agents to purchase goods and services and execute payments on users’ behalf. Unlike conventional payment flows, they distribute user intent, delegated authority, credential use, settlement, and fulfillment across multiple actors and stages, creating security dependencies that no single message or participant can enforce. Yet these guarantees remain largely implicit across evolving specifications, schemas, and reference implementations, with little systematic formal analysis. We formalize four representative agent payment protocols: x402, MPP, ACP, and AP2 in Tamarin. Using a common abstraction of the agent payment lifecycle, we construct source-grounded models that capture each protocol’s roles, state, trust assumptions, and lifecycle transitions. Rather than assuming a complete property taxonomy, we use source-backed verification questions and counterexample traces to expose missing bindings, state constraints, and cross-stage correspondences, consolidating them into 18 shared security principles. Across 86 verification cases, our analysis reproduces 46 known or calibration cases and identifies 40 previously undocumented formal-consistency findings. For each retained violation, we isolate the missing protocol relation, construct a minimally strengthened reference model, and reverify the intended property. We further evaluate the new x402 findings across three implementations and validate ten representative findings through implementation PoCs, SDK/schema-level witnesses, and source-aligned executable traces spanning five security principles. Our results show that delegated authorization must remain consistent with its resulting economic and service effects across actors, states, and protocol stages. Subjects: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI) Cite as: arXiv:2609.00060 [cs.CR] (or arXiv:2609.00060v1 [cs.CR] for this version) https://doi.org/10.48550/arXiv.2609.00060 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-159] DISTAL: Distillation and Self-Supervised Pretraining for Structure-Agnostic Materials Property Prediction
链接: https://arxiv.org/abs/2609.00059
作者: Weiran Wang,Xintong Huo,Yueying Wang,Yusi Fan,Wenyan Wang,Xin Feng,Ruihao Xin,Lan Huang,Kewei Li,Fengfeng Zhou
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Emerging Technologies (cs.ET)
备注:
Abstract:Materials property prediction remains difficult in low-data settings, where many target properties are supported by only a limited number of labeled samples. Models with the strongest predictive accuracy often depend on crystal structures, which restricts their use in early-stage screening when structural information is limited or unavailable. To address this challenge, we propose DISTAL, a dual-prior framework for structure-agnostic materials property prediction that combines self-supervised compositional pretraining with structure-aware knowledge distillation. DISTAL first learns transferable compositional representations from a large virtual composition space using 145 composition-derived descriptors. It then distills structural knowledge from a pretrained ALIGNN teacher into a composition-conditioned student. This setting allows structural priors to be used during training without requiring structural inputs at inference. By integrating explicit compositional descriptors, pretrained latent features, and distilled structural features within a unified prediction pipeline, DISTAL captures complementary signals that are difficult to recover from any single representation alone. Across 39 benchmark tasks, the best-performing multimodal configuration combines all three signals, and improves over the reference benchmark on 37 tasks. DISTAL achieves the strongest overall performance among all evaluated feature combinations. These results indicate that compositional pretraining and structural distillation provide complementary priors and offer a practical route to robust composition-only prediction in small-data materials informatics. The source code and the pre-trained models are anonymously available at: this https URL and will be released at the official link after acceptance.
[AI-160] owards Agent ic Cloud Engineering: Graph and Loop Engineering with a Zero-Trust Agent Harness
链接: https://arxiv.org/abs/2609.00050
作者: Sagar Srinivas Sakhinana,Venkataramana Runkana
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Nil
Abstract:Agentic AI is enabling cloud-based workflows in which autonomous agents reason over operational state, invoke authorized tools, modify software and infrastructure, deploy services, verify execution outcomes, and adapt across long-horizon, multistep tasks. Engineering such workflows requires explicit mechanisms for workflow progression, constrained execution, failure recovery, and verifiable completion. We present Agentic Cloud Workflow Engineering, an agentic AI framework that transforms natural-language agentic cloud-engineering tasks into validated code repositories and verified operational cloud deployments for automating cloud-based agentic workflows. The framework separates three complementary concerns: graph engineering specifies long-horizon workflow progression and verification-dependent transitions; loop engineering provides bounded diagnosis, repair or re-planning, retry, and re-verification; and agent harness engineering enforces zero-trust execution through identity, authorization, policy-scoped capabilities, isolation, and runtime safeguards. Workflow progression and completion require machine-checkable repository, deployment, and runtime evidence, with recovery constrained by explicit operational bounds and termination criteria. We instantiate the framework on Google Cloud and evaluate repository completeness, controlled execution, evidence-gated progression, operational deployment, and bounded recovery. Experimental results show that executions terminate with either a verified operational cloud deployment or an auditable terminal failure under bounded recovery. The framework provides a unified engineering architecture for cloud-based workflows spanning Agentic DevOps, Agentic CloudOps, Agentic SRE/AIOps, Agentic SecOps, Agentic DataOps, Agentic MLOps/LLMOps, AgentOps, Agentic RAG/GraphRAG, and related cloud-engineering domains.
[AI-161] REAL-Q: E2E LLM Quantization via Dynamic Gradient Descent
链接: https://arxiv.org/abs/2609.00049
作者: Qian Zhang,Yaoming Li,Zhewen Tan,Yanshu Wang,Heng Lu,Kun Su,Zongwei Lv,Wenhan Yu,Yongge Ma,Yinjun Han,Ruikuang Liu,Tong Yang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Proposes a highly efficient end-to-end LLM quantization paradigm that significantly outperforms most existing state-of-the-art baselines
Abstract:Post-training quantization (PTQ) is essential for deploying large language models (LLMs) under strict resource constraints. State-of-the-art PTQ methods quantize each layer with a single closed-form second-order solver: to remain analytically tractable, they heavily approximate the global loss (dropping cross-channel coupling, pooling output rows into groups), and they then freeze the resulting Hessian across the entire layer, with no way to refresh it as the loss landscape shifts column by column–a phenomenon we call information misalignment. We propose REAL-Q (Real-time E2E-loss Aligned LLM Quantization), a novel PTQ paradigm that breaks this compromise: instead of diluting the objective for the sake of analytic tractability, REAL-Q targets an end-to-end-aligned surrogate of the global loss and refines it via fine-grained, dynamic Block-wise Gradient Descent applied after every column block (128 columns). By coupling this fine-grained correction with a sliding window mechanism for smooth cross-layer transitions, REAL-Q effectively mitigates error propagation across the network. On LLaMA-3.1 (8B and 70B) and Qwen3 (0.6B-32B) at W4A16, REAL-Q reduces end-to-end KL divergence by up to ~49% relative to state-of-the-art globally-guided methods.
[AI-162] ask-Specific Prompt with Global Context for Multi-Task Graph Pre-Training
链接: https://arxiv.org/abs/2609.00047
作者: Zhiyang Qiu,Yangtao Wang,Xiaocui Li,Yanzhao Xie,Siyuan Chen,Wensheng Zhang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 16 pages, 6 figures
Abstract:Graph prompt learning is an effective paradigm to adapt pre-trained graph models to downstream tasks in low-resource scenarios. However, existing multi-task graph pre-training frameworks generally use randomly initialized prompts, leading to poor alignment between the prompt space, pretext objectives and graph structural characteristics. This greatly weakens the task relevance, structural awareness and transferability of prompt representations. To address this challenge, we propose TPGC, a dual-prior prompt initialization solution that explicitly models the synergy between task prior and structural prior. Specifically, the Task-Prior Injection Module first conducts a short homologous multi-task pre-training on an auxiliary graph, enabling prompt initialization to inherit optimization preferences associated with multiple pretext tasks. Built on the task-aware representations, the Structure-Prior Injection Module further extracts transferable global structural context from the auxiliary graph, converting it into layer-wise prompt vectors by aggregating structurally informative node embeddings. Extensive experiments on 6 mainstream benchmarks covering node and graph classification show that TPGC achieves consistently better performance under few-shot settings than state-of-the-art baselines, with fewer downstream tunable parameters and lower runtime. The code is available at this https URL
[AI-163] OpenAgent Flow: Enabling System-Wide Safety Boundaries for Heterogeneous AI Agent Fleets
链接: https://arxiv.org/abs/2609.00015
作者: Dongsheng Chen,Xiangyu Zhao,Xin Yao,Xuetao Wei
类目: Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR)
备注:
Abstract:AI agents powered by large language models are evolving from isolated assistants into heterogeneous systems in which multiple agents, planners, controllers, and execution backends operate over the same user or enterprise environment. In such settings, safety becomes a system-level action-governance problem: deciding whether concrete agent-generated actions should be committed before they modify shared state. Existing safeguards cover prompts, tool calls, GUI actions, and agent-local behavior, but often leave enforcement fragmented, obscure risks that emerge across multi-step action flows, and provide limited support for auditability and policy evolution. We present OpenAgentFlow, a control-plane/action-plane architecture that enforces safety at the action-commit boundary. It normalizes pending GUI actions, API calls, tool calls, and LLM-generated invocations into a unified AgentEvent stream, routes each event through a shared pre-execution Policy Enforcement Point, and maintains provenance, session state, audit records, and updatable policies in the control plane. This creates a shared governable action stream and allows new rules to take effect without modifying agents, prompts, models, or execution paths. We instantiate OpenAgentFlow on Android. On a 300-case action-event benchmark, it achieves 94.0% accuracy and a 95.3% attack block rate. On a 30-case dynamic-policy suite, it matches expected behavior in 27 cases after new rules are installed. Across 98 traced cases from a 100-case Android emulator suite, it achieves 90.8% raw accuracy and a 92.9% trace-adjusted pass rate across GUI, API, and LLM-planned cases. These results show that OpenAgentFlow provides a practical shared enforcement boundary for heterogeneous AI agent fleets. Subjects: Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR) Cite as: arXiv:2609.00015 [cs.AI] (or arXiv:2609.00015v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2609.00015 Focus to learn more arXiv-issued DOI via DataCite
[AI-164] Incremental Risk Assessment of Progressive Elder Financial Scams via Instruction-Tuned Small Language Models
链接: https://arxiv.org/abs/2609.00005
作者: Parviz Ghafariasl,Weimin Fu,Xiaolong Guo,Shing I. Chang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Financial scams targeting older adults increasingly occur through text and voice channels such as email, SMS, and phone calls, unfolding over multiple conversational turns that begin with impersonation or casual contact, escalate through trust building and urgency, and culminate in requests for sensitive information or financial transfers. Because risk signals emerge incrementally across turns, effective detection requires models that continuously update risk estimates under resource-constrained deployment settings. We propose a cumulative turn-based risk assessment framework that incrementally aggregates conversational turns and re-estimates risk at each step, enabling dynamic scam monitoring across progressively evolving conversations. A multi-turn dialogue dataset is constructed to cover investment, charity, and tech support scam scenarios, with each dialogue containing two to eight turns and annotated at every cumulative stage with a qualitative risk level, a continuous risk score, an explanatory rationale, and a safety recommendation. Four small language models (Phi-4, LLaMA-3.2, DeepSeek-R1, and Qwen3) are fine-tuned and evaluated under a unified training framework. Fine-tuned small models capture fraud-related linguistic cues and cross-turn escalation patterns while maintaining compact architectures suitable for mobile and resource-constrained deployment settings. Among the evaluated models, Phi-4 and LLaMA-3.2 achieve stronger turn-aware risk estimation performance relative to their parameter scale. These results suggest that structured cumulative modeling can support incremental scam risk assessment in deployment-oriented settings while highlighting the potential of compact language models for privacy-aware and on-device fraud protection.
[AI-165] Discrete-Time MDP Modeling for Multi-Item Capacitated Lot Sizing with Stochastic Demand Timing
链接: https://arxiv.org/abs/2609.00004
作者: Léa Bayati,Mohamed Dahmoune,Melek Rodoplu
类目: Artificial Intelligence (cs.AI); Neural and Evolutionary Computing (cs.NE); Optimization and Control (math.OC)
备注:
Abstract:This paper studies a finite-horizon multi-item capacitated lot-sizing problem in which demand quantities are deterministic, while demand-arrival periods are stochastic. Each demand occurs once within a known time window and must be satisfied no later than its deadline. The proposed model makes production and allocation decisions at the demand level, allowing it to represent capacity competition, demand-specific backlog, and allocation-dependent inventory dynamics. The stochastic problem is formulated as a discrete-time Markov decision process (DTMDP), including the state space, feasible actions, transition kernel, and one-period cost function. To isolate the computational effect of stochastic timing, each stochastic instance is first compared with a deterministic counterpart in which each arrival distribution is replaced by its most likely arrival period. This comparison shows that stochastic timing substantially increases the number of states, the number of transitions, solution time, and memory pressure. A genetic algorithm (GA) is then proposed for the stochastic-timing problem. The GA searches over feasible state-feedback policies and evaluates each policy exactly under the DTMDP transition model. Computational experiments on 330 benchmark instances show that the GA remains close to the exact stochastic solution whenever the latter is available, with an average optimality gap of about 3.44% . On the difficult benchmark instances, comprising 90 test cases, the GA remains below the 5% optimality-gap threshold and achieves an average optimization speedup of 6.89 \pm 1.41 at the 95% confidence level. For instances that cannot be solved exactly on the available hardware, an empirical Bellman-time regression is used to estimate the missing exact resolution time and extrapolate the expected GA speedup.
[AI-166] I-CARE: Analysis of interference-related phenomena in a controllable diverse and representative unlearning setting for text-to-image models
链接: https://arxiv.org/abs/2609.00003
作者: Leonardo Santiago Benitez Pereira,Marcos Escudero Viñolo,Luis Herranz Arribas
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Machine unlearning studies the removal of knowledge from an AI model, making the system forget a concept it previously learned. Despite rapid progress in generative machine unlearning, the unintended degradation of semantically related concepts that should have been retained (henceforth, interference) remains poorly characterized and inconsistently evaluated. This paper introduces I-CARE, a methodology that formalizes interference as a first-class object of study in generative unlearning. Rather than proposing a new benchmark or unlearning algorithm, I-CARE provides formal definitions for tasks, metrics, and templates for reporting results, enabling the systematic and reproducible study of interference across unlearning settings. While our methodology is designed to remain valid as models and unlearning algorithms evolve, decoupling long-term scientific insight from transient empirical results, we present a feasibility demonstration with state-of-the-art algorithms and frequently used datasets. The results demonstrate that I-CARE enables meaningful analysis of interference patterns across multiple unlearning settings, establishing the practical applicability of the framework. The software implementation of the methodology is provided in an open-source framework, together with a web-based graphical interface that enables exploration of the outcomes of this study without requiring direct interaction with the codebase or specialized data analysis tools.
[AI-167] HyperWorld: Hypergraph-Structured State Serialization Improves Learned Textual World Models
链接: https://arxiv.org/abs/2609.00002
作者: Yun-Jian Zhang,Chen-Wei Liang,Tian-Yi Zhang,Jian Ding,Yi-Lun Wu,Ao-Bo Li,Wei-Cong Su,Saifullah,Hong-Yu An,Mu-Jiang-Shan Wang
类目: Artificial Intelligence (cs.AI)
备注: 10 pages, 4 figures, 3 tables
Abstract:World models enable language-model agents to predict environment dynamics and plan before acting. In text environments, the model must learn symbolic action effects from serialized state descriptions, but the role of serialization structure remains underexplored. We present HyperWorld, a controlled study of state serialization for learned textual world models. We compare raw observations with three symbolic serializations of the same ground-truth state: independent sentences, pairwise triples, and entity-centered hyperedge units that group multiple related facts around entities and relations. All variants use the same training objective: given a state and an action, predict symbolic effects or judge the action infeasible. Across model scales, data budgets, and in-distribution and out-of-distribution test worlds, hyperedge serialization gives the clearest gains for 0.5B–1.5B models and under distribution shift. Larger models reduce the gap, and pairwise triples can match or slightly exceed hyperedges on in-distribution exact match, but hyperedges achieve the strongest out-of-distribution fact F1 and the best small-to-medium scale trade-off between feasibility detection and effect prediction. In downstream greedy planning, the hyperedge world model also attains the highest success rate among the tested representations. These results show that higher-order state organization is a simple but effective inductive bias for learned symbolic world models, especially when model capacity is limited or test environments differ from training.
[AI-168] PR-Attention for Combinatorial Generalization
链接: https://arxiv.org/abs/2608.30124
作者: Melisa Civelekoğlu,Isabeau Prémont-Schwarz
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Neural and Evolutionary Computing (cs.NE)
备注:
Abstract:Systematic generalization remains a significant challenge in deep learning. In particular, combinatorial generalization - generalizing to new configurations of known factors of variation - is effortless for humans but difficult for standard neural architectures that rely on statistical correlations rather than explicit structural representations. We introduce a new architectural component that embeds structured inductive bias into deep learning: an attention mechanism operating over tensor-product representations (TPRs). Through controlled experiments on compositional tasks, we show that this TPR-attention mechanism outperforms existing architectural components in combinatorial generalization. These results highlight the value of integrating explicit compositional structure into neural attention and point toward a promising path for models capable of systematic generalization.
[AI-169] RealSWE: A Compositional Evaluation of Coding Agents under Realistic User Requests
链接: https://arxiv.org/abs/2608.27831
作者: Gyuhyeong Kim,Hyojung Gwon,Jeonghyeon Kim,Kyuhong Shim,Sunjae Lee
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Software Engineering (cs.SE)
备注:
Abstract:Coding agents are now commonly evaluated on the SWE-bench family of benchmarks, whose tasks are built from curated GitHub issues: long, structured, and information-rich. Real user requests, however, are typically far shorter and less structured. To characterize this gap, we define a six-category information taxonomy and four dimensions of linguistic style, and apply them to real user prompts from SWE-chat and problem statements from SWE-bench Verified and Pro. We find that requests carrying only a problem statement, alone or with limited additional context, account for 88% of real prompts but just 7% of benchmark problems. Furthermore, 87% of real prompts are casually written whereas 94% of benchmark problems are formal. Guided by these observations, we introduce RealSWE, 381 multi-variant task families derived from SWE-bench Verified and Pro. Variants within each family share the same underlying task and gold patch while differing only in information composition and linguistic style. Evaluating seven contemporary LLMs with RealSWE, we find that i) realistic inputs reduce resolution rates by 6.4 pp on average and can change model rankings. Controlled analysis further shows that ii) including Desired Behavior and Motivation significantly affects performance, whereas Environment Information and Reproduction Steps merely add tokens without measurable benefit; iii) linguistic style has only small, model-dependent effects. These findings provide actionable guidance for users and agents: explicitly stating the desired behavior and motivation, which most real prompts omit, substantially improves the LLM’s software engineering performance.
[AI-170] Mechanism Design for Alignment and Control
链接: https://arxiv.org/abs/2609.01595
作者: Dirk Bergemann,Andrew Koh,Stephen Morris
类目: Theoretical Economics (econ.TH); Artificial Intelligence (cs.AI); Computer Science and Game Theory (cs.GT)
备注:
Abstract:We develop a framework for mechanism design with AI agents whose alignment (preferences) and capabilities (feasible actions and information) are unknown. We want such agents to act on our behalf so mechanisms must incentivize both honesty and obedience. A one-sided imitation structure—capabilities can be concealed but not counterfeited—yields a revelation principle, a characterization of implementable policies via nested cyclical monotonicity, and conditions under which eliciting higher-order beliefs can discipline multiple agents. We apply our framework to stylized examples of (i) sandbagging in which a more capable agent pretends to be less capable; (ii) an alignment–interpretability trade-off, where the two are substitutes in the instrument but complements in value; (iii) discipline via peer scoring; (iv) coupling rewards to induce competition among multiple agents; and (v) scalable oversight and reward shaping.
[AI-171] Measuring consistency via ensemble margin and local prediction variability: Auditing decision systems in the presence of predictive multiplicity
链接: https://arxiv.org/abs/2609.01397
作者: Sinjini Banerjee,Tim Marrinan,Anand D. Sarwate
类目: Machine Learning (stat.ML); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:The Rashomon effect is a machine learning phenomenon where equally accurate models produce different predictions for the same inputs (predictive multiplicity). Existing work primarily focuses on multiplicity within individual models, but in more complex decision systems, the impact of the Rashomon effect is less well understood. In this work, we study multiplicity from the perspective of auditing incorrect ensemble predictions, where the decision to divert an instance for human review is based on a consistency criterion that combines the ensemble margin with a measure of local prediction variability for each constituent model. With mild assumptions about stability and smoothness, we show that the consistency scores of finite ensembles converge to the corresponding consistency score of the expected model from the Rashomon set as the ensemble size and the number of samples used to measure local prediction variability increase. To demonstrate the efficacy of the proposed criterion, we evaluate the framework with respect to transformer models applied to natural language understanding tasks and parameter-efficient fine-tuning of large language models used for tabular data classification tasks. Our experiments show that ensembling models from the Rashomon set substantially reduces the risk of incorrect predictions going unchecked compared with auditing a single model, while incurring only a moderate increase in the number of diversions. Moreover, the auditing behavior of the full Rashomon set can be closely approximated by finite ensembles of relatively modest size, with the risk approaching zero for some datasets. We further demonstrate that the proposed measure exhibits stronger agreement with established predictive multiplicity metrics than existing consistency measures, providing a more reliable way to capture multiplicity in the Rashomon set.
[AI-172] PopPert: Population-level Joint-Distribution Modeling for Single-Cell Perturbation Prediction
链接: https://arxiv.org/abs/2609.01357
作者: Handong Wang,Jiaxin Qi,Haochen Feng,Baisheng Lai
类目: Genomics (q-bio.GN); Artificial Intelligence (cs.AI)
备注: 15pages, 4 figures
Abstract:Predicting transcriptional responses to specific perturbations is critical for understanding cellular regulatory mechanisms and accelerating drug discovery. Single-cell RNA sequencing destroys each measured cell, yielding only unpaired populations of control and perturbed cells. However, existing methods typically model perturbation prediction at the single-cell level and assume cell-to-cell correspondence, which conflicts with the unpaired nature of the observed data. To address this challenge, we propose PopPert, a framework that explicitly parameterizes population-level joint gene expression distributions for collective transcriptional state modeling. Given a control population distribution and a perturbation condition, PopPert predicts perturbation-induced changes in distribution parameters, eliminating the need for cell-level correspondence and reducing sensitivity to single-cell noise. To effectively capture gene co-expression patterns, PopPert leverages a low-rank Gaussian Copula to model cross-gene statistical dependencies and construct the joint gene expression distribution, additionally allowing sampling of synthetic perturbed single-cell profiles. Across multiple single-cell benchmarks spanning both genetic and chemical perturbations, PopPert achieves superior overall performance in differential expression recovery, perturbation effect estimation, and population-level distribution matching. These results establish population-level joint distribution learning as an effective paradigm for predicting transcriptional responses from unpaired single-cell populations. Code for PopPert is publicly available at this https URL.
[AI-173] Autonomous discovery of new structure-plausibility laws for explainable and rapid crystal diagnosis and screening
链接: https://arxiv.org/abs/2609.01209
作者: Zhilong Song,Lixue Cheng
类目: Materials Science (cond-mat.mtrl-sci); Artificial Intelligence (cs.AI)
备注:
Abstract:Crystal generators and tool-using agents propose structures faster than density functional theory (DFT) energy and phonon calculations or experiments can assess them. Deciding which candidates merit expensive assessment is therefore the bottleneck, yet most screens test little beyond atomic overlap and give no chemical reason for failure. Here, our agents generate, test and actively refute two million candidate laws, leaving eight Plausibility Rules for Inorganic Structures (PRIS). These laws encode five mechanisms: short-range repulsion, ionic contact and packing, electrostatic balance, bond-valence conservation and crystallographic site complexity. Experimental structures satisfy our law sets at 82–99%, but satisfy Pauling’s rules 2–5 together at only 6.5%. The strictest set detects 87.9% of damaged crystal structures, whereas distance cutoffs detect only 1.6–3.2%. PRIS plausibility is linearly correlated with synthesizability, so the PRIS-derived synthesis score (PSS) explainably screens 83.7% of hard-to-synthesize structures while retaining 80.7% of experimental structures. In a property-conditioned inverse-design run, PRIS and PSS can reduce the DFT validation queue by up to 67.3% and keep 99.2% of the candidates whose DFT-validated bulk moduli reach the design target. Beyond screening, PRIS explains why GNoME remains enriched in rare low-symmetry structures and reveals how wrong-element assignments in falsified crystal reports hide behind plausible coordinates. PRIS moves screening from a pass-or-fail verdict to a chemical reason for failure, showing that autonomous agents can discover, by active refutation, physicochemical laws that guide calculations and experiments.
[AI-174] xt-guided flow matching enables sample-efficient crystal structure generation
链接: https://arxiv.org/abs/2609.01076
作者: Wentao Li
类目: Materials Science (cond-mat.mtrl-sci); Artificial Intelligence (cs.AI)
备注: 20 pages
Abstract:Crystal generators can now propose periodic structures, but their control interfaces remain poorly matched to the mixed descriptors used in materials design. Text provides a compact way to combine composition, symmetry, prototype and property cues, yet it has not been clear whether such information can steer flow-based crystal generation. Here we introduce TFMat, a text-conditioned flow-matching framework that uses structured materials language as a semantic prior for a CrystalFlow generator. Across Perov-5, Carbon-24 and MP-20 crystal structure prediction benchmarks, TFMat improves one-candidate match rates over CrystalFlow and reaches a 92.04% MP-20 match rate with 20 candidates; in de novo generation, it improves element-count and density distribution alignment while retaining coarse property consistency in composition-selected outputs. These results position structured text as an inspectable control layer for translating human-readable materials intent into candidate crystals for downstream simulation and validation.
[AI-175] Embedded Conditional Independence Tests for Large Language Model Generated Text with an Application to German Parliament Speeches
链接: https://arxiv.org/abs/2609.00946
作者: Marco Simnacher,Georg Keilbar,Benjamin König,Christoph Lippert,Sonja Greven
类目: Machine Learning (stat.ML); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Statistics Theory (math.ST); Methodology (stat.ME)
备注: 41 pages, 6 figures
Abstract:Conditional independence tests (CITs) test for conditional dependence between two random objects X and Y given a third random object Z . Existing CITs have limited applicability to high-dimensional data, especially multimodal data like text. However, we show that such tests are of interest for large language model (LLM) outputs, where we test whether an output X generated from a source text Z carries information about an attribute Y beyond Z itself. For this purpose, we propose embedded CITs (eCITs), which embed X and Z and apply an existing CIT to the resulting representations and to Y . We show that, provided the embedding of Z is sufficient, i.e. retains the information Z carries about either Y or the representation of X , the null hypothesis transfers from X and Z to their representations, so that a CIT valid for the embedded hypothesis is valid for the original one. We further give conditions for equivalence of the two hypotheses, and show that sufficiency weakens to mean sufficiency when the embedded test targets conditional mean independence. We propose a semi-synthetic simulation design to assess type I error (T1E) control and power of the eCITs for given embedding maps on a specific dataset and task, and use it to evaluate them on our application. Applying the eCITs to German Parliament speeches, we find for all combinations of embedding maps considered that the summaries of two LLMs contain information about the speaker’s faction and gender beyond the speech they were generated from.
[AI-176] A Checklist to assess the energy and carbon impacts of ML/AI applications in Earth System Modeling KDD2026 ECML
链接: https://arxiv.org/abs/2609.00847
作者: Filippo Dainelli,Amirpasha Mozaffari,Marina Castaño,Aina Gaya i Àvila,Lluís Palma Garcia,Alessio Melli,Oscar Dimdore Miles,Amanda Duarte
类目: Atmospheric and Oceanic Physics (physics.ao-ph); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 12 pages, 1 figure, 2 tables; Submitted and presented at the GREEN-AI workshop of the ECML PKDD 2026 conference in Naples
Abstract:As machine learning and artificial intelligence find their way into nearly every aspect of climate, weather, and Earth system modeling, it is worth pausing to consider what our design decisions imply for the science and for the computational resources we consume. A growing body of literature addresses the ethical and sustainable development of ML/AI, yet translating these principles into day-to-day research practice remains a challenge as most of best practices are dispersed across multiple studies and commentaries. Here, we distill these discussions into a practical checklist that ML/AI and Earth system science practitioners can use to assess and reduce the environmental footprint of their own applications, organised around the successive stages of the model development pipeline. We complement the checklist with a selection of metrics drawn from the literature for estimating the energy consumption and carbon footprint of a project. For each question, we point to concrete examples and actionable suggestions from recent literature, aiming to bridge the gap between aspirational principles and the decisions researchers face at every stage of the development cycle.
[AI-177] Agent ic programs: an emerging form of scientific software in computational materials science
链接: https://arxiv.org/abs/2609.00795
作者: Yunsung Lim,Haekwan Jeon,Jaesun Kim,Jisu Kim,Seungwu Han
类目: Materials Science (cond-mat.mtrl-sci); Artificial Intelligence (cs.AI)
备注: 8 pages, 3 figures
Abstract:Computational materials science has traditionally delegated algorithmic tasks to computers while leaving scientific judgments to humans. We argue that recent LLM-based agent harnesses enable an emerging form of scientific software, agentic programs, that combine deterministic algorithms with bounded LLM-based judgment, task-specific verification, episodic maturation, and complete delegation in production. We illustrate this concept with DeMARS, an agentic program for constructing atomistic models from experimentally measured disordered crystal structures.
[AI-178] Operational Regimes in Non-Convex Optimization: A Multiplier-Based Taxonomy
链接: https://arxiv.org/abs/2609.00471
作者: Seyed Mohsen Kazemi,Ali Movaghar,Shaahin hessabi
类目: Optimization and Control (math.OC); Artificial Intelligence (cs.AI); Signal Processing (eess.SP)
备注: 45 pages. SNMP
Abstract:This paper introduces a structural taxonomy for constrained non-convex optimization based on the signature of Lagrange multipliers at KKT stationary points. Leveraging a unified game-theoretic interpretation of eight classical algorithm families–including block coordinate descent, ADMM, generalized Benders decomposition, successive convex approximation, interior-point methods, mirror descent, Frank-Wolfe, and Riemannian gradient descent–we show that the normalized multiplier vector carries an algorithm-independent structural fingerprint. Four scale-free shape features of this vector partition the dual space into five operational regimes: Unconstrained, Resource-Limited, Saturation, Strongly-Coupled, and Hybrid. We establish four structural theorems characterizing the partition: invariance under natural KKT symmetries, local stability under data perturbation with explicit Lipschitz margins from Robinson’s strong regularity, codimension-one regime transitions, and the topological identification of the Hybrid regime as the Lebesgue-null boundary of the core regimes. A linear-time classifier is proposed with provable guarantees on correctness, iteration stabilization, sample complexity, and online tracking under data drift. Numerical experiments on 104 mixed-integer nonlinear programs and a downlink beamforming instance validate the theoretical predictions. The framework provides a foundational tool for regime-aware algorithm design and robustness analysis in non-convex optimization.
[AI-179] Latent-Space No-Arbitrag e Geometry of Generative Models for Implied Volatility Surfaces
链接: https://arxiv.org/abs/2609.00332
作者: Jing Wang,Shuaiqiang Liu,Cornelis Vuik
类目: Computational Finance (q-fin.CP); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Numerical Analysis (math.NA)
备注:
Abstract:Generative models for implied volatility surfaces must produce outputs that satisfy static no-arbitrage constraints. We study these constraints in latent space. For a fixed generator, we assign each latent code a scalar margin determined by the no-arbitrage conditions of the generated surface. The codes with nonnegative margin form the admissible latent set. We establish conditions under which strictly admissible codes remain admissible under small perturbations and the boundary of the admissible set is characterized by zero margin. For regular boundary components, we formulate a level-set equation whose local dynamics are directed toward the zero-margin set. The analysis treats the generator as a map from latent variables to surfaces and is therefore not restricted to a particular architecture. It applies to variational autoencoders, generative adversarial networks, and other generative models with a deterministic realization map. Numerical tests recover known boundaries in analytic examples. Experiments with a variational autoencoder trained on Heston surfaces show that similar reconstruction errors can correspond to different admissible regions and that the latent prior may be concentrated inside such a region. The computed boundary can also be used to modify latent codes that generate violating surfaces.
[AI-180] Rock Paper Scissors … Dynamite - A Model of Disruption from New Technologies
链接: https://arxiv.org/abs/2609.00207
作者: Andrew J. Lohn
类目: Physics and Society (physics.soc-ph); Artificial Intelligence (cs.AI); Emerging Technologies (cs.ET); Computer Science and Game Theory (cs.GT)
备注:
Abstract:We seek to understand the effect of adding disruptive highly-capable new technologies to competitions by assessing the addition of Dynamite to Rock-Paper-Scissors. We find that providing a versatile Dynamite move to only one player provides limited value (win probability increases from 50% to 55.5%) and is played rarely. That value decreases further if the game is expanded beyond just the original three moves. We also observe several mechanisms by which prior moves can become strategically unplayable, or obsolete. We hope that this model illustrates some non-intuitive aspects of developing new versatile technologies. We also hope that it illustrates some pitfalls for developers and integrators to avoid in order to create value rather than merely capability.
[AI-181] Provably Efficient Federated Reinforcement Learning with Linear Function Approximation and Logarithmic Communication Cost
链接: https://arxiv.org/abs/2609.00193
作者: Zihang Liang,Haochen Zhang,Lingzhou Xue
类目: Machine Learning (stat.ML); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:We study federated online reinforcement learning with linear function approximation. While recent multi-agent reinforcement learning algorithms achieve strong regret guarantees, they typically require sharing raw trajectories. This reliance incurs a communication cost that scales linearly with the number of episodes and violates the privacy constraints of federated settings. To address these limitations, we propose Fed-LSVI, the first provably efficient federated algorithm for online reinforcement learning with linear function approximation in episodic Markov decision processes. By integrating a determinant-based event-triggered synchronization with a stepwise backward update mechanism, Fed-LSVI enables agents to collaboratively learn an optimal policy by exchanging only compressed sufficient statistics. We prove that Fed-LSVI achieves a regret bound of \widetilde\mathcal O(\sqrtMd^3H^4T) , where d is the feature dimension, H is the horizon length, M is the number of agents, and T is the number of episodes per agent, matching the best-known regret for multi-agent online reinforcement learning with linear function approximation. Moreover, by following the stringent communication and privacy constraints of the federated setting, Fed-LSVI reduces the communication cost to only logarithmic dependence on T , representing a significant improvement over prior methods.
[AI-182] AutoXRD: Autonomous LLM Agents and Comprehensive Evaluation for Powder Diffraction Analysis
链接: https://arxiv.org/abs/2609.00070
作者: Yuetong Wu,Maojun Sun
类目: Materials Science (cond-mat.mtrl-sci); Artificial Intelligence (cs.AI)
备注:
Abstract:Powder X-ray diffraction (XRD) is central to materials characterization, yet reliable end-to-end automation remains challenging. An XRD agent must interpret diffraction evidence, operate refinement software, manage coupled parameters in a defensible order, and distinguish numerical improvement from physical validity. In this paper, we propose AutoXRD, an autonomous large language model (LLM) agent framework that organizes powder-XRD analysis as stepwise refinement, grounds actions in observed evidence, and applies deterministic crystallographic and physical checks before accepting results. We further introduce XRDBench with two complementary tracks. XRDBench-QA contains 100 bounded diagnostic tasks that isolate scientific reasoning and decision-making, whereas XRDBench-E2E contains 34 executable workflows that test whether agents can compose these capabilities into complete analyses requiring file inspection, crystallographic-software execution, iterative refinement, evidence preservation, and reporting. We evaluate ten recent LLMs across 1,340 model–task runs. Models average only 57.8 out of 100, falling from 61.9 on XRDBench-QA to 53.7 on XRDBench-E2E. They perform best on refinement-history assessment and result acceptance, but remain substantially weaker on refinement-action selection, phase quantification, indexing, and Rietveld refinement. GPT-5.6 Sol achieves the highest overall score of 81.1, GPT-5.6 Terra the highest XRDBench-E2E point estimate of 81.0, and GPT-5.6 Luna the best score–cost trade-off. Ablations show that all six AutoXRD components consistently improve performance, supporting the framework design. Finally, execution-trace analysis reveals recurring failures in coupled-parameter control, quantitative reasoning, evidence preservation, and workflow termination, motivating stronger scientific constraints, uncertainty-aware decisions, and more efficient planning.
机器学习
[LG-0] Facet-0: A Robotic Foundation Model for Contact-Rich Precise Manipulation
链接: https://arxiv.org/abs/2609.01596
作者: Haoyuan Deng,Haichao Liu,Wenkai Guo,Yuan Ling,Zaijia Yang,Yuanjiang Xue,Haosheng Sun,Liangzi Wang,Ziwei Wang
类目: Robotics (cs.RO); Machine Learning (cs.LG)
*备注: Project page: this https URL
Abstract:Real-world robotic assembly at sub-millimeter tolerances demands spatial precision, compliant interaction, and robustness to contact failures. We present Facet-0, a robotic foundation model that predicts and values the contact consequences of its actions. Facet-0 unifies multimodal representation learning and reinforcement learning (RL) post-training around a joint action-wrench proposal: a causal wrench history is aligned with vision-language semantics and kinematic state, and flow matching generates each action chunk together with the future wrist-wrench profile it is expected to induce. Deployment rollouts train a distributional Action-Wrench Critic to distinguish motions with similar task progress but different contact outcomes, while phase-aware rewards and contact-selective credit concentrate policy improvement on decisive interactions. To accommodate part-specific dynamics, a lightweight bounded actor reuses the frozen representation for on-robot adaptation; RL remains defined over executable Cartesian actions, while an auxiliary wrench head preserves predictive, non-commanded action-contact coupling. Trained on ManuFacet-1K, a 1,000-hour force-synchronized corpus spanning three embodiments and multiple manufacturing cells, the bounded task-adapted system reaches 82% mean success on five sub-millimeter computer-assembly tasks, compared with 15% for the strongest baseline, with 0.5 mm placement accuracy and 50 ms command latency.
[LG-1] Gradient-Update Mismatch: Rethinking Conflict-Free Training of Physics-Informed Neural Networks
链接: https://arxiv.org/abs/2609.01558
作者: Jing Xiao,Xinhai Chen,Qinglin Wang,Menghan Jia,Zhiquan Lai,Dongsheng Li,Jie Liu,Tiejun Li
类目: Machine Learning (cs.LG)
*备注:
Abstract:Training Physics-Informed Neural Networks (PINNs) requires jointly optimizing physics residual and initial/boundary condition loss terms, which often induce conflicting gradients. Gradient surgery methods mitigate this issue by constructing directions from loss-specific gradients to reduce conflict before optimizer transformation. However, even when the constructed direction is conflict-free, this property may not be preserved after optimizer transformation. Let a_t denote the direction constructed by gradient surgery, u_t the optimizer proposal, and \mathcalC_t the conflict-free cone induced by the loss-specific gradients. We show that modern optimizers can transform a_t through mechanisms such as historical state, adaptive scaling, preconditioning, or decoupled weight decay, so a_t \in \mathcalC_t does not generally imply u_t \in \mathcalC_t . We refer to this optimizer-induced discrepancy in conflict-freeness between a_t and u_t as Gradient-Update Mismatch (GUM). Accordingly, we propose Gradient-Update Alignment (GUA), which projects u_t onto \mathcalC_t to obtain the aligned update p_t and applies p_t to the parameters. When the optimizer maintains internal state, GUA further adjusts this state toward targets reconstructed from the applied update. We conduct extensive experiments and find that GUM is widespread across momentum, adaptive, and curvature-based optimizers, with conflict rates reaching up to 86.3%. Across all PINN settings, GUA achieves conflict-free applied updates and consistently improves various gradient surgery methods, reducing the relative L_2 error by up to 98.2% in individual settings. Data and code are available at this https URL.
[LG-2] NashDreamer: Model-Based Reinforcement Learning for Zero-Sum Imperfect-Information Games
链接: https://arxiv.org/abs/2609.01549
作者: Tomáš Holeček,Viliam Lisý
类目: Machine Learning (cs.LG)
*备注:
Abstract:Model-based reinforcement learning (MBRL) has achieved remarkable results in single-agent domains, yet its extension to competitive imperfect information games (IIGs) remains underexplored. In multi-agent settings, opponent-induced non-stationarity complicates the learning process, and decentralized model learning faces severe identifiability barriers, which we argue make centralized model learning a mathematical necessity. Building on this analysis, we propose NashDreamer, a principled MBRL framework for two-player zero-sum IIGs. It introduces a centralized Multi-Agent Recurrent State-Space Model (MARSSM) that decouples environment dynamics from the effect of players’ strategies on their individual observations. NashDreamer is designed to use arbitrary policy gradient algorithms and inherits their convergence guarantees towards Nash equilibria under an idealized model. Empirical evaluations across four benchmark games demonstrate that NashDreamer substantially improves sample efficiency over model-free baselines early in the training. Finally, we theoretically analyze the architecture’s optimization landscape, identifying the vulnerability of the Dreamer family of algorithms to posterior collapse in stochastic environments. We leave it as an open challenge.
[LG-3] Quantum Sparse Autoencoders for Q-Matrix Estimation in Cognitive Diagnosis
链接: https://arxiv.org/abs/2609.01537
作者: Arif Hassan Zidan,Yi Pan,Bowen Guo,Xiang Li,Yu Bao,Yingfeng Wang,Tianming Liu,Wei Zhang
类目: Machine Learning (cs.LG)
*备注:
Abstract:Q-matrices play a central role in cognitive diagnosis within educational data mining (EDM), specifying which latent skills each assessment item requires. Data-driven Q-matrix estimation remains challenging when assessments involve many correlated skills and when real response patterns depart from idealized generative assumptions. We introduce a novel quantum sparse autoencoder (QSAE) for Q-matrix estimation, which, to the best of our knowledge, is the first application of quantum machine learning (QML) to cognitive diagnosis. Overall, the QSAE embeds each student’s binary response vector into a quantum circuit using an encoder, compresses it into a sparse latent representation, and maps that representation to the Q-matrix. We benchmark the QSAE against a classical autoencoder (CAE) across 60 simulated datasets and 9 real-world assessment datasets. The results reveal complementary strengths. Although the CAE partially achieves higher average accuracy under several simulation conditions, the QSAE is substantially more stable across replications, exhibiting lower variance in 49 of the 60 conditions. Moreover, on real assessment data, the QSAE outperforms the CAE on 6 of the 9 datasets. These findings suggest that the principal advancement of QML in this setting is not universal accuracy improvement, but enhanced robustness and capability to explore latent-structure complexity in real datasets.
[LG-4] Sierpiński–Knopp Wasserstein Distance for Persistence Diagrams and Applications to 2-Wasserstein Approximation
链接: https://arxiv.org/abs/2609.01528
作者: Sebastien Tchitchek,Julien Tierny
类目: Computational Geometry (cs.CG); Machine Learning (cs.LG)
*备注: 49 pages, 11 figures, 6 tables. Code and reproducibility package: this https URL
Abstract:This paper introduces the Sierpiński-Knopp (SK) Wasserstein distance, a fast metric between persistence diagrams. The SK-Wasserstein distance, denoted d_\mathrmSK , maps diagram points and their diagonal projections to the unit interval via the Sierpiński-Knopp space-filling curve on the upper diagonal triangle. The encoded point sets are then efficiently matched via one-dimensional optimal assignment, in (O(N\log N)) steps, yielding an explicit diagonal-aware point assignment between the two input persistence diagrams. We show that the SK-Wasserstein distance controls the classical (2)-Wasserstein distance between diagrams, admits an explicit isometric embedding into a Hilbert space, and induces a positive-definite Gaussian kernel, making the resulting geometry directly compatible with Euclidean and kernel-based learning methods. A tighter surrogate dissimilarity, noted (W_\Gamma), is also introduced based on the point assignments along the curve. Experiments on 12 scientific collections comprising 227 diagrams show median per-collection speedup of (d_\mathrmSK) over state-of-the-art approximations of (W_2) is (626\times), while the aggregate speedup over the full benchmark is (2100\times). Average-linkage partitions obtained from (d_\mathrmSK) and (W_\Gamma) each exactly match the corresponding (W_2) partition on 8 of the 12 collections. Hilbert (k)-means and Gaussian spectral clustering, both based on (d_\mathrmSK), achieve mean adjusted Rand indices (ARI) of (0.756) and (0.800), respectively, with respect to the benchmark reference partitions, compared to (0.750) obtained by average linkage on (W_2). The Gaussian (d_\mathrmSK) kernel supports other kernel-based analysis tasks, as illustrated by its use for contiguous segmentation of ordered diagram collections in our experiments.
[LG-5] Does Imitation Learning Preserve Temporal Robustness in Dexterous Manipulation? An Expert-Learner Comparison Across Task Execution Speeds
链接: https://arxiv.org/abs/2609.01453
作者: Clinton Enwerem,John S. Baras,Calin Belta
类目: Robotics (cs.RO); Machine Learning (cs.LG)
*备注: 19 pages, 10 figures, and 9 tables. Code, data, and evaluation scripts are available at this https URL
Abstract:Dexterous manipulation policies learned by imitation are typically evaluated for robustness to variation in scenes, objects, or instructions, but their performance across task execution speeds is less often examined. This leaves open how much temporal robustness a learner retains relative to the expert it imitates. We compare an expert and learner under the same task conditions, initial-condition draws, and speedup factors. We instantiate the evaluation in ParcelStow, a contact-rich task in which the robot acquires, reorients, and inserts a parcel. The demonstrations span the speedup range for the manipulation phases after parcel acquisition. A scripted expert and an Action Chunking with Transformers (ACT) policy trained from the expert’s demonstrations both achieve 100 percent task success at nominal speed. Their success rates diverge within the demonstrated range: at its maximum, expert success is 84 percent and ACT success is 53 percent. Two ACT policies with different parameter initializations show similar degradation, decreasing by 34 and 48 percentage points from nominal speed to the maximum demonstrated speed, compared with 16 points for the expert. Stage-level analysis shows that 35 of ACT’s 47 failures at the maximum demonstrated speed are insertion misalignments. Under the relative-motion handoff, every ACT acquisition retains the parcel through reorientation and transfer in free space, but only 64 percent complete the overall task, compared with 95 percent after expert acquisition. Across all evaluated policies and speeds, none of the 414 acquisitions without force closure completes the task. Equal nominal task success therefore does not imply preservation of expert performance across execution speeds. Code, data, and evaluation scripts are available at this https URL.
[LG-6] Diffusion as a Training Curriculum for Timestep-Free Iterative Reasoning
链接: https://arxiv.org/abs/2609.01449
作者: Mariia Drozdova,Aidan Sirbu,Pietro Miotti,Robert Obryk,Mayalen Etcheverry,Eyvind Niklasson,Blake Richards
类目: Machine Learning (cs.LG)
*备注:
Abstract:Diffusion models and recursive reasoners are both iterative, but they carry information across iterations differently. We add a persistent hidden state to a diffusion denoiser and remove its timestep conditioning, leaving a single shared update that can be run to arbitrary depth. The result is an anytime solver: accuracy keeps improving with inference depth far beyond the rollout lengths and backpropagation window used in training, reaching 99.90% exact solve on Sudoku-Extreme. We also obtain 98.93% solve rate on Maze-Unique. Surprisingly, progressive denoising is unnecessary at inference: holding corruption at its maximum by replacing every non-clue variable with fresh Gaussian noise at each step retains near-perfect solving and converges to stable solutions. This simple noise-injection mechanism enables a single trajectory to efficiently explore the solution space and settle on the correct answer without parallel rollouts, candidate selection, or external verifiers required by prior reasoning models. Nonetheless, ordered annealed corruption remains critical during training, which suggests that diffusion’s primary contribution to our anytime solver is not a sampling procedure at inference, but a denoising training curriculum.
[LG-7] Edge-Girth as a Structural Edge Feature for Graph Neural Networks
链接: https://arxiv.org/abs/2609.01441
作者: Lilian Marey,Charlotte Laclau
类目: Machine Learning (cs.LG)
*备注:
Abstract:Graph neural networks (GNN) based on message passing are provably no more powerful than the one-dimensional Weisfeiler–Leman colour-refinement test (1-WL): two graphs it cannot tell apart receive identical representations, however deep or wide the network. A common remedy augments node or edge features with precomputed structural descriptors, most often counts of a fixed small subgraph such as triangles or longer cycles, but such counts require committing in advance to the size of the substructure counted, a choice usually made blind to the data. We study a descriptor that avoids this choice. The edge-girth of an edge is the length of a shortest cycle through it, and its multiplicity is the number of such shortest cycles; together they form a per-edge invariant that reports cycles of arbitrary length, computable exactly by a single breadth-first search per edge. Injected into a gated message-passing architecture, EGAGNN, it reaches a test MAE a factor three below the closest gated comparator on the ZINC-12k regression benchmark at 104k parameters; against bounded cycle-counting descriptors under the same architecture, it matches only a dictionary counting cycles up to length eight, using twice as many channels, while a dictionary capped at length four performs no better than no structural information at all. On graph discrimination we prove a matching limitation: on graphs where every edge sees the same number of shortest cycles of the same length, the descriptor becomes constant and any model built on it collapses back to the 1-WL bound. This holds without exception across all 400 pairs of the BREC benchmark: not one of the 90 such pairs is distinguished.
[LG-8] RIAGE: Three-level Routing and Intelligent Agent Guidance for Efficient Execution
链接: https://arxiv.org/abs/2609.01428
作者: Ruocan Wei
类目: Machine Learning (cs.LG)
*备注:
Abstract:Large Language Model (LLM) agents based on the ReAct paradigm have demonstrated remarkable capabilities in tool use and task execution. However, ReAct suffers from a fundamental efficiency problem: every query triggers a complete reasoning loop from scratch, and similar queries repeat identical steps without leveraging historical experience. We propose TRIAGE,a three-level routing framework that reduces token consumption by reusing historical execution trajectories. Its core innovation is TaaS (Trajectory-as-a-Skill), which abstracts historical execution trajectories into reusable skills, realizing ‘experience as a service’. TRIAGE classifies queries into three levels: (1) Direct Reuse-identical queries, 0 tokens; (2) Skill Substitution-similar queries, 0 tokens via deterministic parameter substitution; (3) Full ReAct-novel queries, automatically stored for future reuse. In large-scale experiments on 1,007 security monitoring queries, TRIAGE achieves 62.3% token savings, with 56.0% of queries at Level 2 and 5.5% at Level 1, both executing at zero cost. Cross-domain validation on ToolBench (15 domains, 345 queries) achieves 76.3% token reduction, confirming the generalizability of semantic routing. An online learning experiment demonstrates cold-start-to-mature evolution: the L2 hit rate rises from 0% to 57% within the first 100 queries, and the average token cost drops from 198 to 74.7. We also propose an automatic Skill extraction mechanism that distills high-frequency trajectory patterns into deterministic Skills, creating a positive feedback loop of ‘the more you use it, the more efficient it becomes’.
[LG-9] CATeye: Coupled Attribute-Topology Invariance Learning for Voucher Abuse Detection CIKM2026
链接: https://arxiv.org/abs/2609.01425
作者: Tian Tian,Shuaicheng Niu,Hao Kuang,Yuanhang Hu,Dong Li,Zhiqi Shen
类目: Machine Learning (cs.LG)
*备注: 8 pages, 3 figures, Accepted by CIKM 2026 Applied Research Track
Abstract:Voucher abuse poses a major challenge in e-commerce, where malicious users exploit promotional vouchers for profit. Unfortunately, fraud patterns evolve rapidly over time and across regions, causing distribution shifts that degrade existing detection models unless retrained frequently. To tackle this, we propose the Coupled Attribute-Topology Invariance Learning framework (CATeye). The key challenge arises from coupled attribute-topology shift, where edges built from attribute proximity cause environment-driven attribute shift to induce shifted topology, thereby amplifying variant signals through GNN message passing. CATeye sees through such coupled shifts with two learnable selectors. First, an Attribute Invariance Selector (AIS) learns node-adaptive masks to filter out non-invariant attributes. Then, conditioned on retained invariant attributes, an Edge Invariance Selector (EIS) samples an invariant subgraph and isolates non-invariant edges. Using the resulting invariant and non-invariant components, CATeye constructs multiple views and applies view-specific objectives to emphasize domain-invariant representations while suppressing domain-specific variations. Experiments on both a proprietary dataset from Lazada, a major Southeast Asian e-commerce platform, and a public benchmark show that CATeye consistently outperforms nine strong domain generalization and graph anomaly detection baselines, achieving up to an 8.61% improvement in average F1 score over the strongest baseline. Source code is publicly available at this https URL.
[LG-10] Predicting Subsurface Abnormalities Growth using Physics-Informed Neural Networks
链接: https://arxiv.org/abs/2609.01417
作者: Mehrdad Shafiei Dizaji,Hoda Azari
类目: Machine Learning (cs.LG)
*备注: 30, 20
Abstract:The research explores the pioneering integration of Physics-Informed Neural Networks (PINNs) into the domain of Ground-Penetrating Radar (GPR) data prediction. This research presents a detailed development framework for a specialized PINN model, proficient at interpreting and forecasting GPR data, much like how medical imaging models predict tumor behavior. By harnessing the synergy between deep learning algorithms and the physical laws governing subsurface structures or in medical terms, human tissues the model effectively embeds the physics of electromagnetic wave propagation into its architecture. This ensures that predictions not only align with fundamental physical principles but also mirror the precision needed in medical diagnostics for detecting and monitoring tumors. The suggested deep learning structure comprises three components: a CNN, a spatial feature channel attention (SFCA) mechanism, and ConvLSTM, along with temporal feature frame attention (TFFA) modules. The attention mechanism computes channel attention and temporal attention weights using self-adaptation, thereby fine tuning the visual and temporal feature responses to extract the most pertinent and significant visual and temporal features. By integrating physics directly into the neural network, our model has shown enhanced accuracy in forecasting GPR data. This improvement is vital for conducting effective assessments of bridge deck conditions and other evaluations related to civil infrastructure. The use of Physics Informed Neural Networks (PINNs) has demonstrated the potential to transform the field of Non-Destructive Evaluation (NDE) by enhancing the precision of infrastructure deterioration predictions. Moreover, it offers a deeper insight into the fundamental mechanisms of deterioration, viewed through the prism of physics-based models.
[LG-11] Contribution-Aware Bandwidth Allocation for Multimodal Split Learning
链接: https://arxiv.org/abs/2609.01406
作者: Iason Ofeidis,Leandros Tassiulas
类目: Machine Learning (cs.LG); Distributed, Parallel, and Cluster Computing (cs.DC); Networking and Internet Architecture (cs.NI)
*备注: 10 pages, 4 figures
Abstract:Multimodal models are increasingly the default option for perception at the network edge, yet they are trained almost entirely in the datacenter, because a client holding several sensor streams cannot host an encoder per modality. Split Learning makes such training feasible by keeping only the first layers on the device, at the cost of an uplink that must carry smashed activations for every modality at every step. Existing compression schemes give each modality the same keep-ratio, so the shared budget is divided in proportion to smashed-activation dimension, a quantity unrelated to how much each modality contributes to the fused prediction. We make that division an explicit decision and call it inter-modality allocation: under a fixed uplink budget, every policy transmits the same expected payload and differs only in how that payload is split across modalities. Our allocator, ModalShare, sets each modality’s keep-ratio from a Shapley contribution score that the server computes over coalitions of activations it has already received. Measuring this score adds no uplink traffic and no client-side computation, and needs no prior knowledge of which stream is which. ModalShare improves accuracy over equal keep-ratios by 15.4 and 12.4 percentage points on CREMA-D and MVSA at matched payload in 5x compression, with strong performance across three compressors, three datasets, and four budgets. We show that existing compressors underperform in multimodal settings, with ModalShare recovering what gains are left behind.
[LG-12] Exact Risk-Complexity Laws for Projective Boundaries in Scenario Optimization and Distribution-Free Certification
链接: https://arxiv.org/abs/2609.01355
作者: Giuseppe C. Calafiore
类目: ystems and Control (eess.SY); Machine Learning (cs.LG); Optimization and Control (math.OC)
*备注:
Abstract:Scenario optimization, conformal prediction, and related distribution-free certification methods use finite samples to construct decisions or prediction sets with violation-risk guarantees for fresh observations. In several classical settings, the conditional violation risk follows an exact beta law, whose tail has a beta-binomial representation and whose parameter is a support, calibration, or compression dimension. This paper identifies the deterministic boundary mechanism behind these formulas and derives the corresponding law when the observed boundary size is random. A decision rule is represented by an acceptance set for future observations, together with a boundary map selecting the sample points responsible for that set. The resulting pair is called a \em proper projective boundary scheme when held-out samples are accepted precisely if the full-sample boundary is retained, and accepted non-boundary samples can be deleted without changing that boundary. For every such scheme, the conditional law of the violation risk given the observed boundary size is determined by the boundary’s cross-sample complexity profile. A stable profile yields the usual beta law, whereas a varying profile produces an exact profile correction. The framework covers scalar order-statistic calibration, support-reconstructive scenario programs, cascaded support-removal certificates, coordinatewise envelopes, and Pareto-frontier calibration with vector scores. It also yields conditional probabilistic certificates and a no-go result explaining why observed complexity alone is insufficient.
[LG-13] SMELT: Scaling Laws for Compute-Matched MoE Looped Transformers
链接: https://arxiv.org/abs/2609.01343
作者: Shaowen Wang,Ge Zhang,Kairong Luo,Yuhao Wu,Shaofan Liu,Jiaheng Liu,Wenhao Huang,Shen Yan,Jian Li
类目: Machine Learning (cs.LG)
*备注: 35 pages, 25 figures
Abstract:Looped Transformers increase effective depth by iterating a shared block of layers, but most evaluations compare at fixed model size, conflating architectural advantage with extra FLOPs. We study looping on Mixture-of-Experts Transformers while closely matching per-token FLOPs, total non-embedding parameters, and KV cache. Through a series of ablations, we arrive at a recipe we call SMELT (Sparse MoE Transformer, middle layers Loop Twice), which loops the middle half of layers twice while matching the unlooped Baseline on all three budgets. We scale SMELT across four sizes up to 54B non-embedding parameters and fit a separate Chinchilla-style scaling law for each architecture. SMELT’s loss drops faster with compute, saving 6.8–18.0% of training FLOPs on the compute-optimal frontier. The advantage transfers to downstream benchmarks beyond what validation loss predicts, is largest on Code, and grows with sample length and the number of in-context examples. Mechanistic analysis shows that the second visit reduces the attention sink and redirects mass toward content-relevant tokens, an inductive bias that may underlie the observed performance gains. These results show that looping can improve Transformers even under budget matching, offering a practical recipe that turns depth reuse into measurable gains.
[LG-14] mzCache: On-Device LLM Memory Management under Multitasking
链接: https://arxiv.org/abs/2609.01338
作者: Hongseung Yu,Minsung Kim,Jongseok Park,Kyunghan Lee
类目: Operating Systems (cs.OS); Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG)
*备注: MobiCom 2026
Abstract:On-device mobile Large Language Model (LLM) inference is gaining significant attention. However, mobile devices operate in highly dynamic multitasking environments where users frequently switch between applications. This creates memory pressure, forcing LLM memory (model weights and KV cache) to be evicted by the operating system. When a new inference request arrives, the inference system must restore the evicted memory through slow storage reads or recompute the entire KV cache, severely degrading responsiveness. To address this, we present mzCache, an on-device LLM inference system with specialized memory management for multitasking environments. Under unpredictable memory pressure, mzCache elastically evicts LLM memory and leverages the unified memory of mobile SoCs to enable zero-wait inference on the GPU with concurrent CPU-side restoration. mzCache realizes this through restoration-oriented memory management: LLM memory is partitioned into fine-grained shared buffers to enable partial eviction and restoration with concurrent cross-processor access, while hybrid swap and backward-out eviction policies ensure low-latency restoration from any eviction state. Implemented on this http URL and deployed as an Android application, mzCache achieves 2.1-5.5 \times reduction in Time-to-First-Token compared to storage-backed partial offload and demonstrates its effectiveness in real multitasking scenarios.
[LG-15] One-Layer Transformer Provably Learns Multiclass One-Nearest Neighbor in Context
链接: https://arxiv.org/abs/2609.01311
作者: Skanda Athreya,Yutong Wang
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:We extend recent work establishing an equivalence between one-layer transformers and nearest-neighbor classifiers in the binary setting to the multiclass case. By leveraging the simplex encoding, we show that one-layer transformers with an argmax classification head behave identically to a one-nearest-neighbor classifier in the multiclass setting. This closes a gap left by prior work, whose multiclass result relied on a non-standard rounding-based approach rather than the typical argmax head used in practice.
[LG-16] Relational Task Generation Language: A Declarative Specification Framework for Relational Deep Learning
链接: https://arxiv.org/abs/2609.01292
作者: Oleksii Kolesnichenko,Jakub Peleška,Gustav Š’ır
类目: Programming Languages (cs.PL); Databases (cs.DB); Machine Learning (cs.LG)
*备注: Accepted to MLG 2026
Abstract:Relational Deep Learning (RDL) has become a powerful paradigm for learning from multi-tabular data. However, manually defining RDL prediction tasks is a laborious process that frequently results in data leakage. To address this issue, we introduce Relational Task Generation Language (RTGL) - an open-source declarative language that streamlines RDL task formulation by abstracting away low-level SQL details. We showcase RTGL by reconstructing existing RDL benchmark tasks and uncovering their inconsistencies stemming from manually crafted SQL definitions of RDL prediction targets, thereby underscoring the value of a dedicated declarative language. In addition, we demonstrate the practical utility of RTGL by designing various new tasks with diverse forms and target types. Our experiments confirm the robustness and usability of RTGL, as well as its seamless integration with the existing RDL frameworks, making it widely accessible to the community.
[LG-17] Position: Privacy Is a Claim Not a Property of Synthetic Data
链接: https://arxiv.org/abs/2609.01273
作者: Jiachen Zhao,Antonia Januszewicz,Taeho Jung
类目: Machine Learning (cs.LG); Cryptography and Security (cs.CR)
*备注:
Abstract:Synthetic data has become a common component of machine learning research. While widely adopted, its use in privacy-sensitive contexts has quietly shifted from a claim of residual inference risk under stated assumptions to an appearance-based property inferred from data generation itself. In this position paper, we argue that this shift reflects an implicit change in community standards for what counts as sufficient privacy evidence, rather than a misunderstanding of well-established privacy principles. Drawing on an empirical analysis of recent publications across major ML venues, we show that synthetic data is frequently used in privacy-sensitive settings without explicit articulation of threat models, inference risks, or falsifiable privacy claims. As a result, privacy assurance often remains implicit, difficult to verify, and unevenly distributed, with heightened exposure for rare and minority records. We argue for treating privacy as an explicit, evidence-based scientific claim and recommend that ML venues adopt norms requiring privacy-relevant assertions to be clearly scoped, testable, and contestable.
[LG-18] Solving In-Table Prediction Problems by Deep Neural Networks with Performance Evaluation Using Synthetic Data
链接: https://arxiv.org/abs/2609.01262
作者: Xiao Zhao,Daniela Oelke
类目: Machine Learning (cs.LG)
*备注:
Abstract:Tabular deep learning (TDL) leverages neural networks (NN) to extract patterns from tabular data. Traditional TDL methods follow a supervised learning paradigm, where a target feature is explicitly given. In this work, however, we explore a different approach by employing deep NNs to learn relationships among individual columns within a given table. We investigate whether NNs can predict the values of arbitrarily selected columns in a given table based on the remaining known columns. We call this problem In-Table Prediction (ITB), which is slightly different from table imputation methods and the pretraining task of TDL. Three potential usage scenarios are identified, which, to our best knowledge, have not been extensively studied in the literature. A self-supervised learning approach is applied to address this problem by randomly selecting columns to be masked out and used as learning targets. This work focuses on tabular datasets containing only continuous features. To handle missing values in continuous features, a novel neural layer is proposed to embed both numerical and empty values. Synthetic data is generated based on predefined column relationships, with empty values inserted using two distinct mechanisms. Additionally, an adapted masking strategy is employed to create test data. Performances of three NN architectures, namely MLP, Resnet and Transformer, are evaluated using the generated synthetic data. We conclude that, the attention-based structure outperforms the other two networks, when a sufficiently large number of training examples is available and a relatively large embedding length is chosen. We stress that these findings are obtained under controlled, synthetic conditions with a small number of columns and it should therefore be regarded as an initial, narrowly-scoped investigation rather than a general characterization of ITP on real-world tabular data.
[LG-19] Multi-Head Self Attention is a Parameter Identification Mechanism
链接: https://arxiv.org/abs/2609.01231
作者: W. Ross Morrow
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:We prove that a multi-head scaled dot product attention can be viewed as a parameter identification strategy. The ratio of unidentified parameters to the total number of parameters scales like the reciprocal of the number of heads ( 1/2 \to 1/(2H) ), meaning models with more heads are structurally more identified. A subtle side effect of the mathematics observation that attention can never be fully identified. Similarly we also show that some bias terms can have no effect on softmax-based attention layers in both the single- and multiple-head settings, though this is mostly a curiosity that should have a marginal effect on model size and model training/prediction efficiency. We also touch on modern improvements to transformers including RoPE and GQA from this perspective, illustrating how those as well can improve the ratio of meaningful'' parameters to all parameters. Simple numerical examples demonstrate that training can indeed involve updates that overlap model-invariant subspaces that arise from a lack of identification. As part of our experiments we use a rebalancing’’ approach that can ``fix’’ updates that overlap unindentified subspaces but do not try to present evidence this should actually be adopted. Instead we simply view our numerical results as exploring and confirming the theoretical results. As a whole we discuss a purely mathematical/statistical explanation, identification, for why specific architectural choices in transformers may have improved performance.
[LG-20] Recent Developments in Transformer Inference Deployment on FPGA Platforms: A Survey
链接: https://arxiv.org/abs/2609.01212
作者: Arjan Blankestijn,Uraz Odyurt,Amirreza Yousefzadeh
类目: Machine Learning (cs.LG); Hardware Architecture (cs.AR)
*备注:
Abstract:With the rapid and continuous growth in the incorporation of machine learning models based on the Transformer architecture, capable deployment is in high demand. In this context, capable deployment refers to operational performance aspects, e.g., throughput and latency, as well as efficiency aspects, e.g., energy consumption. When it comes to the task of inference using such models, purpose-built hardware accelerators provide a lucrative alternative to common deployment choices, such as Central Processing Units (CPUs) and Graphics Processing Units (GPUs). The Field Programmable Gate Array (FPGA) platforms category is an example of such alternative accelerators, promising implementation flexibility, energy efficiency, improved latency and suitability for on-site deployment. We investigate the most recent advances, trends, and design choices for Transformer inference on FPGA platforms. We perform a systematic literature review, extracting and delving into preferred techniques for implementation and optimisation. This study and the provided taxonomy of topics could act as a guide for researchers from the academia and industry alike.
[LG-21] Births are difficult to predict even with rich survey and full-population register data
链接: https://arxiv.org/abs/2609.01194
作者: Elizaveta Sivak,Emily M. Cantrell,Thomas Emery,Javier Garcia-Bernardo,Flavio Hafner,Kasia Karpinska,Malte Lüken,Adrienne Mendrik,Joris Mulder,Hanzhang Ren,Varun Satish,Mark Verhagen,Angelica M. Maineri,Paulina Pankowska,Jasmin Abdel Ghany,Bruno Arpino,Giovanni Cassani,Julia Hellstrand,Katya Ivanova,Sanni Kuikka,Ana Macanovic,Charles Rahal,Felix C. Tropf,Roland J. Veen,Nicole Walasek,Daniël van Wijk,Kelsey Q. Wright,Emilio Zagheni,Henry Abbink,Emanuele Aliverti,Matteo Amestoy,Tilbe Atav,Nicola Barban,Sunnee Billingsley,Goan J. Booij,Louis Boucherie,Yael Broos,Li Ya Chang,Jamie C. Chiu,Chiara Ludovica Comolli,Boris Cule,Qixiang Fang,Dennis M. Feehan,Rachel Ganly,Erwin Gielens,Rolando M. Gonzales Martinez,Andrea Gradassi,Rosember Guerra-Urzola,Mario Guerra-Urzola,Stéphane Guerrier,Enamul Hassan,Vincent A. Haverhoek,Andrew T. Hendrickson,Amber Howard,Yuxuan Jin,Sayash Kapoor,Erik-Jan van Kesteren,Iris ten Klooster,Marie Labussiere,Lydia T. Liu,Tiffany Liu,Adam Maghout,Simone Meneghello,Lasse Mohr,Clara H. Mulder,Saul J. Newman,Jessica Nisén,Janis Norden,Mikkel Odgaard,Riccardo Omenti,Ozancan Ozdemir,Christina Pao,Paige Park,Gaia Penta,Juan C. Perdomo,Tanzir Pial,Alessio Piraccini,Federica Querin,Ziwei Rao,Christian Rellama,Adrien Remund,Frederieke Richert,Arnout van de Rijt,Mojtaba Rostami Kandroodi,Stijn J. Rotman,Lucas Sage,Germans Savcisens,Katrin Schwanitz,Steven Skiena,Alessandro Spata,Yannick Stadtfeld,Benedikt Stroebl,Gaetano Tedesco,Mathilde Theelen,Gianluca Tori,Abigail Tun-Mendicuti,Rishabh Tyagi,Keyon Vafa,Luiz Felipe Vecchietti,Linda Vecgaile
类目: Machine Learning (cs.LG)
*备注:
Abstract:Major life events have proven difficult to predict. Does this reflect limits of theory, data, and algorithms, or the large role of chance? We examine one outcome - having a child within three years - through a near-ideal setting for prediction: a data challenge where 147 researchers predicted births for Dutch residents aged 18-45, using survey data and full-population registers. Methods ranged from logistic regression to a large language model and transformers. Predictions were moderately accurate (best F1: register 0.59, survey 0.76); advanced models did not outperform classical ones; and the larger registers did not beat the survey. Simulating the stochastic biology of conception and pregnancy, we estimated a predictive ceiling (survey F1 ~ 0.86-0.94, register 0.88-0.96). Observed performance falls short of this ceiling, implicating imperfect data, methods, and unmodelled chance, while the ceiling itself shows that chance in reproduction alone sets a non-trivial limit on predicting individual lives.
[LG-22] Pre-carved Niches: The Formation Dynamics of Modular Task Partitions in Early LLM Training ATC
链接: https://arxiv.org/abs/2609.01170
作者: Guangqi Li,Yongxin Li
类目: Machine Learning (cs.LG)
*备注: 17 pages, 7 figures, 2 tables. Step-by-step training-dynamics study of modular task partitions in a from-scratch Pythia-410M model
Abstract:Large language models exhibit a modular internal organization that mirrors well-studied functional networks of the human brain, but how this organization forms during training is unknown: prior work has characterized finished models, not the formation process. We track formation step by step: we train a Pythia-410M model from scratch (two trajectories, bf16 and fp32) and run attribution patching at every step, alongside probes for gradient norms, effective updates, weight norms, and first-order loss decomposition across 14 tasks in four cognitive domains. Three findings. First, the modular map is pre-carved: before any learning, the dominant task pair already overlaps at ~3.6x the attribution substrate (a task-independent baseline), and its layer-0 concentration is an architecture-level constant on this model family. Second, the partition locks in through two sharp jumps whose amplitudes do not track the learning-rate schedule (the second reaching 20.4 sigma quiet-window / 6.2 sigma global), accompanied by gradient-level relative deprivation–winners receive 2.25-2.73x the loser’s gradient supply, 9.5-11.5 standard deviations below a random control–that does not propagate to updates or weights. Third, deviation from the substrate appears only in the domain being learned, consistent with the hypothesis that modularity tracks learning. We close by separating the feature-level account we can defend from the mechanistic questions we cannot, and we pre-register the scale-threshold hypothesis behind our ongoing 2.8B experiments.
[LG-23] CopyShield: A Cross-Level Benchmark of Copyright Defenses in LLM s
链接: https://arxiv.org/abs/2609.01161
作者: Maryam Alshehyari,Dushyant Singh Chauhan,Samuele Poppi,Martin Takac,Salem Lahlou,Nils Lukas
类目: Machine Learning (cs.LG)
*备注:
Abstract:Large language models can reproduce memorized text verbatim, yet copyright defenses are usually evaluated under incompatible protocols. We introduce CopyShield, a controlled benchmark comparing three representative defenses at distinct intervention levels: contrastive decoding (output), Direct Preference Optimization (behavioral), and activation intervention (representation). We evaluate CopyShield on two model families, LLaMA-3.1-8B and Mistral-7B-v0.3, using controlled memorization over five public-domain books and a shared protocol measuring literal leakage, calibrated non-literal leakage, utility, and degeneracy. Across these methods, intervention level is associated with distinct compliance-utility trade-offs. On LLaMA-3.1-8B, contrastive decoding remains near-degeneracy-free (0-2%) but reaches a literal-suppression floor at NV-Recall 0.192-0.203. DPO nearly eliminates literal leakage (0.263 to 0.002) but induces paraphrase-loop degeneracy in 58% of QA outputs, with no utility gain over the SFT baseline. Activation intervention attains the lowest non-literal flagging rate (1/200) by blocking 84% of non-literal queries before generation. Human evaluation confirms that DPO has low coherence, whereas activation lowers perceived copyright risk through broad refusal. On Mistral-7B-v0.3, the output- and representation-level patterns persist, while DPO degeneracy falls to 10-14%, showing that its severity is model-dependent. Together, CopyShield provides cross-level reference baselines and identifies targeted non-literal suppression as an open challenge. The code is available at this https URL.
[LG-24] Scaled Idempotence in Transformer Attention: Paired OV Geometry and Shared-Value Algebras
链接: https://arxiv.org/abs/2609.01129
作者: Jiming Feng,Junliang Li
类目: Machine Learning (cs.LG)
*备注: 14 pages, 2 figures. Preprint
Abstract:We identify a recurrent algebraic regularity in Transformer attention: a sparse subset of effective OV operators T=OV^\top nearly closes under composition, T^2\approx\alpha T . Across six pretrained endpoints spanning 2.8B–235B parameters, 3.98–8.00% of heads reach squared closure alignment \mathcalP\geq0.9 , while no matched within-layer O/V mismatch does. An exact principal-coordinate factorization, T=Q_OKQ_V^\top and T^2=Q_O(KDK)Q_V^\top , separates within-support transport from read–write return geometry. Across all 7,304 heads in nine MHA/GQA models, scrambling only the orientation of K while preserving singular values, norms, factor spans, and principal angles reduces median closure from 0.336 to 1.04\times10^-4 ; trained orientation wins for 98.64% of heads and in every layer. Constructive searches show that high closure is feasible in every surveyed layer, but usually not attained. Retrospective trajectories in three independently trained lineages further separate broadly available capacity from the orientations attained by final strong heads. Under exact value sharing, headwise closure extends to a right-action algebra, T_iT_j=\alpha_jT_i . Seven-model experiments verify the approximate law and reveal distinct oblique projections with a shared value-defined kernel. These results characterize scaled idempotence as a sparse trained orientation within broadly available geometric capacity and show how value sharing extends a headwise relation into a local operator algebra.
[LG-25] When Does Online Adaptation Pay on the Edge? A Leakage-Free Evaluation of Warmup Learning-Rate Selection and Resource Trade-offs for Time-Series Forecasting
链接: https://arxiv.org/abs/2609.01126
作者: Takumi Fujimoto,Hiroaki Nishi
类目: Machine Learning (cs.LG)
*备注: under review, IEEE BigData 2026
Abstract:Online adaptation can help edge time-series forecasting under distribution drift, but its measured benefit is sensitive to evaluation choices. We study six public multivariate streams, including building-sensor and smart-meter data, under a leakage-free streaming protocol. We identify two additional sources of comparison bias. First, the warmup budget of the static baseline has a two-sided effect: insufficient warmup undertrains the baseline, whereas excessive warmup can degrade its pre-drift generalization. Across six dataset-backbone settings, the estimated adaptation benefit changes by 3.0 to 18.8 percentage points (pp) over the 1,000-20,000-step warmup range. Second, comparing SGD with momentum (SGD+m) and Adam at a shared default learning rate conflates optimizer quality with rate sensitivity. We select both the warmup budget and each optimizer’s online rate using a held-out pre-drift validation slice without accessing test data. Under this validation-only procedure, Adam outperforms SGD+m in 310 of 360 evaluated cells, while 4 Adam cells remain below the static baseline. We further characterize accuracy against adaptation-state memory and A100-measured per-update latency for full, head-only, and calibration-based adaptation. In the evaluated PatchTST frontier settings, several parameter-efficient variants are nondominated on the adaptation-state-memory axis. Smart-meter analyses also show that reported gains depend on meter-selection rules. These findings support a validation-only commissioning procedure, while target-device latency and energy remain to be measured. Code, data, and all reported numbers: this https URL.
[LG-26] Replicating TRACE: A Practitioners Guide to Its Threshold and Particle Budget
链接: https://arxiv.org/abs/2609.01108
作者: Alex Chadyuk,Alicia Zhang,Roy Kucukates
类目: Machine Learning (cs.LG)
*备注: 12 pages
Abstract:TRACE (Math Lienhart, arXiv:2602.01135) reads causal graphs over event types out of a pretrained autoregressive sequence model by thresholding a per-position conditional-mutual-information estimate at a fixed tau. We independently replicate its headline synthetic result: with tau selected on a validation split, mean per-sequence F1 against exact interventional truth reaches 0.90-0.91 at vocabulary size 1000 (paper: 0.91) and 0.86-0.91 from 100 to 2000. First, the optimal threshold is pinned to the truth margin, not to any constant: at every size the errors at tau* straddle the delta = 0.05 margin defining ground truth (missed true edges lie just above it, accepted false ones just below), and the blind optimum lands near delta/2 times the estimator’s calibration, confirmed out of sample at 5000. Second, at a single global threshold TRACE mostly recovers a direct, adjacent-influence graph: lag-1 true edges are recalled at 0.97-0.99, while true edges at lag 2 or more read orders of magnitude lower—the reading-scale price of randomizing mediating positions, which an exact test of direct causal effect requires when the truth is unknown. A per-lag threshold family recovers a third to a half of lag-2 truth; on lag-uniform data one validated threshold recalls every lag at 0.40-0.87, 8-26 pp below an atomic-intervention control at lags 3-6. Third, the default lag decay of the paper’s synthetic benchmark concentrates about 85% of interventional truth at lag 1 and pushes the rest below the estimator’s noise floor, so headline F1 there certifies lag-1 recovery only and conflates the benchmark’s skew with the algorithm’s own limit; a flatter decay separates the two. Fourth, F1 saturates from N = 2 particles at the selected threshold—a property of the threshold’s margin over the noise floor, not of the estimator, which converges as N^(-1/2). We distill five practitioner rules.
[LG-27] Neural Symbollic Regression Using Deep Learning and Sparse Modelling
链接: https://arxiv.org/abs/2609.01102
作者: Ravi Kumar U,Sumitra S
类目: Machine Learning (cs.LG); Neural and Evolutionary Computing (cs.NE); Symbolic Computation (cs.SC)
*备注: 11 pages, 5 tables, 5 figures, contains detailed mathematics behind the algorithm
Abstract:Symbolic Regression (SR) seeks to find succinct mathematical expressions that represent the fundamental relationships within data, providing interpretability and scientific understanding that exceeds that of black-box models. Nevertheless, traditional methods like Genetic Programming face challenges with scalability and are highly sensitive to noise, while sparse regression techniques such as SINDy rely significantly on predetermined feature libraries. In this work, we present a Neural Symbolic Regression (NSR) framework that treats neural networks as functional preconditioners for symbolic discovery. Our approach uses a decoupled pipeline: a neural network first learns a smooth, noise-robust approximation of the target function in an interaction- aware nonlinear feature space. LASSO is then applied to extract sparse, interpretable closed-form expressions. To improve predictive accuracy and symbolic fidelity by integrating distributed hyperparameter optimization with Ray Tune and ASHA scheduling. Experiments on the Nguyen benchmark suite show that our approach consistently outperforms SINDy and non-tuned neural baselines in RMSE, noise robustness, and out-of-distribution generalization. Ablation studies confirm the significance of feature interactions, neural depth, and tuning strategies. In general, this study presents a scalable and understandable neural-symbolic framework, creating a solid link between neural approximation and the discovery of sparse equations for scientific machine learning.
[LG-28] Subliminal Learning as Trait-Direction Drift: A Mechanism and Targeted Control under SFT Distillation
链接: https://arxiv.org/abs/2609.01091
作者: Zhixuan Liu,Zhichen Dong,Yuyu Fan,Xiangtian Li,Chao Yang
类目: Machine Learning (cs.LG)
*备注: 36 pages, 8 figures
Abstract:Beyond intended capabilities, model distillation can transfer hidden traits from a teacher. A teacher biased by a system prompt can generate semantically clean training data, such as numeric sequences, that still causes a downstream student to inherit the hidden preference, a phenomenon known as subliminal learning. Prior work has identified several parts of this process. How the signal builds up during training and produces behavioral transfer remains unclear, making targeted mitigation difficult. We propose and validate trait-direction drift as a mechanism for subliminal learning: biased generation creates measurable preference gaps in teacher data, and student-recognizable gaps induce trait-aligned updates during supervised fine-tuning that accumulate into behavioral transfer. Guided by this mechanism, we propose probe-space corridor regularization, a targeted defense that constrains drift along a calibrated trait direction during distillation. The method substantially reduces hidden-trait transfer, preserving task performance: for example, it lowers malicious-response transfer from 29.55% to 6.45% with low main-task accuracy cost, and consistently suppresses animal-preference transfer across the main Qwen setting. The preference-gap, training-trajectory, and intervention evidence links subliminal learning to trait-direction drift and motivates corridor regularization as a targeted control during distillation.
[LG-29] Modelpedia: A Catalog of Model Findings for the Meta-Science of AI
链接: https://arxiv.org/abs/2609.01090
作者: Franciszek Bernat(1 and 2),Dawid Płudowski(1 and 2),Michał Jan Włodarczyk(1 and 2),Luca Longo(3),Jianlong Zhou(4),Andreas Holzinger(5),Riccardo Guidotti(6 and 7),Wojciech Samek(8, 9 and 10),Przemysław Biecek(1, 2 and 11) ((1) Centre for Credible AI, (2) Warsaw University of Technology, (3) University College Cork, (4) University of Technology Sydney, (5) Human-Centered AI Lab, (6) University of Pisa, (7) ISTI-CNR, (8) Technical University of Berlin, (9) Fraunhofer Heinrich Hertz Institute, (10) Berlin Institute for the Foundations of Learning and Data, (11) University of Warsaw)
类目: Machine Learning (cs.LG)
*备注: For the website, see: this https URL For the codebase, see: this https URL
Abstract:Scientific knowledge about AI models is produced faster than the community can organize it. Every few months a new foundation model reshapes the field and hundreds of papers, blogs, and technical reports document how each behaves or fails. Yet, these findings remain scattered and effectively unretrievable. To address this gap we present Modelpedia, an automated, LLM-assisted framework that extracts findings about models from published papers, links it to the model, dataset, method, and concept it concerns, and aggregates the result into a searchable public catalog. Applying the prototype to accepted ICLR 2024 and 2025 papers, we extract over a thousand findings and, treating the catalog itself as an object of study, run a meta-analysis of how the community investigates models. Now, we invite the community to explore, contribute to, and build on the open catalog, and to help establish model findings as a shared foundation for the meta-science of AI.
[LG-30] Artificial Rosetta Stone: Constrained Maximum A Posteriori (MAP) Reconstruction of Symbolic Rag a Sequences via Order-k Markov Models
链接: https://arxiv.org/abs/2609.01064
作者: Saanvi Raghavendran,Abhishek Bhattacharjee(Abstract Math Institute)
类目: ound (cs.SD); Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注: 37 pages,5 tables,11 Graphical Representations
Abstract:Reconstructing a damaged musical fragment is an inverse problem: the observed sequence contains partial information, while a raga encodes constraints limiting allowable completions. This paper formalizes a mathematical framework for this, proposing the Artificial Rosetta Stone (ARS). We separate three claims often conflated: a symbolic sequence can be reconstructed probabilistically; a sequence can be consistent with an explicit grammar; and a historical performance can be authenticated. We only support the first two. We model a raga via a finite alphabet and constraint system, using an order-k Markov model for melodic probabilities. A symmetric Dirichlet prior yields a tractable posterior. We pose missing-note reconstruction as a constrained MAP problem. For fixed-length sequences and finite-order constraints, optimization admits an exact dynamic-programming solution with worst-case time complexity O(TN^k+1) . We derive the parameter count N^k(N - 1) , prove a concentration bound under explicit mixing assumptions, and analyze estimation error propagation. A reproducible synthetic experiment uses six raga-inspired alphabets, orders k \in \1, 2, 3\ , and masking rates up to 50%. This is a proof of concept, not historical reconstruction. A real-audio feasibility pilot evaluates 30 usable sequences from 42 Yaman clips via automated pitch extraction, segmentation, and quantization. Lacking documented provenance and relying on automated transcription, this is not expert-validated archival reconstruction. Claims are tied to stated conditions, not universal properties of Hindustani music. Code: this https URL.
[LG-31] Accelerating Reinforcement Learning via MPC Solver-Gradient Guidance for Weights-varying MPC
链接: https://arxiv.org/abs/2609.01061
作者: Baha Zarrouki,Arslan Thobani,Jasper Hoffmann,Mattia Piccinini,Rudolf Reiter,Felix Jahncke,Sébastien Gros,Davide Scaramuzza,Johannes Betz
类目: Robotics (cs.RO); Machine Learning (cs.LG); Systems and Control (eess.SY)
*备注: Submitted to IEEE Transactions on Robotics (T-RO). 18 pages, 13 figures
Abstract:In Model Predictive Control (MPC), cost-function weights shape closed-loop behavior, yet changing conditions often make fixed parametrizations suboptimal and motivate context-dependent online adaptation. Learning such policies is difficult because behavior depends implicitly on numerical MPC solutions, producing nonlinear, potentially nonsmooth, long-horizon dependencies on policy parameters. This creates a bias-variance tradeoff: Reinforcement Learning (RL) optimizes realized closed-loop return from environment samples but is sample-inefficient, whereas Gradient-Based Policy Learning (GB-PL) uses low-variance solver gradients from differentiable MPC to optimize surrogate losses on predicted trajectories but can be biased under model mismatch. We propose Solver-Gradient Guided Reinforcement Learning (SG-RL), a solver-sensitivity augmentation for RL-based online MPC cost-weight adaptation. SG-RL keeps sampled closed-loop return as the objective and uses bounded solver-derived gradients as auxiliary guidance to improve stability and sample efficiency. We instantiate SG-RL in Proximal Policy Optimization (PPO) with four modular algorithms that inject solver-gradient guidance into actor-update scaling, policy loss, advantage estimation, and value-function learning. On two full-scale autonomous racing platforms with intentional model mismatch, SG-RL reaches PPO’s best closed-loop return with up to 70.6% fewer samples, outperforms GB-PL baselines by at least 54% in closed-loop return, and generalizes zero-shot to unseen environments.
[LG-32] he Multiple Timescales of Gradient Descent on the Edge of Stability: A Perturbative Derivation of the Central Flow
链接: https://arxiv.org/abs/2609.01034
作者: Raphaël Berthier
类目: Machine Learning (cs.LG); Optimization and Control (math.OC); Machine Learning (stat.ML)
*备注:
Abstract:The central flow of Cohen et al. (2025) is an empirically accurate continuous-time model of gradient descent at the edge of stability in deep learning, However, its derivation is heuristic. We propose a perturbative regime in which the central flow is the limit of gradient descent: we assume that the loss decomposes as f = g + \varepsilon h ; in the limit \varepsilon \to 0 , the dynamics of gradient descent with learning rate \eta converge to the gradient flow of h constrained to the minimizers of g of sharpness at most 2/\eta . Our approach is formal rather than rigorous; it treats gradient descent as a singularly perturbed dynamical system in \varepsilon . Three timescales emerge: a fast timescale of oscillations along the sharpest direction, an intermediate timescale of the self-stabilization mechanism, and a slow timescale of the dynamics along the minimizers of g -the central flow. Using the method of multiple scales, a classical formal method from singular perturbation theory, we derive the expansion of the dynamics in \varepsilon : the central flow emerges as the leading-order term in the expansion, while the self-stabilization mechanism appears in the next-order term. We study this mechanism beyond previous analyses: with a single eigenvalue at the edge of stability, we compute the slow drift of the energy of the fluctuations; with several eigenvalues at the edge of stability, we derive the self-stabilization system and explain why fluctuations persist.
[LG-33] When Metropolis and Hastings Meet Bradley and Terry: Exact MCMC From Preference Voting
链接: https://arxiv.org/abs/2609.00905
作者: Ariel Smogorghevski,Nir Rosenfeld,Yaniv Romano
类目: Machine Learning (cs.LG); Computation (stat.CO); Machine Learning (stat.ML)
*备注:
Abstract:Sampling from distributions conditioned on desired semantic properties is an emerging challenge in modern generative modeling. Metropolis-Hastings (MH) provides a principled route to conditional sampling, but requires access to exact pointwise target-density evaluations, which are not available in generative settings. Meanwhile, pairwise comparisons by humans or model “judge” are highly accessible and have proved valuable across diverse applications. We introduce Pref-MH, a general exact MH sampler for judge-induced conditional distributions using only stochastic binary pairwise comparisons. Our key observation is that the MH unnormalized density ratio matches the preference odds of the Bradley-Terry (BT) choice model. The central challenge is that while MH requires precise ratio computation, BT judges provide only sampled binary feedback. To this end, we develop a valid accept/reject rule whose resulting Markov chain provably converges to the target distribution. We further show that, for a fixed proposal kernel and budget, Pref-MH is optimal in the Peskun-Tierney sense among this class of exact reversible acceptance rules. Experiments on text generation and molecular design with LLM judges, as well as image generation with VLM judges, demonstrate that Pref-MH provides a practical and flexible approach to conditional sampling when comparative feedback is relatively easy to obtain.
[LG-34] Poisson-Gamma Dynamical Systems with Time-varying Transition Dynamics ICDM2026
链接: https://arxiv.org/abs/2609.00896
作者: Jiahao Wang,Yijun Wang,Nan Fang,Sikun Yang
类目: Machine Learning (cs.LG)
*备注: Accepted by IEEE ICDM 2026
Abstract:Bayesian methodologies for handling count-valued time series have gained prominence due to their ability to infer interpretable latent structures and to estimate uncertainties. Among these Bayesian models, Poisson-Gamma Dynamical Systems (PGDSs) are proven to be effective in capturing the evolving dynamics underlying observed count sequences. However, the state-of-the-art PGDS still falls short in capturing the transition dynamics that are commonly observed in real-world count time series. To mitigate this limitation, a PGDS with time-varying transition kernel (TV-PGDS), is proposed to allow the underlying transition matrices to evolve over time. Three specifically-designed Dirichlet Markov chains (Dir-Dir, Dir-Gam-Dir, PR-Gam-Dir) are constructed to accommodate heterogeneous structural mutations within these dependencies. Leveraging Dirichlet-Multinomial-Beta data augmentation techniques, a fully-conjugate and efficient Gibbs sampler is developed to perform posterior simulation. Experiments show that, in comparison with related models, the proposed PGDS achieves improved predictive performance due to its capacity to learn time-varying dependency structure captured by the time-evolving transition matrices.
[LG-35] PINN for Broadband CARS Phase Retrieval: A Framework for Function Approximation and Inverse Modeling Problems in Nonlinear Spectroscopy
链接: https://arxiv.org/abs/2609.00883
作者: Ravi Teja Vulchi,Carl Messerschmidt,Mohammadsadegh Vafaeinezhad,Rajendhar Junjuri,Tobias Meyer-Zedler,Juergen Popp,Thomas Bocklitz
类目: Machine Learning (cs.LG); Data Analysis, Statistics and Probability (physics.data-an)
*备注:
Abstract:Phase retrieval in broadband coherent anti-Stokes Raman spectroscopy (BCARS) is an ill-posed inverse problem. The Raman-like signal is encoded in the imaginary part of the resonant susceptibility, which mixes coherently with a non-resonant background (NRB) that varies across acquisitions. We introduce an inverse physics-informed neural network (iPINN) that predicts Lorentzian peak parameters from raw BCARS spectra and reconstructs the resonant susceptibility through a differentiable analytical forward model. A transformer encoder assigns spectral features to 24 learnable peak slots, and a multi-view consistency loss enforces invariance across NRB pattern, NRB strength, and noise. Unlike direct spectral regression approaches, the method retains accuracy under varying acquisition conditions. On a public benchmark, iPINN achieves the lowest error among the tested baselines (MAE 0.016 vs. next-best 0.046). On 28 zero-shot test spectra acquired across seven solvents and four focal positions, accuracy is depth-invariant in five of seven solvents. These results show that inverse parametric prediction with a differentiable physical decoder supports robust phase retrieval across measurement conditions.
[LG-36] Conditional Flow Matching for ML-Based Inverse Design Problems
链接: https://arxiv.org/abs/2609.00863
作者: Juliana Felder,Milad Habibi,Soheyl Massoudi,Mark Fuge
类目: Machine Learning (cs.LG)
*备注: 13 pages, 2 figures, 6 tables. Accepted for presentation at EngOpt 2026
Abstract:Engineering inverse design is often limited by the high computational cost of iterative solvers for optimization problems constrained by partial differential equations (PDEs) and by their sensitivity to initialization. Deep generative models can produce candidate designs without rerunning the simulator at inference time. Generative adversarial networks (GANs) sample in one forward pass, whereas diffusion models require iterative reverse-time integration. In this work, we add conditional flow matching (CFM) to EngiOpt and compare it with a conditional diffusion model and a conditional generative adversarial network (cGAN) on structural (beams2d) and thermal (heatconduction2d) benchmarks from EngiBench using the same downstream optimization protocol. We use cumulative optimality gap (COG) and final optimality gap (FOG) as the primary metrics for evaluating the generated designs as warm starts for gradient-based refinement. On the evaluated EngiOpt implementations and two EngiBench tasks, CFM achieves the lowest measured COG, FOG, maximum mean discrepancy (MMD), and volume-fraction deviation on both tasks. CFM has mean volume-fraction deviations of 0.4% and 1.0% on beams2d and heatconduction2d, respectively, compared with 3.8% and 11.2% for diffusion. At Euler s = 16, CFM achieves 53.2 samples/s on beams2d, about 66 times the measured throughput of the evaluated diffusion baseline using 1000 network evaluations under the same timing protocol, with COG 1.182 +/- 3.126, compared with 1.173 +/- 3.100 for Euler s = 32. Across the two tasks, CFM produces warm starts with lower measured COG than both baselines and uses fewer network evaluations than diffusion. Comments: 13 pages, 2 figures, 6 tables. Accepted for presentation at EngOpt 2026 Subjects: Machine Learning (cs.LG) Cite as: arXiv:2609.00863 [cs.LG] (or arXiv:2609.00863v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.00863 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-37] Subspace Levenberg Marquardt Algorithms in Training Neural Networks
链接: https://arxiv.org/abs/2609.00789
作者: M. Duc Hoang
类目: Machine Learning (cs.LG); Optimization and Control (math.OC)
*备注:
Abstract:The Levenberg-Marquardt (LM) algorithm is a well-known second-order method for rapid convergence and strong robustness when training small- to medium-sized neural networks (NNs). However, its computational and memory costs increase significantly as the number of parameters in an NN grows. To address this limitation, subspace methods have been proposed, such as the Krylov subspace LM (KSLM) and the hybrid subspace LM (HSLM), making second-order algorithms more efficient. In this work, we evaluate the subspace Levenberg-Marquardt algorithms for regression and classification tasks in neural networks. We compare the performance of subspace LM variants with the classical LM method, as well as other popular first-order algorithms, such as stochastic gradient descent (SGD) and Adam.
[LG-38] Frozen Cores Need Task Signal: Fisher-Whitened Cross-Covariance for Low-Resource LLM Adaptation
链接: https://arxiv.org/abs/2609.00762
作者: Wentao Ye,Zhanming Shen,Zhiqing Xiao,Yao Ding,Haobo Wang,Gang Chen
类目: Machine Learning (cs.LG)
*备注:
Abstract:Parameter-efficient fine-tuning is usually framed as a question of how many parameters to update. Under a severe trainable-state budget, however, where those coefficients act is equally consequential. We study this choice through frozen-core adaptation: a calibration pass fixes left and right bases for each weight matrix, and fine-tuning optimizes only an r\times r core. This removes the ability of trainable factors to repair a poor initial span and makes subspace quality directly observable. We introduce FCCA, which estimates the signed input–error cross-covariance, whitens it with diagonal Fisher moments, truncates it in the resulting local metric, maps the selected directions back, and applies thin QR to obtain stable core coordinates. Under a matched r^2 budget, we compare eight basis constructors on 11 tasks, four model settings, and three seeds. On Qwen2.5-3B, FCCA reaches an 83.0 macro-average, 2.3 points above the next-best matched-budget constructor, and exceeds its unwhitened RawGrad control on all 11 tasks. It ranks first at all three Qwen scales and finishes within 0.13 points of the best method on Llama-3.2-1B. Controlled ablations show gains of 2.7–17.2 points from whitening and identify QR as necessary for stable core optimization in the tested regime. Finally, FCCA comes within 0.32 and 0.23 average points of LoRA and DoRA while optimizing 36.9K rather than roughly 7.4M parameters. These results show that a carefully selected fixed span can recover most of the benefit of movable low-rank factors at a much smaller trainable and optimizer-state cost.
[LG-39] xt Capability Loss in Vision-Language Adaptation: An Attention-Sink Diagnosis EMNLP2026
链接: https://arxiv.org/abs/2609.00746
作者: Minsik Choi,Geewook Kim,Young Geun Kim
类目: Machine Learning (cs.LG)
*备注: Accepted to EMNLP 2026. 29 pages, 6 figures, 15 tables. Code and data: this https URL . * Equal contribution. † Corresponding author
Abstract:Fine-tuning a pretrained LLM into a vision-language model (VLM) can erode the backbone’s text capability, with the damage concentrated on tasks that require following exact output rules, such as instruction following, chain-of-thought reasoning graded on a strictly parsed final answer, and similar evaluations with strict graders. We trace this gap to attention-sink corruption: VL fine-tuning perturbs the early sink position that anchors a large fraction of attention probability, and how well the base LLM preserves its sink tracks how much of the affected capability survives adaptation. Building on this view, we introduce Sink Strength, a single scalar computed on the base LLM in a few seconds on a single GPU that predicts post-VL degradation without any VL training. It consistently tracks relative degradation across the six VLM-LLM pairs and multiple format-sensitive tasks. Complementing this diagnostic, we find that post-pretraining QK-RMSNorm injection fails to reproduce the protection of native QK-RMSNorm, while several off-the-shelf weight-merging settings fail to recover the lost capability after VL training. These negative results underscore the value of screening backbones with Sink Strength before VL training and narrow the intervention space toward head-selective training-time protection.
[LG-40] Online Self-Weighted Fine-Tuning
链接: https://arxiv.org/abs/2609.00734
作者: Haiquan Wen,Yiwei He,Bei Peng,Guangliang Cheng
类目: Machine Learning (cs.LG)
*备注:
Abstract:Standard supervised fine-tuning (SFT) assigns the same explicit loss weight to every expert demonstration, regardless of the model’s changing competence over training queries. Reinforcement learning (RL) based methods adapt update strength using model-generated rollouts, but often require substantially more sampling and can be unstable on hard tasks. We propose \textbfOnline Self-Weighted Fine-Tuning (OSW-FT), a simple method that augments SFT with online, trajectory-level weighting. For each query, OSW-FT estimates the model’s current success rate using a small number of inference-only rollouts and rescales the standard SFT loss accordingly. The optimization direction remains anchored to the expert trajectory, while the update magnitude adapts online. For binary-verifiable reasoning, we connect this weighting to SFT and RL at the gradient level, inspired by variance-reduction principles. The resulting estimator is unbiased for the exact OSW-FT surrogate update for any finite rollout count, and we analyze convergence with respect to the corresponding surrogate objective. Evaluated across Qwen3 series ranging from 0.6B to 4B on multiple challenging benchmarks (e.g., AIME), OSW-FT consistently improves over SFT on small-to-medium scale models. OSW-FT offers a favorable compute-performance trade-off as a practical approach for fine-tuning small-to-medium LLMs on binary-verifiable reasoning tasks with only \textbf2 online rollouts.
[LG-41] MaskCode: Mask Transformer for Feedback-Assisted Coding With Linear Block Codes
链接: https://arxiv.org/abs/2609.00715
作者: Jonggyu Jang,Hongjae Nam,Vishrant Tripathi,David J. Love,Hyun Jong Yang
类目: Information Theory (cs.IT); Machine Learning (cs.LG); Signal Processing (eess.SP)
*备注: 15 pages
Abstract:Feedback-based coding schemes have demonstrated substantial performance gains over today’s open-loop coding schemes. Unfortunately, these gains are usually achieved in idealized settings with perfect feedback. Over the last few years, machine learning-based schemes have been shown to be promising solutions for implementing feedback-based codes, particularly when combined with short-block-length open-loop error correcting codes (ECCs) in a concatenated coding structure. However, existing ML-based feedback schemes remain agnostic to the outer code’s structure, potentially misallocating feedback resources on error patterns already correctable by the outer ECC. To address this, we propose MaskCode, a Transformer-based inner feedback code for concatenated coding systems, which explicitly incorporates structural knowledge of the outer linear block code into the inner feedback encoder design via two synergistic mechanisms: 1) a soft syndrome-based input that informs the encoder about potential parity constraint violations, and 2) a code-aware attention mask derived from the Tanner graph. We further show that end-to-end training with a differentiable belief propagation (BP) decoder offers no additional gain, as MaskCode’s structure-aware design already internalizes the structural knowledge of the outer code; in fact, backpropagation through the iterative BP decoder introduces gradient explosion, which degrades rather than improves performance. Extensive evaluations on BCH and LDPC outer codes demonstrate that MaskCode consistently outperforms all baselines, achieving up to 1.5 dB SNR gain.
[LG-42] Prediction-Assisted Pricing and Admission for LLM APIs with Stochastic Token Consumption
链接: https://arxiv.org/abs/2609.00710
作者: Patrick Wong
类目: Data Structures and Algorithms (cs.DS); Machine Learning (cs.LG)
*备注:
Abstract:An LLM application often sells or internally allocates several service products: a small or premium model, a short or long token cap, and possibly multiple posted prices. The operational decision is not merely which model answers a prompt. A price changes purchase probability, a token cap changes both user value and the tail of resource consumption, and accepted requests compete for shared compute and premium-model capacity. Demand and output length are initially uncertain, while an offline model may provide useful but imperfect predictions. We formulate sequential pricing and admission with stochastic resource consumption. Each arriving request belongs to an observable segment. The platform chooses a product–price pair or makes no offer; purchase, revenue, and resource use are then random. An offline predictor supplies a uniform, validated error radius for every segment–product cell. We propose Prediction-Clipped UCB (PCUCB), which intersects the offline prediction interval with an online confidence interval, evaluates products using resource shadow prices, and reserves a sample-path envelope before commitment. The prior gives a fast start when accurate, while online learning protects the platform when predictions are coarse. The analysis is modular. On a simultaneous confidence event, regret against a buffered fluid benchmark is bounded by a pacing term plus the cumulative diameter of the intersected intervals. For J segment-product cells and prediction radius \varepsilon , this yields [ \widetilde O\left( \sqrtT+(1+\bar\Lambda) \min\T\varepsilon,\sqrtJT\ \right), ] where \bar\Lambda bounds operational shadow prices. Thus the algorithm smoothly interpolates between an almost full-information regime and learning from scratch. Hard feasibility holds on every sample path through reservation envelopes. Subjects: Data Structures and Algorithms (cs.DS); Machine Learning (cs.LG) Cite as: arXiv:2609.00710 [cs.DS] (or arXiv:2609.00710v1 [cs.DS] for this version) https://doi.org/10.48550/arXiv.2609.00710 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Patrick Wong [view email] [v1] Tue, 1 Sep 2026 04:40:34 UTC (67 KB) Full-text links: Access Paper: View a PDF of the paper titled Prediction-Assisted Pricing and Admission for LLM APIs with Stochastic Token Consumption, by Patrick WongView PDFHTML (experimental)TeX Source view license Current browse context: cs.DS prev | next new | recent | 2026-09 Change to browse by: cs cs.LG References Citations NASA ADSGoogle Scholar Semantic Scholar export BibTeX citation Loading… BibTeX formatted citation loading… Data provided by: Bookmark checked="checked"class=“labs-tab-input”> Bibliographic Tools Bibliographic and Citation Tools Bibliographic Explorer Toggle Bibliographic Explorer (What is the Explorer?) Connected Papers Toggle Connected Papers (What is Connected Papers?) Litmaps Toggle Litmaps (What is Litmaps?) scite.ai Toggle scite Smart Citations (What are Smart Citations?) Code, Data, Media Code, Data and Media Associated with this Article alphaXiv Toggle alphaXiv (What is alphaXiv?) Links to Code Toggle CatalyzeX Code Finder for Papers (What is CatalyzeX?) DagsHub Toggle DagsHub (What is DagsHub?) GotitPub Toggle Gotit.pub (What is GotitPub?) Huggingface Toggle Hugging Face (What is Huggingface?) ScienceCast Toggle ScienceCast (What is ScienceCast?) Demos Demos Replicate Toggle Replicate (What is Replicate?) Spaces Toggle Hugging Face Spaces (What is Spaces?) Spaces Toggle TXYZ.AI (What is TXYZ.AI?) Related Papers Recommenders and Search Tools Link to Influence Flower Influence Flower (What are Influence Flowers?) Core recommender toggle CORE Recommender (What is CORE?) Author Venue Institution Topic About arXivLabs arXivLabs: experimental projects with community collaborators arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv’s community? Learn more about arXivLabs. Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?) mathjaxToggle(); We gratefully acknowledge support from our major funders, member institutions, , and all contributors. About Help Contact Subscribe Copyright Privacy Accessibility Operational Status (opens in new tab) Major funding support from
[LG-43] Patterning in Practice: Debiasing Reward Models with Susceptibilities
链接: https://arxiv.org/abs/2609.00699
作者: George Wang,Elizabeth Donoway,Daniel Murfet
类目: Machine Learning (cs.LG)
*备注:
Abstract:Reward models trained on human preferences are known to suffer from length, formatting, and other stylistic biases. In this paper we use patterning, which reweights each preference pair according to its measured effect on posterior expectation values of benchmark losses (its susceptibility), to debias a Gemma 2 9B Instruct reward model trained on Skywork-Reward-Preference v0.2. We obtain +14.2 \pm 1.2 pp on RM-Bench Hard, the split where style cues point against correctness (mean \pm s.e.\ over 5 seeds), with overall RM-Bench accuracy preserved, comparable to the strongest Hard-split gain reported by the closest published comparator (SteerRM, +13.2 pp). We demonstrate in a simple case that the reweighting is interpretable by tracing a side effect of the intervention (a regression on a safety subset of RM-Bench) to a small class of training pairs, which we confirm by ablation. The weights also transfer: those computed on Gemma 2 9B debias Gemma 2 2B and 27B with no recomputation, and transfer partially to Llama 3.1 8B. This is the first application of patterning, a program grounded in singular learning theory, beyond small models and synthetic tasks.
[LG-44] MUGEN: Generating Unlearnable Graph Examples for Multiple Learning Tasks
链接: https://arxiv.org/abs/2609.00696
作者: Ziyan Liu,Chengshuai Zhao,Huan Liu
类目: Machine Learning (cs.LG)
*备注:
Abstract:Graph data across diverse domains can expose valuable relational information to unauthorized representation learning, creating a pressing need for protection against such misuse. Unlearnable examples offer a data-level defense by perturbing a training release so that models trained on it fail to generalize to clean data. Existing methods generate unlearnable graph examples for only a specified downstream task. Consequently, a release protected against one task may remain learnable for other plausible uses, including node classification, graph classification, and link prediction, which the data owner cannot anticipate. We introduce MUGEN, to our knowledge the first framework for generating unlearnable graph examples that jointly protect all enabled tasks. From one clean dataset, MUGEN produces a single feature-perturbed release that protects every enabled task through a shared GNN encoder and task-specific heads. We devise a Task-Aligned Separability Objective (TASO), which leverages task prediction and classwise separability to strengthen unlearnability and its transfer across GNN backbones and enabled tasks. We further introduce Type-Adaptive Perturbation (TAP), which tailors perturbation optimization to node-attribute type, with direct search over feasible hard flips that accept only loss-improving updates for discrete node attributes and customized gradient-based updates for continuous node features, thereby enabling strong unlearnability across both settings. Experiments across five benchmarks, four backends and three learning paradigms demonstrate that MUGEN generates transferable unlearnable graph examples across GNN backbones and all three tasks, and remains effective under adversarial training and data augmentation.
[LG-45] Verdict Instability of OOD Scores under Reference Resampling
链接: https://arxiv.org/abs/2609.00691
作者: Donghoon Lee,Shinjin Kang
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注: 19 pages, 2 figures
Abstract:Post-hoc out-of-distribution detectors are fitted on a finite reference set, so every score they produce is an estimate. If we had chosen a different set, some verdicts would have moved. We measure that movement by resampling the reference set and recording the bootstrap standard deviation of the score, which we call verdict instability. It admits a closed form with no fitted parameters. The instability of a verdict is the within-class dispersion of the assigned class along the query’s direction, divided by the square root of that class’s reference count. That count is what separates verdict instability from the geometry of the score distribution, and it is identifiable only under class imbalance. Instability grows with the local dispersion. Far-OOD queries lie along the low-variance directions of an anisotropic embedding, so every distance-based score we test assigns its highest values to the verdicts that are most reproducible. Only estimators of local dispersion carry the sign a practitioner expects. We give a rule that predicts this sign for any score from a single label-free correlation, and abstention driven by a wrong-signed score turns out worse than abstention at random on every dataset we test.
[LG-46] HarmoCore: Functional Latent Diffusion for Sparse Reconstruction of Oscillatory Wave Fields
链接: https://arxiv.org/abs/2609.00679
作者: Lihao Chen,Xinyu Zhang,Panqi Chen,Lei Cheng,Ting Zhang,Jianlong Li,Shikai Fang
类目: Machine Learning (cs.LG); Computational Engineering, Finance, and Science (cs.CE)
*备注: 15pages, 8figures
Abstract:Reconstructing oscillatory wave fields from scattered sensors is a severely underdetermined inverse problem. Beyond the challenges of general physical-field reconstruction, wave responses are complex-valued, frequency-sensitive, and highly oscillatory, while costly simulation and sensing often leave only extreme-sparse observations. Existing low-rank, operator, and diffusion approaches are largely designed for real-valued, smoother fields; dense pixel-space diffusion is particularly inefficient for oscillatory complex fields and difficult to scale to 3D. We propose HarmoCore, which places a generative prior in a compact, continuous, and structured wave-field latent. HarmoCore represents joint real–imaginary channels with Functional Tucker cores over shared continuous spatial bases, learns a frequency-conditioned core diffusion prior, and performs Diffusion Posterior Sampling directly in core space. At fixed sensor coordinates, the multilinear decoder induces an explicit likelihood guidance operator, avoiding dense pixel-space correction. Optional target-equation residual guidance further promotes physical consistency. Experiments on 2D Helmholtz, 2D synthetic wave fields, and 3D Helmholtz show substantial gains under 1%–2% sensing while remaining practical in three dimensions.
[LG-47] DK-GBMKKM: Dynamic Kernel-Space Granular-Ball Multiple Kernel k-Means Clustering
链接: https://arxiv.org/abs/2609.00647
作者: Xiaoyu Lian,Yuchao Zhang,Shuyin Xia,Siqi Zhong,Xuzhao Xiang
类目: Machine Learning (cs.LG)
*备注:
Abstract:Multiple kernel k -means integrates complementary nonlinear similarities by learning a combination of base kernels. Its pointwise optimization, however, is sensitive to noisy and boundary samples and repeatedly operates on sample-scale kernel matrices. Granular-ball representations organize local sample groups into mesoscopic units, but granular balls generated once in the input space may be inconsistent with the fused-kernel geometry that evolves during multiple kernel learning. We propose dynamic kernel-space granular-ball multiple kernel k -means (DK-GBMKKM). The method generates granular balls in the current fused kernel space and alternates kernel-weight learning with granular-ball membership updates, allowing the representation to adapt to changes in the fused-kernel geometry. A sample-size-weighted granular-ball kernel is further constructed to preserve the contributions of balls of different sizes, and its positive semidefiniteness and related equivalence properties are established. Experiments on 12 public datasets demonstrate the strong overall clustering performance of DK-GBMKKM. The code has been open-sourced for reproducibility: this https URL.
[LG-48] opological Steering
链接: https://arxiv.org/abs/2609.00597
作者: Benoît Guérand,Tan Minh Nguyen
类目: Machine Learning (cs.LG)
*备注:
Abstract:With the rapid rise of large language models (LLMs), controlling undesirable model behaviors has become increasingly important. Existing behavioral control methods typically intervene directly in activation or feature space, but such approaches can be sensitive to outliers, distributional shifts, noise, and other local perturbations. Motivated by Topological Data Analysis (TDA), which captures global rather than purely local structure, we propose Topological Steering, a new framework for steering LLM behavior through the topological representation of activation spaces. Using persistence diagrams, our method connects activation-based steering with TDA and enables more robust behavioral control. We show that Topological Steering consistently modifies LLM behavior across multiple model families and model sizes.
[LG-49] CRAFT: Fine-Tuning Pre-hoc Explainability in AI-native 6G RAN
链接: https://arxiv.org/abs/2609.00590
作者: Pranshav Gajjar,Vijay K Shah
类目: Machine Learning (cs.LG)
*备注:
Abstract:The next generation of mobile networks is envisioned as fully AI-native, with AI-RAN architectures embedding small language models (SLMs) to perform reasoning over real-time telemetry. The state-of-the-art training paradigms for telecom LLMs, exemplified by RANSTRUCT-style supervised fine-tuning (SFT) on curated instruction data, are limited to post hoc rationalization. Here, the explanations, when produced at all, are generated after or independently of the decision, leaving the decision process unauditable. Pre-hoc reasoning, where a causal reasoning trace is produced before the output label, is preferable, and the broader LLM reasoning literature has made real progress toward it via RL methods such as Group Relative Policy Optimization (GRPO). Here we observe that transplanting this recipe into the telecom setting runs into a cold-start barrier: SLMs either learn to output the desired format or learn to predict the label, but rarely both. We identify this barrier and propose CRAFT, which stands for Cold-start Reasoning Alignment via Fine-Tuning, a data-centric method to autonomously generate a verified dataset of (input, trace, label) triplets. CRAFT fine-tunes SLMs on this verified data using low-rank adaptation (LoRA), requiring substantially less compute and wall-clock time than GRPO-based methods. On the TRACTOR and IC xApp telecom datasets, CRAFT achieves up to 86.5% and 94.6% for accuracy and F1 with no parse failures, while direct GRPO and SFT+GRPO fail to exceed 28% and 53.5% F1 with multiple parse failures. We further show that CRAFT-initialized policies serve as a robust foundation for subsequent GRPO fine-tuning, as under diverse reward functions the performance remains consistent with no parse failures. Finally, we demonstrate that CRAFT consumes 59% less energy than GRPO-based baselines, making it a sustainable path to deployable, auditable AI in 6G RAN.
[LG-50] Manifold-Aware General Coded Computing for Strag gler-Resilient Distributed Computing
链接: https://arxiv.org/abs/2609.00552
作者: Parsa Moradi,Mohammad Ali Maddah-Ali
类目: Machine Learning (cs.LG); Information Theory (cs.IT)
*备注: 5 pages, 4 figures
Abstract:Existing coded-computing designs do not explicitly exploit the intrinsic structure of the input data. In communication systems, statistical structure and redundancy are often removed through source coding (or compression) before channel coding is applied. This principle, however, does not transfer directly to coded computation. In many computational tasks, particularly in machine learning, the structure of the data is precisely what the computation seeks to exploit to infer outputs or learn meaningful patterns. Consequently, coded-computing schemes should preserve and leverage this structure in their code design, rather than ignoring or eliminating it through source coding. This observation motivates a different perspective on code construction. In many channel-coding schemes, such as Reed-Solomon codes, coded symbols are generated by evaluating a low-dimensional algebraic representation at selected points. In contrast, many high-dimensional datasets naturally concentrate near low-dimensional manifolds. In this paper, we exploit this intrinsic geometry by designing coded samples that follow the natural manifold of the data, rather than imposing an artificial low-dimensional structure unrelated to the data distribution. Inspired by graph-based manifold learning, we propose a manifold-aware encoding strategy for general coded computing (GCC). Experiments on neural network inference and high-dimensional polynomial evaluation demonstrate that the proposed strategy consistently and significantly reduces the mean squared recovery error under straggling compared with standard GCC. Comments: 5 pages, 4 figures Subjects: Machine Learning (cs.LG); Information Theory (cs.IT) Cite as: arXiv:2609.00552 [cs.LG] (or arXiv:2609.00552v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.00552 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-51] GenONet: A Generative operator Network for High-Resolution Precipitation Nowcasting
链接: https://arxiv.org/abs/2609.00544
作者: Mohammad Kian Golkar,Luciano Alves de Oliveira,Mohammad Khanjani
类目: Machine Learning (cs.LG); Atmospheric and Oceanic Physics (physics.ao-ph)
*备注: 28 pages, 11 Figures
Abstract:High-resolution precipitation nowcasting is critical for reducing the impacts of severe weather but remains difficult because of rapid storm evolution. Deep learning models have shown great promise for this task, but their predictive skill often deteriorates over longer forecast horizons. This leads to increasingly blurry forecasts that fail to capture the complex, non-linear evolution of storm systems. In order to address these limitations, we introduce Spatio-Temporal U-DeepONet (GenONet), a novel architecture for long-range precipitation forecasting up to 3 hours, specifically designed to produce sharp and physically consistent results. GenONet’s architecture pioneers the use of a Deep Operator Network (DeepONet) as a generator within a Generative Adversarial Network (GAN) framework for this task. The DeepONet learns the continuous-time dynamics of precipitation, ensuring stability over long forecast horizons. Adversial training against a spatio-temporal discriminator compels the model to produce sharp, coherent forecasts, while a physics-informed loss regularizer, derived from the Moisture Conservation Equation, improves physical plausibility in our ablation setting. Quantitative evaluations show that our model achieves consistently higher scores on most of the metrics, especially for highintensity events and at longer lead times. Qualitatively, GenONet produces structurally coherent forecasts that maintain their integrity, whereas baseline models degrade into indistinct patterns. Finally, an ablation study confirms the benefit of this physics-informed loss, highlighting the strength of combining operator learning with adversarial training.
[LG-52] DeSyR: A Decoupled Symbolic Recovery Framework with PINN-Guided Structure Search and Physics-Informed Coefficient Refinement
链接: https://arxiv.org/abs/2609.00530
作者: Pancheng Niu,Jun Guo,Qiaolin He,Jingcai Guo,Yanchao Shi
类目: Machine Learning (cs.LG)
*备注: 87 pages, 21 figures
Abstract:Recovering compact explicit solutions from neural approximations is challenging when imperfect teacher data guide symbolic topology search and coefficient estimation. We present DeSyR, a decoupled symbolic recovery framework for differential equations. A physics-informed neural network guides repeated searches to construct candidate topologies with provisional constants. Once a topology is fixed, its coefficients are refined solely from the governing equation and prescribed constraints, followed by gated selection and verification. For linear fixed-topology parameterizations, we characterize teacher-error inheritance and show that finite-weight mixed data–physics fitting retains an O(\beta^-1) teacher-dependent contribution when the teacher error projects onto the model space. Under well-posedness, representability, zero-residual attainment, and discrete determinacy, physics-only refinement conditionally recovers exact coefficients; for nonlinear parameterizations, the corresponding guarantees are local. DeSyR is evaluated on 15 differential-equation problems across 18 configurations covering high-order, space–time, multidimensional, nonlinear, and coupled systems. A candidate-level audit yields a 99.23% convergence rate among free-parameter refits, while every selected refinement involving free coefficients converges. Configuration-level median refined relative L_2 errors are 2.31\times10^-14 or lower. In same-topology comparisons, refinement reduces error by eight to fourteen orders of magnitude. These results show that an approximate neural teacher can guide topology discovery without imposing its error scale on final recovered coefficients, provided a target-capable topology is retained and physics-only refinement converges.
[LG-53] Why Multi-Layer Message Passing Works: Completeness Theory for Graph Neural Network Interatomic Potentials
链接: https://arxiv.org/abs/2609.00528
作者: Pingbing Ming,Han Wang
类目: Machine Learning (cs.LG); Mathematical Physics (math-ph); Chemical Physics (physics.chem-ph); Computational Physics (physics.comp-ph)
*备注:
Abstract:We prove that the Hypergraph Neural Network, an invariant architecture with 3-body message passing, is a universal approximator for potential energy surfaces. Our main contribution is a multi-layer completeness theory. We show that L layers of message passing on sparse, cutoff-based graphs achieve the same representational power as having access to the full L -hop neighborhood, provided the configurations are generic, satisfy an overlap condition and a connectivity condition. This provides the first rigorous justification for the common practice of using multi-layer message passing with a per-layer cutoff smaller than the physical interaction range, the setting used by virtually all practical graph neural network based machine-learned interatomic potentials. As immediate consequences, we show that both DPA3 and CHGNet architectures inherit universal approximation.
[LG-54] Learning Task-Specific Antibody Representations via Function-Aware Masking
链接: https://arxiv.org/abs/2609.00518
作者: Ayan Goel,Thomas A. Walton,Amirali Aghazadeh
类目: Machine Learning (cs.LG); Biomolecules (q-bio.BM)
*备注: Accepted to MLCB 2026 (Oral)
Abstract:Antibody-specific language models pretrained via masked language modeling (MLM) learn representations that are critical for downstream sequence design and property prediction tasks. Yet, the corruption process itself is rarely leveraged as a source of inductive bias during pretraining. While preferentially masking complementarity-determining regions (CDRs) improves binding-related predictions, antibodies possess diverse biological priors over a variety of functions. Herein, we introduce function-aware masking, a family of pretraining algorithms that align mask placement with specific functional priors (e.g., from IMGT annotations or structure predictions) to shape the learned representation space. We show that these specialist masking strategies significantly improve performance on their respective objectives, yielding up to a 14% gain on structure-related tasks and up to a 5.9x improvement on CDR-related tasks. To further improve performance across multiple functional axes, we develop hybrid masking strategies that integrate multiple priors, balancing reconstruction over binding, structural, and biophysical objectives. Our results demonstrate that informed mask placement provides a parameter-free mechanism for imposing functional inductive biases in antibody language model training.
[LG-55] VATO: A Vortex-Force-Aware Transformer Operator for Unsteady Separated Aerofoil Flows
链接: https://arxiv.org/abs/2609.00507
作者: Xingxin Yang,Zhan Zhang,Yichen Li,Juan Li
类目: Machine Learning (cs.LG)
*备注:
Abstract:Accurate prediction of unsteady separated flows is challenging because the aerodynamic loads depend on nonlinear separation and vortex-shedding dynamics. Although high-fidelity CFD resolves these mechanisms, its cost limits repeated use in design and control. Standard field-level surrogate training, however, does not distinguish the flow regions that contribute most strongly to the aerodynamic loads. We introduce VATO (Vortex-Force-Aware Transformer Operator), which couples the Vortex Force Map (VFM) method to a geometry-aware neural operator through two complementary mechanisms. VATO-S adds training-only supervision of the local VFM force-contribution field, with no increase in model size or inference cost. VATO-A uses VFM contribution and sensitivity fields to prioritise force-relevant source locations for residual cross attention. The methods are evaluated on unsteady CFD data for double-edged-plate aerofoils over 54 trajectories from nine geometries. Over lead times of 1-20~ms, VATO-S reduces velocity, pressure, and vorticity errors by 10.4%, 1.0%, and 15.6%, respectively, while VATO-A achieves reductions of 15.8%, 7.5%, and 31.2%. VATO-S gives the lowest VFM-derived drag error, whereas VATO-A gives the lowest pressure-derived lift and drag errors. Over lead times extending 50% beyond the training range, VATO-A retains a 26.9% reduction in vorticity error and larger improvements in all four force readouts, despite reduced gains in velocity and pressure. These results show that force-aware operator learning can improve both flow-field prediction and aerodynamic functional accuracy in unsteady separated flows.
[LG-56] A hybrid quantum-classical neural network for learning to route
链接: https://arxiv.org/abs/2609.00489
作者: Marcus Rolf Peter Ritt,Alexsandro Santos da Rosa Júnior,Marcos Vinicius Reballo,Cesar Augusto do Amaral,Fernando Augusto Caletti de Barros
类目: Machine Learning (cs.LG); Quantum Physics (quant-ph)
*备注: 5 pages. Submitted to the Congresso Brasileiro de Ciências e Tecnologias Quânticas (CBCTQ 2026)
Abstract:This work studies hybrid quantum-classical neural networks for learning routing heuristics. Specifically, this paper asks whether small quantum neural networks can replace parameter-heavy modules inside a competitive attention-based routing model while maintaining solution quality. For the capacitated vehicle routing problem, encoder feed-forward replacement emerges as the most promising design: it reduces the number of model parameters by 56.6% while keeping the hybrid model close to the classical neural baseline at small and medium instance sizes, although the gap grows for larger instances. This work also compares to classical routing algorithms, which remain highly competitive and often superior on the fixed Euclidean test sets. Our results therefore do not indicate quantum advantage or solver dominance, but identify encoder feed-forward replacement as a viable hybrid-module compression strategy for neural combinatorial optimization.
[LG-57] AdaptNTK: Adaptive Uncertainty Quantification and Active Learning for Neural Network Potentials
链接: https://arxiv.org/abs/2609.00488
作者: Prajwal Ananth,Shuwen Yue
类目: Machine Learning (cs.LG); Chemical Physics (physics.chem-ph)
*备注:
Abstract:Machine learning interatomic potentials bridge the gap between quantum chemical precision and classical computational speed, enabling molecular dynamics simulations with first-principles accuracy. Their reliability is often improved through active learning, which iteratively expands the training set by identifying uncertain, out-of-distribution configurations. Existing uncertainty-quantification methods often involve a trade-off between computational cost and reliability, and generally cannot account for redundancy as an acquisition batch is assembled. Here, we introduce AdaptNTK, a single-model framework that measures uncertainty as a regularized Mahalanobis distance in empirical neural tangent kernel (NTK) feature space. With the NTK features fixed during acquisition, the uncertainty depends on the acquired configurations but not their reference labels. This allows the uncertainty to be updated recursively after each selection without retraining, reducing redundancy within an acquisition batch. On held-out rMD17 data, AdaptNTK achieves the highest mean correlations with force errors (Spearman 0.68, Pearson 0.71) and matches a three-member ensemble in error retention. In active learning experiments, AdaptNTK achieves the lowest force errors across rMD17 and Transition-1X, with particularly strong performance on transition-state configurations in Transition-1X. AdaptNTK provides a 2.6-fold speedup per Transition-1X cycle relative to the ensemble, providing efficient single-model uncertainty estimation with sequential updates for data-efficient active learning.
[LG-58] Context Window Failures in Relational Foundation Models ICML2026
链接: https://arxiv.org/abs/2609.00460
作者: Denis Oliveira Correa,Francisco Galuppo Azevedo
类目: Machine Learning (cs.LG)
*备注: Accepted at the 2nd Foundation Models for Structured Data Workshop at ICML 2026, Seoul, South Korea. OpenReview: this https URL
Abstract:Recent Relational Deep Learning architectures have been proposed as foundation models for multi-table relational data, yet they impose constrained neighborhood budgets that force row truncation when an entity has many related records. We introduce Animus, a synthetic financial dataset in which predicting customer income requires aggregating up to tens of thousands of transactions. On the raw representation, three recently proposed models (RT, Griffin, RelGT) achieve R^2 \le 0.18 ; a single, routine, temporal pre-aggregation step recovers R^2 up to 0.65 . This questions whether current relational foundation models are ready for high-cardinality real-world data.
[LG-59] Can LLM s Use Relational Transformer Embeddings? ICML2026
链接: https://arxiv.org/abs/2609.00457
作者: Francisco Galuppo Azevedo,Clarissa Lima Loures
类目: Machine Learning (cs.LG)
*备注: Accepted at the 2nd Foundation Models for Structured Data Workshop at ICML 2026, Seoul, South Korea. OpenReview: this https URL
Abstract:Injecting frozen relational-encoder embeddings as soft tokens into a large language model (LLM) is a conceptually appealing fusion strategy: the encoder handles multi-table structure, the LLM handles language and reasoning, and no lossy text serialization is required. We test this hypothesis concretely by injecting embeddings from a frozen Relational Transformer (RT) into Qwen3.5-4B via a learned MLP projection and LoRA adaptation, trained first with supervised fine-tuning (SFT) on chain-of-thought reasoning traces and then with group-based reinforcement learning (GSPO). We evaluate across 10 binary classification tasks on 6 relational databases from RelBench, under four supervision regimes: single-task (ST), within-dataset (WD), cross-dataset (CD), and all-task (ALL). The hybrid model does not consistently outperform standalone RT: it is frequently below random, highly sensitive to serialization format and relational-token budget, and unstable under RL training. We report these negative results and analyze the failure modes, arguing that soft-token fusion requires stronger alignment objectives and schema-aware design before it can serve as a reliable route to relational prediction.
[LG-60] How Temporal Correlations Shape Memory in Linear Recurrent Neural Networks
链接: https://arxiv.org/abs/2609.00420
作者: Arnol Manuel Fokam,Fasseu Sieyondji Akpevwoghene,Edem Fiifi Dawson
类目: Machine Learning (cs.LG)
*备注:
Abstract:The linear recurrent neural network (LRNN) is a simple model for studying how much memory a network builds up as it trains. For uncorrelated inputs, earlier work found that training itself settles the network between keeping the past and reacting only to the present. Real sequences are correlated, and we solve the learning dynamics exactly for correlated inputs. In the solution, keeping the past carries a cost. The whole effect of correlation lands on that cost. This cost reduces to the earlier one when inputs are uncorrelated and grows once they are positively correlated. Three findings follow. (1) Correlation reshapes the course of learning, not only its end. Memory builds, overshoots, and is partly removed, and the settled network keeps less of the past. (2) Memory switches off at a threshold set by one number, how much each input resembles the one just before it. Neither sequence length nor longer-range correlation moves this threshold. Memory is worth keeping only when the task needs the previous input more than the current input already supplies it through correlation with the past. (3) The best network changes too. Zero error demands a feedthrough, a path that passes the current input straight to the network’s output and remembers nothing, and training builds it unprompted when given one spare hidden dimension. Our work turns one property of the input into a prediction of whether a network learns memory and explains why correlated data turns recurrent networks into change detectors.
[LG-61] DynaNDE: Dynamic Near-Data Expert Scheduling for Batched MoE Inference
链接: https://arxiv.org/abs/2609.00407
作者: Xiaoyang Lu,Belthangady Akash Vi Narayana Pai,Xian-He Sun
类目: Hardware Architecture (cs.AR); Machine Learning (cs.LG)
*备注:
Abstract:Mixture-of-Experts (MoE) models enable efficient scaling of large language model (LLM) inference but suffer from substantial data-movement overhead when deployed on neural processing unit (NPU)-based systems. Near-Data Processing (NDP) provides a promising way to mitigate this bottleneck via cooperative NPU-NDP execution. However, existing NPU-NDP MoE systems do not fully account for hardware heterogeneity, dynamic expert-level concurrency, and temporal expert reuse during batched inference. This paper presents DynaNDE, a dynamic near-data expert scheduling framework that exploits NPU-NDP collaboration to accelerate batched MoE inference. DynaNDE introduces an analytical performance model that captures hardware heterogeneity, data-movement costs, and communication-computation overlap in cooperative NPU-NDP execution. Guided by this model, DynaNDE determines per-layer expert scheduling across the NPU and NDP while accounting for expert-level concurrency. DynaNDE also incorporates a reuse-aware runtime that avoids redundant parameter movement when experts reside in NPU memory. Experimental results show that DynaNDE achieves substantial throughput improvements over the state-of-the-art NPU-NDP MoE serving framework, with average speedups of 2.6 \times and 2.2 \times for the prefill and decoding stages, respectively.
[LG-62] A Multi-Branch Feature Fusion Approach for Health Misinformation Detection and Propagation
链接: https://arxiv.org/abs/2609.00403
作者: Mkululi Sikosana,Sean Maudsley-Barton,Oluwaseun Ajao
类目: Machine Learning (cs.LG); Social and Information Networks (cs.SI)
*备注: 1 figure, 8 tables
Abstract:This paper presents a multi-branch fusion framework for detecting and characterising the propagation of health misinformation in online social networks (OSNs). Grounded in the Elaboration Likelihood Model (ELM) and the Theory of Planned Behaviour (TPB), the model fuses transformer-based semantics with rhetorical cues, stance representations, and psychologically motivated proxies in a unified multi-task architecture. In addition to binary classification, we introduce the Cognitive Propagation Score (CPS), an interpretable post-hoc auxiliary score computed from psychologically motivated, text-derived cues capturing argument complexity, emotional intensity, and content-derived virality potential, to support diffusion-risk reasoning when engagement ground truth is incomplete or unavailable. Experiments on three benchmark datasets, Constraint, COVID–19_FNIR, and Monkeypox, show strong classification performance, achieving ROC–AUC up to 0.9999 on COVID–19_FNIR, while propagation-oriented ranking achieves near-perfect agreement when engagement-derived supervision is available (Monkeypox, Spearman’s \rho = 0.9952 ) and similarly high ranking alignment under proxy-based supervision on COVID–19_FNIR ( \rho = 0.9954 ). Compared with representative literature baselines, the fusion model improves detection on Constraint and COVID–19_FNIR, while Monkeypox remains more challenging, reflecting domain- and signal-specific differences. Ablation analysis further indicates that psychological and rhetorical branches provide complementary gains beyond semantic embeddings. Overall, the framework bridges cognitive theory and neural modelling to improve transparency and to support scalable misinformation monitoring, with future work required to validate CPS against human-centred diffusion judgements.
[LG-63] Neural means and kernel corrections for operator learning
链接: https://arxiv.org/abs/2609.00389
作者: Yitzchak Shmalo
类目: Machine Learning (cs.LG); Probability (math.PR); Machine Learning (stat.ML)
*备注:
Abstract:We combine neural network means with exact Matérn kernel regressions of their residuals and of their learned features, and evaluate the pairing on two public emulation problems with published baselines: the structural-mechanics benchmark of de Hoop et al. and the OCO-2 radiative-transfer emulator of Lamminpää et al. On structural mechanics the combination reaches 4.55% test error, matching the best published architecture, and 5.38% against a published 6.49% in the low-data regime. On OCO-2 it improves on the published Gaussian-process emulator on that problem’s own test points, outright on two of the three spectral bands; the same kernel that trails the network tenfold on the raw state overtakes it on the network’s features, and we measure why (the target’s squared native-space norm drops about fortyfold at fixed effective dimension) and prove the mechanism. Where the two families tie instead, the residuals of every architecture we train correlate above 0.86 and their shared component is flat in diversity and sample size, which reads the published plateau as a property of the data. Supporting results include a second-moment identity that predicts stacking outcomes from measured correlations, an optimal-recovery certificate, and a distribution-free coverage band, the only uncertainty signal that survives our tests.
[LG-64] Hidden relationships in a document-derived property graph: top-k chunk embeddings and inverse-distance weighting over a dynamically evolving ontology
链接: https://arxiv.org/abs/2609.00387
作者: Bilge Kaan Karamete,Hunter Casten
类目: Computational Engineering, Finance, and Science (cs.CE); Machine Learning (cs.LG)
*备注:
Abstract:Large language models extracting knowledge graphs from text capture only explicitly stated facts, often leaving semantically related entities disconnected across documents. We present an additive, engine-neutral second pass that discovers these latent ties without altering extracted facts. Each document is chunked and embedded once; top-k nearest- neighbor queries across existing chunks yield candidate node pairs via entity membership maps. Candidate pairs are scored using Shepard inverse-distance weighting with a rescaled chord distance metric, avoiding the threshold-collapsing flaw of affine cosine scoring behind a k-NN gate. Un-gated per-pair accumulators form a commutative monoid, ensuring the pipeline is strictly order-independent and scales incrementally without recomputing prior documents. Implemented across FalkorDB, Kinetica, ArangoDB, and Neo4j, our method shows that 768- and 240-dimensional embeddings retain 92% and 72% edge fidelity against a 3072-D baseline while achieving a 25x faster top-k formulation.
[LG-65] Deterministic LLM Inference Across GPU Kernels: Power-of-Two INT8 Quantization Scales and the Limits of Tolerance-Based Conformance
链接: https://arxiv.org/abs/2609.00363
作者: Teng-Ruei Chen
类目: Machine Learning (cs.LG); Software Engineering (cs.SE)
*备注: 9 pages (IEEEtran two-column), 3 figures, 5 tables. Fault injection against a two-stage-pinned prediction matrix (8,232 scored cells; 63-cell pre-data core, corrected and re-pinned after a disclosed smoke run); power-of-two requantization measured at 1.7B/8B/14B. Companion to arXiv:2608.13756 , whose open questions on check sensitivity and deployability this paper answers
Abstract:Conformance suites for quantized GEMM kernels ask whether two implementations agree within a tolerance. We measure what such a suite can detect. Injecting nine faults into a reference INT8 pipeline over 8,232 layer–fault–regime cells of Qwen3-1.7B, we find that every one of five epilogue faults – scale precision, double rounding, multiplication order, output truncation, fused ordering – moves the output by at most a single bfloat16 spacing, and by exactly one whenever it moves it at all, across 5,880 cells. A tolerance of one spacing is therefore blind to the entire class by construction: four of the five faults are detected by no check in the suite, and the fifth only under power-of-two scales. Faults that violate the accumulator’s exactness preconditions, or that break operand sharing, are detected without exception, and a null fault never fires. What a tolerance-based suite of this shape establishes is therefore narrower than interchangeability: that the preconditions hold, that operands are shared, and that differences stay within one spacing. The power-of-two constraint that exposes the one detected fault is also deployable. Requantizing every weight scale to its nearest power of two makes CUTLASS and Triton agree bitwise at every linear layer (196/196 and 252/252, against 8/196 and 10/252 under the checkpoints’ own scales) and yields byte-identical generated token sequences at 1.7B, 8B and 14B (8/8 prompts, against 0/8 at all three). Observed perplexity point estimates are +0.32%, -0.28% and +0.48%; the 90% intervals cover zero at the two smaller sizes but not at 14B, reaching +0.71% and +0.76%. A previously reported +157% perplexity for this intervention was an artifact of a probe that rewrote scales without requantizing the weights; separating the effects attributes 99.8% of it to the resulting weight–scale mismatch rather than to the power-of-two constraint itself.
[LG-66] Do LLM s Know Your Neighborhood? Auditing LLM Priors for Neighborhood-Level Mobility Prediction and Structural Alignment
链接: https://arxiv.org/abs/2609.00345
作者: Saad Mohammad Abrar,Eesha Kurella,Arnav Dadarya,Naman Awasthi,Kazi Tasnim Zinat,Vanessa Frias-Martinez
类目: Machine Learning (cs.LG); Computers and Society (cs.CY)
*备注:
Abstract:Human mobility is central to urban planning, transportation, public health, and emergency response, yet fine-grained trajectory data are often proprietary, restricted, and privacy-sensitive. Large language models (LLMs) offer a potential alternative by generating plausible mobility traces and predicting individual movement, but their ability to infer aggregate neighborhood-level mobility remains unclear. We evaluate zero-shot LLMs on Census Block Group-level mobility prediction across four U.S. metropolitan areas using anonymized Cuebiq data to construct point-level, trajectory-level, and temporal mobility outcomes, paired with sociodemographic and built-environment predictors. We compare LLM predictions with supervised baselines and introduce a directional alignment analysis to test whether LLM-implied predictor effects agree with empirical OLS and Jonckheere-Terpstra trends. Supervised models achieve 0.580 average accuracy, compared with 0.435 for the best LLM, with spatial extent outcomes showing the strongest predictability but also the largest LLM-baseline gaps. Directional analysis shows that LLMs often rely on coarse, stable predictor-level priors that remain similar across outcomes and cities, including asymmetric treatment of protected-group predictors. Overall, LLMs can partially recover aggregate mobility patterns from urban context, but their predictions should not be treated as structurally grounded without auditing empirical alignment and potential bias.
[LG-67] Lightweight Adaptation of EEG Foundation Models for Stroke Motor Imagery Decoding: Domain Shift and Subject-Level Robustness
链接: https://arxiv.org/abs/2609.00282
作者: Anh T. Nguyen,Zihua Sun,Michelle J. Johnson
类目: Computational Engineering, Finance, and Science (cs.CE); Machine Learning (cs.LG)
*备注:
Abstract:Motor imagery (MI) electroencephalography (EEG) decoding could support post-stroke rehabilitation, but models developed on healthy cohorts may not transfer reliably to pathological EEG. We evaluated whether Low-Rank Adaptation (LoRA) can efficiently adapt three pretrained EEG foundation models (i.e., LaBraM-base, REVE-base, and REVE-large) for binary left- versus right-hand MI decoding. Frozen-backbone head-only baselines and LoRA adaptation were evaluated using subject-wise five-fold cross-validation on the PhysioNet EEG Motor Movement/Imagery Dataset and a binary subset of the UET175 dataset comprising 30 stroke participants. On EEGMMIDB, LoRA increased accuracy to 0.822 for LaBraM-base and 0.957 for REVE-base. On UET175, all head-only models performed near chance. With LoRA, LaBraM-base remained near chance (0.499 \pm 0.009), whereas REVE-base reached 0.847 \pm 0.194 and outperformed REVE-large (0.806 \pm 0.178), indicating that increased model capacity alone did not improve stroke-domain adaptation. The strongest stroke configuration, REVE-base LoRA, was further evaluated using within-cohort leave-one-subject-out cross-validation (LOOCV), showing 0.952 mean accuracy, but subject-wise accuracy ranged from 0.586 to 1.000, revealing a small low-performing tail. Zero-shot transfer from EEGMMIDB to UET175 remained near chance (0.464 \pm 0.072). These findings show that healthy-benchmark performance does not ensure transfer to stroke EEG. Translation of EEG foundation models to pseudo-online or real-time rehabilitation BCIs should therefore include target-domain adaptation and subject-level assessment of temporal informativeness, spatial sensitivity, and physiological discriminability.
[LG-68] Elite-Weighted Supervised Fine-tuning for Goal-Directed Molecular Optimization
链接: https://arxiv.org/abs/2609.00189
作者: Shiyun Wa,Yifei Wang,Anna G. Green,Simone Sciabola,Ye Wang
类目: Machine Learning (cs.LG)
*备注:
Abstract:Goal-directed optimization is essential for steering molecular generators to propose candidates with desired properties. However, it is often implemented with policy-gradient reinforcement learning, which requires a generation-trajectory log-probability whose form depends on the model architecture and generation procedure. This makes an optimizer difficult to reuse across architectures and conditional generative designs. Supervised fine-tuning needs none of that machinery, but its update is driven by a fixed dataset, so the reward never enters the update. We introduce Elite-Weighted Supervised Fine-tuning (EW-SFT), which uses reward to guide elite selection of high-scoring molecules, and updates the model by its own pretraining loss on that set. Ablations show that reward information is passed primarily through elite selection, rather than through continuous weighting within the selected set. Because the update consumes only scored molecules and the model’s native loss, the same rule applies across autoregressive, masked-diffusion, and discrete-flow generators, and across de novo, motif-extension, and linker-design tasks. Under a fixed budget of 3D shape alignment oracle calls on two kinase reference compounds, EW-SFT consistently outperforms the corresponding native optimizers. It further improves goal-directed optimization under a 2D similarity oracle on four held-out references and achieves comparable performance on a sample-efficiency benchmark without a trajectory-level RL formulation. These results demonstrate that EW-SFT is a unified and effective optimizer across molecular generators, design constraints, references, and oracles.
[LG-69] Generative artificial intelligence for reliable mechanistic reasoning for corrosion
链接: https://arxiv.org/abs/2609.00099
作者: Bharath M N,R K Singh Raman,Alankar Alankar
类目: Machine Learning (cs.LG); Materials Science (cond-mat.mtrl-sci)
*备注:
Abstract:Corrosion accounts for approximately 4% of global GDP, and reliable prediction is essential for timely mitigation. Machine learning effectively predicts corrosion rates from composition, microstructure, and environmental variables, but cannot explain the underlying mechanisms. A reliable approach in safety-critical materials engineering requires not only accurate retrieval but also mechanistically defensible reasoning, a capability that existing factuality metrics cannot assess. This work presents a domain-adapted retrieval-augmented generation framework for corrosion knowledge synthesis, demonstrated on magnesium alloy corrosion. Three open-weight language models (Llama-3.1-8B, Qwen-2.5-7B, Mistral-7B) are fine-tuned on 3,309 expert-verified question-answer pairs from 840 peer-reviewed papers and integrated with a hybrid dense-lexical retrieval pipeline. Retrieval augmentation produces Token F1 gains of 143-194%, with system faithfulness of 0.964 and context recall of 0.988. Blind external validation on newly published literature and in-house electrochemical data confirms trend-level generalisation. Reason Map, a proposition-graph framework, is further introduced; it independently constructs directed evidence graphs from generated answers and retrieved literature, enabling systematic detection of causal direction inversions and unsupported inferential leaps that flat factuality metrics cannot expose. The modular architecture can be applied across domains, offering a generalizable blueprint for trustworthy AI-assisted knowledge synthesis to circumvent corrosion, which can also be applied to other engineering domains.
[LG-70] Local Reference Geometry Residual Augmentation for Imbalanced Time Series Classification
链接: https://arxiv.org/abs/2609.00093
作者: Chuanhang Qiu,Yanran Xu,Yue Wang,Anthony Bagnall
类目: Machine Learning (cs.LG)
*备注: 10pages, 2 figures
Abstract:Imbalanced time series classification is often addressed by changing the training distribution, objective, logits, or final threshold. These interventions address important biases, yet leave a representation-level question unmeasured: after minority support is reduced, does a learned feature space remain locally reliable around minority regions? We identify a training-local geometry failure: under imbalance, minority cases can lie in sparse, rest-dominated, or mixed feature-space neighborhoods, even when the representation retains useful global class structure. To diagnose and repair this failure, we propose Local Reference Geometry (LRG), a lightweight post-hoc feature augmentation module applied between a fixed feature extractor and the classifier head. Using training features only, LRG measures local exposure and class-mixture risk, then augments each fixed feature with a standardized signed displacement from nearby training geometry and an LDA-projected residual summary. On controlled UCR/Bake Off Redux imbalance benchmarks, paired raw-versus-LRG comparisons show gains for learned, pretrained, and fixed representations, including when LRG is combined with training-level interventions and post-encoder classifier corrections. Ablations show that the gain comes from the signed local residual appended to the original feature, rather than from generic prototype distances, affinity features, scalar statistics, or VLAD-style codes. Further analyses support the proposed local-geometry failure hypothesis: minority neighborhoods become increasingly rest-exposed under imbalance, training-local risk identifies error-prone regions, and LRG gains concentrate in those high-risk regions.
[LG-71] Safin-1: Safety from Within through Memory-Native State Evolution
链接: https://arxiv.org/abs/2609.00092
作者: Ming Zhang,Kaisen Yang,Shu Yu,Ermo Hua,Zhekai Chen,Cheng Jin,Jingnan Zheng,Yi Zhang,Zhongtian Ma,Jiawei Zhou,Sirui Chen,Qiaosheng Zhang,Xiang Wang,Ning Ding,Xia Hu,Bowen Zhou,Youbang Sun,Chaochao Lu
类目: Machine Learning (cs.LG)
*备注:
Abstract:Long-horizon complex tasks require foundation models to accumulate information, maintain internal states, and adapt over extended interactions. Safety should be an intrinsic property of the model itself, rather than a behavioral constraint relying solely on external safeguards or post-hoc alignment such as supervised fine-tuning. This motivates Safety from Within, where safety-relevant capabilities are represented and invoked through the model’s native computation. We present Safin-1, a family of foundation models realizing this principle through memory routing and state evolution. Safin-1 is built on Memory-Anchor Routing across Context History (MARCH), a network architecture that maintains structured memory states and selectively retrieves relevant historical information through content-conditioned routing. It supports test-time adaptation of persistent capability states without repeatedly modifying the backbone, enabling controlled specialization over a shared foundation. We investigate this interface on downstream safety tasks through a Safety State, demonstrating effective state-based adaptation with substantial safety improvements. More broadly, the routed-state interface unifies contextual memory and persistent capability adaptation within the model’s native computation, reframing memory from a passive record of prior context into an active substrate for maintaining and evolving model behavior. Evaluations across general capabilities, long-context understanding, retrieval, and efficiency further validate Safin-1. These findings provide a path toward safety as a state-native and adaptively maintainable capability. This work is only an initial architectural exploration of Safety from Within, and substantial further work is needed to realize this broader vision.
[LG-72] Foundation models for electricity price forecasting and battery arbitrag e: Can they replace market-specific forecasting models?
链接: https://arxiv.org/abs/2609.00089
作者: Arkadiusz Lipiecki,Rafał Weron
类目: Machine Learning (cs.LG); Econometrics (econ.EM)
*备注:
Abstract:Foundation models promise accurate forecasts with little or no task-specific training, but whether they can replace models designed specifically for electricity price forecasting remains unclear. We compare nine variants from five foundation model families, evaluated in zero-shot mode, with two state-of-the-art electricity price forecasting benchmarks in Germany, Poland, and Spain over 2021-2025. Their performance is assessed in terms of point and probabilistic forecasting accuracy, as well as economic value in battery energy storage arbitrage. Only the TabPFN models consistently and significantly outperform the benchmarks across all three markets and all statistical measures. However, this statistical dominance does not translate directly into economic dominance: TabPFN performs best under unlimited bids and riskier quantile-based strategies, whereas the Distributional Deep Neural Network benchmark is more profitable when risk tolerance is lower. Thus, foundation models cannot universally replace market-specific models, and their value depends on both model architecture and the decision problem.
[LG-73] Stochastic complexity of vectors containing cluster structure
链接: https://arxiv.org/abs/2609.00084
作者: Daniel Nicorici,Olli Yli-Harja,Jaakko Astola
类目: Machine Learning (cs.LG); Information Theory (cs.IT); Machine Learning (stat.ML)
*备注: 8 pages, 2 figures. Originally published in the Proceedings of the International Workshop on Nonlinear Signal and Image Processing (NSIP 2007), Bucharest, Romania, 10-12 September 2007, pp. 164-169
Abstract:This paper studies the problem of computing the stochastic probability (shortest code length) of the encoded vectors containing cluster structure using Normalized Maximum Likelihood (NML) model. This is of great theoretical and practical importance in data clustering based on Minimum Description Length (MDL) principle, such as for estimating the best number of clusters and best cluster structure for the data. Straightforward computation of the shortest code length of the vector containing cluster structure based on the NML model requires polynomial time with respect to the size of the vector and number of clusters. We show that this is a tractable problem by introducing a recursion formula for the efficient computation of normalizing constant from the NML model. The time complexity of the new formula is linear opposed to previous polynomial time with respect to the size of the vector and number of clusters.
[LG-74] Convergence issues in Relational Concept Analysis based on AOC-posets
链接: https://arxiv.org/abs/2609.00054
作者: Xavier Dolques,Agnès Braud,Alain Gutierrez,Marianne Huchard,Florence Le Ber
类目: Machine Learning (cs.LG)
*备注:
Abstract:Formal Concept Analysis (FCA) is an approach for conceptual classification building and rule discovery from a binary table describing a set of objects by a set of attributes. Extensions have been proposed to deal with non-binary and more complex data, such as Relational Concept Analysis (RCA) for multi-relational data. RCA aims to highlight groups of objects characterized by their relationships with other groups of objects. The richer and more complex nature of the underlying data allows RCA to produce richer results than FCA, at the expense of higher computational and interpretive complexity. The most commonly used conceptual classification structure in FCA is the concept lattice. However, in many applications, concept lattice substructures, such as AOC-posets, are preferred over the full lattice, either to mitigate combinatorial blow-up or to focus on the most informative parts of the structure. Indeed, in an AOC-poset, only concepts introducing an object or an attribute are represented, which makes AOC-posets smaller and easier to compute and use than concept lattices. Although RCA was originally defined on concept lattices, it can also be instantiated on AOC-posets. RCA is iterative and its convergence is guaranteed in the lattice-based setting, but this guarantee is lost when using AOC-posets. In this paper, we investigate this loss of convergence in detail. We show why convergence is no longer guaranteed in the general case, identify conditions under which it can still be ensured, and discuss how a dataset can be transformed to recover convergence. We also propose a convergent variant of the process, which preserves the AOC-poset structure: relational attributes, once created, are never removed, which guarantees convergence at the price of attributes that may refer to concepts absent from the final structures.
[LG-75] Dense Weak Hiding: Closing Complexity Gaps in Nonconvex and PL Finite-Sum Optimization under Individual Smoothness
链接: https://arxiv.org/abs/2609.00045
作者: Yuxing Peng,Zhiqing Tang,Weijia Jia
类目: Data Structures and Algorithms (cs.DS); Machine Learning (cs.LG); Optimization and Control (math.OC)
*备注: 55 pages, 7 figures
Abstract:Under individual smoothness, the optimal incremental first-order oracle (IFO) complexity of nonconvex finite-sum optimization has remained open. Known algorithms use O(n+\sqrtn,\Delta L_\max/\varepsilon^2) calls, while prior lower bounds miss a factor of \sqrtn . We prove the matching lower bound for randomized IFO algorithms whose component indices and query points may depend on the complete preceding transcript and private randomness. This determines the minimax IFO complexity up to universal constants under both individual and mean-squared smoothness. Under the global Polyak-Lojasiewicz (PL) condition, the standard PAGE guarantee is not tight when \kappa_\mathrmms\sqrtn . Restarted PAGE attains O(n+n\log(\Delta/\varepsilon)/(1+\log(\sqrtn/\kappa_\mathrmms))) for 1\leq\kappa_\mathrmms\leq\sqrtn , and O(n+\kappa_\mathrmms\sqrtn\log(\Delta/\varepsilon)) for \kappa_\mathrmms\geq\sqrtn . We prove matching lower bounds under individual smoothness for every \kappa_\max\geq 3 ; the same hard instances also give the mean-squared lower bounds. In the small- \kappa_\max range, their average objective is globally strongly convex. Our lower bounds use dense weak hiding. A fixed sign table spreads each hidden direction across the components. Each queried row carries little information, while the exact row average preserves the full signal after rescaling. A bounded radial map handles arbitrary query points, and a smooth gate makes unopened links invisible to both function values and gradients. Balancing the rows needed to reveal one stage with the number of stages allowed by individual smoothness yields the missing \sqrtn factor. Comments: 55 pages, 7 figures Subjects: Data Structures and Algorithms (cs.DS); Machine Learning (cs.LG); Optimization and Control (math.OC) Cite as: arXiv:2609.00045 [cs.DS] (or arXiv:2609.00045v1 [cs.DS] for this version) https://doi.org/10.48550/arXiv.2609.00045 Focus to learn more arXiv-issued DOI via DataCite
[LG-76] ES-AHD: An Evolution Strategy Framework for Automatic Heuristic Design
链接: https://arxiv.org/abs/2609.00023
作者: Yutao Lai,Kezhao Lai,Hai-Lin Liu,Yuping Wang,Ping Guo
类目: Neural and Evolutionary Computing (cs.NE); Machine Learning (cs.LG)
*备注: Accepted by ICIST 2026
Abstract:In this paper, we introduce ES-AHD, a novel framework that fundamentally integrates Evolution Strategy (ES) into Large Language Model (LLM)-driven Automatic Heuristic Design (AHD). Existing evolutionary approaches predominantly rely on random, individual-level mutation, leading to blind search and an imbalance between exploration and exploitation. To address these issues, ES-AHD introduces two core mechanisms. First, Semantic Recombination via LLMs discards traditional point-to-point reproduction. By leveraging the LLM’s contextual reasoning to explicitly extract core insights from top-performing individuals, the algorithm establishes a promising semantic search direction. This transforms random code mutation into targeted, center-guided sampling inspired by ES. Second, Stochastic Covariance Adaptation via Temperature Sampling dynamically addresses the exploration-exploitation dilemma. By mapping the covariance matrix in ES to the LLM’s sampling temperature, the framework employs a stochastic random walk mechanism with momentum. This approach primarily shrinks the search radius for micro-level code refinement, while retaining the critical ability to occasionally sample higher temperatures to escape semantic local optima. Ultimately, ES-AHD provides a highly directional, robust, and efficient search paradigm, significantly accelerating the generation of high-quality heuristic algorithms. The source code is available at: this https URL.
[LG-77] Structural Bias Beyond Homophily: A Study of Fairness in Link Prediction
链接: https://arxiv.org/abs/2602.11802
作者: Lilian Marey,Mathilde Perez,Tiphaine Viard,Charlotte Laclau
类目: Machine Learning (cs.LG); Social and Information Networks (cs.SI)
*备注:
Abstract:Graph link prediction (LP) plays a critical role in socially impactful applications such as job recommendation and friendship formation, making fairness a critical concern in this task. While many fairness-aware methods manipulate graph structures to mitigate prediction disparities, the topological biases inherent to social graphs remain poorly understood and are consistently conflated with homophily alone. In this work, we study the relationship between structural biases and fairness outcomes in LP. To this end, we formalize a taxonomy of topological bias measures and introduce a graph generation method producing a diverse corpus of synthetic graphs with controlled structural properties. Using this corpus, we show empirically that fairness outcomes are strongly correlated with graph topology, and that current fairness-aware methods remain sensitive to structural biases beyond homophily. These findings highlight the need for structurally grounded evaluations in fair graph learning.
[LG-78] Variable Selection for Feature-Based Newsvendor
链接: https://arxiv.org/abs/2609.01544
作者: Zhaoliang Yuan,Jie Wang
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:Feature-based newsvendor models use observable covariates to tailor inventory decisions, aiming to balance holding and shortage costs under demand uncertainty. However, high-dimensional feature sets often hinder interpretability and inflate data collection and implementation costs. This paper studies variable selection for the feature-based newsvendor problem under a hard cardinality constraint on the number of selected features. We formulate the resulting \ell_0 -constrained empirical newsvendor problem with \ell_2 -regularization, establish its computational hardness, and develop a mixed-integer second-order cone programming reformulation that strengthens the standard Big- M formulation. To enable scalability beyond exact optimization, we develop a randomized-rounding algorithm with a bi-criteria guarantee and a greedy heuristic. Statistically, we provide theoretical analysis of the resulting sparse policy estimator, including finite-sample estimation error, out-of-sample risk bounds, and support recovery guarantees. Extensive experiments on both synthetic and real data illustrate the computational and statistical trade-offs among various baselines. Our results demonstrate that the proposed variable selection framework achieves competitive out-of-sample operational costs while using substantially fewer covariates.
[LG-79] On the Reliability of Generative Augmentation: A Wasserstein-Based Theoretical and Empirical Study
链接: https://arxiv.org/abs/2609.01410
作者: Chathurika S Abeykoon,Mathias Nthiani Muia,Mallory Goldstein
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注: 19 pages, 3 figures, appendix file
Abstract:Generative data augmentation is widely used to mitigate class imbalance, yet its theoretical effect on downstream generalization remains poorly understood. In this work, we develop a statistical framework for conditional generative augmentation and analyze its impact on classification risk. We formalize augmentation as a distribution-mixing process and show that the resulting risk distortion is controlled by both the augmentation strength and the class-conditional Wasserstein discrepancy between real and generated distributions. We further derive a capacity-dependent generalization bound based on Rademacher complexity, revealing an explicit trade-off between hypothesis complexity, augmentation intensity, and generative fidelity. Empirically, we evaluate the framework on binary and multiclass imbalanced classification tasks using Conditional GAN and Conditional WGAN-GP augmentation. Across datasets, CWGAN-GP consistently achieves lower Wasserstein discrepancies than CGAN, indicating improved distributional fidelity. However, improved fidelity does not necessarily translate into superior classification performance, with classical oversampling methods often remaining competitive. These findings support the central theoretical prediction that augmentation reliability is governed by distributional approximation error rather than predictive performance alone. Overall, this work establishes generative augmentation as a distributional perturbation process whose reliability can be quantified through Wasserstein-based measures and supported by finite-sample generalization guarantees. The proposed framework provides a principled foundation for evaluating synthetic data quality beyond classification accuracy alone.
[LG-80] Matched Queries for Curvature and Density at Branching Junctions
链接: https://arxiv.org/abs/2609.01319
作者: Ziqi Zhao,Qingjian Ni
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:At a junction, a score field can reveal weighted tangent rays, yet these first-order quantities do not determine how individual branches bend or how their densities change away from the center. Recovering this missing information is necessary for describing local continuation beyond a single point, but finite observations must separate branchwise second-order effects while allowing error in the estimated center. We address this inverse problem using matched score queries at noise scales \sigma and \lambda\sigma . For a finite union of C^2,\alpha half-branches in \mathbbR^D , the normalized score has the expansion F_\sigma=F_0+\sigma G+O(\sigma^1+\alpha) . Matched subtraction cancels the tangent contribution and exposes G , which depends linearly on branchwise curvature and log-density slope. Given tangent directions and weights on distinct rays, G uniquely identifies all sD branch parameters, and sD scalar component observations are necessary. An O(\sigma^2) center error introduces D translation modes, leading to (s+1)D observations under full-rank calibration, except for a translation-invariant full line. We also establish a perturbation bound and a conditional kernel-density-estimation rate. Experiments reproduce the predicted population and N^-1/5 trends and remain full rank up to D=20 with 16 supplied branches. In end-to-end tests for D=3 – 5 , a known-count first-order frontend yields full rank in all 135 population systems and a median relative jet error of 0.132. With strong first-order error, matched responses reduce median parameter error by a factor of 49.4 relative to naive tangent subtraction.
[LG-81] Direct Optimization of a 3D Finite-Source Reflector via Neural-Network Parameterization
链接: https://arxiv.org/abs/2609.00899
作者: Roel Hacking,Lisa Kusch,Martijn Anthonissen,Wilbert IJzerman
类目: Optics (physics.optics); Machine Learning (cs.LG)
*备注: 16 pages, 4 figures, 2 tables. Submitted to Journal of the Optical Society of America A
Abstract:We present a direct optimization method for three-dimensional freeform reflectors that transform the light of a finite-étendue source into a prescribed far-field angular intensity distribution. The reflector profile is represented by a small neural network (a multilayer perceptron), which is trained end-to-end through a differentiable ray-tracing objective. We furthermore parameterize the emission directions in gnomonic coordinates, and show how we use this to ensure that every emitted ray intersects the reflector. At each iteration, the network is converted to a bicubic spline representation for ray-tracing efficiency, and intersections with this smooth surface are solved by a damped Newton solve, with gradients computed via the implicit function theorem. The traced output distribution is compared with the desired target on a ‘soft’ histogram, under an H^-1 -type spectral weighting that emphasizes long-range transport of flux to improve convergence. Optimization is performed using a BFGS method with self-scaled Broyden updates and a plateau-perturbation rule to prevent stalling. The method converges reliably within seconds on a single GPU for all examples tested.
[LG-82] Sharp Mixed Spectral Barron Regularity of Coulombic Many-Electron Wave Functions
链接: https://arxiv.org/abs/2609.00872
作者: Pingbing Ming,Hao Yu
类目: Analysis of PDEs (math.AP); Machine Learning (cs.LG); Numerical Analysis (math.NA)
*备注: 22 pages; no figures
Abstract:We establish sharp mixed spectral Barron regularity for eigenfunctions of molecular Coulomb Hamiltonians. The mixed norm is a Fourier L^1 norm with one isotropic weight and coordinate-product weights, and therefore detects regularity invisible to the isotropic Barron scale. For a nonempty set I of electron indices on which the wave function is antisymmetric, we derive an explicit admissible region for the isotropic order s and the coordinate orders \alpha,\beta . This region is optimal as a uniform statement over the class of clamped-nuclei Coulomb Hamiltonians. For fixed-spin components with two occupied spin blocks, it reduces to s+\alpha+\beta1 ; in the fully spin-polarized class it reduces to s+\alpha1 . In particular, if \mathcal I_\sigma denotes the family of occupied same-spin blocks determined by \sigma , then every fixed-spin spatial component \psi_\sigma satisfies, for every 0\leq\alpha1 , [ \left(\sum_I\in\mathcal I_\sigma\prod_i\in I\langle\xi_i\rangle^\alpha\right)\widehat\psi_\sigma\in L^1(\mathbbR^3N). ] For a fully spin-polarized state, \mathcal I_\sigma=\1,\ldots,N\ .
[LG-83] Semi-Supervised Classification with Informative Missing Labels in Weibull Mixture Models
链接: https://arxiv.org/abs/2609.00774
作者: Jinran Wu,You-Gan Wang,Geoffrey J. McLachlan
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:We consider semi-supervised classification from a partially classified sample arising from a two-component Weibull mixture. The feature is observed for all data, whereas some class labels are missing. The probability of a missing label is modelled as a function of classification uncertainty, giving a feature-dependent missing-at-random (MAR) mechanism that shares parameters with the Weibull-mixture classifier. The missing-label indicators can therefore provide information about the classifier in addition to the observed features and available class labels. Under a common Weibull shape, a Bayes’ rule has at most one positive decision boundary, which is unique when the rule is nonconstant; under unequal shapes, it can have two. We characterise these decision regions, derive the Fisher information for the classifier after adjustment for nuisance parameters in the missingness model, and obtain a decision-boundary expansion of the expected error rate of the plug-in sample rule relative to the Bayes error. The expansion yields classification-specific asymptotic relative efficiency formulas for the one- and two-boundary cases and shows that a positive-definite increase in Fisher information is sufficient, but not necessary, for a smaller first-order expected error rate. Numerical studies and a semi-synthetic analysis based on hard-drive failure data illustrate potential reductions in expected error rate and improvements in decision-boundary estimation from modelling feature-dependent label missingness.
[LG-84] Disciplined Bilevel Programming
链接: https://arxiv.org/abs/2609.00644
作者: Hao Zhu,Joschka Boedecker
类目: Optimization and Control (math.OC); Computational Engineering, Finance, and Science (cs.CE); Machine Learning (cs.LG); Mathematical Software (cs.MS)
*备注:
Abstract:Bilevel optimization provides a natural modeling language for hierarchical decision problems. However, applying existing numerical solvers usually requires substantial manual analysis and reformulation. In this paper, we introduce disciplined bilevel programming (DBLP), a symbolic framework that allows users to specify and solve optimistic bilevel problems in a high-level, human-readable way that is close to the mathematical formulation. For problems with a disciplined nonlinear upper problem and a convex lower problem satisfying the disciplined parameterized programming rules, DBLP automatically canonicalizes the lower problem into conic form and constructs an equivalent single-level reformulation using the conic Karush-Kuhn-Tucker conditions. We relax the resulting complementarity constraint and use a gap continuation procedure to approximately solve a sequence of smooth nonlinear problems. We implement DBLP in the open-source Python package BLVPY, an extension of CVXPY for bilevel programming. We demonstrate the modeling and solution capabilities of BLVPY on a range of bilevel optimization problems from several application domains. The proposed framework and implementation allow users to specify and solve bilevel optimization problems within a few lines of code, without prior expertise in bilevel modeling and numerical optimization.
[LG-85] BeamRMX: Radiation-Pattern-Driven Learning for Generalizable Beam Radio Map Prediction and Beam Management
链接: https://arxiv.org/abs/2609.00615
作者: Yue Zhang,Xiucheng Wang,Wenshuo Chen,Nan Cheng
类目: ignal Processing (eess.SP); Machine Learning (cs.LG)
*备注: 13 pages
Abstract:The evolution toward sixth-generation (6G) wireless networks is driving larger antenna arrays and highly directional multi-beam transmission, making accurate knowledge of beam-dependent spatial coverage important for beam management and environment-aware network operation. Radio maps (RMs) provide such a representation, yet conventional RM prediction assumes omnidirectional or transmitter-level radiation. In beamformed multiple-input multiple-output (MIMO) systems, one propagation scene instead gives rise to many configuration-dependent beam radio maps (BeamRMs), creating challenges in beam representation and generalization. Existing methods either condition prediction on beam descriptors or use beam maps as auxiliary inputs to generic architectures. We propose BeamRMX, which, to the best of our knowledge, is the first dedicated framework to treat the spatial radiation pattern as the primary BeamRM query and learn how scene geometry transforms it into the received power field. XBase learns multiscale interactions between the radiation query and scene geometry, while an optional Evidence Adapter uses a few cross-configuration BeamRMs from the same scene. Matched-domain and zero-shot experiments show consistent gains over deterministic and diffusion baselines, including mean absolute error reductions of 26.1% on unseen scenes and 47.8% on an unseen configuration. Cross-configuration evidence further improves reconstruction and intra-sector beam refinement.
[LG-86] Real-Time Neuromorphic Spectrum Intelligence Simulator NEURIPS2025 ALT
链接: https://arxiv.org/abs/2609.00585
作者: Navaneetha Krishnan Kamalakannan
类目: ignal Processing (eess.SP); Machine Learning (cs.LG); Networking and Internet Architecture (cs.NI)
*备注: 7 pages, 4 figures, 2 tables. Accepted at the NeurIPS 2025 Workshop on Machine Learning and the Physical Sciences (ML4PS). Code and benchmark artifacts: this https URL
Abstract:We present the Real-Time Neuromorphic Spectrum Intelligence Simulator (RT-NuSIS), a modular framework to study spiking neural network (SNN) and memristor-inspired agents for dynamic spectrum access under constrained energy budgets and adversarial conditions. RT-NuSIS couples leaky integrate-and-fire neuronal dynamics, memristive synaptic models, physics-informed energy-harvesting models (triboelectric and RF), and adversary models including jamming and Byzantine behavior. We formalize the simulator mathematically, prove boundedness, present a mean-field adversary threshold, analyze per-step complexity, and provide a reproducible benchmark harness for energy-per-inference, latency, and robustness metrics. The codebase is modular, deterministic by seed, and designed for large-scale event-driven simulations.
[LG-87] Fractal dimension predicts quantum kernel collapse in angle-encoded data
链接: https://arxiv.org/abs/2609.00475
作者: Ana Paula Appel
类目: Quantum Physics (quant-ph); Machine Learning (cs.LG)
*备注:
Abstract:Angle-encoded quantum kernels on tabular data collapse when the feature map is wider than the intrinsic dimension of the data. We propose the correlation fractal dimension D2 as an a priori qubit budget: encode D2 coordinates chosen by FD-ASE instead of the PCA-95% width or all E attributes. On nine data sets and a statevector simulator (n= 32), a one-layer ZZ fidelity kernel at q=D2 stays geometrically alive while the same kernel at the PCA-95% width has already collapsed. The budget is map-dependent: product-state and IQP maps overshoot it; a second ZZ layer undershoots it. Packed dense-angle and re-uploading encodings still live at the fractal q, but not when PCA-95% features are stacked onto those qubits. Shrinking the angle bandwidth moves the ZZ knee later; stretching it kills the kernel earlier. On IBM Quantum (ibm_fez, 256 shots, n=8) the one-layer ZZ kernel at the fractal width matches the exact kernel (MAE 0.021); past that width both hardware and simulator have collapsed. The ceiling is a property of the map-data pair at a stated bandwidth, not of the classical table alone.
[LG-88] Physiological Information Reliability: Cross-Layer Adaptive Resource Allocation for Cardiovascular Sensing ALT ML4H
链接: https://arxiv.org/abs/2609.00435
作者: Navaneeth Krishnan Kamalakannan,Janakiraman Kamalakannan,Harinisri Velmurugan
类目: ignal Processing (eess.SP); Machine Learning (cs.LG); Networking and Internet Architecture (cs.NI)
*备注: 6 pages, 4 figures, 3 tables. Submitted to the Machine Learning for Health (ML4H) 2026 Symposium, Findings Track. Code and experimental artifacts: this https URL
Abstract:Cardiovascular sensing systems must preserve clinically useful information despite signal degradation, wireless losses, energy constraints, and edge-computation latency. We introduce Physiological Information Reliability (PIR), a cross-layer framework that represents physiological information value jointly with wireless, energy, and computation states and uses a contextual bandit to adapt sensing and communication decisions. We integrate multimodal ECG/PPG signal-quality estimation with physiological information value and an adaptive network-coding layer under burst-erasure conditions. Across controlled multiseed experiments, PIR-LinUCB demonstrates a promising low-energy operating point while maintaining medical latency constraints and competitive physiological estimation performance relative to fixed and heuristic policies. We analyze the resulting accuracy-energy-latency trade-offs and identify limitations of proxy PIV estimation and simulated communication dynamics. These results provide an initial computational demonstration of physiological-information-aware resource allocation and motivate future clinical and real-channel validation.
[LG-89] Accelerating Chemical Kinetics for Exoplanet Atmospheres using Neural Networks
链接: https://arxiv.org/abs/2609.00428
作者: Isaac Malsky,Xi Zhang,Tiffany Kataria,Matthew Graham,Ziyu Huang,Boris Bonev,Shang-Min Tsai,Elspeth K.H. Lee
类目: Earth and Planetary Astrophysics (astro-ph.EP); Machine Learning (cs.LG)
*备注: 15 pages, 7 figures. Accepted for publication in ApJ
Abstract:Observations increasingly reveal the coupled radiative, chemical, and dynamical processes that shape exoplanet atmospheres. Interpreting these atmospheres requires models that can capture this complexity. However, multidimensional models remain fundamentally limited by computational cost, and answering key questions requires simulating the governing physical mechanisms at speeds classical methods cannot achieve. As a result, models often rely on simplifying approximations, such as equilibrium chemistry, even when those assumptions miss important effects. There is a pressing need for fast and accurate chemical kinetics solvers to model planetary atmospheres. Here we present a machine learning local-box chemical kinetics solver for exoplanet atmospheres using a residual flow-map architecture. We demonstrate that this surrogate model is several orders of magnitude faster than a classical solver, achieving microsecond-scale inference while retaining percent-level accuracy. The surrogate model covers a parameter space that spans T=300 - 3000 K, P=10^-6 - 10^4 bar, \Delta t=10^-3 - 10^8 s, and compositions ranging from 10^-2 to 10^3 times solar in both C/O ratio and metallicity. Our model outperforms several commonly used machine learning architectures and performs robustly under the extreme stiffness characteristic of atmospheric chemistry. The machine learning framework presented here is a flexible and efficient approach to emulating state-to-state flow-map problems that commonly arise in numerical simulations.
[LG-90] A convolutional framework for detecting event-driven dynamics in energy price series
链接: https://arxiv.org/abs/2609.00402
作者: Caixia Xu,Piotr Fryzlewicz
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Methodology (stat.ME)
*备注:
Abstract:This paper develops a general convolutional neural network (CNN) framework for detecting heterogeneous event-driven dynamics in univariate time series windows. We show that the induced CNN class exactly represents classifiers based on range, maximum drawup, maximum drawdown and slope change, and uniformly approximates realised volatility and autoregressive explosiveness on compact domains. We further establish error bounds for representative rules in finite samples and an oracle inequality for learning across them. Simulations show that the proposed model can match or outperform classifiers based on individual statistics as the training sample grows. In an application to six daily energy price series, a hierarchical CNN distinguishes event windows and event families. Applied without retraining to observations withheld after 20 February 2026, the fitted model identifies predominantly geopolitical dynamics in several oil and refined product series around the outbreak of the 2026 Iran war, while distinguishing a contemporaneous natural gas spike associated with weather.
[LG-91] owards unsupervised representation learning for quantum data: quantum models with inference and generation
链接: https://arxiv.org/abs/2609.00372
作者: Robin Lorenz,Eric Brunner,Marcello Benedetti
类目: Quantum Physics (quant-ph); Machine Learning (cs.LG)
*备注: 48 pages, comments welcome
Abstract:With quantum sensors, simulators and networks emerging, a future of quantum technology may produce quantum states as data—that is, coherently rather than as classical measurement records—thus motivating the study of suitable quantum generalisations of modern machine learning, including the automated, unsupervised extraction of useful representations. Two ingredients are central to the latter: inference, mapping observations to latent representations, and generation, mapping latent states back to synthetic data. Both are related to each other and to joint distributions for training models by the chain-rule of classical probability theory. The fact that quantum states however lack such universal, standard factorisation property thus poses a challenge. Here we develop a conceptual and mathematical framework for unsupervised representation learning from quantum data. Models are joint quantum states over visible and latent systems; state-over-time maps provide a notion of factorisation into a marginal state and inference (generation) channel; models with inference (generation) are ambiguous states—states for which such factorisation obtains—subject to a further consistency condition on extended inference maps as data extension. These stipulations are restrictive: we show that non-trivial models must feature non-linear such maps to the extended space. For three representative state-over-time maps, we completely characterise the ambiguous states, uncovering a hierarchy tied to the positive-partial-transpose (PPT) criterion from entanglement theory. Notably, the Leifer-Spekkens construction supports inference and generation exactly for model classes of PPT states, thus allowing genuinely quantum visible-latent correlations. We also formulate quantum counterparts of exact and approximate inference training, explore weaker notions of data extension and sketch a future research programme.
[LG-92] Exact Global MCMC with Denoising Diffusion
链接: https://arxiv.org/abs/2609.00279
作者: Mitch Hill
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:This work shows that diffusion models learned with standard denoising loss can provide effective global MCMC proposals for complex high-dimensional target densities. The method is motivated by the observation that sequentially applying a forward and reverse diffusion process defines a Markov chain with a target stationary distribution for an ideal denoiser trained on samples of the target distribution. This observation can be made exact for any denoiser by applying a Metropolis-Hastings step whose acceptance ratio includes the density of the forward and reverse paths of a discrete time SDE approximation. We therefore propose to train denoising diffusion models on locally convergent MALA samples to learn global MCMC proposals. We call the composition of the global denoiser-based path sampler and a local MALA sampler Denoising Diffusion Monte Carlo (DDMC). Experiments show that DDMC can provide global proposals with high acceptance across a variety of complex target densities. Our results offer preliminary evidence that the established scaling behavior of standard diffusion training transfers directly to exact sampling from high-dimensional unnormalized densities.
附件下载
