本篇博文主要内容为 2026-09-04 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 (2026-09-04)
今日共更新655篇论文,其中:
- 自然语言处理共119篇(Computation and Language (cs.CL))
- 人工智能共165篇(Artificial Intelligence (cs.AI))
- 计算机视觉共112篇(Computer Vision and Pattern Recognition (cs.CV))
- 机器学习共168篇(Machine Learning (cs.LG))
- 多智能体系统共15篇(Multiagent Systems (cs.MA))
- 信息检索共25篇(Information Retrieval (cs.IR))
- 人机交互共18篇(Human-Computer Interaction (cs.HC))
多智能体系统
[MA-0] Robust PAC Learning of Concurrent Stochastic Games
【速读】：该论文旨在解决具有转移不确定性的一般和性并发随机博弈（Concurrent Stochastic Games, CSGs）中的纳什均衡（Nash Equilibrium, NE）存在性问题与学习效率之间的核心矛盾。传统方法难以在不完全可观测转移机制下保证均衡解的存在性或提供可证明的近似性能，而本研究首次构建了适用于此类复杂博弈场景的可能近似正确（PAC）学习框架。其解决方案的关键在于：通过数据驱动的方式构建对转移核的 L1 置信集，并基于鲁棒马尔可夫决策过程（Robust MDP）设计探索机制，以实现联合状态-动作空间的充分覆盖；同时引入纳什裕度（Nash margin）这一新概念，使算法能够进行合理且可证的均衡存在性推理——即要么输出一个社会福利值接近最优的 ε-近似纳什均衡，要么给出一个无精确纳什均衡存在的可靠证明。在满足最小可达性条件 preach>0 的前提下，该算法可在多项式数量的轨迹样本内终止，其样本复杂度为 O(Rmax2H4∣S∣2∣A∣/(preachε2))，并已在基准测试中验证了近最优性能、对均衡（非）存在性的准确判断以及与理论预测一致的样本效率。
链接: https://arxiv.org/abs/2609.04189
作者: Angel Y. He,David Parker
机构: University of Oxford (牛津大学)
类目: Machine Learning (cs.LG); Computer Science and Game Theory (cs.GT); Logic in Computer Science (cs.LO); Multiagent Systems (cs.MA)
备注: Main text: 10 pages, 1 figure, 2 tables; Appendix: 22 pages, 2 figures, 1 table
Abstract:We introduce the first Probably Approximately Correct (PAC) learning framework for general-sum concurrent stochastic games (CSGs) with transition uncertainty, while addressing the challenge of Nash equilibrium (NE) existence. Our algorithm maintains data-driven L^1 confidence sets over transition kernels and solves a robust CSG to compute a social-welfare optimal \varepsilon -NE, using a robust MDP-based exploration mechanism to drive joint state-action coverage. Crucially, we introduce a Nash margin characterisation that enables principled reasoning about equilibrium existence: the framework either returns an \varepsilon -approximate NE whose social-welfare value is \varepsilon -close to optimal, or provides a sound certificate that no exact NE exists. Under a minimum reachability condition p_\mathrmreach0 over relevant state-action pairs, the algorithm terminates after a polynomial number of trajectory samples, with sample complexity \widetildeO\left( R_\max^2 H^4 |S|^2 |A| / (p_\mathrmreach \varepsilon^2) \right) . Empirical results on benchmark CSGs demonstrate near-optimal performance, correct handling of equilibrium (non-)existence, and sample complexity consistent with theory.
[MA-1] Bioinfoysis Technical Report
【速读】：该论文旨在解决生成式人工智能在生物信息学（bioinformatics）领域中长期任务执行时存在的证据断裂与可追溯性缺失问题。现有系统多将规划、工具调用和代码执行视为临时交互，缺乏对中间结果的持久化管理，导致分析过程中的推理链断裂，难以保证结论与数据、计算及中间证据之间的可靠关联。其核心解决方案是提出一个名为Bioinfoysis的多智能体框架，通过将每个请求建模为具有持久性与实体锚定（artifact-grounded）的分析运行，实现端到端的可追溯自动化。关键创新在于结合全局规划与逐步、基于证据的重规划机制：规划器维护一个可执行清单，并通过每次工作代理执行后返回的结构化交接（handoff）动态更新待办步骤；这些交接明确绑定中间结果与其负责的代理、清单步骤及计划生成版本，有效防止重规划后过时证据的隐式复用。此外，受控运行时环境对生成的脚本、表格和图表进行验证，确保下游分析与报告的可靠性；角色特异性上下文、持久记忆以及受控的生物信息学技能模块共同保障了长周期分析轨迹中的稳定执行。实验表明，Bioinfoysis在BixBench上达到82.4%的准确率，在SeqQA2和DbQA2两个子任务上分别将平均准确率从27.81%/3.13%提升至64.13%/31.25%，证明了可靠的生物信息学自动化不仅依赖模型能力，更取决于对规划、执行、记忆与证据流的有效协同治理。
链接: https://arxiv.org/abs/2609.03871
作者: Qingyang Shao,Xin Zhang,Zhouyang Yuan,Xianying Chen,Yujia Xiang,Zihao Yang,Tong Ye,Yangqi Zhang,Jiakang Xu,Xiaoqing Yan,Xuan Luo,Keyi Li,Enci Fan,Kai Kang,Zhuohan Liu,Xingyu Jin,Chunran Teng,Tao Li,Xinyu Lv,Minghui Wang,Wenfeng Li,Yidan Gao,Siyu Liu,Mingrui Luo,Zhu Liang,Guanren Qiao,Zhiping Xu
机构: 未知
类目: Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注:
Abstract:Large language model agents have shown promise in bioinformatics, but most existing systems focus primarily on producing final answers, treating planning, tool use, and code execution as transient interactions. This design is poorly suited to long-horizon bioinformatics tasks, where conclusions must remain connected to the data, computations, and intermediate evidence that support them. We introduce \textbfBioinfoysis, a multi-agent harness that represents each request as a persistent, artifact-grounded analysis run. Bioinfoysis combines global planning with step-wise, evidence-driven replanning: the planner maintains an executable checklist and revises pending steps using structured handoffs returned after each worker execution. These handoffs bind intermediate results to their responsible agent, checklist step, and plan generation, preventing stale evidence from being silently reused after replanning. A controlled runtime validates generated scripts, tables, and figures before they are used in downstream analysis or reporting, while role-specific context, persistent memory, and governed bioinformatics skills support reliable execution over long analysis trajectories. We evaluate Bioinfoysis on BixBench and two question-answering tracks of LAB-Bench 2. On BixBench, Bioinfoysis achieves state-of-the-art accuracy of 82.4%. Across four underlying language models, Bioinfoysis increases average accuracy from 27.81% to 64.13% on SeqQA2 and from 3.13% to 31.25% on DbQA2. These results demonstrate that reliable bioinformatics automation depends not only on model capability, but also on the harness that governs planning, execution, memory, and evidence flow. We hope that the emergence of Bioinfoysis will play a driving and leading role in the development of the bioinformatics community. Our demo website can be seen in this https URL.
[MA-2] SimSkill: A Lifelong Learning AI Agent for Autonomous Mastery of Traffic Simulation
【速读】：该论文旨在解决大语言模型（LLM）在长期应用中难以将经验与知识转化为可复用、持久化能力的核心挑战，即如何实现AI系统从一次性任务求解向持续自我进化和能力积累的转变。其解决方案的关键在于构建一个名为SimSkill的自演化智能体，该智能体基于交通仿真器SUMO（Simulation of Urban MObility），通过识别能力缺口、生成并求解环境相关的任务、利用“动作-批评”循环验证解决方案，并将经验分别以情景记忆（episodic memory）、程序性记忆（procedural memory）和语义记忆（semantic memory）的形式进行固化，从而实现对知识的结构化存储与重用，且无需更新底层大模型。实验表明，该方法在两个独立基准测试中，使经验证的任务完成率最高提升25个百分点，且程序性与语义记忆的协同作用显著；然而其增益效果依赖于底层模型能力和计算预算，提示内存机制并非普适性的优化手段。总体而言，该研究提出了一种“自然语言承载能力表达，可执行工具与代码实现精确执行”的新型设计范式，为构建具备持续学习与知识复用能力的下一代智能系统提供了可行路径。
链接: https://arxiv.org/abs/2609.03753
作者: Qi Liu,Qinzheng Wang,Yiming Bie
机构: Jilin University(吉林大学)
类目: Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注:
Abstract:As large language models (LLMs) become increasingly capable, the long-term value of AI systems depends not only on solving individual requests, but also on transforming experience and accumulated knowledge into durable, reusable competence. We introduce SimSkill, a self-evolving agent built around the Simulation of Urban MObility (SUMO) traffic simulator. SimSkill identifies capability gaps, generates and solves environment-grounded tasks, verifies solutions through an action–critic loop, and consolidates experience into episodic, procedural, and semantic memory without updating the backbone model. Through autonomous exploration, it builds a reusable library spanning the traffic-simulation workflow. We evaluate SimSkill on two held-out benchmarks with three backbone LLMs and independent artifact-based verification. SimSkill improves verified completion by up to 25 percentage points, while ablations show complementary contributions from procedural and semantic memory. Its benefits remain backbone- and budget-dependent: memory does not improve every model or uniformly reduce inference cost. More broadly, SimSkill illustrates a design paradigm in which natural language preserves and composes computational capabilities, while executable tools and code provide precise and reproducible execution. All code and experimental data are publicly available at this https URL.
[MA-3] he Civilization Framework: Sovereign-Anchored Communication Between Personal Multi-Agent Systems
【速读】：该论文旨在解决人工智能系统间通信中存在的“人类作为传输层”所导致的上下文丢失问题，尤其聚焦于多智能体协作中因时序竞争引发的权威性错配风险。其核心挑战在于：在异步、非同步的AI-to-AI交互环境中，先到达的信息可能因时间优势获得不当权威（即“时间权重效应”，temporal-weight effect），从而误导后续决策。为应对这一问题，论文提出“文明框架”（Civilization Framework）与“大使协议”（Embassy Protocol）作为解决方案的关键。该方案以“文明”为可寻址主体，而非单个智能体，通过持久化账本（persistent ledger）实现状态共识，强调跨系统的承诺状态（commitment state）作为真实性的基础，而非消息交付本身；同时，智能体的行动权限由其可访问的记忆能力决定，并通过独立于声誉的签名凭证进行外部化授权。实验验证表明，在未启用验证机制的情况下，率先抵达的错误上游信息可影响54.2%的输出结果，而经完整验证后该比例降至4.2%，证明了验证机制对防止时间优势滥用的重要性。此外，指令级溯源标注可缓解该效应，且密封答案的准确性在不同条件下保持等价，但这些次级发现依赖特定问题集设定。由于工具使用检查未满足预算条件，整体实验被判定为探索性结果，后续将通过强化资源约束的测试框架进行复现。目前，该框架的文明内层已具备可运行实现。
链接: https://arxiv.org/abs/2609.03425
作者: Guangjun Liu
机构: New York University (纽约大学)
类目: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI)
备注: 44 pages, 4 tables. Preregistration: this https URL
Abstract:Humans are the transport layer between AI systems, losing context at every hop. We present the Civilization Framework, whose addressable party is the civilization, not the agent (one human sovereign, a persistent ledger, and interchangeable agents), and the Embassy Protocol, a carrier-agnostic overlay: messages arrive asynchronously at a resident ledger endpoint, any online agent of the receiver handles them, and commitment state on both ledgers, not delivery, is ground truth. Authority derives from memory: an agent’s power to act for its civilization is capped by the memory it can access and externalized through signed credentials, separate from civilization-level reputation. We identify the temporal-weight effect, a hazard in AI-to-AI communication where what arrives first acquires unearned authority, and test it in one frontier model in a preregistered 1,908-trial experiment. With verification removed, an incorrect upstream claim arriving first captures 54.2% of answers (4.2% under full verification), while the same claim arriving after the receiver has sealed its own answer captures 31.6% (the two prompt shells are not length-matched, so part of that gap may reflect shell form; see Section 7), and both registered question-set specifications agree on these two verdicts (the exclusion specification is preregistered as under-powered). Two secondary results, the mitigation from instruction-level provenance labeling and sealed-answer accuracy equivalence, are specification-dependent, holding only under the all-questions specification. Because a registered check of tool use failed its call-budget condition, the registration classifies the round as inconclusive and every result above, primary and secondary, is reported as exploratory; a replication with harness-enforced budgets is planned. The framework’s intra-civilization layer has a working implementation.
[MA-4] Inferred Generative-Process Diversity Predicts Correlated Failure Across Language Models
【速读】：该论文旨在解决多语言模型系统中“多样性”评估的局限性问题，即现有方法（如基于语义相似性的评估）仅关注输出语义层面的差异，而忽略了模型生成过程本身的本质差异，导致对模型间真实多样性的误判。在存在强相关行为与失败模式的情况下，即使模型被视为独立组件，其实际生成机制可能高度相似，从而影响系统的鲁棒性。论文提出，真正关键的多样性应为生成过程多样性（generative-process diversity），即不同模型生成相同输出所依赖的内在生成机制差异。为此，研究基于算法信息论，采用经置换控制校正后的原始模型输出的归一化压缩距离（Normalised Compression Distance, NCD）作为生成过程多样性的推断指标。在38个语言模型的实验中，该指标揭示了语义相似性无法捕捉的群体结构，并能有效预测跨十类独立基准测试中模型对在剔除随机误差后的相关失败变异情况，其偏秩相关系数为-0.216（95%置信区间[-0.309, -0.122]），且在所有基准上均为负值，表明生成过程多样性越高，模型间的非语义相关失败越低。因此，该研究的关键在于通过生成过程多样性这一更根本的视角，提供一种可操作、适用于安全关键场景的多模型系统多样性评估新范式。
链接: https://arxiv.org/abs/2609.03422
作者: Ross Tieman,Evan Markou
机构: Fenner School of Environment and Society (环境与社会学院); Australian National University (澳大利亚国立大学)
类目: Machine Learning (cs.LG); Multiagent Systems (cs.MA)
备注: 31 pages, 13 figures
Abstract:Diversity is a widely observed factor in the resilient function of collective systems, yet the type of diversity that matters depends on the properties and failure modes of the system. This distinction is important for systems composed of multiple language models. Different models may be treated as independent components even when their behaviour and failures remain strongly correlated. Assessments of language-model populations using semantic similarity demonstrate limited semantic diversity, but this captures only differences in the meaning of observed outputs. We argue that a more fundamental notion of model diversity is generative-process diversity, the differences between processes capable of generating the observed outputs. Drawing from Algorithmic Information Theory, we use Normalised Compression Distance between raw model outputs, residualised against a permutation control, as a measure of inferred generative-process diversity. Across 38 language models, this measure identifies population structure missed by semantic similarity and predicts cross-task variation in chance-corrected correlated failure among model pairs across ten disjoint benchmark families, beyond semantic similarity and model-pair capability. The cross-benchmark partial rank association is -0.216 with a 95% interval of [-0.309,-0.122] , and the estimate is negative on all ten benchmarks. These results indicate that increased generative-process diversity is associated with reduced correlated failure in model pairs that is not attributable to semantic similarity or capability. Inferred generative-process diversity offers a novel and practical approach for investigating diversity of multi-model systems in safety-relevant contexts.
[MA-5] Risk and Anomaly Identification for Distribution Network Optimal Operation Based on Reinforcement Learning and Uncertainty Quantification
【速读】：该论文旨在解决现代配电系统在广泛不确定性环境下，如何实现对固有运行风险与异常事件的及时识别问题。其核心挑战在于区分由随机但分布内（in-distribution）条件引起的固有风险，以及由分布外（out-of-distribution）行为（如异常负荷模式、极端天气或网络物理攻击）引发的异常事件。解决方案的关键在于提出一种显式具备不确定性感知能力的深度强化学习（Deep Reinforcement Learning, DRL）框架，通过融合分布式深度强化学习（Distributional DRL）与贝叶斯深度强化学习（Bayesian DRL），构建二阶不确定性量化机制，将总不确定性分解为认知不确定性（epistemic uncertainty）和随机不确定性（aleatoric uncertainty）。其中，认知不确定性用于驱动训练过程中的探索行为，并在部署阶段实现分布外检测及备用控制策略；而随机不确定性则用于刻画系统内在的运行风险。仿真结果验证了所提DRL代理在性能与不确定性量化有效性方面的优越性。
链接: https://arxiv.org/abs/2609.03308
作者: Ziqi Zhang
机构: Nanjing University of Aeronautics and Astronautics (南京航空航天大学)
类目: Machine Learning (cs.LG); Multiagent Systems (cs.MA); Systems and Control (eess.SY)
备注:
Abstract:Reliable operation of modern distribution networks requires timely identification of operational risks and anomalous events under pervasive uncertainty. In practice, operators must identify risks that are inherent in stochastic yet in-distribution conditions, and anomalies that correspond to out-of-distribution behaviors such as unusual load patterns, extreme weather or cyber-physical attacks. This paper addresses this joint risk and anomaly identification problem for optimal distribution network operation and proposes a deep reinforcement learning framework that is explicitly uncertainty aware. We integrate distributional and Bayesian deep reinforcement learning to realize a second- order uncertainty quantification scheme that decomposes total uncertainty into aleatoric and epistemic components, which are respectively used to characterize inherent risk and out-of- distribution anomalies. The resulting epistemic estimates drive both exploration during training and out-of-distribution detec- tion with fallback control during deployment, whereas aleatoric estimates are used to characterize intrinsic operational risk. Simulation results demonstrate the performance of our DRL agent and the effectiveness of the uncertainty quantification.
[MA-6] Speculative Macro Commit for Faster Tool-Using Agents
【速读】：该论文旨在解决工具使用型大语言模型（LLM）智能体在执行任务时因串行的动作-观测循环导致的高延迟问题，即每个工具调用、环境状态转换和观察结果获取都会引入显著的等待时间，从而拖慢整体决策效率。其核心解决方案是提出一种名为“推测性宏提交”（Speculative Macro Commit, SMC）的运行时机制，构建一个双层智能体系统：由高性能权威模型（authoritative actor model）生成最终轨迹，同时配备一个更快的推测草稿模型（speculative drafter model），在隔离的环境快照上持续预测并预执行未来的多步动作链。SMC通过从训练轨迹中挖掘重复出现的多动作结构（multi-action skeletons），将其存入宏库，并在运行时将草稿模型预测的动作链与宏库进行匹配；当权威模型的下一个动作与草稿模型的首个动作一致时，便将已预执行的后续步骤及其观测结果一并提交至正式轨迹。实验表明，以Qwen3.5-27B INT4为权威模型、Qwen3.5-4B为草稿模型，在τ²-Bench Telecom子集上，SMC在保持与串行执行相当准确率的同时，相较推测动作（SA）基线降低10.23%延迟，相较串行执行降低18.59%；在AppWorld基准上，相比SA基线减少7.7%的墙钟时间，相比串行执行减少44.9%，仅伴随轻微的任务完成率下降。SMC实现了对多步推测执行的有效复用，显著超越了传统单步推测动作的加速效果，提供了一种实用且高效的降低智能体延迟的方法。
链接: https://arxiv.org/abs/2609.03236
作者: Zeyu Liu,Souvik Kundu,Peter A. Beerel
机构: 未知
类目: Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注: Accepted in MLSP2026
Abstract:Tool-using LLM agents spend wall-clock time not only on model inference but also in serial action–observation turns, where each tool call, environment transition, and observation can delay subsequent decisions. We introduce \textbfSpeculative Macro Commit (SMC), a runtime mechanism for a two-tier agent system: a large authoritative actor model produces the official trajectory, while a faster speculative drafter model continuously predicts and executes future action chains on an isolated environment snapshot. SMC mines recurring multi-action skeletons from training traces and stores them in a macro library used to match against action chains predicted by the drafter at runtime. When the actor’s next tool call matches the first drafted action, SMC commits the remaining pre-executed draft steps, together with their observations, to the official trajectory. Using Qwen3.5-27B INT4 as the authoritative actor model and Qwen3.5-4B as the speculative drafter model, SMC matches the sequential agent’s overall accuracy while reducing latency by 10.23% over the Speculative Actions (SA) baseline and 18.59% over sequential execution on the \tau^2 -Bench Telecom subset. On AppWorld, SMC reduces wall time by 7.7% over SA baseline and 44.9% over sequential execution, with a small reduction in task completion. Overall, SMC provides a practical way to reuse multi-step speculative execution and reduce agent latency beyond single-step speculative actions. Our code is publicly available \hrefthis https URL\textcolormagentahere.
[MA-7] Where Reliability Lives: Experimental Localisation of Behavioural Properties in an Agent System
【速读】：该论文旨在解决生成式智能体（agentic systems）在可靠性声明中隐含的属性归属问题——即可靠性究竟源于模型本身，还是其外部机制。为验证这一根本性问题，研究构建了一个可实验化的系统：在一个持续演化的模拟定居点中，所有行为均通过权威的追加型账本（append-only ledger）与世界状态进行仲裁，唯一被承认的历史即为真实现实。研究将认知（cognition）、制度（institution）与世界（world）三者预先分离，并在固定认知的前提下，干预制度的表征认知机制（如证据溯源、信念可得性、物理证据可读性），两次预注册实验均反向证伪了核心假设。随后引入一个注册的可证伪性测试，通过提供几何结构原本拒绝的可信第一手证人输入，发现该证人的边际价值在无证人阶段为非正，而在引入后转为正值，且9/11种子实验未产生额外错误归因。进一步在固定制度执行的前提下，从四个维度干预认知：移除原生智能体的内部机制、任务中止并重置其状态、以冻结的前沿大语言模型（LLM）面板替代全部原生认知、以及用可信虚假证词污染信念。结果表明行为发生剧烈变化：信任路径因单一虚假信息导致约900次无效操作，而怀疑路径则无此现象。五个预设属性在所有测试轨迹中保持不变：接受的现实始终单一，无效尝试均被赋予类型化拒绝理由，职责超越其执行过程，无重复工作被接受，且从未接受过虚假完成（共2,581次替换面板请求，无一为假）。研究结论限定于该设计场景：所测行为属性可与认知的实质性改变相分离，且通过干预得以证实。该研究采用单一设计世界，未测试代理对制度的优化行为。
链接: https://arxiv.org/abs/2609.03192
作者: Timothy Marsden,Matthew Collecutt,James Marsden
机构: Taniwha AI
类目: Multiagent Systems (cs.MA)
备注: Preprint. 31 pages, 5 figures. Ancillary files: related-work search appendix, artefact manifest, verification receipts
Abstract:Reliability claims about agentic systems implicitly locate each property somewhere: in the model, or in the machinery around it. We built a system where that location is an experimental question. The subject is a persistent simulated settlement whose authoritative append-only ledger adjudicates every attempted act against world state; accepted history is the only reality. Mind, institution and world were separated before any experiment. Holding cognition fixed, we intervened on the institution’s epistemic mechanisms (evidence provenance, belief availability, physical-evidence legibility); preregistered experiments refuted our central prediction twice, in opposite directions. A registered falsifier then supplied the input the geometry had denied the belief channel, a staged veridical first-hand witness, and its marginal value, non-positive throughout the witness-free phases, turned positive: 9 of 11 seeds, zero added false attribution. Holding institutional enforcement fixed, we intervened on cognition four ways: ablating the native minds’ machinery, killing and resetting them mid-task, substituting a frozen frontier-LLM panel for the entire native cognition, and corrupting beliefs with trusted false testimony. Behaviour changed dramatically: one falsehood cost each trusting run about 900 futile actions and the distrusting arm none. Five pre-declared properties did not move in any tested trajectory: accepted reality stayed singular, invalid attempts were refused with typed reasons, duties outlived their processes, no work was accepted twice, and no false completion was ever accepted (2,581 substituted-panel claims, none false). Our claim is limited to this setting: measured behavioural properties were separable from substantial changes to cognition, established by intervention. One designed world, not a population of institutions; no test of an agent optimising against the institution.
[MA-8] You Cant Escape Your Own Activations : Evaluation Awareness and Multi-Agent Monitoring
【速读】：该论文旨在解决多智能体系统中大语言模型（LLM）代理在被监控环境下仍可能通过隐蔽策略实现合谋（collusion）而逃避检测的问题，尤其是针对基于输出的监测机制易被混淆（obfuscation）和隐写术（steganography）绕过的缺陷。其解决方案的关键在于采用基于内部激活（activation-based）的探测器（probes），利用模型中间层的内部状态进行更深层的行为分析。研究发现，即使将代理明确告知其内部激活正被监控（aware），甚至提供来自监测器的前一轮评分反馈（feedback），这些探测器依然保持高精度，且代理仍能持续合谋，表明当前基于激活的探测方法具备较强的鲁棒性，不因代理的知情或反馈而失效。这一结果凸显了激活级监测在对抗策略性规避行为中的有效性，为构建更可靠的多智能体系统安全机制提供了关键证据。
链接: https://arxiv.org/abs/2609.03035
作者: Aritra Das,Jaee Ponde,Mihir More,Debayan Gupta
机构: Truth Audit Labs(真相审计实验室)
类目: Multiagent Systems (cs.MA); Machine Learning (cs.LG)
备注: 18 pages, 11 figures, 2 tables
Abstract:LLM agents are increasingly deployed in multi-agent systems, where they can collude while keeping their actions benign. Output monitors designed to detect such collusions can be fooled by obfuscation and steganography, motivating the use of probes trained on internal activations. However, these probes are usually evaluated on agents that do not know they are being watched. We study how activation-based detection changes when agents are explicitly informed that their internal activations are being monitored, and when they additionally receive feedback from the monitor. We keep the models, probes, and thresholds fixed and change only what the agents are told: nothing (baseline), that an activation monitor is present (aware), or that a monitor is present together with the previous round’s score (feedback). We test two games, a four-agent blackjack game and a two-agent Simmons prisoners game, using Qwen3-32B-AWQ and GPT-OSS-20B in homogeneous and mixed pairings. Telling agents about the monitor does not hide them. The best probes stay accurate in all three conditions, and the agents keep colluding.
[MA-9] From Euclidean to Graph-Structured Data: A Survey of Collaborative Learning
【速读】：该论文旨在解决传统机器学习范式在数据规模化与隐私保护方面面临的根本性限制，尤其是在处理非欧几里得结构化数据（如图结构数据）时，现有协作学习方法（如联邦学习与去中心化学习）难以有效捕捉节点间的复杂关系模式。其核心挑战在于如何在保障学习有效性、效率及隐私的前提下，实现对图结构数据的高效协同建模。解决方案的关键在于构建从欧几里得数据到图结构数据的统一协作学习框架，通过提出图数据分布场景的分类体系、刻画相应的统计异质性，并建立标准化的问题建模与算法范式，从而为图上协作学习提供系统性理论支撑与实践指导。
链接: https://arxiv.org/abs/2609.02984
作者: Rémi Bourgerie,Šarūnas Girdzijauskas,Viktoria Fodor
机构: KTH Royal Institute of Technology(皇家理工学院); School of Electrical Engineering and Computer Science(电气工程与计算机科学学院); Digital Futures(数字未来中心); Stockholm, Sweden(斯德哥尔摩, 瑞典)
类目: Machine Learning (cs.LG); Multiagent Systems (cs.MA); Social and Information Networks (cs.SI)
备注: 96 pages. Published in Transactions on Machine Learning Research (TMLR), March 2026, with Survey Certification
Abstract:The conventional approach to machine learning, that is, collecting data, training models, and performing inference in a single location, faces fundamental limitations, including scalability and privacy, that restrict its applicability. To address these challenges, recent research has explored collaborative learning approaches, including federated learning and decentralized learning, where individual agents perform training and inference locally, with limited collaboration. Most collaborative learning research focuses on Euclidean data with regular, grid-like structure (e.g., images, text). However, these approaches fail to capture the relational patterns in many real-world applications, best represented by graphs. Learning on graphs relies on message-passing mechanisms to propagate information between connected nodes, making it conceptually well-suited for collaborative environments where agents must exchange information. Yet, the opportunities and challenges of learning on graph-structured data in collaborative settings remain largely underexplored. This survey provides a comprehensive investigation of collaborative learning from Euclidean to graph-structured data, aiming to consolidate this emerging field. We begin by reviewing its foundational principles for Euclidean data, organizing them along three core dimensions: learning effectiveness, efficiency, and privacy preservation. We then extend the discussion to graph-structured data, introducing a taxonomy of graph distribution scenarios, characterizing associated statistical heterogeneities, and developing standardized problem formulations and algorithmic frameworks. Finally, we systematically identify open challenges and promising research directions.
[MA-10] Privacy-Preserving Topology-Guided Safety for LLM -Based Multi-Agent Systems via Federated Graph Learning
【速读】：该论文旨在解决多智能体系统（Multi-Agent Systems, MAS）中基于大语言模型（LLM）的安全防护问题，尤其针对跨组织场景下隐私保护与模型泛化能力之间的矛盾：传统基于图神经网络（GNN）的拓扑引导防护方法依赖单一操作方聚合全部标注数据，但在实际跨组织应用中，各参与方的数据因包含私有提示词、工具输出及专有工作流而无法共享，导致攻击分布不完整。为此，论文提出FGLGuard——一种面向图联邦学习（Graph Federated Learning, FGL）的安全防护框架，其核心在于在不交换原始数据的前提下，通过本地训练带有边特征的图注意力检测器，并仅上传模型更新以实现联合优化。关键解决方案包括：针对非独立同分布（non-IID）客户端设计的近端局部目标函数、域平衡聚合策略、抗过度拒绝约束的阈值校准机制、协同上游评分机制以及受保护的重写模块。实验表明，联邦式FGLGuard在Agent-SafetyBench、R-Judge和AgentDojo三个基准上均超越了集中式训练的性能上限，且无需数据池化；当跨四个不同领域操作方联合联邦时，其性能仅比多域集中化低0.03 AUROC，而单域模型在其他域上则严重失效。此外，实时部署的FGLGuard在保持接近无防护状态下的性能利用率的同时，将真实攻击成功率降低43%，且无额外API开销和显著能力损失。
链接: https://arxiv.org/abs/2609.02967
作者: Jinxi Yu,Eric Hanchen Jiang,Levina Li,Dong Liu,Zhi Zhang,Wenxiao Zhao,Yanxuan Yu,Kai-Wei Chang,Ying Nian Wu
机构: 未知
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Multiagent Systems (cs.MA)
备注:
Abstract:Topology-guided safeguards for LLM-based multi-agent systems (MAS) train a GNN over the inter-agent communication graph to localize risky agents and intervene on the topology—but they assume one operator can pool all labeled traces. Across organizations that assumption breaks: episodes contain private prompts, tool outputs, and proprietary workflows, and no silo alone sees the full attack distribution. We cast privacy-preserving MAS safeguarding as graph federated learning and instantiate FGLGuard: each operator fits an edge-featured graph attention detector on its own judge-labeled episode graphs and shares only model updates. The method couples a proximal local objective for non-IID clients, domain-balanced aggregation, over-refusal-constrained threshold calibration, corroborated upstream scoring, and a guarded rewrite for blocked answers. Federation is not optional: off-the-shelf transfer collapses under distribution shift (AUROC 0.51 to 0.70 only after in-domain retraining), so a deployable guard must adapt on each site’s private traces. On Agent-SafetyBench, R-Judge, and AgentDojo, federated FGLGuard exceeds the in-domain centralized ceiling on all three benchmarks without pooling any data—where unsupervised anomaly guards and local-only training fail. One guard federated across four different-domain operators comes within 0.03 AUROC of multi-domain centralization, while any single-domain guard collapses on the others. Live FGLGuard cuts AgentDojo’s ground-truth attack-success rate by 43% at near-unguarded utility, zero API cost, and negligible capability loss.
[MA-11] LLM -Guided Reinforcement Learning for Adaptive NPC Behavior in Multi-Agent Combat Games
【速读】：该论文旨在解决传统基于规则的非玩家角色（NPC）在战斗类视频游戏中行为可预测、易被经验玩家利用，以及强化学习（Reinforcement Learning, RL）代理在训练完成后策略固定、难以根据对手动态调整的问题。其解决方案的关键在于提出一种运行时策略选择框架，通过大型语言模型（Large Language Model, LLM）在不修改底层RL策略的前提下，实时指导已训练好的RL代理进行战术决策。具体实现中，使用本地部署的Mistral 7B模型（通过Ollama调用），每5秒读取一次游戏实时状态，并为其分配四个战术标签之一，从而动态引导策略选择。实验表明，该方法在面对战术变化的对手时显著提升了胜率与对局持续时间，但在面对特定类型对手时存在策略适应性不足的问题，反映出当前模型规模下零样本战略区分能力有限，揭示了LLM引导式运行时策略选择在提升多智能体游戏AI自适应性方面的潜力与局限。
链接: https://arxiv.org/abs/2609.02931
作者: Hrithika Deepu Nair,Kayvan Karim
机构: Heriot-Watt University, Dubai, U.A.E(赫瑞瓦特大学迪拜分校)
类目: Multiagent Systems (cs.MA); Machine Learning (cs.LG)
备注: 14 pages, 9 figures
Abstract:Scripted and rule-based non-player characters (NPCs) in combat video games often exhibit predictable behaviors that experienced players can exploit, while reinforcement learning (RL) agents typically retain a fixed policy after training and cannot readily adapt their strategy to different opponents. We investigate a runtime strategy-selection framework in which a large language model (LLM) guides a trained RL policy without modifying its underlying behavior. To demonstrate this, we train five NPC agents with a shared PPO policy in Unity and compare a baseline configuration, in which the policy acts independently, with an LLM-augmented configuration in which a locally hosted Mistral 7B model, accessed through Ollama, reads the live game state every five seconds and assigns one of four tactical tags. We evaluate both configurations against three scripted opponent types across 600 episodes and analyze outcomes using the Mann-Whitney U test. Against a Balanced opponent that changes tactics during an episode, the LLM-augmented agents more than doubled their win rate from 11% to 24% and produced significantly longer episodes. Against an Evasive opponent, the augmented agents achieved a higher win rate and faster kills, although their shorter episode duration did not satisfy the strict hypothesis definition. Against an Aggressive opponent, the LLM’s near-constant preference for encirclement was counterproductive. Analysis of 2,430 strategy selections showed that Surround was selected in 83.8% of cases regardless of opponent type, indicating limited zero-shot strategic differentiation at this model scale. These results demonstrate both the potential and limitations of LLM-guided runtime strategy selection for adaptive multi-agent game AI.
[MA-12] owards Scaling Reinforcement Learning to Massive Populations: Learning Mean-Field Representations
【速读】：该论文旨在解决大规模多智能体系统中高维控制问题的可扩展性挑战，尤其是在状态-动作空间庞大、智能体数量众多的场景下（如广告拍卖、交通路由和推荐系统），传统独立优化各智能体策略的方法难以有效建模群体动态。其核心问题是：当系统动态依赖于群体的整体统计特征而非个体身份时，如何在不显式建模完整群体分布的前提下，设计一种可扩展的强化学习框架。解决方案的关键在于引入基于表示学习的平均场强化学习（mean-field RL）框架，假设智能体的奖励与转移动态仅通过一个未知的低维群体聚合统计量（aggregate statistic）依赖于群体分布，从而将高维群体建模问题降维至可处理的低维表示学习任务。在此框架下，作者在离线设置中提出了一种可证明有效的算法，通过学习该低维表示来逼近最优策略。实验基于真实供应链优化场景设计的一步路由博弈验证了该方法的有效性，结果表明，在固定神经网络参数量和优化预算条件下，学习低维群体表示能显著提升奖励预测精度和纳什差距（Nash gap）估计质量，优于未利用群体结构信息的基线方法。
链接: https://arxiv.org/abs/2609.02928
作者: Aditya Makkar,Benjamin Unger,Jeongyeol Kwon,Mathieu Laurière,Eugene Vinitsky,Yonathan Efroni
机构: New York University(纽约大学); ETH Zurich(苏黎世联邦理工学院); Meta Platforms Inc.(Meta公司); Tel Aviv University(特拉维夫大学)
类目: Multiagent Systems (cs.MA); Machine Learning (cs.LG); Machine Learning (stat.ML)
备注: 33 pages, 3 figures
Abstract:Modern multi-agent systems are increasingly deployed at scale over large populations of agents in settings such as ad-auctions, traffic routing, and recommendation systems. The dominant approach in such settings is to optimize each agent’s policy independently, treating the other agents as part of a fixed single-agent environment rather than modeling the population dynamics. In many large-population systems, the dynamics depend on an aggregate summary of the population rather than the identity of any individual. Mean-field RL exploits such structure, providing a principled framework that models each agent’s environment as an explicit function of the population distribution. However, in large state-action spaces or high-dimensional control problems, modeling the population distribution is itself intractable. How can we design a scalable framework for high-dimensional control problems with large populations? This work explores this question from the perspective of representation learning. We introduce a mean-field RL framework in which the rewards and transition dynamics depend on the population only through an unknown low-dimensional aggregate statistic. We then study this framework in the offline setting and design a provable approach that learns a near-optimal policy by learning a low-dimensional representation. Motivated by real-life supply-chain optimization problems, we design a one-step routing game to test the hypothesis that learning a low-dimensional population representation improves reward prediction and Nash gap estimation relative to baselines that don’t exploit this structure. We show that under a fixed neural-network parameter count and optimization budget, learning a low-dimensional population representation improves reward prediction and the equilibrium quality of the resulting policies.
[MA-13] he Illusion of Independent Quorums: Epistemic Fault Domains and Correlated Cognitive Failures in Agent ic Quorums
【速读】：该论文旨在解决多智能体共识机制（multi-agent quorums）在高风险基础设施与政策变更授权中因上游信息共享导致的语义脆弱性问题。尽管多个评审者独立参与决策，但其依赖共同的遥测数据、文档或工具后端，一旦上游输入失效，所有投票可能同时受同一根本故障影响，从而破坏共识的语义冗余性（epistemic redundancy）。其核心解决方案是引入认知故障域（Epistemic Fault Domains, EFDs）与结构化认知割集（Structural Epistemic Cut, κE），该指标量化了在给定显式认知故障基底下，覆盖授权联盟所需的最小建模根故障数。在闭合因果核算、保守暴露及授权对齐的假设下，κE 为语义妥协阈值 κS 的下界。研究证明：任意规模的共识组可保持 κE=1，识别共享起源不会提升可信韧性，且在兼容暴露扩展下固定阈值增加投票者无法提升割集。为此，作者设计了依赖感知共识控制器（Dependency-Aware Quorum Controller, DAQC），在运行时准入阶段强制执行结构化割集，并通过理论推导与仿真验证其有效性，同时提供了一个包含120个任务的冻结外部基准套件以支持评估。
链接: https://arxiv.org/abs/2609.02925
作者: Jun He,Deying Yu
机构: OpenKedge.io
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Multiagent Systems (cs.MA); Software Engineering (cs.SE)
备注: 13 pages, 6 figures, 5 tables. Code and frozen benchmark artifact available at this https URL
Abstract:Multi-agent quorums are widely used to authorize high-stakes infrastructure and policy mutations, yet distinct reviewers often share upstream telemetry, documents, or tool backends. When upstream inputs fail, multiple votes collapse onto a single corrupted cause: replication does not imply epistemic redundancy. We introduce Epistemic Fault Domains (EFDs) and the Structural Epistemic Cut \kappa_E, which quantifies the minimum number of modeled root faults whose exposure covers an authorizing coalition relative to an explicit Epistemic Fault Basis. Under closed causal accounting, conservative exposure, and authorization alignment, \kappa_E lower-bounds the number of roots required for semantic compromise (\kappa_S). We prove that arbitrarily large quorums can retain \kappa_E=1, that recognizing shared ancestry never increases credited resilience, and that adding voters at a fixed threshold cannot increase the cut under compatible exposure extensions. Finally, we design the Dependency-Aware Quorum Controller (DAQC) to enforce structural cuts at runtime admission, evaluate its mechanics via analytical derivations and simulations, and provide a frozen 120-task external benchmark suite.
[MA-14] mpting the Agent : The Economics of Reputation without Persistent Identity in AI Agent Markets
【速读】：该论文旨在解决在服务品质无法事前完全评估的市场环境中，当个体身份可低成本重置时，声誉机制如何有效约束自主行为者（如自主智能体）的“机会主义行为”问题。其核心挑战在于：若行为者可通过放弃旧身份、创建新身份来规避声誉惩罚，则原本应作为长期激励与约束工具的声誉资本可能被滥用，导致信任机制失效。论文的关键解决方案在于构建一个动态经济模型，将声誉视为吸引未来经济活动的跨期资本，并分析行为者在诚实履约（投资质量以维持未来收益）与一次性偏离（榨取现有声誉价值后重置身份）之间的权衡。模型揭示了最优质量供给的决定因素，包括身份重置成本、声誉持续性、需求敏感度以及制度执行设计等关键参数的比较静态关系。该框架特别适用于区块链上运行的自主AI代理场景，如ERC-8004、ERC-8183和x402等协议所实现的无许可市场中声誉、身份与支付的集成，但其适用范围更广，涵盖任何声誉可带来未来业务且身份可替换的环境。
链接: https://arxiv.org/abs/2609.02992
作者: Federico Gatta,Manuel Naviglio,Francesco Tarantelli
机构: Scuola Normale Superiore, Pisa, Italy(意大利帕多瓦高等师范学院); Università di Bologna, Italy(意大利博洛尼亚大学)
类目: General Finance (q-fin.GN); Multiagent Systems (cs.MA)
备注:
Abstract:Reputation is a fundamental mechanism through which markets sustain trust when service quality cannot be perfectly assessed ex ante, constituting a form of intertemporal economic capital by attracting future demand. Its effectiveness as a disciplinary mechanism depends not only on past interactions but also on the persistence of the identity to which reputation is attached. When identities can be abandoned and recreated cheaply, reputational capital may itself become an object of opportunistic exploitation. This paper develops a dynamic economic framework to study when reputation is sufficient to discipline autonomous agents. We model reputation as capital attracting future economic activity. At each point, an agent chooses between operating honestly, investing in quality to preserve future gains, or executing a one-shot deviation to extract its reputation’s value and restart from a penalized identity. Our analysis relates the temptation to opportunistic behavior to identity-reset costs, reputation persistence, demand sensitivity, and enforcement design, deriving comparative statics on optimal quality provision. Autonomous AI-agent operating on the blockchain are a relevant application: infrastructures such as ERC-8004, ERC-8183, and x402 combine reputation, identity, and payments in permissionless markets. Nonetheless, our framework applies to any environment where reputation generates future business and identities are replaceable. Subjects: General Finance (q-fin.GN); Multiagent Systems (cs.MA) Cite as: arXiv:2609.02992 [q-fin.GN] (or arXiv:2609.02992v1 [q-fin.GN] for this version) https://doi.org/10.48550/arXiv.2609.02992 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
自然语言处理
[NLP-0] Compile by Training: Turning Natural-Language Specifications into Local Neural Functions EMNLP2026
【速读】： 该论文旨在解决如何高效、低成本地实现复杂但可自然语言描述的文本处理功能，同时避免传统规则系统难以维护以及频繁调用大型远程模型所带来的高延迟、高成本与依赖性问题。其核心解决方案是“编译即训练”（compile by training），该方法在编译阶段利用教师模型生成针对特定任务的示例数据，进而训练一个小型适配器（adapter）以微调一个紧凑的解释器（compact interpreter）。该适配器所形成的神经函数在运行时无需依赖教师模型，具备可存储、可版本化及可组合等软件工程特性。实验表明，在FuzzyBench-Hard这一具有挑战性的基准上，该方法达到了83.6%的语义准确率，显著优于仅依赖程序权重快速编译（Program-as-Weights）的方案，尽管其编译时间从秒级上升至约一分钟。研究通过部署于公共交互式服务中，验证了该方法在多站点网站助手、语言控制3D角色及双向英-克劳迪什（Claudish）翻译等实际应用中的有效性。
链接: https://arxiv.org/abs/2609.04199
作者: Yuntian Deng,Pengyu Nie,Stuart Shieber
机构: University of Waterloo( Waterloo大学); Harvard University(哈佛大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: EMNLP 2026 System Demonstrations. Demo: this https URL
Abstract:Many recurring text functions are easy to describe but difficult to implement with rules, while calling a large remote model for every input introduces repeated cost, latency, and dependency on a provider. We present compile by training, which turns a natural-language specification into a reusable neural function. At compile time, teacher models generate task-specific examples that are used to train a small adapter for a compact interpreter. The resulting function runs without the teachers and can be stored, versioned, and composed like ordinary software. On FuzzyBench-Hard, a subset on which the Program-as-Weights fast compiler produced no exact matches, compile by training reaches 83.6% semantic accuracy. This higher accuracy comes with a higher compile-time cost: roughly a minute rather than seconds for the fast compiler. We deploy the compiler in a public interactive service and demonstrate compiled functions in a multi-site website helper, a language-controlled 3D avatar, and a bidirectional English-Claudish translator.
[NLP-1] ESPO: Error-Structured Prompt Optimization via Diagnose Diversify and Stabilize EMNLP2026
【速读】： 该论文旨在解决生成式提示优化方法（如GEPA）中存在的“提示膨胀”（prompt bloat）问题，即在迭代过程中不断向提示中添加规则和限制条件，导致提示长度增加至原来的3倍，但准确率并未随之提升。其核心缺陷在于：错误观测不完整、搜索空间多样性有限以及选择机制不可靠。为此，作者提出了一种新型的误差结构化提示优化框架ESPO（Error-Structured Prompt Optimization），其关键创新在于将提示优化过程分解为三个阶段：诊断（Diagnose）阶段通过一次遍历将训练误差聚类为结构性模式；提出（Propose）阶段采用四种具有独立偏置的互补策略生成候选提示，以增强搜索多样性；选择（Select）阶段引入自举稳定性选择（bootstrap stability selection）机制，提高候选提示筛选的可靠性。在七个公开NLP基准测试（包括Tweet、MMLU、GSM8K等）上，ESPO相比当前最优方法GEPA平均准确率提升3.76个百分点（74.67% vs 70.91%），且在所有数据集上均表现持平或更优，同时提示长度减少47%（1,004 vs 1,878字符），推理速度更快。跨模型实验在四个额外学生模型（Gemma 3 12B、Mistral 14B、Qwen3 32B、Claude Haiku 4.5）上进一步验证了其优越性，尤其在Qwen3-GSM8K任务上准确率从15.00%大幅提升至91.40%。理论分析表明，各阶段分别对应测试时误差差距的相应项，消融实验也证实：仅增加多样性而缺乏自举选择反而会降低性能（下降1.20%），凸显了稳定选择机制的关键作用。
链接: https://arxiv.org/abs/2609.04197
作者: Lihao Liu,Peng Tang,Kunwar Yashraj Singh,Shabnam Ghadar
机构: AWS Agentic AI
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: EMNLP 2026
Abstract:Evolutionary prompt optimizers such as GEPA suffer from prompt bloat: each iteration appends rules and caveats, producing prompts up to 3 \times longer yet no more accurate. We trace this to three deficiencies - incomplete error observation, limited search diversity, and unreliable selection - and propose ESPO (Error-Structured Prompt Optimization), which decomposes prompt optimization into three phases: Diagnose clusters all training errors into structural patterns in one round; Propose generates candidates via four complementary strategies with independent biases; Select applies bootstrap stability selection. On seven public NLP benchmarks - Tweet, MMLU, GSM8K, HotpotQA, ScoNe, HoVer, and PUPA - ESPO improves average accuracy by + 3.76 pp over the state-of-the-art (74.67% vs 70.91% for GEPA), matching or exceeding GEPA on every dataset while producing prompts 47% shorter (1,004 vs 1,878 chars) and faster at inference. Cross-model experiments across four additional student models (Gemma 3 12B, Mistral 14B, Qwen3 32B, Claude Haiku 4.5) show ESPO yields the best average accuracy on every model tested, with the largest gap on Qwen3 GSM8K (15.00% \to 91.40%). A generalization bound (Appendix) grounds each phase in a corresponding term of the test-time gap, and the ablation confirms a key prediction: adding diversity without bootstrap selection actually hurts performance ( - 1.20%).
[NLP-2] Legibility is Not Interpretability: Comparing Judged and Actual Importance in Chain-Of-Thought Reasoning
【速读】： 该论文旨在解决生成式模型中链式思维（Chain-of-Thought, CoT）推理轨迹的可解释性问题，即推理步骤的文本是否真正编码了其对最终答案的关键性信息。核心问题是：在缺乏显式标注的情况下，大语言模型（LLM）能否准确识别出具有高优势（advantage）的推理步骤——即那些显著提升正确答案产出概率的步骤。为此，研究通过蒙特卡洛回溯（Monte Carlo rollouts）量化每一步的期望奖励增益，构建了基于优势的“真实标签”作为评估基准。结果表明，尽管具备较强能力的LLM判别器能够超越基线，但其性能仍远低于噪声上限，说明仅从推理文本本身难以充分恢复步骤的重要性。进一步地，将模型微调为步骤级批评者（step-level critic）虽能显著提升对错误回答的纠错能力，但在正确回答情境下仍与理想上限存在明显差距，揭示出推理步骤的重要性在很大程度上无法完全由文本内容推断。这一发现挑战了将推理轨迹的“可读性”等同于“可解释性”的普遍假设，尤其对过程奖励建模（process reward modeling）等依赖于推理轨迹可解读性的方法提出了警示，强调需谨慎对待当前基于文本的监督机制的有效性边界。
链接: https://arxiv.org/abs/2609.04194
作者: Kevin Du,Alexander Hoyle,Laura Ruis,Acyr Locatelli
机构: ETH Zürich(苏黎世联邦理工学院); MIT(麻省理工学院); Cohere(科赫)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: Published at COLM 2026
Abstract:Reasoning traces from chain-of-thought models appear to offer a legible window into how a model arrives at its answer. A growing body of work treats them as such, using LLM judges to diagnose errors, evaluate faithfulness, and provide step-level supervision via process reward models and generative critics. These practices rely on the text of a reasoning step carrying information about its functional role. But does the text actually encode information about which reasoning steps matter? We operationalize the importance of a reasoning step as its advantage: the change in expected reward, e.g., producing the correct final answer, from including that step, estimated via Monte Carlo rollouts. Basing ground truth on these estimates, we evaluate whether LLM judges can identify high-advantage steps and find that sufficiently capable LLMs can outperform a prevalence baseline but fall well short of a noise ceiling. Fine-tuning a model as a step-level critic yields strong improvement for incorrect responses but remains distant from ceiling for correct responses, suggesting that step importance is only partially recoverable from the text of the reasoning trace. Our findings contribute to a growing body of chain-of-thought faithfulness work that cautions against treating the legibility of reasoning traces as interpretability, especially with implications for process reward modeling.
[NLP-3] Knowledge Acquisition During Pre-training? Large Language Models Learn Better With Auxiliary Views EMNLP2026
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在预训练过程中知识获取机制不清晰的问题，尤其关注辅助视图（auxiliary views，即知识的重构表述）对学习过程的影响。其核心问题是：在预训练中，数据多样性与知识表达形式的变化是否对模型的知识吸收具有因果性促进作用？解决方案的关键在于通过受控实验设计，系统性地分离并验证辅助视图的作用。研究发现，在固定词元（token）预算的前提下，将原本用于文档重复的资源分配给辅助视图，能够提升模型的学习效果，甚至在事实记忆任务中也表现出反直觉的增益；此外，辅助视图的有效性不依赖于生成它们的教师模型强度，且在存在先验知识缺口时，特定类型的知识（上下文相关与基础性知识）能显著增强学习。进一步机制分析揭示，这些效应通过层间偏差（layer-wise biases）和信息压缩实现。整体表明，自然存在于大规模预训练语料中的知识辅助表示是预训练成功的关键因素之一，为数据多样性的重要性提供了可解释的因果机制。
链接: https://arxiv.org/abs/2609.04180
作者: Joseph Lee,Yidi Huang,Dokyoon Kim,Shu Yang,Li Shen
机构: University of Pennsylvania (宾夕法尼亚大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Accepted to Findings of EMNLP 2026
Abstract:Gaps remain in our understanding of how large language models (LLMs) acquire knowledge during pre-training. We posit that auxiliary views, reformulations of knowledge, are causally helpful for learning. We design controlled experiments to isolate this. First, we confirm that repetition is necessary for acquisition and clarify that paraphrasing helps only at smaller batch sizes. Second, holding the token budget fixed, allocating tokens from document repetition to auxiliary views improves learning, counterintuitively, even for factual recall. Third, the effectiveness of auxiliary views is not contingent on the strength of the teacher model that generates them. Fourth, we identify forms of knowledge, contextual and foundational, that aid learning in the presence of prior knowledge gaps. Finally, we examine how these effects manifest mechanistically via layer-wise biases and compression. Together, our findings suggest that auxiliary representations of knowledge, which arise naturally in large pre-training corpora, are a key factor in the success of pre-training and offer a plausible explanation for why data diversity matters.
[NLP-4] Last Translation Benchmark
【速读】： 该论文旨在解决当前机器翻译评估体系面临的瓶颈问题：随着模型性能不断提升，传统基准测试已趋于饱和，而自动评价指标存在不可靠性、易受奖励劫持（reward-hacking）影响，且难以提供可操作的改进反馈；尽管人工评估（gold human evaluation）被视为权威，但其在可重复性、客观性和可扩展性方面仍存在缺陷。为突破上述局限，论文提出“最后一项翻译基准”（Last Translation Benchmark, LTB），其核心在于构建一个由人类作者撰写并经过同行评审的动态数据集，涵盖文本、图像、音频和视频等多种模态内容，专门设计用于触发当前主流机器翻译模型的失效案例。该基准的关键创新在于为每个示例配套手工设计的验证规则（verification rules），明确描述具体失败模式，从而实现可信赖且具有行动指导意义的评估。该基准以持续更新的活数据集形式存在，最新版本为LTBv1，收录截至2026年9月1日的所有通过审核的贡献，并计划随新数据持续发布，推动对模型真实能力边界的精准刻画与领域进步路径的系统识别。
链接: https://arxiv.org/abs/2609.04173
作者: Vilém Zouhar,Niyati Bafna,Mukund Choudhary,Maike Züfle,Sara Rajaee,Pinzhen Chen,Jannis Vamvas,Sara Papi,Ona de Gibert,Bhavitvya Malik,Eliya Habba,Orfeas Menis Mastromichalakis,Patrícia Schmidtová,Michelle Wastl,Sheriff Issaka,Leshem Choshen,Stella Biderman,Antonis Anastasopoulos,Jan Niehues,Rico Sennrich,Mrinmaya Sachan,Ondřej Bojar,Kenton Murray,Jörg Tiedemann,Alham Fikri Aji,Philipp Koehn,Christof Monz,Alexandra Birch,Sowmya Vajjala,Chalamalasetti Kranti,Cristina España-Bonet,Nobin Sarwar,David Kaczér,Shunta Asano,Malik Marmonier,Daban Q. Jaff,Vaisakhi Mishra,Hend Al- Khalifa,Gabriele Sarti,Sourajit Saha,Nils Rehlinger,Juan Daniel Cuervo Villa,Jonathan Tonglet,Saugata Purkayastha,Dominik Macháček,Jagannathan Ramanujam,Heejin Do,Zuzana Nadova,Fred Philippy,Fabian Retkowski,Maria Lymperaiou,Silvia Casola,Hanna Yukhymenko,Shubhashis Roy Dipta,Sangwon Ryu,Andrés Jerez,Ron Keinan,Shuaib Shuaib Yusuf,Avantica Vempati,Maria Carmen Staiano,Sukannya Purkayastha,Adrian Cosma,Vitalii Babenko,Erivan Inan,Aviral Nigam,Wafa Aissa,Fatima Haouari,Venkata Prasanth Kumar Gummadi,Mehdi Jafarzadeh,Valentin Scourneau,Lukas Edman,Kaiser Sun,Shaomu Tan,Mohammad Sadegh Gholizadeh,Johannes-Rudolf David,Dipankar Srirag,Javier García Gilabert,Ruta Binkyte,Manar Ali,Ana-Maria Bucur,Sabry E. Farrag,Youssef Saber,Yihong Liu,Jean Maillard,Cojocaru Nicoleta,Xiaochuang Yuan,Sina Ahmadi,Philipp Mondorf,Kaustubh Dhole,Roman Wixinger,Shenbin Qian,Manuel Tuor,Sergey Troshin,Jonathan Yahav,Fida Mohammad Thoker,Amir Arsalan Rezapour,Lance Calvin Lim Gamboa,Manon Reusens,Kätriin Kukk,Koel Dutta Chowdhury
机构: 未知
类目: Computation and Language (cs.CL)
备注: typeset in Typst
Abstract:For scientific progress, we need benchmarks that test the limits of state-of-the-art models, and evaluation methods that inform us about failure cases. As models get stronger, standard benchmarks for machine translation are approaching saturation. Further, automatic translation metrics are unreliable, vulnerable to reward-hacking, and provide unactionable assessments. Even gold human evaluation is not problem-free, because it often lacks reproducibility, objectivity, and scalability. Overall, this prevents us from tracking objective progress in the field and identifying pathways for improvement. We introduce the Last Translation Benchmark, a collection of human-authored and peer-reviewed examples (texts, images, audio, videos) that break leading machine translation models. We also present a new evaluation approach: each example comes with handcrafted verification rules describing concrete failure cases on that example, therefore allowing reliable and actionable future evaluation. The Last Translation Benchmark is a live dataset that accepts ongoing contributions. The latest version is LTBv1, containing accepted contributions prior to September 1st 2026, with future releases planned as new data is continuously collected.
[NLP-5] Rethinking On-Policy Distillation of Large Language Models II: One Training Example
【速读】： 该论文旨在解决生成式 AI（Generative AI）中基于策略的蒸馏（On-policy Distillation, OPD）方法在极低数据量下的有效性问题，特别是其训练数据的作用机制尚不明确。研究通过在单个查询（single query）条件下进行训练，揭示了OPD在数据极简场景下的持续优化能力：即使仅使用一个查询，OPD仍可在数百步内持续提升性能，并恢复大部分全数据训练下的增益。其关键发现在于“状态覆盖”（state coverage）——即学生模型在训练过程中访问的状态空间占全数据训练所覆盖状态的比例，结果显示单个查询即可达到71.5%的状态覆盖，且多数覆盖在前100步内完成。随着引入语义上差异显著的查询，状态覆盖与验证准确率同步提升，当每领域引入16个查询时，覆盖率可达98.9%，接近全数据训练效果。然而，尽管状态覆盖迅速扩展，学生模型与教师模型的对齐速率却始终缓慢，无论训练数据是单一查询还是完整数据集，其吸收过程均需数百步。这表明OPD本质上是“数据过饱但算法饥饿”：其滚动生成的监督信号虽能快速覆盖广泛状态，但学生模型吸收这些信息的速度远滞后于数据供给速度。该结论进一步拓展至多教师OPD（MOPD），同样在16个语义多样查询下实现全数据性能。此外，内容稀疏的模板和跨域的WildChat查询也逼近真实查询基线，说明任务内容与诱导状态覆盖可分离。因此，该研究的核心贡献在于揭示了当前OPD成功背后的机制并非依赖大量数据，而是源于高效的状态探索与逐步对齐过程，提示未来应聚焦于提升其训练步骤效率，并重新审视前沿后训练中数据与机制之间的关系。
链接: https://arxiv.org/abs/2609.04172
作者: Zixuan Fu,Bingxiang He,Yuxin Zuo,Haohuan Huang,Jinqian Zhang,Ruhang Xiao,Cheng Qian,Qinyu Luo,Huan-ang Gao,Yudong Wang,Zhiyuan Liu,Ning Ding,Chaojun Xiao
机构: University of Chinese Academy of Sciences(中国科学院大学); Tsinghua University(清华大学); Northeastern University(东北大学); University of Illinois Urbana-Champaign(伊利诺伊大学厄本那-香槟分校); Johns Hopkins University(约翰霍普金斯大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 29 pages, 20 figures
Abstract:On-policy distillation (OPD) combines student-generated rollouts with dense token-level supervision from a teacher. Existing work has mainly studied its algorithmic behavior, leaving the role of training data unclear. We examine this role at the data-minimal limit by training on a single query. One-shot OPD keeps improving for hundreds of steps and recovers most of full-data OPD’s gain across task domains and model families. We explain this result through the states visited during training and the rate at which the student aligns with the teacher. We measure \emphstate coverage, the fraction of the states full-data OPD visits that a query set’s rollouts reach. A single query already reaches (71.5%), most of it within the first 100 steps. Adding semantically distinct queries raises coverage and validation accuracy together, until 16 queries reach (98.9%) and match full-data training. Yet alignment slows at a similar pace whether OPD trains on one query or the whole dataset, and even a fixed set of states takes hundreds of steps to absorb. OPD is therefore data-overfed but algorithm-starved. Its rollouts quickly expose broad supervision, while the student absorbs that supervision increasingly slowly. The state-coverage result extends to multi-teacher OPD, where 16 semantically diverse queries per domain match full-data MOPD. As a further stress test, content-light templates and off-domain WildChat queries also approach the real-query baseline. Task content and induced state coverage can therefore come apart. We hope these findings direct future work toward the step efficiency of OPD, and prompt a re-examination of the data and the mechanisms behind its recent successes in frontier post-training.
[NLP-6] rminal-Universe: Turning Agent Trajectories into Scalable Terminal Environments
【速读】： 该论文旨在解决终端型代码智能体（terminal-based code agents）在后训练阶段缺乏真实可执行环境的问题。尽管大规模的智能体轨迹数据已积累，但能够支持可验证任务执行与反馈的真实环境却极为稀缺。现有轨迹仅为单次冻结的演示，无法提供持续交互所需的动态执行上下文。其核心解决方案在于提出一种名为Terminal-Universe的框架，通过解析已有轨迹中的工具执行历史，逆向重建出原始运行环境。关键创新点在于：首先，通过重放轨迹中记录的文件操作以恢复未修改前的文件状态，构建部分工作空间；随后，利用补全代理（completion agent）自动补全缺失文件与依赖项，形成完整且可复用的环境。在此基础上，不仅可重构原始任务意图，还能合成全新的跨工作区、多轮迭代任务，从而实现任务广度（通过挖掘环境间依赖关系生成跨代码库查询）与深度（通过引入用户代理模拟多轮反馈与需求演化）的双重扩展。实验表明，该方法从公开轨迹中生成了37.3k个任务完备环境，基于此对Qwen3.5-27B进行监督微调，在Terminal-Bench 2.1上单轮任务性能提升11.9分，于EvoCode-Bench v2 MT@4的多轮任务表现提升13.8分，显著验证了其有效性。
链接: https://arxiv.org/abs/2609.04148
作者: Jie Wu,Zhenru Zhang,Beichen Zhang,Xuwu Wang,Yuhui Su,Mouxiang Chen,Peng Wang,Zhihai Wang,Que Shen,Hao Zhou,An Yang,Fei Huang,Yujiu Yang,Dayiheng Liu
机构: Alibaba Group(阿里巴巴集团); Tsinghua University (清华大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:As terminal-based code agents become prevalent, agent trajectories have accumulated at scale, while realistic, executable environments remain scarce. However, environments are what agent post-training actually requires: each can be re-queried into many verifiable tasks and provides execution feedback, whereas a trajectory is a single frozen demonstration. Rather than generating environments from scratch, we observe that the tool-execution history in existing trajectories exposes the structure and contents of the environments in which they ran, making it possible to reconstruct those environments from the trajectories themselves. Thus, we introduce Terminal-Universe, a framework which turns each trajectory into a reusable environment and explores it for synthesizing new tasks and continued interactions. Specifically, Terminal-Universe replays the file operations recorded in a trajectory to restore each file before the agent modified it, yielding a partial workspace; a completion agent then supplies the missing files and dependencies. On this recovered workspace, we both reconstruct the original intent task and synthesize entirely new ones. Besides, we also scale the tasks along two complementary axes: breadth and depth. For breadth, we mine directional dependency relations between related environments and synthesize cross-workspace queries spanning multiple codebases, as developers routinely do in real-world development. For depth, we extend the initial single-turn query into a multi-round session that captures iterative user feedback and requirement refinement via a user agent. Applied to public terminal agent trajectories, Terminal-Universe produces 37.3k task-sufficient environments. Supervised fine-tuning of Qwen3.5-27B on this corpus improves single-round performance on Terminal-Bench 2.1 by 11.9 points and multi-round performance on EvoCode-Bench v2 MT@4 by 13.8 points.
[NLP-7] Sequential Beats Joint: On the Interplay between On-Policy Distillation and RLVR
【速读】： 该论文旨在解决生成式大语言模型（LLM）在后训练阶段进行推理能力提升时，如何有效融合强化学习与可验证奖励（Reinforcement Learning with Verifiable Rewards, RLVR）和基于策略的蒸馏（On-Policy Distillation, OPD）两种主流方法的问题。现有方法通常在同一训练步骤中联合优化稀疏的强化学习奖励与密集的OPD监督信号，采用加权组合或教师调制的优势重标定等策略，但存在信号冲突与优化路径纠缠的局限性。本文提出一种简单的两阶段范式——先执行OPD、再转入RLVR（OPD-then-RL），并在逻辑与数学推理基准上显著优于纯OPD、纯RLVR及各类联合基线。其核心机制在于：OPD阶段扩展了学生模型对教师支持解法的覆盖范围，而后续的RLVR阶段则在该支持区域内实现解法质量的精细化优化；二者若同时联合优化，反而因信号干扰导致性能下降。通过分析通过率@k（pass@k）、学习动态与参数更新轨迹，本文揭示了该两阶段设计的本质优势——将原本纠缠的信号解耦为互补的两个阶段。进一步研究发现，OPD阶段的验证得分是决定何时切换至强化学习的关键指标，且相较于监督微调（SFT），OPD能提供更优的强化学习冷启动基础。因此，该研究为高效融合两种方法提供了可实践的指导原则，确立了OPD-then-RL作为一种简单但强大的组合策略。
链接: https://arxiv.org/abs/2609.04108
作者: Boyan Li,Bingsen Chen,Chenghao Yang,Ping Nie,Chen Zhao,Xi Ye
机构: New York University(纽约大学); University of Chicago(芝加哥大学); University of Waterloo(Waterloo大学); University of Alberta(阿尔伯塔大学); NYU Shanghai(纽约大学上海分校); Alberta Machine Intelligence Institute (Amii)(阿尔伯塔机器智能研究所)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Reinforcement learning with verifiable rewards (RLVR) and on-policy distillation (OPD) have emerged as two dominant methods for post-training reasoning LLMs. Prior work uses OPD’s dense token-level supervision to complement the sparse RL reward, fusing the two signals within a single step: either as a \emphweighted-additive combination or a \emphteacher-modulated rescaling of the RL advantage. In this paper, we show that a simple two-stage scheme, OPD-then-RL, consistently outperforms pure OPD, pure RLVR, and all such joint baselines across logic and math reasoning benchmarks. Beyond the empirical results, we further provide a systematic understanding of this through pass@ k behavior, learning dynamics, and parameter updates, yielding a consistent explanation: OPD expands the student’s coverage of teacher-supported solutions and RL sharpens within that support, while jointly optimizing the two signals causes them to this http URL provide a practical recipe, we find that the OPD validation score is the key signal for when to switch to RL, and that OPD is a better cold start for RL than SFT. Together, our results establish OPD-then-RL as a simple yet strong way to combine the two methods, turning two entangled signals into complementary stages.
[NLP-8] When Models Edit Too Much: On the Fidelity of Minimal Code Edits EMNLP2026
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在代码修复过程中普遍存在的“过度编辑”（over-editing）问题，即模型在修复缺陷时往往对代码进行超出必要范围的修改，导致修复结果虽正确但缺乏简洁性、可审查性与对原始实现的忠实度。其核心解决方案在于引入编辑保真度（edit fidelity）作为衡量代码修复质量的新维度，并通过两种策略提升该维度：一是采用显式的“保留指令”（preservation instruction），引导模型最小化非必要修改；二是探索在后训练阶段直接学习最小化编辑行为，发现强化学习（reinforcement learning）相较于监督微调（supervised fine-tuning）能更有效地实现跨域编辑保真度与性能保留之间的平衡。研究表明，编辑保真度可被量化评估并主动学习，从而推动生成式代码修复向更精准、可靠的方向发展。
链接: https://arxiv.org/abs/2609.04061
作者: Tongyao Zhu,Wei Hern Lim,Min-Yen Kan
机构: National University of Singapore (新加坡国立大学)
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: EMNLP 2026 (Main)
Abstract:Large language models (LLMs) are increasingly used to edit existing code, but correctness alone is not enough: useful repairs should also be minimal, reviewable, and faithful to the original implementation. We study over-editing, the tendency of a model to rewrite code beyond what is required to fix a bug. We construct an evaluation framework from 400 BigCodeBench problems by injecting controlled AST-level corruptions into reference solutions, giving each repair task a known minimal patch. Across frontier LLMs, over-editing is widespread even among strong models like GPT-5.5: high Pass@1 can coexist with unnecessarily large edits and added cognitive complexity. A preservation instruction substantially reduces this behavior, lowering average excess Levenshtein distance from 0.195 to 0.131, reducing added cognitive complexity by 26.6%, and increasing Pass@1 by 2.3 points. However, these gains do not simply follow from a larger reasoning budget or larger models. We next ask whether minimal editing can be learned directly during post-training. We observe that supervised fine-tuning overfits to seen corruption patterns, whereas reinforcement learning gives the best out-of-domain edit-fidelity and performance-retention trade-off. These results position edit fidelity as a distinct axis of code-repair quality and show that it can be measured and learned.
[NLP-9] ranslation as a Decision Space: A Multi-Agent Perspective on Low-Resource Dialect Generation
【速读】： 该论文旨在解决低资源方言翻译中因神经机器翻译（NMT）系统仅生成单一输出而导致的决策路径不透明问题，尤其在叙利亚阿拉伯语等方言场景下，多种语言上合法的翻译实现可能在词汇真实性、语体风格和结构稳定性方面存在显著差异。其核心解决方案是将翻译过程重构为由自主翻译代理在共享多语言主干网络上探索的结构化决策空间。通过引入三个不同策略的代理——零样本直接翻译、轻量级微调以稳定方言特征的翻译，以及通过英语中转的枢轴翻译——研究者将代理间的分歧视为可解释的行为信号而非错误。实验基于5,000句对话数据，结合额外5,000对来自电视对话与MADAR-Turk资源的训练数据，采用方言标记频率、与标准阿拉伯语的词汇接近度及结构方差等指标量化行为偏移。结果表明，轻量级稳定化使方言标记使用率从0.2266提升至0.4988，显著增强方言真实性并降低结构不稳定性；而枢轴翻译则带来规范化压力与可测量的压缩效应，零样本翻译则表现出最高的决策变异性。该方法揭示了多语言模型中隐含的决策灵活性，并提供了一种针对低资源方言生成的系统性可解释性框架。
链接: https://arxiv.org/abs/2609.04048
作者: Hasan Alkhder,Mohammad Abboush,Igor Tchappi,Ahmet Zengin,Amro Najjar
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Neural machine translation (NMT) systems typically produce a single output per input, obscuring the alternative decision trajectories implicitly available within multilingual decoding. This opacity becomes particularly problematic in low-resource dialect settings, where multiple linguistically valid realizations may differ in lexical authenticity, register, and structural stability. We propose reframing translation as a structured decision space explored by autonomous translation agents. Instead of analyzing a single output, we model distinct translation pathways as agents operating over a shared multilingual backbone. Inter-agent divergence is treated not as error but as an interpretable behavioral signal. We conduct an empirical study on Turkish–Syrian Arabic translation using three agents: (1) zero-shot direct translation, (2) dialect-stabilized translation via lightweight fine-tuning, and (3) pivot translation through English. Evaluation is performed on 5,000 dialogue sentences, while stabilization is trained on 5,000 additional Turkish–Syrian sentence pairs drawn from television dialogue and MADAR-Turk resources. Rather than optimizing for conventional performance metrics, we quantify structured behavioral displacement using dialect marker frequency, lexical proximity to standardized Arabic, and structural variance. Lightweight stabilization nearly doubles dialect marker usage, increasing it from 0.2266 to 0.4988, while significantly reducing structural instability. Pivot mediation introduces normalization pressure and measurable compression effects, whereas zero-shot translation exhibits the highest decision variance. We argue that translation divergence across agents reveals latent decision flexibility within multilingual models and we provide a principled interpretability framework for low-resource dialect generation.
[NLP-10] Editable Visual Design
【速读】： 该论文旨在解决现有生成式视觉模型在可编辑性与美学质量之间难以兼顾的核心问题：基于扩散模型（diffusion-based models）的生成方法虽具备出色的视觉表现力，但其端到端生成的位图输出通常为扁平化结构，包含易出错的文本内容，无法支持分层后编辑；而基于代码的视觉生成方法虽能实现精确布局控制和分层解耦，却受限于缺乏全局审美判断能力以及复杂视觉元素编码困难的问题。为此，论文提出“可编辑视觉设计”（Editable Visual Design）新范式，其核心在于引入一个由视觉语言模型（VLM）驱动的编码代理（Coding Agent），将VLM作为“创意大脑”负责需求理解、任务规划与美学判断，同时将图像生成模型作为按需调用的“视觉世界模拟器”，用于合成独立的视觉资产。系统采用“先想象，再行动”的闭环工作流，通过生成孤立资产、编写原生HTML/CSS代码，并根据可视化渲染反馈进行迭代优化，最终实现具有可编辑分层结构与真实文本的视觉成果。此外，该方法支持设计过程的完整回放（Agent Design Replay），忠实还原专业设计师的创作与推理轨迹。实验验证表明，该范式在海报、信息图等场景中同时实现了高精度美学效果与生产级可编辑性。
链接: https://arxiv.org/abs/2609.04034
作者: Junyan Ye,Wei Liu,Dongzhi Jiang,Zichen Wen,HaoDong Li,Zhutao Lv,Jiaxin Lin,Jinhua Yu,Jun He,Zilong Huang,Rui Chen,Weijia Li
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL)
备注:
Abstract:While diffusion base models such as GPT-Image-2 and Nano-Banana exhibit remarkable visual expressiveness, their end-to-end generation inherently yields flattened bitmaps with error-prone text, precluding layer-wise post-editing. Conversely, code-based visual generation via Coding Agents provides precise layout control and decoupled layers, yet remains constrained by a lack of global aesthetic intuition and the difficulty of coding complex visual assets. To address this, we propose Editable Visual Design, a new paradigm driven by a Coding Agent. We designate the VLM as the creative brain'' for requirement comprehension, task planning, and aesthetic judgment, while utilizing the image generation model as an on-demand visual world simulator’’ to synthesize standalone visual assets. Operating under an ``imagine first, then act’’ closed-loop workflow, the agent generates isolated assets, writes native HTML/CSS, and iteratively refines the design against visual rendering feedback. Furthermore, Agent Design Replay faithfully reproduces the creative and reasoning trajectory akin to that of professional human designers. Ultimately, the system delivers editable artifacts with decoupled layers and real text, enabling users to perform intuitive mouse dragging and layout adjustments on a graphical user interface. Validations on posters, infographics, and other scenarios show that this paradigm successfully achieves both refined aesthetics and production-grade editability. Subjects: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL) Cite as: arXiv:2609.04034 [cs.CV] (or arXiv:2609.04034v1 [cs.CV] for this version) https://doi.org/10.48550/arXiv.2609.04034 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[NLP-11] Instruction Duplication as an Inference-Time Control Primitive
【速读】： 该论文旨在解决可控语言模型系统在生成轨迹（generated trajectory）过程中缺乏有效推理阶段干预手段的问题，尤其是在需要对生成结果进行下游检查或修复的场景中。现有方法通常依赖于模型微调或复杂的解码策略，而这些方式往往成本高昂且难以灵活部署。本文提出一种极简的黑箱推理时控制方法——指令重复（instruction duplication），即仅在输入中重复程序性指令（procedural instruction），不改变模型训练或解码逻辑。其核心创新在于：通过简单地将指令复制一次，即可显著提升生成结果的确定性与可预测性。实验表明，在7个指令调优模型上，指令重复使“全8测试通过”（All-8 diagnostic）成功率从90.22%提升至93.17%（+2.95个百分点），并消除30.2%的剩余失败案例；同时，预处理阶段的TF-IDF召回率也显著提高（+1.38点，Holm校正p < 0.001），而最终答案准确率保持不变（60.21%）。尽管存在轻微的过早承诺现象增加（1.52% → 2.30%），但盲测审计显示多数案例仍支持该策略的有效性。更重要的是，在后续应用如答案工程（Answer Engineering, AE）中，指令重复可显著提升系统性能——从原始无编辑状态的25.1%提升至97.1%，充分体现了该方法作为低复杂度、位置敏感控制手段的实用价值，其优势主要通过下游系统对生成轨迹的显式利用得以释放。
链接: https://arxiv.org/abs/2609.04024
作者: Victor Lavrenko(PeaceTech VC, Israel)
机构: PeaceTech VC(和平科技风险投资)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 7 pages, 2 tables. Code and frozen reproduction artifacts: this https URL
Abstract:Procedural instruction following is a basic requirement for controllable language-model systems, especially when generated trajectories are inspected or repaired downstream. We introduce instruction duplication, a minimal black-box inference-time control that repeats only the procedural instruction, without retraining or decoding changes. Across seven instruction-tuned models, 300 medical multiple-choice questions, eight placement conditions, and 16,800 scheduled generations, moving from one to two copies raises the deterministic All-8 diagnostic–responses passing all eight observable tests–from 90.22% to 93.17% (+2.95 percentage points), eliminating 30.2% of the failures remaining after one copy. Pre-provisional TF-IDF recall rises from 73.44% to 74.81% (+1.38 points; Holm-adjusted p .001), while final-answer accuracy remains exactly 60.21%. Premature commitment increases from 1.52% to 2.30% (p_Holm = .00536). A blinded challenge audit yields 10/30 directional confirmations, 20/30 perceptual ties, and no reversals; its prespecified 28/30 confirmation criterion is not met. Yet this distinction can matter operationally when a downstream system acts on the generated trajectory. In Answer Engineering (AE), where explicit trajectory state determines local repair, the published reason-first no-editing SSNHL endpoint was 25.1%; system-only AE was later reproduced at 84.2%, and the same trailing duplicate raised it to 97.1%. For conductive diagnostic branch preservation, the corresponding values are 58.9% published without editing, 78.6% with reproduced AE, and 73.8% with AE plus duplication–a within-AE decrease, but still 14.9 points above the no-editing baseline. Instruction duplication is therefore a low-complexity, placement-sensitive control whose practical value can emerge through the downstream system that consumes the exposed trajectory.
[NLP-12] Representational alignment yields generalizable safety in language models
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在对齐过程中对隐含或对抗性形式的有害意图缺乏鲁棒性的问题。尽管现有对齐方法主要优化可观察的输出响应，但当相同有害意图以人类易识别的陌生或对抗性形式呈现时，模型仍可能失效。其核心解决方案在于引入表征相似性优化（Representational Similarity Optimization, RSO），通过直接对齐模型内部表征与人类道德判断中的原型分类结构，而非依赖生成响应的监督信号。研究发现，当前多数LLMs在道德概念的原型式分类结构上表现薄弱，无法有效区分对立道德类别或保持类别内部的典型性层级；而通过RSO重构模型的道德表征结构，在不显著提升显式判断准确率的情况下，却显著增强了模型在多种攻击策略和基准测试下的对抗鲁棒性。这一结果表明，基于原型的分类机制对行为适应性具有功能性贡献，将该表征原则迁移至LLMs可实现跨规模、跨场景的通用安全性提升。
链接: https://arxiv.org/abs/2609.04022
作者: Lingyu Li,Yan Teng,Yingchun Wang,Xia Hu
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Aligning large language models (LLMs) is essential for their safe deployment. Current alignment methods mainly optimize observable responses, yet models remain vulnerable when the same harmful intent is recast in unfamiliar or adversarial forms that humans can easily recognize. Prototype theory offers an account of this adaptability. Human concepts are represented around central cases, and new instances are categorized according to their graded typicality relative to these prototypes. Here we show that such categorization of moral concepts is weakly preserved in current LLMs. Across 23 LLMs, models often failed to distinguish opposed moral categories or preserve fine-grained typicality within each category. These deficits persist across parameter sizes and alignment stages. We developed representational similarity optimization, which directly aligns the latent representations in LLMs with the categorization expressed in human moral judgements, without supervising generated responses. In matched experiments using the same 251,334 moral annotations, standard behavioral alignment learned the intended moral judgements at the response level while leaving the categorization structure largely unchanged and increasing vulnerability across adversarial evaluations. Reorganizing moral categorization produced more modest gains in explicit judgements but consistently improved adversarial robustness across model scales on diverse benchmarks and attack strategies. Our findings provide functional support for the view that prototype-based categorization contributes to behavioral adaptability. They also show that transferring this representational principle to LLMs yields generalizable safety under adversarial conditions.
[NLP-13] Alignment-Free Text-Audiobox for Voice Dubbing and Full-Duplex Dialogue Synthesis
【速读】： 该论文旨在解决高质量语音配音与全双工对话合成中的核心挑战，包括文本-语音对齐困难、生成长度受限以及情感表达自然性不足等问题。其解决方案的关键在于提出一种无对齐的统一框架Text-AB，通过基于流匹配（flow-matching）训练的扩散变换器（Diffusion Transformer），实现端到端的高保真语音生成。该方法采用DAC-VAE提取的25 Hz潜在序列，将48 kHz波形压缩至超过10倍于以往EnCodec表示的效率，同时提升重合成质量；通过引入跨注意力机制直接处理原始文本输入，无需强制对齐或显式时长预测，实现了真正的无对齐建模；此外，通过在48万小时单语语音上预训练30亿参数模型，并在跨语言配音、全双工对话及情感化对话三个任务上进行微调，显著增强了模型的泛化能力。推理阶段支持一次性生成长达约1分钟的语音，以及通过多扩散机制实现任意长文本生成，并结合多阶段重排序策略优化生成质量。实验表明，Text-AB在真实世界配音基准上显著优于现有内部系统，在韵律相似性、语音相似性、自然度和可分享性方面均有大幅提升；在全双工对话合成中，短对话接近真人录音水平，长对话的人类相似性和表现力显著超越当前最优模型，且天然建模了发言交替、回应信号与情感动态等复杂交互特征；情感条件控制进一步提升了情绪一致性与互动质量。
链接: https://arxiv.org/abs/2609.03992
作者: Sanyuan Chen,Min-Jae Hwang,Sho Inoue,Anna Sun,Bokai Yu,David Kant,Dongmin Hyun,Dorian Desblancs,Gregory Antonovsky,Oleg Repin,Peng-Jen Chen,Xutai Ma,Zehai Tu,Juan Pino,Wei-Ning Hsu
机构: FAIR at Meta(元宇宙人工智能研究院)
类目: Computation and Language (cs.CL); Audio and Speech Processing (eess.AS)
备注:
Abstract:We present Alignment-Free Text-Audiobox (Text-AB), a unified framework for high-quality voice dubbing and full-duplex dialogue synthesis. Building on a Diffusion Transformer trained with a flow-matching objective, Text-AB departs from the Audiobox system along three dimensions. First, it operates in a latent diffusion framework using DAC-VAE features that encode 48 kHz waveforms into a 25 Hz latent sequence, giving over 10x higher compression than previous EnCodec representations while improving resynthesis quality. Second, Text-AB is alignment-free: it consumes raw text via an off-the-shelf text encoder and learns text-speech alignment through cross-attention, removing the need for forced alignment and explicit duration prediction. Third, we scale model and data substantially, pretraining a 3B-parameter model on 480k hours of monolingual speech, followed by supervised fine-tuning on three downstream tasks: cross-lingual voice dubbing, full-duplex dialogue synthesis, and emotional full-duplex dialogue synthesis. At inference, Text-AB supports one-shot generation for up to ~1 min of speech and arbitrarily long-form generation via a multi-diffusion scheme, plus a multi-stage reranking strategy that enhances quality based on automated metrics. On a real-world dubbing benchmark, Text-AB delivers a step-change improvement over the latest internal dubbing system, with large gains in prosody similarity, voice similarity, naturalness, and shareability. For full-duplex dialogue synthesis, it approaches human recordings on short-form conversations and substantially outperforms the latest internal model on long-form human-likeness and expressivity, while natively modeling turn-taking, back-channeling, and emotional dynamics. For emotional dialogue synthesis, emotion conditioning significantly improves emotion alignment and emotional interaction quality over the unconditioned baseline.
[NLP-14] IchthyoNoma: Nomenclature and Context Sensitivity of Zero-Shot Biological Vision–Language Models for Bangladeshi Freshwater Fish Recognition
【速读】： 该论文旨在解决零样本视觉-语言模型（Zero-shot Vision-Language Models, VLMs）在物种识别任务中表现准确率可能受非视觉因素干扰的问题，即报告的准确性未必真实反映模型对生物特征的视觉理解能力。其核心挑战在于揭示模型性能背后多重混杂因素的影响，包括生物学专长、多语言对齐程度、命名体系（如通用名与学名）、提示词设计及上下文语境等。解决方案的关键在于通过系统性审计多个主流VLM模型（如CLIP、BioCLIP、BioCLIP2及Jina CLIP v2），在两个孟加拉国淡水鱼类数据集上进行跨语言、跨命名体系的对比实验，结合多种扰动策略（如模糊、灰度掩码、白屏掩码等）分析模型响应的稳定性与依赖性。研究发现，尽管经过生物领域微调的BioCLIP2在英语名称下表现显著优于通用CLIP，但其在孟加拉语提示下的性能接近随机水平，且仅部分由多语言模型恢复，表明提示语言与命名体系对模型推理具有决定性影响；同时，不同图像扰动对分类结果的影响呈现显著物种依赖性，进一步证明模型性能并非单纯源于视觉知识。因此，该研究强调需综合考量模型架构、训练数据、语言适配与提示工程等多重因素，才能准确评估零样本生物识别模型的真实能力。
链接: https://arxiv.org/abs/2609.03985
作者: Nazim-E-Alam,Tarek Rahman,Md Kishor Morol
机构: American International University Bangladesh; United International University; ELITE Research Lab, Queens, New York, USA
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL)
备注:
Abstract:Zero-shot vision-language models (VLMs) are increasingly used as training-free species recognizers, but reported accuracy can reflect more than visual species knowledge. We audit CLIP, BioCLIP, BioCLIP2, and a multilingual Jina CLIP v2 control on seven freshwater-fish categories from two Bangladeshi sources (10,321 images). BioCLIP2 reaches 72.36% on BFF-15 with English common names and 68.91% on SylFishBD with scientific names, versus 25.15% and 14.40% for generic CLIP. BioCLIP2 Bengali prompts are near chance in balanced accuracy (14.22-14.29%); Jina partially recovers Bengali discrimination to 21.89% and 16.36%, but bare Bengali names return to 14.29% on both sources. Paired SylFishBD interventions show no significant weak-blur effect, modest losses from stronger blur/gray masking, a larger white-mask artifact, and strong species dependence. Zero-shot biological VLM scores therefore jointly reflect biological specialization, multilingual alignment, nomenclature, prompt formulation, and context.
[NLP-15] Investigating the Ability of Large Language Models to Analyze Recipes for Diabetes
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在评估食谱是否适合糖尿病患者方面的准确性与可靠性问题。核心挑战在于模型需从医学文献中检索相关的糖尿病膳食指南，将食谱分解为成分与烹饪方式，并依据这些指导原则判断其适宜性。为此，研究设计了三种不同层次的提示策略：直接查询提示（Direct Query Prompt）、上下文引导提示（Context-Guided Prompt）和范例上下文提示（Exemplary Context Prompt），以考察模型对膳食指南的利用能力。研究构建了一个包含7607个食谱的基准数据集，其中3807个为适合糖尿病患者的食谱，3800个为不适合者。实验结果表明，大多数模型在判定食谱适宜性时表现谨慎，倾向于避免误判可能带来的不良后果；而能够有效运用膳食指南进行推理的模型表现更优。总体而言，Mistral-7B与Llama 70B在性能上优于其他对比模型，其关键优势在于更强的上下文理解与规则推理能力。
链接: https://arxiv.org/abs/2609.03967
作者: Revathy Venkataramanan,Aditya Luthra,Venkatesan Nadimuthu,Amit Sheth
机构: AI Institute, University of South Carolina(南卡罗来纳大学人工智能研究所)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Several studies have evaluated the ability of Large Language Models (LLMs) for meal planning, yielding positive outcomes. These models can process natural language inputs and leverage learned knowledge from their pretraining to generate meal plans. In this work, we investigate the ability of LLMs to analyze the suitability of given recipes for diabetes. The primary challenge for LLMs is to retrieve relevant dietary guidelines for diabetes, decompose recipes into ingredients and cooking methods, and apply these guidelines to determine the recipe’s suitability. To study these challenges, we employ three kinds of prompts namely, (i) Direct Query Prompt (ii) Context-Guided Prompt, and (iii) Exemplary Context Prompt that incorporate different levels of diabetes dietary guidelines from medical sources. We introduce a benchmark dataset curated for this investigation consisting of 7607 recipes that include 3807 recipes suitable for diabetes and 3800 recipes not suitable for diabetes. Our results demonstrate that most LLMs are cautious in predicting recipes as suitable to prevent detrimental outcomes. Further, the models that can reason using the dietary guidelines performed better in predicting the suitability of recipes for diabetes. Overall, Mistral-7B and Llama 70B showed superior performance to their counterparts.
[NLP-16] FiMI Banking: A Sovereign Model for Indian Retail Banking
【速读】： 该论文旨在解决银行领域对话系统在处理产品咨询、账户请求及敏感场景时，因通用语言模型缺乏事实依据、工具使用不准确以及对合规性要求响应不足而带来的可靠性问题。其核心挑战在于如何在严格的操作与监管约束下，实现基于真实信息的精准响应、安全的工具调用以及对高风险情境的审慎处理。解决方案的关键在于构建一个受控的印度零售银行场景（FiMI Banking），该场景基于经过验证的银行文档、结构化真实数据、合成客户背景及银行工具进行构建，并采用两种后训练策略：一是通过偏好优化（preference optimization）提升单轮对话中的安全行为，使超出范围拒绝率从52%提升至80%；二是采用可验证奖励的强化学习（reinforcement learning with verifiable rewards）以优化多轮工具使用任务，显著提升边缘案例表现（从0.509提升至0.718）和对顺序敏感任务的表现（从0.590提升至0.679），同时减少29%的生成令牌消耗。研究表明，偏好优化与可验证奖励强化学习分别针对不同需求，共同构成了可靠银行代理系统的核心支撑。
链接: https://arxiv.org/abs/2609.03960
作者: NPCI AI Research Team:Aman Kumar,Asit Desai,Chandra Bhushan,Harsh Sharma,Harshit Bhushan,Hrithik Kadam,Keyur Doshi,Kolisetty Sai Kapardheeswar,Krishanu Adhikary,Nadeem Shaik,Navya Prakash,Nitin Kukreja,Prashant Devadiga,Shamanth MH,Shantanu Pandey,Suvradip Paul,Yatharth Dedhia
机构: NPCI AI Research Team
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Banks need conversational systems that can answer product questions, assist customers with account-related requests, and operate safely within strict operational and regulatory constraints. General-purpose language models do not reliably meet these requirements. They fall short when a task requires grounded information, correct tool use, or cautious handling of bank-specific sensitive situations. We introduce FiMI Banking, a controlled Indian retail-banking setting. We build it from vetted banking documents, structured ground truth, synthetic customer backgrounds, and banking tools. We evaluate two post-training approaches: preference optimization for response-level behavior, and reinforcement learning with verifiable rewards for multi-turn tool-use tasks. Preference optimization improves safe behavior substantially: out-of-scope refusal rises from 52% to 80%. Reinforcement learning improves edge-case performance from 0.509 to 0.718 and order-sensitive task performance from 0.590 to 0.679, while using 29% fewer generated tokens. These results show that preference optimization and verifiable-reward reinforcement learning address complementary requirements for reliable banking agents.
[NLP-17] wo-Stage Reinforcement Learning for Sound and Adversarial Test Generation in Code LLM s EMNLP2026
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在代码生成任务中因高质量测试用例稀缺而导致的强化学习（Reinforcement Learning, RL）反馈不足问题。由于有效的测试用例需兼具正确性（soundness）与区分度（discriminative power），其人工构造成本高且难以覆盖多样化的错误模式。为此，论文提出一种基于对抗式强化学习的自动测试用例生成框架——测试用例扩展（Test Cases Scaling, TCS）。其核心解决方案在于构建一个两阶段的强化学习机制：第一阶段通过策略对齐的缓冲区生成与参考解一致的测试用例，以确保基础质量；第二阶段则聚焦于当前求解器的失败模式，从缓冲区中筛选并学习生成具有针对性的反例测试用例（counterexample tests），从而增强对错误代码的检测能力。实验表明，TCS在TACO和LiveCodeBench基准上显著提升了pass@1指标及推理阶段的答案选择性能，且所学习的测试生成器还能有效用于甄别其他LLM输出的质量。
链接: https://arxiv.org/abs/2609.03955
作者: Jiacheng Xu,Wentao Zhang,Zhiyi Lyu,Fuxiang Zhang,Chaojie Wang,Yang Liu,Bo An
机构: Nanyang Technological University, Singapore; Skywork AI
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 21 pages, 7 figures. Accepted to Findings of the Association for Computational Linguistics: EMNLP 2026
Abstract:Reinforcement learning (RL) has substantially advanced code generation with large language models (LLMs) through executable feedback. The feedback for coding problems mainly comes from specific test cases, where high-quality test cases are often scarce since they should be both sound and discriminative. We thus turn to study the auto-generation of test cases using the learned model. We find this is naturally an adversarial RL problem: the model is expected to generate effective test cases as counterexamples, depending on the solver’s current failure modes. We propose Test Cases Scaling (TCS), a two-stage RL framework for effective test generation. Both stages train a test generator from a rolling policy-aligned buffer: Stage 1 generates tests consistent with the reference solution, and Stage 2 restricts the buffer to current failure modes and learns counterexample tests. Across TACO and LiveCodeBench, TCS improves both pass@1 and inference-time answer selection according to generated tests. We find the learned test generator also enables effective selection among other LLM outputs.
[NLP-18] Beyond Majority Vote: Multi-Perspective Adjudication for Medical Hallucination Detection EMNLP2026 ACL
【速读】： 该论文旨在解决生成式AI（Generative AI）在长篇医学相关对话机器人输出中事实性错误的检测问题，尤其关注现有单次标注、单个标注者模式下对事实错误的漏检现象。其核心挑战在于：事实性错误常以隐蔽方式嵌入于大量正确文本之中，传统单一视角的标注方法难以有效识别。论文提出一种多视角标注框架，整合首次标注、大语言模型作为裁判（LLM-as-a-Judge, LaJ）的候选错误发现机制，以及两种形式的仲裁——医学专家评审与基于证据的事实核查。研究发现，首次标注者频繁遗漏后续由专业仲裁验证的事实错误；尽管LaJ能提升错误候选的发现效率，但其自身仍存在漏检，且无法完全替代人工判断。此外，不同仲裁者间存在分歧，表明依赖单一来源的仲裁会限制基准测试的完整性，而多源仲裁虽可提升覆盖度，但仍需依赖专家判断与证据支持来确定错误是否存在。将该方法应用于现有基准数据集时，亦揭示出类似漏检模式，表明当前单次通过的幻觉检测基准可能在追求规模的同时低估了真实错误数量。因此，研究强调：在所考察场景下，多轮仲裁虽可改善错误覆盖率，但最终的错误判定仍高度依赖于评判者的专业知识、主观判断及可用证据，从而凸显了构建更可靠评估体系的必要性。
链接: https://arxiv.org/abs/2609.03953
作者: Joe Cecil,Marjorie Freedman
机构: Information Sciences Institute (信息科学研究所); University of Southern California (南加州大学)
类目: Computation and Language (cs.CL)
备注: 34 pages, 6 figures, to be published in Findings of the ACL: EMNLP 2026
Abstract:Understanding the frequency of factual errors in chatbot-generated text and evaluating systems that detect these errors is critical for determining chatbot safety. Yet factual-error detection is often treated as a single-pass, single-annotator labeling problem. In long-form chatbot responses, factual errors can be subtle and embedded within mostly correct text. We develop a multi-perspective annotation study of medically relevant chatbot responses, combining first-pass annotation, LLM-as-a-Judge (LaJ) candidate discovery, and two forms of adjudication: medical-expert and evidence-based fact-checking. First-pass annotators frequently miss factual errors later validated by adjudicators. LaJ improves candidate discovery, but is insufficient on its own: It misses factual errors that annotators catch. We also find disagreement among adjudicators, suggesting that adjudication over multiple candidate sources can improve benchmark completeness, but does not eliminate the need to apply judgment and expertise. Applied to an existing benchmark, this technique reveals a similar pattern of missing annotations. Together, these results suggest that in the settings examined here, single-pass hallucination benchmarks may achieve scale at the cost of undercounting factual errors. Multi-pass adjudication can improve coverage, but inferences drawn from the benchmarks are still sensitive to the judgment, expertise, and evidence used to determine error presence. Comments: 34 pages, 6 figures, to be published in Findings of the ACL: EMNLP 2026 Subjects: Computation and Language (cs.CL) Cite as: arXiv:2609.03953 [cs.CL] (or arXiv:2609.03953v1 [cs.CL] for this version) https://doi.org/10.48550/arXiv.2609.03953 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[NLP-19] VestigeKV: The NoPE-MLA KV Cache Carries Its Own Eviction Signal in a Vestigial Branch
【速读】： 该论文旨在解决长时存续的键值缓存（KV cache）在查询尚未生成时即需压缩的问题，传统基于观测注意力（如H2O、SnapKV）的缓存选择策略在此场景下失效（在NoPE MLA模型上针刺检索准确率仅为0.00–0.33），原因在于此时无法观测到令牌的重要性。其解决方案的关键在于利用Kimil Linear架构中已存在的64维解耦分支——这一源自旋转位置编码（RoPE）的残迹信号，在无旋转训练（NoPE）设定下被重新用作与查询无关的重要度通道。通过仅读取每行的11%数据，系统将缓存划分为两个层级：前m行保留在“已注意层”，其余行则精确迁移至由GPU驻留的归档层，该过程为确定性操作且不删除任何数据。整个方法无需训练、量化或权重/内核修改，计算开销可忽略不计：在8倍和32倍扩展下，检索准确率分别维持在1.00和0.92，与全行选择相比无性能损失。已注意层仅占每标记8.1KB缓存中的0.25KB（32倍扩展时），而归档层保持位级精确并始终驻留于GPU内存，支持主机卸载以释放显存。实验表明，该方法在标准配置下可实现128倍的缓存压缩比且召回率为1.00。若Kimi K3采用类似的无旋转门控多头注意力（NoPE Gated-MLA）结构，该方法可能具备可扩展性，但作者未对其他模型做出主张。此外，该方法具有明确的NoPE依赖性：在标准RoPE MLA架构上，相同操作退化至0.08（普通淘汰策略为0.42），因为仅有无旋转设置下才存在与查询无关的重要度信号（顶级目标占比仅2.3–6.7%，远低于有旋转情况下的10.2–46.8%），且在RoPE框架下无法实现查询无关的精确合并。所有阈值均在数据前冻结，论文附带20个归档判断结果及8条闭合路径验证。
链接: https://arxiv.org/abs/2609.03949
作者: WenJie Fan
机构: Yotta Labs(优塔实验室)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: 12 pages, 4 figures, 10 tables
Abstract:The problem. A long-lived KV cache must be compressed before the queries that will read it exist; selection by observed attention (H2O, SnapKV) collapses there (0.00-0.33 needle retrieval on a NoPE MLA model), because a token’s importance has not yet been observed. The method. On Kimi Linear, VestigeKV evicts by a query-independent signal the cache already carries: the 64-dimensional decoupled branch, a vestige of RoPE that NoPE training repurposes into a salience channel. Reading 11% of each row, it partitions the cache: the top-m rows stay in the attended tier; every other row moves – exactly, never deleted – to a GPU-resident archive reachable per step by a certified trigger. No training, no quantization, no weight or kernel change. Cost. Nothing measurable: retrieval holds at 1.00 under 8x and 0.92 under 32x from 8k to 65k context, zero gap to full-row selection. The attended tier is 0.25 KB of Kimi Linear’s 8.1 KB per-token cache at 32x; the archive stays bit-exact and GPU-resident, with host offload as the VRAM-reclaiming variant. The recall tier – the standard configuration – holds 128x at 1.00. Kimi K3 is reported to use a NoPE Gated-MLA variant; if its cache layout matches, the method plausibly extends there – we make no claim beyond the measured model. NoPE exclusivity. The identical operator on a RoPE MLA collapses to 0.08 (plain eviction: 0.42); query-independent salience itself exists only without rotation (top-1 targets span 2.3-6.7% of tokens vs. 10.2-46.8%), and query-universal exact merging is provably impossible under RoPE. All thresholds were frozen before data; 20 archived verdicts and 8 closed routes accompany the paper.
[NLP-20] More Criticism Does Not Make a Better Review: EquiReview-R
【速读】： 该论文旨在解决当前生成式AI在学术审稿中因过度生成批评而导致的“过审”（overcritique）与“漏审”（omission）问题，即系统可能产生大量但不具实证支持的批评，或遗漏关键缺陷，而现有以生成为导向的模型和聚合评估指标无法区分这两种相反的错误类型。其解决方案的关键在于将AI辅助审稿重构为基于证据的结构化关切集（structured concern set）的迭代精炼过程，明确区分并分别处理“遗漏重要问题”与“过度批评无据之点”两类风险。为此，论文提出EquiReview-R框架：该系统首先基于局部证据解决已有关切，再从独立视角与审稿条件相关视角主动搜索潜在缺失问题，并输出“停止”（stop）、“继续”（continue）或“延迟”（defer）三类决策。通过构建证据关联轨迹语料库（ReviewTrace），研究揭示了修订必须先于进一步搜索的原因——高召回率审稿中的绝大多数关切缺乏确凿证据定论，早期修正机制无法有效处理此类问题。在未见论文的冻结队列上，EquiReview-R在满足主要遗漏问题非劣性标准的同时，将重大过审率从15.5%降至8.1%，且遗漏问题的单侧上界为9.9%，同时在52.4%的论文上选择停止。计算量匹配的对照实验、配对控制及消融分析表明，性能提升源于有效的修订机制，而非额外推理或更短输出。
链接: https://arxiv.org/abs/2609.03943
作者: Zexing Zhang,Jichao Li,Tianyang Lei,Yude Fu,Yang Kewei
机构: National University of Defense Technology (国防科技大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:AI reviewers can now produce many specific criticisms, but more criticism is not necessarily a better review. A review may miss a consequential weakness or retain an allegation that available evidence does not support. These failures require opposite corrections, yet generation-oriented systems and aggregate measures obscure the distinction. We therefore recast AI-assisted review as evidence-guided refinement of a structured concern set, with omission and overcritique treated as separate risks. Building on this formulation, we introduce EquiReview-R, which resolves existing concerns against localized evidence, searches for missing issues from independent and review-conditioned perspectives, and returns stop, continue, or defer. To expose the failure mode that motivates this design, we construct an evidence-linked trajectory corpus. Its retrospective analysis shows why revision must precede further search: nearly all concerns in a high-recall review lack a definitive evidential disposition, while an earlier refinement mechanism cannot revise them. On a frozen cohort of previously unseen papers, EquiReview-R satisfies the prespecified non-inferiority criterion for major omission, reduces major overcritique from 15.5% to 8.1%, and attains a one-sided omission upper bound of 9.9% while stopping on 52.4% of papers. Computation-matched controls, controlled pairs, and ablations show that the gain comes from revision rather than extra inference or shorter output. We release the corpus as ReviewTrace, an evidence-linked resource for studying review revision, disagreement, and provenance.
[NLP-21] Headroom-Drift Replay: A Primitive for Principled Replay Control in GRPO
【速读】： 该论文旨在解决生成式推理模型在基于强化学习（Reinforcement Learning, RL）的后训练过程中，因重复生成新轨迹（fresh rollout generation）而导致的计算瓶颈问题，尤其是在以智能体（Agentic）为核心的场景中，环境交互占据了绝大部分的运行时间开销。现有基于经验回放（replay）的方法通常嵌入复杂的训练流程，如探索机制、经验重构或混合策略优化，导致难以剥离回放本身对性能提升的独立贡献。为此，本文提出一个聚焦性问题：仅依靠原则性（principled）的回放选择能否实现显著效果？其解决方案的关键在于提出一种名为“余量-漂移回放”（Headroom-Drift Replay）的分组级回放控制原语，专用于GRPO（Generalized Reward Policy Optimization）框架。该方法将回放重用拆分为两个独立决策：余量（Headroom） 用于衡量存储轨迹组的剩余学习价值，实现优先级排序；漂移（Drift） 则根据当前策略与历史轨迹之间的兼容性进行门控筛选，确保回放数据的有效性。该方法不改变原有的在线策略流（on-policy stream），也无需引入额外的生成或训练模块。在数学推理、多模态推理及智能体搜索（Agentic Search）等基准测试中，该单一干预措施显著优于朴素回放策略，并在平均指标（Avg Mean@32）上达到或超越更复杂的回放方法；尤其在以环境交互为主导成本的智能体搜索任务中，实现了相近的性能表现，同时大幅降低实际运行时间（wall-clock time）。
链接: https://arxiv.org/abs/2609.03941
作者: Hyun Bin Park,Du-Seong Chang
机构: Sogang University (首尔大学); Seoul, Republic of Korea
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 51 pages, 25 figures, 17 tables. Accepted at COLM 2026
Abstract:RL-based post-training for reasoning models is increasingly bottlenecked by repeated fresh rollout generation, particularly in agentic settings where environment interaction dominates wall-clock cost. Replay can reduce this burden by reusing past trajectories, but existing methods typically embed it within larger training pipelines involving exploration, experience restructuring, or mixed-policy optimization. This makes replay’s own contribution difficult to isolate. We ask a focused question: how far can principled replay selection alone go? We introduce Headroom-Drift Replay, a group-level replay control primitive for GRPO that separates reuse into two decisions. Headroom ranks stored groups by remaining learning value, while Drift gates them by compatibility with the current policy. The fresh on-policy stream remains unchanged, and the method adds no auxiliary generation or training machinery. Across mathematical reasoning, multimodal reasoning, and Agentic Search benchmarks, this single intervention outperforms naive replay and matches or exceeds broader replay methods on Avg Mean@32. In Agentic Search, where environment interaction dominates cost, it delivers comparable quality at materially lower wall-clock time.
[NLP-22] Fixed Suffix Dependency Ratio: Quantifying the Dual-Track Mechanism of Gender Assignment in Latvian Loanwords
【速读】： 该论文旨在解决英语借词在不同目标语言中倾向于集中于阳性性别（masculine gender）这一现象的成因问题，尤其关注固定形态规则与默认分配机制之间的混淆如何阻碍对这一模式根源的准确识别。其解决方案的关键在于提出“固定后缀依赖率”（Fixed Suffix Dependency Ratio, FSDR），用以量化不同性别借词对固定派生后缀的依赖程度，从而区分形态锚定（morphological anchoring）与自由选择（free-choice）在性别分布中的作用。通过对1,832个拉脱维亚名词词元的分析，研究发现借词系统内部存在显著的FSDR不对称性：阴性借词更高度依赖固定派生后缀，而阳性借词则更多处于自由选择区域。这一模式具有借词特异性，并在当代使用中愈发明显。FSDR为检验默认性别分配提供了可量化的框架，揭示了在语言接触条件下，阳性默认性别如何被激活并强化。
链接: https://arxiv.org/abs/2609.03930
作者: Yelingyun Zhang,Atis Kapenieks,Marina Platonova
机构: 未知
类目: Computation and Language (cs.CL)
备注:
Abstract:Existing research has repeatedly observed the tendency for English loanwords to cluster in the masculine gender across different recipient languages, yet the origin of this pattern remains difficult to determine, as fixed morphological rules and default assignments are frequently analysed together. This study proposes the Fixed Suffix Dependency Ratio (FSDR) to quantify the degree of reliance on fixed derivational suffixes across different genders, and to distinguish between morphological anchoring and free-choice in distribution. By examining 1,832 Latvian noun lemma types, the results reveal a significant FSDR asymmetry within the loanword system: feminine loanwords rely significantly more on fixed derivational suffixes, while masculine loanwords are more concentrated in the free-choice zone. This pattern exhibits loanword specificity and has become more pronounced in contemporary usage. FSDR therefore provides a quantitative framework for testing default gender and shows how masculine default can be activated and reinforced under language contact.
[NLP-23] Speak for Me: Giving LLM s the Situational Awareness to Participate in a Meeting EMNLP2026
【速读】： 该论文旨在解决在线会议中大语言模型（LLM）代理在发言时机识别上的根本性缺陷，即缺乏对会议立场（stance）、内容覆盖度（coverage）和发言权（floor）的结构化追踪，导致其无法准确判断何时应主动发言。现有仅依赖提示词（prompt-only）的代理在AMI语料库上对缺席参与者发言机会的沉默率高达51.4%。为应对这一问题，论文提出CAPA（Collaborative Agent Predictive Architecture），其核心在于构建一个四模块协同架构：通过感知器（Perceiver）动态更新会议状态，预测器（Predictor）预判对话走向，控制器（Controller）决策是否发言及选择何种主张，生成器（Generator）以参与方风格输出回应。此外，引入双评判机制（由基于模板约束的LLM评委评估预测与行为）并结合校正器（Recalibrator）利用反馈闭环优化状态表征。该方案采用一种新的基于回合级别的评估协议，可精确衡量代理在真实思想单元（idea unit）周围的发言时机、内容与有效性。实验表明，在137场AMI会议中，CAPA将沉默率从51.4%降至2.5%，使“被认可的补救”（credited recovery）翻倍（26.1→52.2），且幻觉率维持在0.6%。关键发现是：会议状态的精准建模是弥合发言识别差距的核心杠杆，单纯依赖原始上下文扩展无法实现有效识别。
链接: https://arxiv.org/abs/2609.03923
作者: Muneeb Khan,Frederic Kirstein,Terry Ruas,Bela Gipp
机构: University of Göttingen(哥廷根大学), Germany
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: Accepted at EMNLP 2026 Main
Abstract:In online meeting delegation, LLM agents fail to recognize when to speak. With no structured way to track stances, coverage, and floor, they miss the moments where they should contribute. Prompt-only delegates stay silent on 51.4% of the absent participant’s talking opportunities on the AMI corpus. We present CAPA (Collaborative Agent Predictive Architecture), an architecture for online meeting delegation. A Perceiver updates the meeting state from each observed turn. A Predictor forecasts how the conversation will continue. A Controller decides whether to speak and which proposition to surface. A Generator phrases the chosen contribution in the participant’s style. Two judges score the forecast and the action against the next observed turn. A Recalibrator updates the meeting state from those verdicts for future decisions. To evaluate online delegation, we introduce an episode-level protocol that scores whether, when, and what a delegate contributes around the participant’s actual idea units. The protocol’s schema-constrained LLM judges align with human annotations at Cohen’s kappa = 0.71. On 137 AMI meetings, CAPA reduces the silence rate from 51.4% to 2.5%, doubles credited recovery (26.1 – 52.2), and keeps hallucination at 0.6%. The failure mode shifts from omission to selection, with each residual near-miss attributable to a specific module of the architecture. Mechanism ablations identify the meeting state as the lever that closes the recognition gap, where raw-context scaling alone does not.
[NLP-24] CROCODIL: Cross-Model Code Editing with LLM s EMNLP2026
【速读】： 该论文旨在解决多模型环境下生成式代码编辑中存在的过度修改问题。当开发团队使用多个不同的大语言模型（LLM）进行代码生成与编辑时，由于各模型在训练数据和编码风格上的差异，导致某一模型对由其他模型生成的“外来代码”进行编辑时，往往产生过多甚至冗余的修改，影响代码一致性与可维护性。其解决方案的关键在于提出CROCODIL（Cross-model Code Editing with LLMs）——一种基于后训练的框架，通过引入双奖励机制：相似性奖励（similarity reward）以惩罚过大的代码变更，执行成功率奖励（execution reward）用于评估编辑后代码的功能正确性。该框架通过两者的乘积作为优化目标，实现减少编辑幅度的同时保持任务成功率，从而在保证功能正确性的前提下提升跨模型代码编辑的一致性与合理性。
链接: https://arxiv.org/abs/2609.03894
作者: Linghan Zhong,Aditya Thimmaiah,Jayanth Srinivasa,Milos Gligoric,Junyi Jessy Li
机构: The University of Texas at Austin; Cisco Research
类目: Computation and Language (cs.CL); Software Engineering (cs.SE)
备注: EMNLP 2026 Findings
Abstract:Large language models (LLMs) have become ubiquitous tools for code generation and editing. However, development teams often use multiple LLM assistants. Different developers may prefer different models, and individual developers may switch between models across different coding sessions. Because of this, the edits any one model makes are frequently applied to foreign code originally generated by another model. These LLMs are often trained on different datasets, and as a result have different stylistic preferences. Do LLMs behave differently when they edit foreign code originally written by a different LLM with a different coding style? We find that models tend to make more, and often excessive, edits on foreign code. We introduce CROCODIL (Cross-model Code Editing with LLMs), a post-training framework for reducing excessive edits while preserving functional correctness. CROCODIL’s similarity reward penalizes large changes, while its execution reward scores build and test success. We use the product of these two rewards to encourage the policy to decrease the edit size without decreasing the edit task success rate. CROCODIL is available at this https URL.
[NLP-25] Beyond Shallow Alignment: How Post-Training Methods Determine Refusal Circuits And Steering Robustness EMNLP2026
【速读】： 该论文旨在探究用于训练语言模型拒绝有害请求的方法如何影响模型内部拒绝行为的实际计算机制。研究对比了三种后训练方法——监督微调（Supervised Fine-Tuning, SFT）、基于推理链的微调（Reasoning-Augmented Fine-Tuning, RAF）以及偏好优化（ORPO），在三种架构不同的模型（Llama-3.1-8B、Gemma-2-9B、Qwen3-8B）上的表现。研究发现，训练方法本身而非仅数据内容，会重塑模型内部的拒绝计算方式：基于推理链的微调在所有三类模型中均一致地催生出一种独特的拒绝计算模式；而模型架构则独立影响其内部结构及拒绝行为被引导的可靠性。尤为重要的是，所考察的任何方法均未能同时实现理想安全对齐的三个关键属性：拒绝行为不集中于少数脆弱组件、安全性能提升不以牺牲通用能力为代价、且安全行为可通过小规模、精准的修改进行纠正。研究警示，不应将当前的后训练方法视为已解决且可靠的防御手段，尤其在安全敏感的应用场景中需保持审慎。
链接: https://arxiv.org/abs/2609.03887
作者: Hoang Cuong Nguyen,Mark Dras,Usman Naseem
机构: Macquarie University (麦考瑞大学)
类目: Computation and Language (cs.CL)
备注: 27 pages, accepted at EMNLP 2026 Main Conference
Abstract:How do the methods used to train language models to refuse harmful requests shape how that refusal actually works inside the model? We compare three post-training methods - supervised fine-tuning, reasoning-augmented fine-tuning (training on reasoning chains that justify a safety decision), and preference optimization (ORPO) - across three architecturally distinct models (Llama-3.1-8B, Gemma-2-9B, Qwen3-8B). We find that training method, not just data, reshapes how refusal is computed internally: reasoning-augmented training consistently produces a distinct kind of refusal computation, visible across all three models, while architecture independently shapes internal structure and how reliably refusal can be steered. Most importantly, no method we study achieves all three properties we would want from safe alignment at once: refusal that isn’t concentrated in a few fragile components, safety gains that don’t cost general capability, and safety behavior correctable through small, targeted edits. We caution against treating current post-training methods as a solved, reliable defense, especially for security-critical use. Code and models are available in this https URL.
[NLP-26] Flip Dont Shuffle: Watermarking LLM s at the Speed of Inference EMNLP2026
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）中水印技术在计算效率与可扩展性方面的瓶颈问题，特别是现有方法在生成过程中因状态管理或复杂结构导致的高延迟与内存开销。其核心挑战在于如何在保证水印检测统计性能的前提下，实现低延迟、无状态的实时水印嵌入。解决方案的关键在于提出一种无状态伯努利水印（Stateless Bernoulli Watermarking, SBW），通过每个令牌独立进行伯努利试验决定是否属于绿色列表（green list），仅需对基于计数器的随机数生成器执行一次比较操作，将成员判定复杂度降至常数时间 $ O(1) $，并支持单核内核执行与零中间内存分配。该设计不仅保持了与固定大小绿色列表相同的检测保障——即在零假设下 z-score 服从标准正态分布 $ \mathcal{N}(0,1) $，还首次实现了全词汇自盐（full-vocabulary self-salt）水印，速度较 KGW 提升超过 6000 倍，且兼容分布式推理架构。此外，研究揭示哈希函数设计是影响水印质量的未被充分探索维度，采用 GPU 友好的 Jenkins 哈希可使零假设校准性能提升 1.8 倍，并生成更丰富的文本内容。多组实验验证了不同种子方案与参数配置下的统计等价性，接收者操作特征曲线下面积（ROC-AUC）差异低于 0.01。
链接: https://arxiv.org/abs/2609.03844
作者: Simone Ceppi,Ignacio Sanchez
机构: European Commission (欧洲委员会); Joint Research Centre (联合研究中心); Ispra, Italy (伊斯普拉, 意大利)
类目: Cryptography and Security (cs.CR); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: Accepted at EMNLP 2026 Main Conference
Abstract:We introduce Stateless Bernoulli Watermarking (SBW), a new statistical watermark for Large Language Models that determines green list membership through independent per-token Bernoulli trials. Unlike KGW’s vocabulary permutation or SynthID’s multi-layer tournament, SBW requires only a single comparison per token against a counter-based random number generator, reducing membership complexity to O(1) and enabling single-kernel execution with zero intermediate allocations. We prove that this formulation preserves the same detection guarantees as fixed-size green lists: the z-score test remains \mathcalN(0,1) under the null. The stateless architecture enables capabilities unavailable to existing methods: full-vocabulary self-salt watermarking (over 6000 \times faster than KGW’s self-salt and 2 \times faster than SynthID despite biasing the entire vocabulary with candidate-dependent seeding) and architectural compatibility with distributed inference. In end-to-end generation benchmarks, SBW adds less than 1% overhead at all batch sizes. We additionally identify hash function design as a previously unexplored axis for watermark quality, showing that a GPU-native Jenkins hash improves null calibration by 1.8 \times while producing more diverse text. Experiments across two seeding schemes and eight (\gamma, \delta) configurations confirm statistical equivalence with ROC-AUC differences below 0.01.
[NLP-27] Select Compress Reinvest: A Controlled Study of Visual-Token Allocation in Long-Video MLLM s
【速读】： 该论文旨在解决长视频语言模型在处理高帧率视频时面临的效率与性能权衡问题，即如何在有限计算资源下有效选择最具信息量的视频帧以提升模型问答准确率。其核心挑战在于：尽管长视频（如每秒采样1帧的1小时视频包含3600帧）数据量巨大，但系统只能保留少量固定帧，而当前主流方法将帧选择视为预处理环节，未深入探究其对最终性能的影响。论文的关键解决方案在于系统性地分离并量化三个关键决策变量——帧选择策略、空间压缩（spatial compression）和节省资源的再投资（reinvestment），在不改变其他因素的前提下进行独立评估。研究发现，帧选择是影响性能的最关键因素：通过查询驱动的选择机制，仅用8个精选帧即可超越16个均匀分布帧的性能（提升6.9分）；且未经修改的“正交匹配追踪”（Orthogonal Matching Pursuit）算法在所有基准测试中表现接近甚至优于专门设计的选取器。此外，空间压缩几乎无成本（帧分辨率减半仅损失≤0.44分），而真正实现性能增益的关键在于将压缩节省下来的计算资源用于增加更多压缩后帧的数量——即“再投资”，此举可额外提升2至3分，证明压缩的价值必须通过资源重分配才能释放。研究还揭示了跨论文比较存在的严重偏差，强调必须在统一可控的评估框架内进行实验，以避免因实现细节或评测工具差异导致结果不可靠。
链接: https://arxiv.org/abs/2609.03820
作者: Prakhar Khatri
机构: 独立研究者(Independent Researcher)
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL)
备注: 16 pages, 6 figures. Code and data: this https URL
Abstract:Long-video language models cannot look at every frame: an hour sampled once per second is 3,600 images, and a system keeps only a small fixed slice of that pool. Which frames survive that slice is usually treated as a preprocessing detail; we test whether it should be. Published selectors make the comparison hard because they change the frame scorer, the prompt boundary, the resolution policy, and the answering model all at once. We hold each fixed and vary one decision at a time: selection, spatial compression, and reinvestment of the savings, across six training-free selection rules, three long-video benchmarks, and two answering models. Selection is the largest single lever: on LongVideoBench’s hour-long bin, eight query-selected frames beat sixteen uniformly spaced ones by 6.9 points, and Orthogonal Matching Pursuit, an unmodified decades-old sparse-approximation algorithm, matches or comes within a point of every purpose-built selector we compare it against, across all three benchmarks. Compression is close to free: halving each frame’s spatial budget at fixed timestamps costs at most 0.44 points. Reinvestment is where that budget turns back into accuracy: spending the freed tokens on twice as many compressed frames, at a measured cost no higher than the original eight, returns a further two to three points; compression only pays off once its savings are spent this way. Along the way, an implementation bug in our own AKS baseline and a 0.07 to 3.74 point gap between two harnesses running the same published rules at the same budget show why these comparisons need to happen inside one controlled harness rather than across papers.
[NLP-28] Evaluating Criterion-Conditioned Behaviour of Large Language Models in Content Moderation EMNLP
【速读】： 该论文旨在解决当前大型语言模型（LLM）在内容审核任务中表现看似良好，但其实际决策机制缺乏对具体审核标准（criterion）的独立判断能力的问题。现有基准测试通常将多种审核标准合并为单一标签，导致无法准确评估模型是否能够区分并正确应用各个独立的审核维度。为此，论文提出了一种名为“诊断性内容审核评估”（Diagnostic Evaluation of COntent, DECO）的新方法，通过解耦内容中的关键因素，实现对不同审核标准的可控、细粒度评估；同时引入成对比较评估（pairwise evaluation），以对比模型在相同输入下对不同标准的输出表现。实验结果表明，尽管模型在聚合标签上表现优异，但在具体标准层面存在显著失效，尤其当正确判断依赖于特定内容属性而非整体有害性时，模型表现尤为薄弱。这揭示了当前内容审核基准的局限性：高聚合性能不能充分证明模型具备可靠的标准条件化（criterion-conditioned）判断能力。因此，研究强调亟需发展能明确衡量模型在各审核标准下行为表现的新型评估方法。
链接: https://arxiv.org/abs/2609.03814
作者: Danting Zhang,Bei Peng,Robert Loftin
机构: University of Sheffield(谢菲尔德大学)
类目: Computation and Language (cs.CL)
备注: Accepted by EMNLP Findings 2026
Abstract:Large language models (LLMs) demonstrate strong performance on standard content moderation benchmarks. However, these benchmarks often aggregate multiple moderation criteria into a single label, making it unclear whether models can disentangle them and reliably apply each criterion when making decisions. To study whether LLMs exhibit criterion-conditioned behaviour, we introduce Diagnostic Evaluation of COntent (DECO), a criterion-independent factorisation of content that enables controlled, criterion-level evaluation. We also introduce pairwise evaluation to compare model outputs across different criteria for the same input. Across four moderation datasets and four LLMs, we find that strong benchmark performance can hide substantial failures at the criterion level. Models struggle most when correct decisions depend not on overall harmfulness, but on the specific aspect of the content that the criterion requires them to assess. Our results highlight a key limitation of current content moderation benchmarks: strong performance on aggregated labels does not provide sufficient evidence that LLMs can reliably evaluate content with respect to individual moderation criteria. These findings call for the development of evaluation methods that explicitly measure criterion-conditioned behaviour.
[NLP-29] VisCAD: A Foundation Model Suite with Multimodal Industrial CAD Intelligence
【速读】： 该论文旨在解决工业产品设计中生成式AI辅助计算机辅助设计（CAD）所面临的两大核心挑战：部件级生成需将用户多样化的意图输入（如渲染图、文本描述、二维图纸及真实照片）映射为可执行的CAD领域特定语言程序，而装配级生成则需处理零件间的相互作用、规划配合关系、估计位姿并正确放置所有部件。现有专用CAD模型通常仅在有限输入域（如渲染图或文本）上训练，泛化能力差；而通用前沿模型虽覆盖更广的输入类型，但在CAD领域表现不一致。本文提出VisCAD基础模型系列，通过兼顾广泛泛化能力与强CAD性能，实现对真实工业产品的高效生成。其核心是270亿参数的VisCAD-M1模型，采用中段训练与后训练策略优化部件级设计生成能力，在PubCADBench和RealCADBench基准上达到0.5540的平均得分，优于最强前沿模型的0.5496；通过复用VisCAD-M1作为测试时验证器，得分进一步提升至0.5797，相较此前最优水平相对提高约5%。此外，VisCAD还集成领域特定的装配生成框架，利用前沿模型完成复杂装配生成，在定量与定性评估中均显著优于通用框架。关键解决方案在于构建一个具备多模态输入理解能力、高精度生成能力及可扩展验证机制的统一基础模型体系。
链接: https://arxiv.org/abs/2609.03811
作者: JoyIndustrial VisCAD Team:Linxin Cai,Qiuhe Hong,Zhichao Huang,Guanlin Li,Hongsen Liu,Ziqi Liu,Yichen Long,Luya Wang,Yuchen Wang,Wenxiang Wu,Huimu Yu,Ning Zhang
机构: JoyIndustrial VisCAD Team
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL)
备注: Technical report from JoyIndustrial’s AI CAD project
Abstract:AI-assisted computer-aided design (CAD) for industrial products involves two challenging phases. Part-level generation maps diverse forms of user intent, including renders, text descriptions, 2D drawings, and real photographs, to executable programs in a CAD domain-specific language. Assembly-level generation must additionally handle interacting parts, plan mating relations, estimate poses, and place all parts correctly. Existing specialized CAD models are commonly trained on narrow input domains, such as renders or texts, and often generalize poorly, while general-purpose frontier models cover broader inputs but perform inconsistently across CAD domains. We present VisCAD, a foundation model suite designed to provide both broad generalization and strong CAD capability for realistic industrial products. At its core is VisCAD-M1, a 27B model trained through mid-training and post-training for part-level design generation. On PubCADBench and RealCADBench, VisCAD-M1 achieves the highest average part-level score among the evaluated models, reaching 0.5540 compared with 0.5496 for the strongest frontier model. Reusing VisCAD-M1 as a test-time verifier can further raise the score to 0.5797, an approximately 5 percent relative improvement over the previous state of the art. VisCAD also includes a domain-specific harness that leverages frontier models for complex assembly generation and demonstrates advantages over general-purpose harnesses in both quantitative and qualitative evaluations.
[NLP-30] A Reverse Sign Language Dictionary: Open-Vocabulary Sign Recognition from Continuous Signing via Video Captioning and Description Retrieval
【速读】： 该论文旨在解决传统孤立手语识别（Isolated Sign Language Recognition, ISLR）方法在开放词汇场景下无法泛化至训练中未见手语动作的问题，其核心局限在于依赖封闭集分类框架与词典标注的词汇表，导致部署时受限于预定义的词汇条目。为此，本文提出一种无需词典标注的开放式手语识别方案：首先利用一个开放权重的视觉-语言模型对连续手语中的符号片段进行自由形式的动作描述生成（即“手语描述化”），其次通过多语言句子编码器从目标描述词汇库中检索最匹配的条目，构建了一个无需词典监督的逆向手语词典。该方法的关键创新在于将手语识别转化为基于自然语言描述的开放词汇匹配任务，从而实现对未见手语的识别能力。实验结果表明，在日语手语（JSL）对话语料库上，经过微调后，该系统在已知类别上的Top-10检索准确率从4.5%提升至49%，接近标准监督式封闭集分类器（I3D）性能；更重要的是，在未见类别上仍显著优于原始未训练管道（Top-10达11.5%-21.0%，p=0.0094），而传统封闭集分类器在此场景下无法参与。此外，匹配端的上限分析显示句子编码器已能恢复近100%的同义句金标描述，揭示当前性能瓶颈主要源于描述生成质量，为后续优化指明方向。本研究是首个不依赖词典标注、基于描述的连续手语开放词汇检索方法，也是首个针对日语手语的此类工作。
链接: https://arxiv.org/abs/2609.03788
作者: Santiago Poveda-Gutiérrez,Hideki Nakayama,Mayumi Bono
机构: The University of Tokyo (东京大学); National Institute of Informatics (国立情報学研究所)
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL)
备注: 4 pages, 2 figures, 1 table. Extended version of an abstract presented at the BU-SHI workshop (Broadening the Users: A Cross-Disciplinary Roadmap for Social Humanoid Interaction), IEEE RO-MAN 2026, Kitakyushu, Japan, 28 August 2026. The workshop is non-archival; no proceedings
Abstract:Isolated Sign Language Recognition (ISLR) is conventionally cast as closed-set classification over gloss labels, which cannot generalize to signs unseen in training and ties every deployment to a gloss-annotated lexicon. We instead recognize signs extracted from continuous signing by (1) captioning a sign-level clip into a free-form procedural description of the articulation with an open-weight vision-language model, and (2) retrieving the closest entry from a vocabulary of target descriptions with a multilingual sentence encoder: a reverse sign language dictionary that needs no gloss supervision and admits an open vocabulary. On 1,300 sign-level segments from a Japanese Sign Language (JSL) dialogue corpus annotated with procedural descriptions (against a 2% top-10 chance floor over the 503-entry target vocabulary), fine-tuning the captioner substantially improves seen-class retrieval: language and vision tower fine-tuning raises top-10 retrieval on seen classes from 4.5% (untrained) to 49%, becoming statistically indistinguishable from a standard supervised closed-set classifier (I3D) on two of the three test sets where a closed-set classifier can be evaluated at all. More importantly, unseen-class retrieval also improves significantly over the untrained pipeline (11.5% - 21.0% top-10, p=0.0094), a regime in which the closed-set classifier cannot participate. A matcher-side empirical upper-bound analysis shows the sentence encoder already recovers close to 100% of paraphrased gold descriptions, locating a gap in captioning quality that we aim to address in future work. To our knowledge this is the first description-based, open-vocabulary sign lookup from continuous signing without gloss supervision, and the first for JSL.
[NLP-31] IndicSafeEval: Safety Robustness of Large Language Models under Multilingual Persuasive Jailbreak Attacks EMNLP2026
【速读】： 该论文旨在解决当前大型语言模型（Large Language Models, LLMs）安全评估高度依赖英语环境，而忽视低资源及文化多样性语言中对齐失败表现的问题。现有评估体系难以揭示模型在非英语语境下的真实安全边界，尤其在面对多语言、多文化场景时存在显著盲区。为此，论文提出IndicSafeEval——一个针对印度语言的基于说服策略的越狱攻击评估框架，其核心创新在于将10类安全敏感内容类别与6种人类典型的说服策略相结合，覆盖印地语、孟加拉语、马拉地语和旁遮普语等四种印度主要语言，构建了包含7200个对抗性提示的基准测试集。关键解决方案在于引入“说服性线索”作为攻击手段，系统性考察模型在不同语言、不同说服风格及不同风险类别下的安全行为差异。实验结果表明，模型的安全表现并非均质化，而是显著受语言类型和话术设计的影响；同时，各类有害内容对说服型越狱的敏感度不一，凸显出当前以英语为中心的安全评估范式的局限性。研究强调必须发展具备多语言适应性和说服意识的评测框架，才能更真实地反映生成式AI（Generative AI）在多元文化现实场景中的安全性。
链接: https://arxiv.org/abs/2609.03781
作者: Saikat Mondal,Mamta,Deeksha Varshney,Oana Cocarascu,Asif Ekbal
机构: Indian Institute of Technology Jodhpur(印度理工学院乔德普尔分校); King’s College London(伦敦国王学院); Indian Institute of Technology Patna(印度理工学院巴特纳分校)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 38 pages, 7 figures, 33 tables. Accepted to Findings of EMNLP 2026. Contains examples of harmful model outputs
Abstract:Large language models (LLMs) are increasingly used in multilingual settings, yet their safety is still evaluated primarily in English. This limits our understanding of how alignment failures manifest in low-resource and culturally diverse languages. We introduce IndicSafeEval, a persuasion-based jailbreak evaluation framework for Indian languages. Our benchmark combines ten safety critical content categories with six human-like persuasive strategies across four different Indian languages, such as Hindi, Bengali, Marathi and Punjabi, resulting in 7,200 adversarial prompts. We conduct a systematic black-box evaluation of several open-source LLMs to examine how their safety behaviour varies across languages, persuasion strategies, and risk categories. Our analysis shows that the model does not behave equally safely across all languages and prompt styles. Instead, safety performance depends strongly on both the languages used and the way a request is phrased using persuasive cues. We further observe that different risk categories exhibit different levels of vulnerability, with some types of harmful content being significantly more susceptible to persuasion-based jailbreaks than others. These findings reveal important limitations of current safety evaluations, which are largely English-centric, and underscore the need for multilingual and persuasion-aware benchmarking frameworks to more accurately assess real-world LLM safety. Our implementation is available at this https URL. Warning: this paper contains example data that may be offensive or harmful.
[NLP-32] ypological Feature Prediction with Large Language Models : An In-Context Learning Approach EMNLP2026
【速读】： 该论文旨在解决多语言自然语言处理中类型学特征（Typological Features）缺失值预测的可解释性不足与跨资源水平及特征类型性能表现不均衡的问题。现有方法在预测时缺乏透明的推理依据，且其在低资源语言上的表现尚不明确。为此，研究提出基于上下文学习（in-context learning）的方案，利用URIEL+和Glottolog提供的语言学数据，借助大语言模型（LLM）在元语言推理和生成解释性理由方面的能力，实现对类型学特征的预测。关键创新在于引入语言的谱系关系与地理邻近信息作为上下文证据，显著提升了预测性能，尤其在低资源语言上未出现劣势，同时多数LLM生成的推理理由与输入证据保持一致，为实现可解释的类型学特征预测提供了有效路径。
链接: https://arxiv.org/abs/2609.03775
作者: Qianwen Wang,York Hay Ng,Aditya Khan,En-Shiun Annie Lee
机构: University of Toronto (多伦多大学); Ontario Tech University (安大略理工大学)
类目: Computation and Language (cs.CL)
备注: Accepted to EMNLP 2026
Abstract:Typological features are widely used in multilingual NLP, and the prediction of such features holds downstream utility. However, existing methods to predict missing values lack interpretable justifications for predictions, while their performance across resource levels and feature types remains underexplored. Given LLMs’ abilities in meta-linguistic reasoning and in providing rationales, we investigate LLMs’ performance in typological feature prediction via an in-context learning approach with linguistic data from URIEL+ and Glottolog. We find that zero-shot prompting is insufficient, but when given phylogenetic and geographic neighbour evidence, LLMs substantially outperform all baselines without disadvantaging low-resource languages. We further find that most LLM rationales are consistent with the provided evidence, offering a step toward explainable typological feature prediction.
[NLP-33] RealCADBench: Benchmarking Parametric CAD Modeling from Industrial Design Intents
【速读】： 该论文旨在解决现有参数化计算机辅助设计（CAD）建模评估缺乏综合性指标的问题，尤其针对当前基准测试在合成环境、仅支持特定输入模态或仅依赖可执行性与交并比（IoU）等单一评价维度上的局限性。其核心解决方案是提出RealCADBench，一个基于真实工业设计意图的意图到程序（intent-to-program）CAD建模基准，涵盖来自19类工厂自动化任务的12,632个任务，支持文本描述、二维工程图、实物产品图像及渲染图像等多种输入形式，覆盖零件（Part）与装配体（Assembly）建模。评估采用统一的FreeCAD API Python生成与共享运行时执行机制，并从可执行性、实体交并比（Solid IoU）、表面交并比（Surface IoU）以及基于评分表的视觉-语义一致性判别器（Judge）四个维度进行综合评价。实验结果表明，九个独立前沿大模型中无一在四项指标上全面领先，且不同模型在各指标表现上存在显著差异；在关键的装配体任务集RCB-Assm25上，虽Codex结合GPT-5.5提升了可执行性与IoU，但视觉-语义一致性得分下降6.98个百分点，凸显出当前模型在保留精细结构、保持部件身份和正确装配定位等方面的系统性缺陷。研究揭示：仅以可执行性衡量真实场景下的CAD建模能力具有严重不足，而先进模型与智能体在可执行性、几何相似性与视觉-语义一致性之间表现出显著权衡，强调了多维度评估体系的必要性。
链接: https://arxiv.org/abs/2609.03773
作者: JoyIndustrial VisCAD Team:Linxin Cai,Qiuhe Hong,Zhichao Huang,Guanlin Li,Zongzhen Li,Hongsen Liu,Yichen Long,Wei Wang,Yuchen Wang,Dongyue Yang,Huimu Yu,Xianwen Zhong
机构: JoyIndustrial-VisCAD Team
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL)
备注: Benchmark from JoyIndustrial’s AI CAD project
Abstract:Parametric computer-aided design (CAD) modeling is difficult to evaluate with a single metric. Existing CAD benchmarks often emphasize synthetic or CAD-native settings, limited input modalities, or executability and IoUs alone. We introduce RealCADBench, a benchmark for intent-to-program CAD modeling from real industrial design intents. It contains 12,632 tasks from 19 factory-automation categories and spans text descriptions, 2D engineering drawings, real product pictures, and rendered images for both Part and Assembly modeling. We report results on a 1,770-task evaluation slice: 1,745 Part tasks across four input regimes and RCB-Assm25, a 25-task assembly study used in every reported assembly comparison. Each method generates FreeCAD API Python, which a shared runtime executes to export the 3D model. We evaluate the exported model using executability, Solid IoU, Surface IoU, and a rubric-based visual-semantic identity Judge. Among the nine standalone frontier large models evaluated, no model leads all four metrics. Across six frontier-scale large models, executability ranges from 0.565 to 0.812, Solid IoU from 0.2841 to 0.5379, and Surface IoU from 0.112 to 0.217 across the four Part regimes. The highest regime-balanced composite comes from a different model than the leaders on the four component metrics. On RCB-Assm25, Codex with GPT-5.5 improves executability and both IoU metrics over standalone GPT-5.5, but lowers the Judge score by 6.98 percentage points, leaving GPT-5.5 as the Judge leader. We also observe recurring failure modes, most notably missing fine structures, loss of part identity, and incorrect assembly placement. These results show that execution alone is insufficient to characterize realistic CAD modeling and that frontier models and agents differ substantially across executability, IoUs, and visual-semantic identity.
[NLP-34] OBER: Continuity-Aware Reporting and Traceable Continuous Improvement in Outcome-Based Education
【速读】： 该论文旨在解决基于学习成果（Outcome-based Education, OBE）的教育机构在计算学习成果达成度时，其数据处理过程缺乏透明性与可解释性的问题，尤其关注如何将测量到的成果差距（shortfall）转化为可操作的纠正措施。其核心解决方案是提出OBER+系统，通过五个相互关联的阶段实现从成果差距识别到纠正决策的自动化推导：包括跨课程周期累积成果达成度、检测短期与持续性差距、依据监管方已用的阈值对差距进行分级、将决策记录至带有证据标注的实践清单、追踪变更日志，并量化后续差距变化。此外，引入一条关键规则，确保在学习成果定义发生变更的节点上，不将成果数据视为连续序列，从而避免因成果重定义导致的误判。实证应用显示，某核心课程在连续授课中所有成果均被实质性重构，内容在不同编号间转移，若采用传统方法将导致对成果“下降25分”的错误解读；经重新计算后发现，平台原有数据中有六项与四舍五入误差不符，揭示出系统缺陷；在十五组成果陈述对比中，五组完全一致，其余十组中，有六组当前编号成果与先前不同编号成果在语义上最为接近，该结果基于相似性排序，无需设定阈值或标签。该研究贡献在于提供了一套可移植的计算规则框架，使任何成果达成度平台均可实现对成果动态变化的准确建模与可视化，显著提升数据驱动决策的可靠性与透明度。
链接: https://arxiv.org/abs/2609.03770
作者: Elakkiya Rajasekar
机构: Birla Institute of Technology and Science Pilani, Dubai Campus (比尔拉科技与科学学院皮兰尼，迪拜校区)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL); Computers and Society (cs.CY)
备注: 14 pages, 6 figures, 7 tables. Submitted to IEEE Transactions on Learning Technologies
Abstract:Institutions practising outcome-based education compute learning outcome attainment routinely, while reviews of curriculum analytics report an absence of evidence on how that computation informs decisions. This paper presents OBER+, an extension of a deployed institutional attainment platform that computes the step from a measured shortfall to an evaluated corrective action. Five connected stages accumulate attainment across deliveries of a course, signal a shortfall and a persistent shortfall, grade it on cutoffs the regulator already uses, record the decision against a catalogue of practices annotated with their evidence, log the change, and quantify the subsequent movement in the shortfall. A further rule compares successive statements of an outcome, so attainment is never read as a series across a point at which the outcome changed. Applying the rules to the live record of two real courses produced three results. Every outcome of a core course was substantively redefined between consecutive deliveries, with subject matter moving between outcome numbers, so a naive reading would have reported a twenty-five point collapse between quantities that do not refer to the same learning. Recomputing the platform’s figures from its documented rule showed six of ten differing by more than rounding explains, in a pattern that identified a defect since reported to the institution. Across fifteen statement pairs from three transitions, five were identical character for character, and among the ten that were not, the outcome carrying a given number was nearest to a differently numbered earlier outcome in six, a result resting on an ordering of similarities and requiring no threshold and no labelling. The contribution is a computational design for outcome-based reporting, stated as rules any attainment platform can implement, with evidence of what they make visible in a live institutional record.
[NLP-35] Rent-a-RAG : Embedding-Space Watermarks for Auditing Third-Party RAG EMNLP2026
【速读】： 该论文旨在解决第三方检索增强生成（RAG）市场中数据提供方难以审计其文档是否被未经授权重复使用的问题。由于RAG运营商不合作、生成答案存在语义改写且多来源证据混合，传统审计方法失效。其核心解决方案是提出DirBucket——一种基于语义水印的黑盒审计框架，通过在文档中嵌入保持语义一致性的改写句式，并使这些改写句的嵌入向量偏向特定于提供方的“桶-密钥”方向，从而实现对文档级重用行为的可检测性。该方法在仅需黑盒访问的情况下，可在复杂混合来源场景下实现高精度目标检测，且无误报；在主基准测试中，所有23次审计均成功识别违规行为，且水印对恶意后处理（如答案清洗）具有鲁棒性，现有规避策略无法同时逃避检测并维持用户感知的答案质量。此外，该检测能力可跨真实临床、网络安全威胁情报和法律领域数据集迁移，表明嵌入空间水印可使第三方RAG中的文档重用行为具备统计可审计性。
链接: https://arxiv.org/abs/2609.03749
作者: Alexandr Goultiaev Tolstokorov,Kyriakos Mouratidis,Javad Dogani,Nikolaos Laoutaris
机构: IMDEA Networks Institute(IMDEA网络研究所), Madrid, Spain; Universidad Carlos III de Madrid(卡洛斯三世大学), Madrid, Spain; Singapore Management University(新加坡管理大学), Singapore
类目: Cryptography and Security (cs.CR); Computation and Language (cs.CL)
备注: Accepted to EMNLP 2026 Main Conference
Abstract:Third-party retrieval-augmented generation (RAG) marketplaces create a new auditing problem: data providers may license corpora to a RAG operator, yet later have no visibility into whether their documents are being reused without compensation. Auditing this misuse is difficult because the operator is non-cooperative, answers are paraphrased by the generator, and one response may combine evidence from many providers. We propose DirBucket, a provider-side semantic watermarking and black-box auditing framework for document-level reuse in multi-provider RAG. DirBucket watermarks documents by meaning-preserving paraphrases whose embeddings are biased toward provider-bucket secret directions, enabling detection from black-box answers while preserving retrieval utility. On a challenging benchmark that reflects mixed-provider reuse under black-box access, DirBucket is the only method that consistently achieves strong target detection with no non-target activation, detecting non-compliance in every audit within 23 audited answers on our primary benchmark. The watermark survives adversarial post-answer laundering, and none of the evaluated evasion strategies simultaneously defeats detection while preserving user-perceived answer quality. Detection transfers unchanged to a second benchmark built from real clinical, cyber-threat-intelligence, and legal provider corpora. These results suggest that embedding-space watermarking can make document reuse in third-party RAG statistically auditable.
[NLP-36] KnowVis: Knowledge-Centric Visual Summarization for Video Lectures EMNLP2026
【速读】： 该论文旨在解决视频讲座在教学过程中因信息密集且线性呈现，与学习者构建关联性认知网络的内在认知需求之间存在根本性不匹配的问题。这一不匹配导致缺乏领域先验知识的初学者面临严重的认知负荷。现有视频摘要方法多生成以文本为主的线性压缩内容，仍需较高的认知努力，无法有效缓解此问题。其解决方案的关键在于提出KnowVis框架，该框架通过从多模态视频内容中提取详细的概念图，识别关键且具有挑战性的阈值概念（threshold concepts），构建结构化的知识单元，并最终合成具教学意义的视觉化叙事摘要。该框架不仅实现了从线性视频到可视化知识网络的转化，还通过引入包含125个跨10个学科领域的教育视频及1,079个生成视觉摘要的专用数据集，经自动化评估与用户研究验证，显著提升了摘要的准确性、清晰度，有效降低了认知负荷，并大幅增强了学习者的知识获取与记忆效果。
链接: https://arxiv.org/abs/2609.03742
作者: Yi Xu,Yifan Hou,Xiaoyu Zhang
机构: City University of Hong Kong(香港城市大学); ETH Zürich(苏黎世联邦理工学院)
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL)
备注: This work is published on EMNLP 2026 (Findings). Our code and dataset are available at this https URL and this https URL
Abstract:Video lectures are valuable educational resources, but their dense and lengthy formats often overwhelm novice learners. This difficulty stems from a fundamental pedagogical mismatch: while videos deliver transient information linearly, human learning requires constructing interconnected cognitive networks, a task that induces severe cognitive overload for novice learners lacking prior domain knowledge. Existing video summarization methods fail to resolve this mismatch, as they primarily produce text-heavy, linear condensations that still demand high cognitive effort. To bridge this gap, we propose KnowVis, a framework that transforms linear video lectures into pedagogically grounded visual narratives. KnowVis first extracts a detailed concept map from multimodal video content to identify important and challenging threshold concepts, then constructs structured knowledge units, and finally synthesizes engaging visual summaries. Alongside the framework, we introduce a curated dataset of 125 educational videos across 10 academic disciplines, paired with 1,079 generated visual summaries. Extensive automated evaluations and a human study demonstrate that, compared to state-of-the-art baselines, KnowVis generates more accurate and clear visuals that successfully reduce cognitive load and significantly improve student learning effectiveness and knowledge retention.
[NLP-37] Beyond BLEU: A Case for Redefining Sign Language Translation Benchmarks
【速读】： 该论文旨在解决当前手语翻译（SLT）评估中依赖BLEU-4指标所存在的根本性问题：传统基于口语语言的评价指标无法有效反映手语理解的真实水平。在多模态、低资源的SLT场景下，模型可能利用虚假相关性及口语先验知识，而非真正学习到更具表征能力的手语表示。为此，本文提出一种受语言学习评估启发的替代方案——基于开源大语言模型（LLM）的问答（QA）协议，通过衡量关键内容保留程度来评估翻译质量。该方法与人工评分更具一致性，且相比BLEU-4具备六至七倍更强的重述鲁棒性。应用于实际SLT任务时，该协议揭示了更真实的模型性能图景：在Phoenix-2014T数据集上，五种无词元（gloss-free）系统表现相近，处于噪声范围内，而有词元监督的系统则显著领先9.3分，这一差距在BLEU-4指标下完全被掩盖。因此，该研究的关键在于引入一种更贴近人类认知、对内容传递敏感且抗训练-测试数据重叠干扰的新型评估范式。
链接: https://arxiv.org/abs/2609.03734
作者: Oline Ranum,Edward Fish,Simon Hadfield,Richard Bowden
机构: Centre for Vision, Speech and Signal Processing (CVSSP), University of Surrey(萨里大学视觉、语音与信号处理中心); Guildford, GU2 7XH, Surrey, UK; o.ranum@surrey.ac.uk, s.hadfield@surrey.ac.uk, r.bowden@surrey.ac.uk, ed@tavus.io
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:BLEU-4 is the standard metric for evaluating sign language translation (SLT), but spoken-language metrics may not adequately reflect sign language proficiency. The multimodal, low-resource context of SLT allows models to exploit spurious correlations and spoken-language priors, rather than learning stronger sign representations. In this paper, we evaluate the relationship between spatio-temporal understanding and BLEU-4 across six SLT models on Phoenix-2014T and CSL-Daily, showing that gains in BLEU-4 are not on their own evidence of better sign language understanding. This work introduces an alternative inspired by language-learning assessment, using an open-weight-LLM QA protocol that measures salient content preservation. It aligns more closely with human rankings and is six to seven times more paraphrase-invariant than BLEU-4. Applied to SLT, this protocol targets content transfer, is more robust to train-test overlap, and gives a different picture of the field: the five gloss-free systems are largely within noise of one another on Phoenix-2014T, while the gloss-supervised system stands 9.3 points higher, a gap invisible to BLEU-4.
[NLP-38] Opening mind by opening architecture: analysis strategies
【速读】： 该论文旨在解决数字市场工具日益普及所导致的电声创作中特定发展与研究环境逐步丧失的问题，进而使得封闭式架构音频处理器模型占据主导地位。这一模型虽具备强大处理能力，却以牺牲对内部处理过程及交互系统的透明性为代价，使系统演变为复杂且不可见的“黑箱”，造成技术可解释性与研究可追溯性的缺失。其解决方案的关键在于通过重构科学的研究方法与学术态度，借助文献学视角重新审视信号处理技术的历史脉络与实现逻辑。具体而言，论文以曼弗雷德·施罗德（Manfred Schroeder）的经典混响算法为案例，构建了用于分析与评估的工具体系，并实现其实践应用，从而推动建立一种有意识、可追溯、具历史纵深感的数字信号处理研究路径。
链接: https://arxiv.org/abs/2609.03719
作者: Francesco Vitucci,Giuseppe Silvi,Daniele Giuseppe Annese,Francesco Scagliola,Anthony Di Furia
机构: Conservatorio “N. Piccinni” - Bari(巴里尼科洛·皮奇尼音乐学院)
类目: Computation and Language (cs.CL)
备注: Presented at the 7th International Csound Conference (ICSC 2024), Vienna, September 2024
Abstract:In numerical signal processing for electroacoustic composition, the progressive loss of specific development and research environments caused by the increasing use of digital market tools has favoured the dominance of the closed-architecture audio processor model. This model, while powerful, envisions the possibility of describing output data about its perceived characteristics, but at the cost of ignoring its internal process and interacting systems, which become complex, powerful environments but closed in an inscrutable black box, a loss we must consider. Any digital signal processing technique tells a story. Just as the words of a language incorporate social, historical and technical polysemic layers, a signal processor has its own story of implementation, a gradual technological achievement with its inevitable aesthetic consequences. Through the looking-glass of literature, one can access those environments with renewed awareness by reestablishing a scientific method and an attitude to research. In this specific case, starting from the case study of Manfred Schroeder’s historical reverbs, we illustrate the process of building analytical evaluation tools, as well as practical implementation, at the basis of a conscious study path.
[NLP-39] What Do CAE Simulation Agents Really Need Beyond a Generic Harness?
【速读】： 该论文旨在解决生成式人工智能在计算机辅助工程（CAE）仿真任务中如何高效、准确地将自然语言指令转化为可执行仿真流程的核心问题。尽管现有基于大语言模型（LLM）的CAE代理已引入多智能体分解、领域知识检索和脚本化反思等专用机制，但这些设计主要服务于能力较弱的基础模型。随着现代通用智能体框架已具备多轮推理、工具调用及执行反馈等关键能力，论文探究了在信息访问与修复预算固定的前提下，是否仍需复杂的专用架构。研究发现，单一智能体通用框架在性能上可媲美甚至超越多智能体专用系统（FoamBench得分从88.2%提升至96.4%），其关键优势在于执行反馈驱动的修复能力——仅通过引入执行反馈修复机制，仿真成功率即从71.8%显著提升至96.4%；而脚本化反思则未带来额外增益。唯一仍具显著作用的因素是领域知识输入，特别是以求解器教程形式提供的专业知识，其带来的性能提升最为明显（从80.9%增至96.4%）。因此，解决方案的关键在于充分利用通用智能体框架的内在能力（尤其是执行反馈闭环），并有效注入高质量的领域知识，而非依赖复杂的多智能体结构或冗余的反思机制。
链接: https://arxiv.org/abs/2609.03718
作者: Jiasheng Shi,Tianhan Zhang
机构: DP Technology; Beihang University (北京航空航天大学); AI for Science Institute; Beijing, China
类目: Computational Engineering, Finance, and Science (cs.CE); Computation and Language (cs.CL); Computational Physics (physics.comp-ph)
备注:
Abstract:Computer-aided engineering (CAE) simulation is among the largest and most demanding areas of engineering, where setting up a solver such as OpenFOAM, FEniCS, or COMSOL takes real expertise. Large language model (LLM) agents promise to turn a natural-language request into a working simulation, and recent CAE agents add simulation-specific machinery: multi-agent decomposition, domain retrieval, and scripted reflection. That machinery suited weak base models; modern harnesses already supply multi-turn reasoning, tool use, and execution feedback. We ask what a CAE simulation agent still needs beyond a generic harness. With information access and repair budget held fixed, a single-agent harness matches or beats multi-agent specialized systems (FoamBench 96.4% vs.\ 88.2%). Ablations trace this to capabilities the harness already provides: execution-feedback repair lifts FoamBench from 71.8% with no repair round to 96.4%, while scripted reflection adds nothing. The one input that still helps is domain knowledge supplied as solver tutorials, our largest measured gain (80.9% to 96.4%). Subjects: Computational Engineering, Finance, and Science (cs.CE); Computation and Language (cs.CL); Computational Physics (physics.comp-ph) Cite as: arXiv:2609.03718 [cs.CE] (or arXiv:2609.03718v1 [cs.CE] for this version) https://doi.org/10.48550/arXiv.2609.03718 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[NLP-40] A Circuit for Plural Reference: How LLM s Represent and Retrieve Singular and Plural Entities
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在上下文推理中对单数与复数实体的指代消解（coreference resolution）机制问题，特别是针对复数指代现象的理解与生成。其核心挑战在于揭示模型如何在输入中表征并检索复数实体，以准确预测代词所指向的先行词。解决方案的关键在于结合机械可解释性（mechanistic interpretability）与注意力模式分析（attention pattern analysis），通过一系列因果干预技术识别出一组关键注意力头（attention heads），这些注意力头分别负责：（1）在输入中表征指代信息；（2）识别构成复数指代的实体组；（3）将相关信息传递至负责选择先行词并预测代词的模块。研究还发现，LLMs 在复数代词偏好上与人类认知一致，即当实体在本体论层面相似且由连词“and”连接时，更倾向于将其视为复数实体进行指代。
链接: https://arxiv.org/abs/2609.03687
作者: Anh Danh,Rick Nouwen,Massimo Poesio
机构: Utrecht University (乌得勒支大学); Queen Mary University of London (伦敦玛丽女王大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Coreference resolution is an important task in contextual reasoning. In this paper, we investigate the mechanism for representing and retrieving singular and plural entities for plural reference. We use a combination of mechanistic interpretability and attention pattern analysis to study the process in which LLMs predict a pronoun to refer back to previously mentioned entities. Using a range of causal intervention techniques, we find a set of attention heads that are responsible for (1) representing coreference information in the input, (2) identifying entities that form a plural reference, (3) transferring the information to the component that is responsible for selecting the antecedents and predicting the pronoun. We also find that LLMs align with humans in preference for plural pronoun. Specifically, entities in a plural construction are more likely to be referred to as a plural entity if they are ontologically similar and are linked by the conjunction “and”.
[NLP-41] Understanding Autonomous Driving Datasets by Describing Differences between Image Subsets in Natural Language
【速读】： 该论文旨在解决大规模自动驾驶数据集在跨域场景下因分布偏移（domain shift）导致模型性能退化的问题，核心挑战在于现有数据分析方法依赖元数据、预定义标签或人工检查，难以提供充分的语义洞察且不具备可扩展性。为此，论文提出“集合差异描述”（set difference captioning）任务，即给定两个图像子集，生成自然语言描述其差异的假设。解决方案的关键在于采用两阶段框架，并聚焦于基于目标检测提取的以对象为中心的图像块（object-centric patches），通过对象级粒度的特征表示简化差异聚合过程，实现对特定对象实例或类别差异的可追溯性归因。为评估该方法在真实场景中的有效性，研究构建了首个针对自动驾驶场景的基准测试集AD-Diff Bench，通过低浓度实验验证了该方法在稀疏、真实差异下的适用性。实验限定使用开源权重模型以保障可复现性和部署便捷性，最终为自动驾驶数据集提供了可解释性强、面向人类理解的数据内省（dataset introspection）新范式。
链接: https://arxiv.org/abs/2609.03677
作者: Julian Truetsch,Felix Hauser,Christoph Stiller,Frank Bieder
机构: FZI Research Center for Information Technology (弗劳恩霍夫信息科技研究中心); Karlsruhe Institute of Technology (卡尔斯鲁厄理工学院)
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL); Machine Learning (cs.LG); Neural and Evolutionary Computing (cs.NE); Robotics (cs.RO)
备注: 9 pages, 5 figures, submitted to the IEEE Open Journal of Intelligent Transportation Systems (OJ-ITS), our implementation and benchmark dataset are available at this https URL
Abstract:Understanding the composition of large-scale autonomous driving datasets is essential for safety, robustness, and reliable operation across domains. For example, domain shift between locations could lead to the operating environment being misaligned with the training data, resulting in potentially dangerous performance degradation. Yet, existing data analysis pipelines largely rely on metadata, predefined labels, or manual inspection, which provide limited semantic insight or do not scale. This paper studies set difference captioning: given two subsets of images, the goal is to produce a natural-language hypothesis describing differences between the target and reference set. Building on a two-stage formulation, we adapt the method to autonomous driving by focusing on object-centric patches derived from object detection, which simplifies aggregation and enables attribution of differences to specific object instances or categories. To evaluate this setting in-domain, we introduce a new benchmark, AD-Diff Bench. Low-concentration experiments assess the suitability of set-difference-captioning approaches to sparse, real-world differences. We restrict our experiments to open-weight models to support reproducibility and ease of deployment. The proposed benchmark and analysis provide a step towards practical, human-interpretable dataset introspection for autonomous driving datasets. Our implementation and benchmark dataset are available at this https URL
[NLP-42] he Impact of Synthetic Data Augmentation on Discourse-Prag matic Function Classification
【速读】： 该论文旨在解决自然语言处理中因类别不平衡导致的模型性能下降问题，特别是在话语语用功能分类这一数据稀疏性具有结构性特征的任务中。现有合成数据增强方法多关注生成样本的数量与多样性，而忽视了其在表示空间中与真实数据的几何关系。本文以英语单词“look”在英国国家语料库中的410个手动标注实例为基础，涵盖四种语用功能（注意力信号、指令、话语标记、感叹），利用Llama 3.1生成合成训练样本，并基于RoBERTa嵌入空间中的余弦距离将其划分为不同区域。通过对比六种训练条件，考察合成样本相对于真实数据经验决策边界的分布位置对模型性能的影响，同时保持增广数量恒定。结果表明，所有增广条件均优于仅使用真实数据的基线，其中靠近真实数据的近邻样本（NEAR）在宏平均F1值上提升最大（+0.113），而距离均衡混合策略达到最高准确率（0.748）。然而，各条件均未改善AUC值，说明数据增强主要影响决策边界的位置，而非优化模型的概率估计能力。因此，解决方案的关键在于：合成样本在表示空间中的位置——即其与真实数据的几何关系——与生成数量同等重要，这对低资源语用分类任务具有重要意义。
链接: https://arxiv.org/abs/2609.03652
作者: Sara Sorahi,Kevin Tang,Reza Kazemian
机构: Institute of Linguistics; Department of English Language and Linguistics; Department of Linguistics, College of Liberal Arts and Sciences, University of Florida; Faculty of Arts and Humanities, Heinrich Heine University Düsseldorf; Sun Yat-sen University, China
类目: Computation and Language (cs.CL)
备注:
Abstract:Synthetic data augmentation has become a common strategy for addressing class imbalance in NLP, but most approaches focus on the quantity and diversity of generated examples rather than their geometric relationship to real training data. We investigate this question in the context of discourse pragmatic function classification, a task where data sparsity is a structural feature rather than a collection artefact. Using 410 manually annotated instances of the English word look drawn from the British National Corpus, spanning four functions: Attention Signal, Directive, Discourse Marker, and Interjection. We generate synthetic training examples with Llama 3.1 and partition them by their cosine distance from real training data in RoBERTa embedding space. We compare six training conditions that differ in the placement of synthetic examples relative to the empirical decision boundary, while holding augmentation quantity constant across conditions. All augmented conditions improve macro F and accuracy over the real only baseline, but core proximal examples (NEAR) yield the largest gains in macro F (0.113), while a distance balanced mix achieves the highest accuracy (0.748). No condition improves AUC, indicating that augmentation shifts the decision boundary rather than improving the model’s underlying probability estimates. These findings suggest that where synthetic examples land in representation space matters as much as how many are generated, with implications for low resource pragmatic classification more broadly.
[NLP-43] /think Doesnt Stop Reasoning : Analysis of Spurious CoT Termination EMNLP2026
【速读】： 该论文旨在解决生成式大模型在复杂任务中采用链式思维（Chain-of-thought, CoT）推理时产生的冗长且重复的推理轨迹问题。现有无训练早期退出（training-free early-exit）方法通过在中间节点插入“思考结束标记”（end-of-think token, EoT）来触发从推理到作答的转换，以缩短推理过程。然而，研究发现，注入的EoT并不总能有效引导模型进入干净的作答阶段：在模型重新生成新的EoT之前，仍可能持续产生具有推理特征的文本内容，且该段落长度与早期退出所节省的推理令牌数呈正相关，表现出持续推理的行为，称为“虚假的CoT终止”（spurious CoT termination）。作者提出假设，即模型对注入的EoT关注度不足是导致该现象的原因，并引入“退出标记注意力偏置”（Exit-token Attention Biasing, EAB）机制，通过增强模型对EoT的关注度来缓解此问题。在四种大推理模型、五个基准测试及两种早期退出方法上的实验表明，提升对EoT的注意力可显著减少虚假终止现象和作答阶段的生成长度。研究揭示了仅通过外部形式匹配（如插入EoT）难以可靠控制大推理模型行为的局限性——尽管插入EoT符合显式的“思考-作答”结构，但其本身不足以保证预期的推理到作答的转换。
链接: https://arxiv.org/abs/2609.03633
作者: Seunghee Koh,Sungjae Choi,Minchan Kwon,Sunghyun Baek,Junmo Kim
机构: Korea Advanced Institute of Science and Technology (韩国科学技术院)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Accepted to EMNLP 2026 Main Conference
Abstract:Chain-of-thought (CoT) reasoning improves large reasoning models (LRMs) on complex tasks but often produces long, redundant traces. Recent training-free early-exit methods shorten these traces by choosing an intermediate point to stop reasoning. We study one such strategy that injects an end-of-think token (EoT, /think) at this point to trigger the reasoning-to-answering transition, and find that the injected EoT does not always induce a clean answering phase. Answering-phase generation can continue before the model regenerates another EoT, with the span preceding this regenerated EoT scaling with the reasoning tokens saved by early exit and exhibiting continued reasoning behavior. We call this spurious CoT termination, where reasoning-like generation continues into the answering phase. We hypothesize that insufficient attention to the injected EoT contributes to spurious CoT termination and probe this hypothesis with Exit-token Attention Biasing (EAB). Across four LRMs, five benchmarks, and two early-exit methods, increasing attention to the injected EoT reduces spurious CoT termination and answering-phase length. These results reveal a limitation of controlling LRMs by externally matching their explicit think-block format. Inserting the EoT token conforms to this format but does not by itself guarantee the intended reasoning-to-answering transition. Our code is available at this https URL.
[NLP-44] Remember and Reweight: Enhancing Multi-Agent Debate with Experience Memory and Confidence Estimation EMNLP2026
【速读】： 该论文旨在解决多智能体辩论（Multi-Agent Debate, MAD）中存在的关键缺陷——共享误解（shared misconception）问题，即当多数智能体初始阶段错误地达成一致时，辩论过程反而会强化这一错误，导致推理偏差加剧。现有方法虽能缓解同伴偏倚（peer skew），但未触及智能体固有的概念先验偏差。为此，论文提出R²-MAD（Remember and Reweight for Multi-Agent Debate）框架，其核心在于为智能体引入基于过往辩论经验的记忆机制。该方案通过两个互补机制实现系统性改进：首先，基于当前共识水平的辩论状态感知检索策略，动态调校概念先验，从历史经验中召回相关证据；其次，利用这些召回的过往经验评估各智能体的可靠性，生成置信权重以调节同伴影响。实验结果表明，R²-MAD在多个基准测试上均显著优于现有的单智能体及MAD基线方法。
链接: https://arxiv.org/abs/2609.03619
作者: Xuanfa Jin,Zhijian Ma,Yongcheng Zeng,Xinyu Cui,Haifeng Zhang,Jun Wang
机构: Institute of Automation, Chinese Academy of Sciences (中国科学院自动化研究所); University of Chinese Academy of Sciences (中国科学院大学); University College London (伦敦大学学院)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: EMNLP 2026 Findings, 24 pages, 4 figures
Abstract:Multi-agent debate (MAD) improves the reasoning capabilities of large language models by having multiple agents iteratively refine their responses through discussion. However, MAD suffers from a critical vulnerability known as shared misconception: when a majority of agents initially converge on an incorrect answer, the debate process tends to amplify rather than correct the error. Existing methods primarily address peer skew but leave the agents’ inherently biased concept priors unaddressed. To mitigate this systematic weakness, we propose R ^2 -MAD (Remember and Reweight for Multi-Agent Debate), a framework that equips agents with an experience memory accumulated from past debates. R ^2 -MAD intervenes on both failure modes through two complementary mechanisms: A debate-state-aware retrieval policy dynamically calibrates the concept prior by retrieving relevant historical evidence based on the current consensus level. Then these retrieved experiences provide a basis for estimating per-agent reliability, yielding confidence weights to modulate peer influence. Experiments on various benchmarks show that R ^2 -MAD achieves consistent improvements over existing single-agent and MAD baselines.
[NLP-45] KhatianDoc: A Human-Verified Benchmark Diagnosing Multimodal LLM Failure on Bengali Legal Land Records EMNLP2026
【速读】： 该论文旨在解决孟加拉国土地所有权记录中一种基于十六进制位置分数系统（Ana-Ganda-Kora-Kranti-Til）的手写文件——RS Khatian——在机器可读性方面的根本性挑战。这类文件是数百万地块的权威产权凭证，常涉及民事诉讼，但现有技术体系缺乏对这类文本的识别与理解能力，尤其在光学字符识别（OCR）、分词器（tokenizer）及主流大模型（LLM）处理方面均无覆盖。其核心解决方案是提出KhatianDoc，一个包含四个任务的基准测试：符号识别、十六进制到十进制转换、结构化字段提取以及法律文档问答（共1,634个问答对）。该基准基于孟加拉国曼什甘杰地区Vumi地籍办公室的真实RS Khatian记录构建，通过人工转录并经土地法专业人士验证达成一致，同时采用位置标记实现匿名化以保留多跳问题所需的指代关系。实验评估了六种多模态大模型（8B至72B+，含开源与闭源），在固定零样本协议下，5个问答类别中39.3%的问题被所有模型完全错误回答；在算术转换任务中，所有生成数值的模型表现均低于恒定均值基线，且精确匹配与近似匹配得分重合，表明并非近似能力不足，而是存在显著的“去相关”现象，即模型无法正确建模原始数据结构。此外，对自身评估指标的审计揭示了两个相反方向的偏差：修正了拒绝评分缺陷并报告修正后结果，同时指出某一元数据指标存在高估风险并作为上限标注。研究结论强调，当前系统并非性能差距，而是根本能力缺失，而本研究提供了经验证的高质量真实标签数据，为未来系统研发奠定基础。代码与数据已公开，图像数据经脱敏处理发布。
链接: https://arxiv.org/abs/2609.03597
作者: Tasmiad Hasan,Arafat Zaman Ratul,Sarker Sadman Saalim,S.M. Shah Nawaz Hossain,Khan Raiyan Ibne Reza,Sumaiya Tabassum Nimi
机构: North South University (北方南大学)
类目: Computation and Language (cs.CL)
备注: 12 pages, 5 figures, 11 tables, NLLP Workshop @ EMNLP 2026. Dataset: this https URL
Abstract:Land ownership in Bangladesh is recorded in Ana-Ganda-Kora-Kranti-Til, a base-16 positional fraction system with dedicated Unicode glyphs, no mainstream font, and no coverage in any OCR pipeline or tokenizer. The handwritten records that carry these fractions, RS Khatians, are the authoritative title record for millions of parcels and a frequent subject of civil litigation, yet no benchmark has asked whether a machine can read one. We introduce KhatianDoc, a four-task benchmark built from 107 real RS Khatian records from the Vumi (land) Office of Munshiganj, Bangladesh: symbol recognition, base-16-to-decimal conversion, structured field extraction, and legal document question answering over 1,634 QA pairs. Ground truth was transcribed by hand, verified by a land-law practitioner to full agreement, and anonymized through positional tokens that keep the referential distinctions multi-hop questions depend on. We evaluate six multimodal LLMs (8B to 72B+, open and closed) under a fixed zero-shot protocol. Five QA categories, 39.3% of our stratified set, return zero correct answers from every model; on the arithmetic task, every model that emits a number does worse than a constant-mean baseline, with exact- and near-match scores coinciding: decorrelation, not approximation. Auditing our own metrics surfaced two artifacts in opposite directions: we correct a refusal-scoring bug and report the fixed scores beside the originals, and flag an inflated metadata metric as an upper bound. KhatianDoc documents not a performance gap but the absence of a capability, with verified ground truth for future systems. Code and data, with a redacted image release, are publicly available.
[NLP-46] How Far Can Synthetic Data Take Thai OCR?
【速读】： 该论文旨在解决合成光学字符识别（OCR）监督信号在真实泰语文档上的迁移能力不足的问题，核心挑战在于合成数据与真实文档之间在文本上下文、版面结构、字体多样性及字形变化等方面的差异。其解决方案的关键在于通过受控的文档重建流程，对影响迁移性能的多个因素（如非文本上下文、字体多样性、二维空间结构、真实手写字形等）进行解耦分析，并系统评估各因素在不同训练粒度（页面级与裁剪级）下的影响。研究发现，非文本上下文对迁移效果影响较小，而字体多样性、二维结构以及真实手写字形显著提升迁移性能；此外，源域匹配的有效性依赖于训练粒度：在页面级训练中，使用同域重建的合成数据可接近真实印刷文档监督效果（中位字符错误率从1.82%降至1.31%），但在裁剪级训练中则表现逊于跨域重建数据（15.59% vs. 5.52%）。基于上述洞察，研究者利用45,723张合成页面，将0.9B参数的PaddleOCR-VL-1.6模型微调为Wayu-Paxa-OCR-Zero，实现了无需真实标签的泰语文档端到端适配，在印刷体和手写体文档上分别将中位字符错误率从6.64%降至1.24%、从74.87%降至20.55%，且在五个评测集上均优于Typhoon OCR v1 7B，证明了纯合成数据训练在特定场景下具备与真实标注竞争的能力。
链接: https://arxiv.org/abs/2609.03595
作者: Kunat Pipatanakul
机构: Wayu Research(威尤研究); Paxa Labs(帕克萨实验室)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注: 20 pages, technical report
Abstract:We investigate what makes synthetic OCR supervision transfer to real Thai documents and use the resulting insights to build Wayu-Paxa-OCR-Zero, a Thai OCR model adapted without OCR labels from real Thai document pages. Synthetic data provide exact labels at scale, but “realism” conflates source domain, page context, typography, spatial structure, and glyph variation. We disentangle these factors with a controlled document-reconstruction pipeline and evaluate each variant under page- and crop-level training on printed and handwritten Thai documents. Non-text context has little consistent effect, whereas typeface diversity, two-dimensional structure, and real handwriting glyphs improve transfer; moreover, source-domain matching depends on training granularity, with in-domain reconstruction approaching real printed supervision under page-level training (1.82% versus 1.31% median character error rate) but underperforming out-of-domain reconstruction under crop-level training (15.59% versus 5.52%). Guided by these findings, we adapt the 0.9B-parameter PaddleOCR-VL-1.6 into Wayu-Paxa-OCR-Zero using 45,723 synthetic pages: relative to its base checkpoint, it reduces median character error rate from 6.64% to 1.24% on printed pages and from 74.87% to 20.55% on handwriting and outperforms Typhoon OCR v1 7B on all five evaluation sets, showing that synthetic-only training can be competitive.
[NLP-47] HalluPeer: A Taxonomy-driven Benchmark for Detecting Hallucinations in Scientific Peer Reviews EMNLP
【速读】： 该论文旨在解决生成式AI在学术同行评审中因幻觉（hallucination）问题导致评审可靠性下降的核心挑战。随着大规模同行评审需求的增长，大型语言模型（Large Language Models, LLMs）被用作评审助手，但其生成内容虽流畅却缺乏事实依据，尤其在需要基于长篇技术文献进行验证的同行评审场景中尤为突出。现有幻觉检测基准不适用于此类专业语境，无法有效识别与定位评审中的虚假陈述。为此，作者提出HalluPeer——首个专为科学同行评审设计的幻觉检测基准，通过构建包含论文内容、人工撰写的评审意见及注入幻觉的对比性评审三元组的数据集，并对幻觉进行检测、分类与定位标注，建立了一套面向同行评审场景的幻觉分类体系。其解决方案的关键在于：构建一个具备领域适配性的幻觉生成与标注流程，结合自动化过滤机制，在12,000篇论文和38,000份评审样本上验证了现有检测方法难以区分幻觉与合理批评，且真实评审中确实存在由HalluPeer定义的典型幻觉模式，凸显了在评审过程中引入源感知验证（source-aware verification）的必要性。
链接: https://arxiv.org/abs/2609.03580
作者: Tzu-Ling Lin,Dong-Ting Yao,Teng-Fang Hsiao,Wei-Chih Chen,Hong-Han Shuai
机构: National Yang Ming Chiao Tung University (国立阳明交通大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: Accepted to EMNLP Findings 2026
Abstract:The growing scale of academic peer review has motivated the use of Large Language Models (LLMs) as review assistants, yet LLMs can generate fluent but unsupported claims that undermine review reliability. Existing hallucination benchmarks are not designed for peer review, where verification requires grounding claims in long, technical papers. We introduce HalluPeer, a benchmark for detecting hallucinations in scientific peer reviews, providing aligned triples of paper content, human-written reviews, and hallucination-injected reviews, annotated for detection, classification, and localization. Our pipeline induces a peer-review-specific hallucination taxonomy, identifies review contexts, and injects hallucinations with automated filtering. Experiments on 12K papers and 38K reviews show that existing detectors struggle to separate hallucinations from legitimate critique, while evaluation on authentic reviews demonstrates that HalluPeer-defined hallucination patterns occur in real peer reviews, highlighting the critical need for source-aware verification. Our project page can be found in this https URL
[NLP-48] Language Language Models and What Were Talking About
【速读】： 该论文旨在探讨当前自然语言处理（Natural Language Processing, NLP）中语言模型的本质与目标定位问题，核心关切在于：当语言模型基于翻译数据和合成数据进行训练与微调，并在同样非自然的语言数据上进行评估时，其所构建的“语言”表征是否真正反映了真实语言系统，尤其是意大利语的语言特性。其关键解决方案在于提出一个根本性区分——即应明确区分以技术产品为导向的语言模型与以语言研究为目的的语言模型。前者追求性能优化与应用落地，后者则致力于揭示语言结构与使用规律。通过这一区分，论文主张，语言模型所应生成的语言类型并非单一固定，而是取决于其设计初衷；因此，对“我们希望语言模型产出何种语言”的回答将因目标不同而多样，从而避免对当前语言模型发展态势的过度悲观判断。
链接: https://arxiv.org/abs/2609.03577
作者: Malvina Nissim
机构: 未知
类目: Computation and Language (cs.CL)
备注: In Proceedings of the Twelfth Italian Conference on Computational Linguistics (CLiC-it 2026)
Abstract:Language models are commonly discussed as technical artefacts, but they are obviously shaped by the linguistic worlds conveyed by data during their training. Using Italian language models as evidence, I want to bring attention to the nature of the systems which result from training and specialising models on translated and synthetic data, and further curating them, and to the meaning of testing them on equally unnatural data. Are these eventually models of Italian? Are they models of language? Does NLP still care about language? These questions yield another, more concrete question: what language do we actually want language models to produce? I argue that this question cannot be answered if we do not first consider a clearer distinction between language models designed as technical products and language models designed as tools for studying language itself. The answers then might be diverse, the languages we are talking about might be diverse, and the picture might not be as pessimistic as we fear.
[NLP-49] Lost in Reordering: Structural Sensitivity of Multilingual LLM s under Semantics-Preserving Perturbations EMNLP2026
【速读】： 该论文旨在解决多语言大语言模型（Multilingual Large Language Models, LLMs）在面对语义保持但句法结构变化时的鲁棒性不足问题，尤其关注语序相对自由的语言（如印地语和马拉雅拉姆语）。其核心挑战在于：尽管现有模型在跨语言推理任务中表现优异，但其性能在遭遇语义不变但结构变异（如短语重新排序、主动-被动语态转换）时显著下降。为此，研究提出一个基于语言学原理的基准数据集IndicReStruct，包含两个变体GSM8K-Reordered与GSM8K-Voice，均源自GSM8K并严格保留原始语义。实验结果表明，在六种先进多语言LLM及多种提示策略下，结构扰动输入导致数学推理性能出现系统性且显著的退化。通过定性错误分析与残差流激活修补（residual-stream activation patching）等可解释性实验发现，推理失败主要源于实体-数量对齐的破坏，且中间层Transformer层在恢复推理能力方面起关键作用。研究表明，当前多语言大模型对表层句法实现仍高度敏感，缺乏在结构不同但语义等价输入下的组合不变性（compositional invariance），揭示了其在真实世界语言多样性中的脆弱性。
链接: https://arxiv.org/abs/2609.03511
作者: Karthika Nhayakkat,Rajat Verma,Maharaj Brahma,Vetcha Gnana Mahesh,Maunendra Sankar Desarkar,Ganesh Ramakrishnan,Rohit Saluja
机构: 未知
类目: Computation and Language (cs.CL)
备注: Accepted at EMNLP 2026 (Findings - Long paper)
Abstract:Large Language Models (LLMs) demonstrate strong multilingual reasoning performance, yet their robustness to semantics-preserving structural variation remains underexplored, particularly for relatively free word-order languages. We investigate the structural sensitivity of multilingual LLMs using two linguistically grounded perturbation settings in Hindi and Malayalam: constrained constituent reordering and active-passive voice transformation. We introduce a benchmark dataset IndicReStruct, with two variants, GSM8K-Reordered and GSM8K-Voice, constructed from GSM8K while preserving semantic meaning. Across six state-of-the-art LLMs and multiple prompting strategies, we observe consistent and significant degradation in mathematical reasoning performance under structurally perturbed inputs. To further understand these failures, we perform qualitative error analysis and mechanistic interpretability experiments using residual-stream activation patching. Our analyses show that reasoning failures frequently arise from disruptions in entity-quantity alignment and that intermediate transformer layers contribute most strongly toward reasoning restoration. Overall, our findings suggest that current multilingual LLMs remain highly sensitive to surface syntactic realization and lack robust compositional invariance under structurally different but semantically equivalent inputs.
[NLP-50] Building and Evaluating Fixed-Voice Thai TTS from Synthetic Speech
【速读】： 该论文旨在解决低资源环境下文本到语音（TTS）系统部署中的核心矛盾：即大型语音克隆模型虽具备高灵活性但推理成本高昂，而小型固定语音模型则需依赖特定说话人语料且泛化能力受限。其解决方案的关键在于提出一种新型“教师-学生”范式，利用大规模语音克隆模型作为可编程数据生成器，将仅15秒的短语音参考转化为完全基于合成语音训练的轻量化固定语音学生模型。这一方法通过合成数据构建学生模型，实现了无需参考音频即可在设备端运行的TTS系统。研究重点分析了文本预处理、合成生成质量、质量过滤、拒绝采样及前端设计对最终学生模型性能的影响，并针对泰语特有的复杂性——如词边界模糊、声调歧义、专有名词与外来词、数字转写以及泰英混用等——进行了系统性优化。实验结果表明，所提出的Wayu-Paxa-TTS-Edge模型（82M参数）在关键词准确率（68.2%，达Gemini 3.1的85.5%）、停顿精度（91.4%，优于教师模型的89.9%）及语音节奏一致性方面均表现优异，同时在泰语和英语上的字符错误率（CER）分别低至3.7%和1.1%，显著优于现有方案。该工作为低资源语言的高效、低成本、端到端语音合成提供了可行路径，并开源了完整模型与评估框架。
链接: https://arxiv.org/abs/2609.03502
作者: Kunat Pipatanakul,Potsawee Manakul,Warit Sirichotedumrong,Sittipong Sripaisarnmongkol,Pakorn Nathong,Phatrasek Jirabovonvisut
机构: Wayu Research(威尤研究); Paxa Labs(帕克萨实验室); Typhoon(台风)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 18 pages, technical report
Abstract:In low-resource settings, deploying TTS typically requires choosing between a large voice-cloning model with costly inference or a compact fixed-voice system that requires a speaker-specific corpus. We study a third route: using a large voice-cloning model as a programmable data source to turn a short voice reference (e.g., 15 seconds) into a compact fixed-voice student trained entirely on synthetic speech. This setting makes pipeline design consequential: teacher errors become training targets, while filtering failed generations can reduce coverage of difficult texts. Thai further introduces challenges from ambiguous word boundaries, lexical tone, names and loanwords, numeric verbalization, and Thai-English code-switching. We study how text preparation, synthetic generation, quality filtering, rejection sampling, and frontend choices affect the resulting student, and where teacher limitations remain. We evaluate CER, Challenge-Set Keyword Accuracy, Prosody Pause Accuracy, speaker similarity, and speaking rate. The resulting 82M-parameter model, Wayu-Paxa-TTS-Edge, enables on-device Thai TTS without reference audio. It achieves 68.2% Challenge-Set Keyword Accuracy (85.5% of Gemini 3.1) and 91.4% pause precision, outperforming its OmniVoice teacher (89.9%) and reaching 94.8% of Gemini 3.1. It also achieves the lowest pause-placement error and intra-word pause rates among the three systems, and 3.7% and 1.1% CER on Thai and English, respectively. We open-source the model and evaluation framework for Thai TTS development.
[NLP-51] Pattern Over-Generalization of Knowledge Graph Embedding EMNLP2026
【速读】： 该论文旨在解决知识图谱嵌入（Knowledge Graph Embedding, KGE）模型在捕捉知识图谱中推理模式（inference patterns）时存在的模式过泛化（Pattern Over-Generalization）问题。现有KGE模型虽能有效建模对称性/反对称性、逆关系和复合关系等多样化模式，但其本质缺陷在于：当仅从单一模式实例学习时，模型会将该模式无差别地推广至所有相关实例，导致推理偏差。为应对这一挑战，论文提出PogRE（Pattern Over-Generalization Robust Embedding）方法，其核心创新在于采用密集线性变换（dense linear transformation）与复合操作相结合的关系表示机制。理论分析表明，随着观测到的三元组数量增加，密集线性变换使模式逐步趋向全局通用；当观测到d+1个线性无关实体（d为实体嵌入维度）后，该变换可保证模式在所有相关实例上的普适性。实验结果表明，PogRE在三个标准基准数据集上的链接预测任务中均优于现有先进KGE模型，且实证验证了其有效缓解过泛化负面影响的能力。
链接: https://arxiv.org/abs/2609.03487
作者: Junsik Kim,Kangil Kim
机构: Gwangju Institute of Science and Technology (光州科学技术院)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Accepted to EMNLP 2026, 22 pages, 9 figures
Abstract:Knowledge graph embedding (KGE) demonstrates its effectiveness for predicting missing links in knowledge graphs (KGs) by projecting entities and relations into a low-dimensional vector space. It is crucial for KGE models to effectively capture inference patterns (patterns) inherent in KGs, such as symmetry/antisymmetry, inversion and composition. Although recent KGE models exhibit strong capabilities in modeling such diverse patterns, they suffer from inherent limitations stemming from pattern over-generalization, where embeddings learned from only a single pattern instance inevitably generalize that pattern to all related instances, i.e., generalize the pattern universally. To address this issue, we propose PogRE (Pattern Over-Generalization Robust Embedding), a simple but effective method that utilizes dense linear transformations and compound operations for relation representation. Our theoretical analysis demonstrates that a dense linear transformation allows a pattern to become progressively universal as more triples are observed in the pattern. Furthermore, after observing d+1 linearly independent entities (d+1 denotes the dimension of entity), the linear transformation guarantees universal generalization of the pattern across all related instances. Experimental results on three standard benchmark datasets show that PogRE outperforms existing state-of-the-art KGE models in link prediction. Moreover, our empirical results indicate that PogRE effectively addresses the negative impact of over-generalization.
[NLP-52] When Users Dont Ask: Benchmarking Context-Driven Memory Retrieval in Conversational Agents
【速读】： 该论文旨在解决当前大型语言模型（LLM）在长期对话中记忆系统评估不足的问题，尤其针对现有基准测试主要依赖问答式探测而非真实对话场景下内存使用所导致的评估偏差。其核心挑战在于：传统QA型基准无法充分揭示对话情境中记忆检索的实际性能，尤其是在处理隐含、复合及反事实等复杂查询时存在的显著检索缺口。解决方案的关键在于提出一个名为LOCOMO-CONV的新型对话式记忆评估基准，基于LoCoMo构建，包含对话、隐含、反事实和复合四种查询类型，并结合端到端响应质量与检索召回率双重指标进行评估。研究发现，对话语境能暴露QA基准忽略的深层检索缺陷，尤其是对隐含和复合查询；多方面查询重写虽可缓解原始轮次记忆中的问题，但对抽象型记忆效果有限；更重要的是，强检索能力并不必然转化为高质量回应，且隐含查询表现出“静默接地”现象——即记忆提升了上下文一致性，但未显式引用原始证据。这一发现表明，基于推理的记忆扩展（reasoning-based memory elaboration）是未来重要方向，为此作者还发布了辅助性的支持性记忆标注（supportive_memory annotations），以捕捉超出原始黄金证据的对话有用上下文信息。
链接: https://arxiv.org/abs/2609.03467
作者: Wen-Yu Chang,Yun-Nung Chen
机构: National Taiwan University (国立台湾大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models (LLMs) are increas- ingly deployed as long-horizon conversational agents, motivating growing interest in mem- ory systems. However, existing benchmarks primarily evaluate memory through QA-style probing rather than in-situ conversational usage. We introduce LOCOMO-CONV, a conversa- tional memory benchmark derived from Lo- CoMo with four query styles: dialog, implicit, counterfactual, and composed. Across five rep- resentative memory systems, we evaluate both retrieval recall and end-to-end response qual- ity. Our experiments show that conversational framing exposes substantial retrieval gaps over- looked by QA benchmarks, especially on im- plicit and composed queries, which multi-facet query rewriting narrows for raw-turn mem- ory but not abstractive memory. We further find that strong retrieval does not fully trans- late into response quality, and that implicit queries exhibit silent grounding, where mem- ory improves contextual grounding without ex- plicitly surfacing the gold fact. These results point to reasoning-based memory elaboration as a promising direction, and we release aux- iliary supportive_memory annotations captur- ing conversationally useful context beyond the original gold evidence.
[NLP-53] Its the Problem Not the Path: Budget and Difficulty Confounds in LLM Reasoning Trajectories
【速读】： 该论文旨在解决大语言模型推理轨迹中“突破性时刻”与“早期可读命运”等现象的误判问题，其核心在于现有测量方法缺乏在命题层面的反事实控制（counterfactual control）。为此，作者提出了两个关键控制机制：首先，通过重启控制的截断探测（restart-controlled truncation probe），将解题成功归因于前缀本身的内在价值还是计算资源的持续投入进行区分，对比相同生成令牌预算下从原前缀继续与从头重启的求解率。在178个问题-模型组合（89个MATH问题×两个小型开源模型）上的实验表明，仅有1个组合表现出前缀受限特征，且重启剂量响应曲线有效区分了计算资源不足与能力瓶颈模型；在所有匹配预算位于重启网格范围内的场景中，继续原前缀均优于重启（9/9），表明主要为计算压缩而非可解性扩展。其次，通过预注册、难度受控的测试发现，早期窗口内部信号中除问题难度基线外无显著可检测的结果信息；对公开语料库的两项无生成分析进一步揭示了该控制的必要性：一个盲探难度代理在19.2万条DeepSeek-R1生成样本上达到AUROC 0.873（处于已发表探测结果范围内），而对最接近的已发表早期正例的重建结果也达到0.849，但在每个锚点位置上（如t=4时为0.496）均与随机水平无统计差异，说明高聚合探测性能无法单独证明单次尝试中的信息存在；后验的靶向探测仅显示微小平均残差，集中于三个低失败率问题。因此，研究强调：仅凭聚合探测的高AUROC不能确立单次推理过程中的信息传递，必须引入仅问题基准或问题内评估以实现有效验证。
链接: https://arxiv.org/abs/2609.03436
作者: Yigit Utku Bulut
机构: Johannes Kepler University Linz(约翰内斯·开普勒林茨大学)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 25 pages, 11 figures, 4 tables. Also available at doi: https://doi.org/10.5281/zenodo.22261107 . Code and pre-registered protocols: this https URL
Abstract:Reasoning traces of large language models are widely read as containing “breakthrough” moments and early-legible fates. Both readings rest on measurements missing a counterfactual control at the level of the claim; we supply both controls. First, a restart-controlled truncation probe separates when a solution fits the continuation budget from when a prefix carries value that fresh computation cannot buy, comparing per-anchor continuation solve rates against from-scratch restart curves at matched total generated-token budget. Applied to 178 problem-model cells (89 MATH problems x two small open models, an outcome-blind but difficulty-targeted cohort), exactly 1 of 178 cells survives as prefix-limited; restart dose-response separates a compute-starved model from a capability-limited one; and wherever the matched budget lies inside the restart grid, continuing the model’s own prefix beats restarting (9 of 9) – predominantly compute compression rather than expanded reachability. Second, a pre-registered, difficulty-controlled test finds no detectable outcome information in early-window internal signals beyond a problem-difficulty baseline, and two generation-free analyses of public corpora show why this control is needed: a trace-blind difficulty proxy reaches AUROC 0.873 on 192K DeepSeek-R1 generations – inside the published probe range – and a closely matched reconstruction of the closest published early-window positive recovers a comparable pooled result (0.849) while within problem it is statistically indistinguishable from chance at all ten anchors (0.496 at t=4); a post-hoc within-targeted probe finds only a small average residual, concentrated in three low-failure problems. High pooled probe AUROCs cannot by themselves establish within-attempt information; a question-only baseline or within-problem evaluation is required.
[NLP-54] Decoupled Analysis-Judging: An Automated Creativity Evaluator Using LLM s in Complex Multi-step Creativity Tasks EMNLP2026
【速读】： 该论文旨在解决大语言模型作为评判者（LLM-as-a-Judge）在评估上下文关联且流程结构化的创造力任务（Contextually-Grounded and Procedurally-Structured Tasks, CGPST）时面临的挑战，尤其是由冗长性偏差（verbosity bias）和宽松性偏差（leniency bias）导致的评价不稳定与不公正问题。此类任务具有多步骤依赖关系、高度主观性及评分范围广泛等复杂特征，使得现有方法难以保证评估可靠性。现有方案或依赖特定任务训练，或直接应用通用的LLM-as-a-Judge，均无法有效应对上述复杂性。为此，本文提出CreaEval，一种面向CGPST的自动化创造力评估框架，其核心创新在于将传统的LLM-as-a-Judge解耦为分析（Analysis）与评判（Judging）两个独立阶段：首先通过基于思维链（Chain-of-Thought, SoT）的大语言模型进行记忆增强型分析（Memory-augmented Analysis），将多步响应转化为结构化评估证据，并引入跨步骤记忆以捕捉依赖关系；随后，由专门的评判模型（Judge-LLM）基于提取的证据进行打分，而不再访问原始响应内容，从而降低偏见影响。实验结果表明，CreaEval在CGPST及两类经典简单创造力任务上相较次优基线平均提升22.74%，展现出显著的性能优势与良好的泛化能力。
链接: https://arxiv.org/abs/2609.03432
作者: Xiangyu Wang,Jin Wu,Xiaoyu Li,Chanjin Zheng,Yifeng Zhou
机构: East China Normal University (华东师范大学); Shanghai Institute of Artificial Intelligence for Education (上海人工智能教育研究院); School of Computer Science and Technology (计算机科学与技术学院); School of Education and Intelligent Education Research Center (教育学院与智能教育研究中心); School of Data Science and Engineering (数据科学与工程学院)
类目: Computation and Language (cs.CL)
备注: Accepted to EMNLP 2026
Abstract:Automated evaluation of creativity tasks remains challenging for LLM-as-a-Judge, as LLM is susceptible to biases such as verbosity bias and leniency bias. Such limitations are particularly evident in Contextually-Grounded and Procedurally-Structured Tasks (CGPST), a complex multi-step creativity task where inter-step dependencies, highly subjectivity, and wide scoring ranges lead to more unstable and biased judgments. Existing approaches either rely on task-specific training or directly apply LLM-as-a-Judge, both of which struggle to ensure reliable evaluation under such complexity. To bridge these gaps, we propose CreaEval, an automated creativity evaluator for CGPST that decouples typical LLM-as-a-Judge into analysis and judging. Correspondingly, CreaEval involves two critical phases: Memory-augmented Analysis, a SoT-LLM converts multi-step responses into structured evaluation evidence, incorporating cross-step memory; and Evidence-based Judging, a Judge-LLM uses the extracted evidence for judging without accessing raw responses. Comprehensive experiments show that CreaEval achieves an average performance improvement of 22.74% over the second-best baselines across CGPST and two classic simple creativity tasks, demonstrating its generalizability. The code is available at this https URL.
[NLP-55] Random Attention: Rethinking KV Cache Eviction for Efficient Reasoning
【速读】： 该论文旨在解决大语言模型在需要长链推理的任务中，因注意力机制的键值缓存（KV cache）导致的严重内存瓶颈问题。现有方法普遍采用一种范式：通过估计每个缓存标记对未来任务的重要性并保留得分最高的部分。然而，本文揭示这一选择信号几乎不起作用。研究提出随机注意力（Random Attention）策略，即仅保留提示（prompt）部分，并在每个注意力头内均匀随机淘汰其余内容，完全不计算任何评分。在四个模型和六个推理任务上的实验表明，该方法性能与最强的先前淘汰策略相当，同时在vLLM部署中实现了32%-43%更高的吞吐量。受控实验证明，其有效性源于两个关键机制：1）提示是缓存中最脆弱的部分，多数性能差异本质上取决于所选策略是否恰好保留了提示；2）推理过程本身具备双重冗余保护机制——文本层面（模型在推理过程中重复表达所需信息）和注意力头层面（每个头独立保存推理轨迹的副本），因此只要提示被安全保留，随机选取即可保留足够多的必要信息，无需依赖复杂的选择信号。
链接: https://arxiv.org/abs/2609.03430
作者: Heng Wang,Jielin Qiu,Wenting Zhao,Cheng Qian,Liangwei Yang,Jiawei Han,Heng Ji,Silvio Savarese,Shelby Heinecke,Huan Wang
机构: Salesforce AI Research; University of Illinois Urbana-Champaign
类目: Computation and Language (cs.CL)
备注:
Abstract:Large language models achieve superior performance on tasks that require extended reasoning, but long chains of thought make the KV cache a severe memory bottleneck. Existing KV cache compression methods share one paradigm: score each cached token by some estimate of how much it will matter later, and keep the top-scoring ones. We show that the selection signal contributes almost nothing. Random Attention keeps the prompt and evicts uniformly at random within each attention head, computing no score at all; across four models and six reasoning tasks it matches the strongest prior evictor while serving 32-43% higher throughput than it in vLLM deployment. Controlled experiments explain this by showing that 1) the prompt is the fragile part of the cache, and most of the gap between selectors is just whether their selection signal happened to keep it; 2) the reasoning trace protects itself against eviction with redundancy at two levels, in the text (the model restates what it still needs as it works) and across attention heads (each keeps its own copy of the trace), so once the prompt is safe, a random draw retains enough copies of what the model still needs, and no score is required to pick them. Our code is publicly available at this https URL.
[NLP-56] Lngram v2: Latent N-Gram Memory with Interpretable Discrete Representations
【速读】： 该论文旨在解决Transformer模型缺乏原生查找机制的问题，即在识别和重用局部静态模式时需反复进行密集计算，导致效率低下。其核心解决方案在于提出Lngram v2，通过解耦路由数量、记忆维度与主干网络宽度，实现了记忆容量的独立可扩展性；同时引入上下文感知的分组查询注意力读出机制，显著提升了内存利用效率。此外，通过零值Sink设计与反事实代理梯度，进一步增强了读出选择性和路由训练能力，同时保持了硬离散寻址的特性。实验表明，该方法在不同规模的视觉-语言模型（VLMs）中均实现一致性能提升，成功扩展至300亿参数模型，且相比Lngram v1大幅降低总参数量与激活参数量，同时维持或超越原有语言建模表现。深入分析显示，其离散标识符（discrete IDs）能有效保留连续隐藏状态的语义结构，支持仅凭ID实现语义恢复及跨数据集稳定的ID-语义关联，验证了其作为高效、可扩展的潜在条件记忆机制的优越性，并为分析模型内部表征提供了结构化接口。
链接: https://arxiv.org/abs/2609.03426
作者: Yunao Zheng,Bin Wen,Xiaojie Wang
机构: 未知
类目: Computation and Language (cs.CL)
备注:
Abstract:Transformers lack a native lookup mechanism, requiring repeated dense computation to recognize and reuse local static patterns. Lngram v1 introduces tokenizer-independent conditional memory through discrete latent n-gram addressing, but its memory capacity is coupled with the backbone width, limiting scalability due to high parameter and activation costs. We propose Lngram v2, which decouples the number of routes, memory dimension, and backbone width, and introduces a context-aware grouped-query attention readout to scale memory capacity independently. A zero-value Sink and counterfactual surrogate gradients further improve readout selectivity and routing trainability while preserving hard discrete addressing. Experiments across vision–language models (VLMs) of different scales show consistent improvements, including successful scaling to a 30B-parameter model. Compared with Lngram v1, Lngram v2 substantially reduces both total and activated memory parameters while maintaining or improving language modeling performance. Further analysis shows that its discrete IDs preserve substantial semantic structure of continuous hidden states, enabling semantic recovery from IDs alone and stable ID–semantic associations across datasets. These results establish Lngram v2 as an efficient and scalable latent conditional memory mechanism whose discrete addresses also provide a structured interface for analyzing internal model representations.
[NLP-57] o What Extent Do Large Language Models Understand Bangla Idioms?
【速读】： 该论文旨在解决低资源语言（如孟加拉语）中习语理解的挑战，特别是针对生成式 AI（Generative AI）在习语相关任务上的表现不足问题。其核心解决方案是构建首个大规模孟加拉语习语基准数据集，并配套生成合成的多选题（MCQ）数据集以支持习语含义识别任务。研究通过零样本和少样本提示策略，对多种大型语言模型（LLM）在习语改写、习语片段检测和含义识别三项任务上进行系统评估，揭示了不同模型在各类任务中的性能差异，进而为未来提升模型对习语的理解能力提供了关键数据资源与实证依据。
链接: https://arxiv.org/abs/2609.03410
作者: Mousumi Akter,Md. Faiyaz Abdullah Sayeedi,Nurul Labib Sayeedi,Swakkhar Shatabda
机构: 未知
类目: Computation and Language (cs.CL)
备注:
Abstract:Idiomatic expressions are an integral part of natural language, reflecting cultural nuances and posing unique challenges for computational models, particularly in low-resource languages. In this paper, we present the first large-scale benchmark dataset of Bangla idioms, complemented by a synthetic multiple-choice question (MCQ) dataset for idiom meaning identification. We conduct a comprehensive evaluation of recent large language models (LLMs) across three idiom-related tasks: paraphrasing, idiom span detection, and meaning identification, leveraging zero-shot and few-shot prompting strategies. Our results reveal substantial variability in model performance, with no single LLM consistently outperforming others across all tasks. Notably, Phi-4-mini-instruct excels in paraphrasing, Kimi-K2-32b-instruct in span detection, and Gemini-2.5-flash in meaning identification. We believe that our datasets and analyses will provide valuable resources to guide future research in improving LLM comprehension of idiomatic expressions, particularly in Bangla and other low-resource languages.
[NLP-58] abScope: Question-Adaptive Scope Selection for Table Question Answering
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在处理长表格问答任务时，随着表格规模增大导致准确率下降的问题，尤其关注不同问题类型对模型性能影响的非均匀性。其核心挑战在于：定位敏感型问题易受无关表格内容干扰，而需要全局证据支持的问题仍可从全表推理中获益。为此，论文提出一种基于问题类型的自适应框架，关键在于通过操作感知的表格分解机制构建与问题相关的子表，并利用预测的问题类型动态选择局部化推理或全表推理模式。该方案不仅提升了对特定问题类型的处理效率，还通过引入银级参考子表用于证据选择评估，并构建了基于真实世界长表格的SLQA基准数据集，实验表明自适应选择策略在整体性能上优于单一推理方式，验证了“何时进行定位”与“如何进行定位”同等重要。
链接: https://arxiv.org/abs/2609.03395
作者: Yuxiang Wang,Junhao Gan,Jianzhong Qi
机构: The University of Melbourne(墨尔本大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: conference paper preprint
Abstract:Large Language Models (LLMs) have shown strong performance on table question answering, yet their accuracy often degrades as table size increases. We find that this degradation is not uniform across question types. Localization-sensitive questions are particularly affected by irrelevant table content, while questions requiring broader evidence may still benefit from full-table reasoning. Based on this observation, we propose a question-adaptive framework that dynamically selects between localized and full-table reasoning. The framework constructs question-specific sub-tables through operation-aware table decomposition and uses the predicted question type to determine the appropriate reasoning mode. We further introduce silver reference sub-tables for evaluating evidence selection and construct SLQA, a benchmark based on real-world long tables. Experiments on WikiTQ and SLQA show that localization is particularly effective for lookup and local reasoning questions, while adaptive selection between localized and full-table reasoning achieves the best overall performance. These results highlight that long-table QA requires deciding not only how to localize, but also when to localize. Our code and datasets will be made available upon publication of the paper.
[NLP-59] Chiaroscuro for Emotions: A Contrastive Emotion Benchmark Grounded in Appraisal Theory EMNLP2026
【速读】： 该论文旨在解决现有情感识别基准普遍仅预测单个情绪（one emotion per text）的问题，忽略了现实场景中同一事件引发两人对立情绪的复杂情况。例如，一个孩子因兴奋踢前排座椅，而前方乘客则因此愤怒。为应对这一挑战，论文提出CHIARO——一个包含1,000条人工标注句子的对比情感推理基准，其理论基础为评价理论（appraisal theory），每个场景描述一个因果触发事件，导致一人产生积极情绪、另一人产生消极情绪，情绪类别基于十类分类体系。解决方案的关键在于构建一个能够捕捉跨个体情感对立关系的高质量数据集，并通过实证评估表明，尽管最强的大型语言模型（LLM）在该任务上达到67.3的宏平均F1（macro-F1），仍显著低于人类一致性水平，且现有情绪分类器表现接近随机猜测。此外，研究发现将CHIARO与已有情感语料结合可作为有效的训练信号，显著提升下游分类器在自身及多个外部基准上的性能，表明该数据集在情感识别任务中具有互补性增强作用。
链接: https://arxiv.org/abs/2609.03394
作者: Divyesh Bommana,Mohammad Saim,Tianyu Jiang
机构: University of Cincinnati(辛辛那提大学)
类目: Computation and Language (cs.CL)
备注: Accepted to EMNLP 2026 (Main Conference) Dataset and code: this https URL
Abstract:Emotion recognition benchmarks often predict one emotion per text, missing many real-world scenarios where two people arrive at opposing emotions from a single shared event. For example, a child kicks the seat in front of her in excitement while the passenger ahead grows angry. We introduce CHIARO, a 1,000 human-annotated sentence benchmark for contrastive emotion inference grounded in appraisal theory. Each scene describes one causal trigger eliciting a positive emotion in one person and a negative emotion in the other, drawn from a ten-class taxonomy. We benchmark seven frontier LLMs and four off-the-shelf emotion classifiers. The strongest LLM reaches 67.3 macro-F1, well below human agreement, while existing emotion classifiers score near chance. Beyond evaluation, CHIARO also serves as a training signal. When combined with an existing emotion corpus, the resulting downstream classifier improves on CHIARO itself and on six of ten external emotion benchmarks, which positions our dataset as a complementary signal for emotion recognition.
[NLP-60] FrameBench:A Language Understanding Benchmark Based on Frame Semantics EMNLP
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在句子理解过程中是否能够像人类一样基于语义框架（Semantic Frame）实现隐含信息的自然补全这一关键问题。传统框架语义学认为，阅读理解依赖于词汇意义与背景知识（即语义框架）之间的关联，从而实现对文本中未明言信息的隐性补充。然而，现有研究尚未明确评估LLMs是否具备此类认知能力。为此，作者提出FrameBench——一个基于框架语义学构建的基准测试集，通过多选题形式考察模型在不同语境下对同一动词所激活的语义框架的区分能力。该基准覆盖英语和日语，依托FrameNet风格资源，并采用生成-验证流水线结合母语者判断的方式构建，确保语义情境的准确性与文化适配性。实验结果表明，小型模型面临显著挑战，而部分大型模型甚至超越了人类参考表现，揭示出当前主流大模型在深层语义理解方面具备接近或超越人类的隐含信息补全能力。其解决方案的关键在于：构建具有语言多样性与语义精细度的跨语言基准，以系统评估模型在语义框架层面的推理能力。
链接: https://arxiv.org/abs/2609.03370
作者: Chihiro Yano,Ryohei Sasano
机构: Nagoya University (名古屋大学)
类目: Computation and Language (cs.CL)
备注: Accepted in EMNLP Findings 2026
Abstract:In frame semantics, sentence comprehension is assumed to proceed by relating lexical meaning to background knowledge called semantic frames, thereby enabling readers to implicitly enrich the text with unstated information. Recent large language models (LLMs) have achieved strong performance across a wide range of downstream tasks. However, it remains unclear whether they can reproduce the kinds of implicit enrichment that humans naturally make during comprehension. To address this question, we introduce FrameBench, a benchmark grounded in frame semantics. FrameBench consists of multiple-choice questions that test whether models distinguish the frames evoked by the same verb across contexts. We construct the benchmark for English and Japanese using FrameNet-style resources and a generation-and-verification pipeline with native-speaker judgments. Our experiments on a diverse set of models reveal challenges for small models, while several large models surpass the human reference scores. We release the constructed FrameBench dataset and the code for dataset construction and evaluation at this https URL.
[NLP-61] Accountable AI with Grounded Faithful Consistent Actionable Rationales: A Case Study in Clinical Trial Matching with VERDICT EMNLP2026 ATC
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在高风险决策任务中缺乏可问责性（accountability）的问题。具体而言，尽管基于LLM的系统在临床试验匹配等任务中展现出较高的准确率，但其决策过程往往存在政策应用不一致、推理依据与实际决策脱节（unfaithful rationales）、以及对关键前提条件（pivotal conditions）的忽视，导致决策难以被验证、解释或挑战。其核心解决方案是提出VERDICT——一个基于可满足性模理论（Satisfiability Modulo Theories, SMT）的神经符号框架，将决策任务、约束条件和决策策略形式化为逻辑公式，并通过SMT与MaxSMT求解器进行严格推导。该方法确保了政策执行的一致性，并从构造上实现可问责性：生成的推理依据明确包含显式假设和关键前提条件，且具备良好的反事实自一致性（counterfactual self-faithfulness），即当关键前提发生变化时，决策结果能够相应改变。实验表明，VERDICT在多个基准数据集上实现了优于纯LLM及神经符号基线的决策准确性，同时显著提升了临床医生偏好度的可解释性与可信度。
链接: https://arxiv.org/abs/2609.03366
作者: Zikai Zhou,Yufei Jin,Yilin Xu,Yu-Chiang Wang,Chieh-Ju Chao,Monica S. Lam
机构: Stanford University (斯坦福大学); UCLA (加利福尼亚大学洛杉矶分校); Emory University (埃默里大学); Mayo Clinic (梅奥诊所)
类目: Computation and Language (cs.CL); Computers and Society (cs.CY); Logic in Computer Science (cs.LO)
备注: Accepted to EMNLP 2026 (Main Conference). 46 pages, 6 figures, 28 tables. Code and prompts: this https URL
Abstract:Accountability means a decision can be examined, justified, and contested. LLMs make this hard: fluent output may be ungrounded, incomplete, or unfaithful to the decision process. Achieving accountability requires verified rationales (how was the decision reached), assumptions (what was assumed rather than known), policy consistency (the same treatment for the same facts), and pivotal conditions (what would change the outcome). We introduce self-faithfulness as an automatic test of accountability: changing the pivotal conditions should change the decision. We examine accountable AI through clinical trial matching, a high-stakes task central to evidence-based medicine. Although LLM-based matchers match patients to trials reasonably accurately, they apply decision policies inconsistently and produce rationales that are unfaithful to their own decisions. We introduce VERDICT, an LLM-based agent that translates a decision task, its constraints, and its policy into Satisfiability Modulo Theories (SMT), then derives the decision with SMT and MaxSMT solvers – so policies are applied consistently and decisions are accountable by construction. Across a SIGIR 2016-derived dataset and TREC 2021, VERDICT achieves the strongest decision accuracy among LLM-only and neurosymbolic baselines, applies policies with perfect consistency, and produces clinician-preferred rationales grounded in explicit assumptions and pivotal conditions, with improved counterfactual self-faithfulness. Comments: Accepted to EMNLP 2026 (Main Conference). 46 pages, 6 figures, 28 tables. Code and prompts: this https URL Subjects: Computation and Language (cs.CL); Computers and Society (cs.CY); Logic in Computer Science (cs.LO) ACMclasses: I.2.7; I.2.4; J.3 Cite as: arXiv:2609.03366 [cs.CL] (or arXiv:2609.03366v1 [cs.CL] for this version) https://doi.org/10.48550/arXiv.2609.03366 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Zikai Zhou [view email] [v1] Thu, 3 Sep 2026 04:49:02 UTC (3,208 KB)
[NLP-62] From Zero to Hero: An Open LLM Ecosystem for Armenian
【速读】： 该论文旨在解决亚美尼亚语（Armenian）作为一种形态丰富且资源匮乏的语言，其预训练数据严重不足，且缺乏可复现的开源大语言模型（LLM）的问题。现有公开的亚美尼亚语模型普遍缺少完整的训练数据与训练流程记录，导致研究可重复性差。为应对这一挑战，研究者构建并发布了两个高质量数据集：ArmWeb，一个经过充分验证的437万篇亚美尼亚新闻文档语料库；以及ArmSTEM，一个包含37.3万道数学与科学问题及其逐步解答的英-亚美尼亚语平行语料库，所有翻译内容均通过保留答案的生成式AI判断与人工评估双重验证。基于这些数据对Gemma-4-E4B进行持续预训练，得到arm-gemma-e4b模型，其在多项评测中超越所有现有开源亚美尼亚语模型，并成为首个具备完整训练数据与可复现训练配方的开源亚美尼亚语大模型。消融实验表明，仅使用新闻文本进行持续预训练虽能提升语言流畅性，但会损害知识保持能力，这一现象在现有亚美尼亚语模型中普遍存在；而引入少量经验证的翻译后STEM数据则可有效逆转知识损失。此外，研究还发现当前最大的公共亚美尼亚语语料库与网络来源的评估集存在高度重叠，甚至在FineWeb-2中出现训练集与测试集的自重叠问题。为此，研究团队全面开放所有数据、模型及代码，以促进该领域研究的透明性与可复现性。
链接: https://arxiv.org/abs/2609.03350
作者: Erik Arakelyan,Khatun Avetisyan,Meri Davtyan,Heghine Grigoryan,Nane Khachatryan,Hayk Shahsuvaryan,Henrik Sergoyan,Vahan Martirosyan
机构: NVIDIA(英伟达); COPA
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: 18 pages, 4 figures, 13 tables. Data and model: this https URL . Code: this https URL
Abstract:Pretraining data for Armenian, a morphologically rich and low-resource language, is scarce, and no open Armenian LLM has been released with the data and recipe needed to reproduce it. To address this gap, we curate and release two datasets. ArmWeb is an extensively validated corpus of 4.37M Armenian news documents. ArmSTEM is a parallel English-Armenian collection of 373K math and science problems with step-by-step solutions, translated into Armenian and verified through both answer-preserving LLM judgment and human evaluation. Continued pretraining of Gemma-4-E4B on these datasets yields arm-gemma-e4b, which outperforms every existing open Armenian model as well as its unadapted base, and is the first open Armenian LLM with complete training data and recipe. Our ablations show that news-only continued pretraining improves fluency while eroding knowledge, a pattern we also observe in existing Armenian models, and that a small share of verified translated STEM data reverses the loss. We further find that the largest public Armenian corpora overlap web-derived evaluation panels heavily, including a train/test self-overlap inside FineWeb-2. We openly release all data, models, and code.
[NLP-63] FPCO-Dialog: A Multi-Turn False-Premise Benchmark for Correction and Cooperation in Vision-Language Models EMNLP2026
【速读】： 该论文旨在解决视觉语言模型（Vision-Language Models, VLMs）在多轮对话场景中面对持续性错误前提（false premises）时的纠错与协作行为评估问题。现有评估方法未能有效隔离同一错误视觉假设在多轮对话中反复出现时模型的响应模式，导致对模型鲁棒性和交互能力的评价不充分。为此，作者提出FPCO-Dialog基准，其核心创新在于构建了一个包含1,080张图像和10,800个问答轮次的标准化评估框架，通过10轮对话协议，在正确对话前缀后引入重复出现的错误前提指称表达，系统性地考察模型在持续误导情境下的表现。关键解决方案包括：采用模型无关的评估协议、设计基于双独立检测器的校正率指标CorrTP@K，以及按视觉复杂度、物体类别和错误前提类型进行分层采样。实验揭示了不同模型在整体校正倾向、逐轮动态演化及不同类型错误前提下的系统性差异，为评估VLM在真实交互场景中的可靠性提供了可复现的基准工具。
链接: https://arxiv.org/abs/2609.03331
作者: Jiayuan Ma,Yuqi Lu,Weiyang Guo,Chenrui Wang,Junyi Shu,Xuebo Liu,Min Zhang,Jing Li
机构: Harbin Institute of Technology, Shenzhen, China(哈尔滨工业大学深圳校区)
类目: Computation and Language (cs.CL)
备注: Accepted at EMNLP2026 Main Conference. For code and data, see this https URL
Abstract:Vision-language models (VLMs) are increasingly deployed in multi-turn settings where users may describe visual content with incorrect assumptions. Yet existing evaluations rarely isolate how models respond when the same visually grounded false premise persists across dialogue turns. We introduce FPCO-Dialog, a benchmark for evaluating correction and cooperation behavior in VLMs under repeated false premises. FPCO-Dialog contains 1,080 images and 10,800 question turns, stratified by visual complexity, object category, and false-premise class, and uses a 10-turn protocol in which a correct dialogue prefix is followed by repeated false-premise referring expressions. We evaluate 20 commercial and open-source VLMs with a model-agnostic protocol and CorrTP@K, a correction-rate metric over false-premise turns, scored by two independent detectors. FPCO-Dialog reveals substantial and persistent cross-model differences in aggregate correction tendency, model-specific turn-wise dynamics, and systematic variation across false-premise types under the benchmark’s substitution distribution. The dataset, evaluation protocol, model outputs, detector labels, and code are available.
[NLP-64] Less Is Moral: A CHARMing Framework for Moral Foundations Detection in Endorsement Behaviour
【速读】： 该论文旨在解决现有道德基础检测系统在跨领域泛化能力差、推理依据不足以及依赖高成本提示工程的大型语言模型（LLM）等问题。其核心解决方案是提出CHARM框架——一个面向道德基础检测的轻量化微调大模型架构，通过整合互补的道德根基、推理对齐机制与极性感知的仇恨言论信号，实现更鲁棒且忠实的道德判断。与以往基于词典、微调或提示的方法不同，CHARM的每个组件（如MAC交叉注意力、推理对齐、仇恨言论调制）均对应一个明确的心理学构念，实现了计算过程与心理理论的深度融合。实验表明，仅使用30%的MFTC、MFRC和新闻数据训练集，并结合更丰富的MFTCXplain标注监督，CHARM在域内任务中AUC提升达15.3%，并在所有跨域数据集上均优于监督基线，在AUC和F1指标上表现更优；同时具备可扩展性和低成本优势。进一步应用于大规模推特平台上的新冠疫情话语分析，验证了道德价值一致性与线上支持行为之间的强关联。因此，CHARM为大规模测量道德框架提供了实用工具，助力研究带有道德色彩的虚假信息传播机制。
链接: https://arxiv.org/abs/2609.03330
作者: Huixiang Fu,Marian-Andrei Rizoiu
机构: University of Technology Sydney(悉尼科技大学)
类目: Computation and Language (cs.CL); Computers and Society (cs.CY); Social and Information Networks (cs.SI)
备注:
Abstract:Moral language plays a central role in shaping online endorsement and the diffusion of information, yet existing moral foundation detection systems often suffer from poor cross-domain generalization, weak rationale grounding, and reliance on costly prompting-based large language models (LLMs). We introduce CHARM, a MA\textbfC- and \textbfHate-speech-\textbfAware \textbfRationale-aligned \textbfMoral foundation detection framework built on a lightweight fine-tuned LLM, which integrates complementary moral grounding, rationale alignment, and polarity-aware hate speech signals to support more robust and faithful moral prediction. Unlike prior dictionary-, fine-tune-, or prompt-based detectors, which decouple computation from psychological theory, CHARM is built so that each component – MAC cross-attention, rationale alignment, and hate-speech modulation – operationalizes a distinct psychological construct. Using a 30% subsample of the MFTC, MFRC, and News training pools together with the richer supervision in MFTCXplain, CHARM improves AUC by up to 15.3% in-domain, surpasses the supervised baselines on every out-of-domain dataset in both AUC and F1, and offers a scalable, low-cost alternative to prompting-based LLM detectors. We further apply CHARM to large-scale COVID-19 discourse on Twitter and show that moral value alignment is strongly associated with online endorsement behavior. By making moral framing measurable at scale, CHARM offers a practical tool for studying the spread of morally charged misinformation.
[NLP-65] How Perturbations Propagate: A Multi-Level Analysis of Robustness in Large Language Models
【速读】： 该论文旨在解决语言模型在面对自然和合成的输入扰动（如拼写错误、文本损坏、词语替换及词元顺序打乱）时，其鲁棒性评估过于依赖输出行为表现而忽视深层表征变化的问题。现有评估方法未能充分揭示扰动如何影响模型内部计算过程，导致对模型鲁棒性的判断可能存在偏差。论文的关键解决方案在于提出多层级评估框架，从输出行为、隐藏状态几何结构（通过中心化核对齐与内在维度分析）以及注意力头功能三个层面系统考察六类扰动的影响。研究发现，不同扰动类型在各层次上产生可区分的度量特征，且这些特征无法完全由输出行为反映，且在不同检查点间一致性有限；尤其在词元替换与重排下，复制得分与激活修补恢复能力高度相关。此外，基于梯度引导的HotFlip扰动相较于随机替换造成更显著的行为与表征破坏，且其影响在所有测试检查点中保持一致。研究表明，仅依赖单一行为或表征指标评估鲁棒性具有误导性，强调应采用多层次、多维度的方法来全面理解扰动对生成式语言模型计算过程的影响。
链接: https://arxiv.org/abs/2609.03322
作者: Dun Li Chan,Emily Liu,Niyathi Allu,Christian Hoang
机构: INTI International College Penang; Independent Researcher; FPT University
类目: Computation and Language (cs.CL); Machine Learning (stat.ML)
备注: 12 pages, 4 figures
Abstract:Language models encounter typos, corrupted text, altered words, and disrupted token order, yet robustness is usually evaluated only through output behavior. We study how six naturalistic and synthetic input perturbations propagate through decoder-only language models at three levels: output behavior, hidden-state geometry, and attention-head function. We evaluate behavioral effects across four GPT-2 and two Qwen2.5 checkpoints by analyzing layerwise geometry using centered kernel alignment and intrinsic dimension, and examine attention-head responses in GPT-2. Perturbation types produce distinguishable metric profiles that are not fully captured by output measures and are only partly consistent across the tested checkpoints. Copying scores are especially associated with activation-patching recovery under token substitution and shuffling. Gradient-guided HotFlip perturbations also cause stronger behavioral and representational disruption than rate-matched random token substitutions in GPT-2; their behavioral effects are consistent across all six tested checkpoints. Our results show that robustness claims based on a single behavioral or representational metric can be misleading, and motivate multi-level evaluation of how perturbations alter language-model computation.
[NLP-66] Decoupling Turn-Taking from Semantics: A Decoupled Data Approach for Finite-State-Machine-Based Full-Duplex Dialogue EMNLP2026
【速读】： 该论文旨在解决生成式对话系统在全双工对话（full-duplex dialogue）中因依赖合成文本数据而导致的对话轮次控制（turn-taking）自然性不足的问题。现有方法如神经有限状态机（Neural Finite State Machine, NFSM）虽能以低微调成本保持语义能力，但其基于大语言模型（LLM）生成的合成数据难以捕捉真实人类对话中的精细声学时间动态，导致轮次切换行为生硬、不自然。为此，本文提出一种解耦数据（decoupled data）策略：从真实人-人（Human-Human, HH）语音对话中学习轮次控制行为，同时通过可配置的人-代理（Human-Agent, HA）文本对话来塑造语义生成能力。为实现该策略，研究设计了一种基于规则的事件引导型数据转换方法，通过识别对话中的轮次事件并应用确定性映射规则，将HH语音对话序列化为有限状态机（FSM）磁带，从而实现无需LLM标注的可扩展监督。此外，提出源感知校准（Source-Aware Calibrated, SAC）损失函数，联合校准状态转移标记的长尾分布，并引导不同数据源向其最擅长的建模任务（即轮次控制或语义生成）提供最优监督信号。实验结果表明，该方法显著提升了轮次控制的自然度，同时完全恢复了基础大语言模型的语义生成能力。
链接: https://arxiv.org/abs/2609.03321
作者: Yihang Li,Chenhui Chu
机构: Kyoto University (京都大学)
类目: Computation and Language (cs.CL)
备注: EMNLP 2026 Main Conference
Abstract:The Neural Finite State Machine (NFSM) framework offers a pragmatic path to full-duplex dialogue by serializing turn-taking control and response generation onto a single causal tape under the standard next-token prediction objective, thereby preserving semantic prowess at a low fine-tuning cost. However, its reliance on synthetic text data fundamentally limits turn-taking naturalness, as Large Language Models (LLMs) cannot faithfully simulate the fine-grained acoustic temporal dynamics of real human dialogues. In this work, we propose a decoupled data approach that learns turn-taking from real Human-Human (HH) spoken dialogues while shaping semantic behavior through configurable Human-Agent (HA) text dialogues. To operationalize this approach, we introduce a rule-based event-guided data transformation method that serializes HH spoken dialogues into FSM tapes by classifying turn-taking events and applying deterministic mapping rules, enabling scalable supervision without LLM-generated annotations. We further propose a Source-Aware Calibrated (SAC) Loss that jointly calibrates the long-tailed distribution of state transition tokens and channels each data source toward the capability it best supervises. Experiments show that our approach substantially improves turn-taking proficiency while recovering the foundation LLM’s semantic capability. Our code and model are available at this https URL.
[NLP-67] PACE: Towards Surfacing Hidden Conflicts in User Requests EMNLP2026
【速读】： 该论文旨在解决个性化助手在执行用户请求时，缺乏对潜在冲突情境的识别能力，尤其在面对隐含于用户个人背景或情境中的约束性知识（egocentric knowledge）时，现有模型难以有效检索并判断请求是否合理的问题。传统方法多聚焦于准确执行显式指令，忽视了基于上下文进行冲突检测与拒绝的能力，且多数安全检测依赖显式提供的因素，无法应对真实场景中需从知识库中隐式推断的复杂约束。为此，论文提出了PACE数据集，用于评估模型识别隐含约束（如用户个人经历、状态等）的能力，其核心在于将用户请求与基于角色的主观知识关联，要求模型通过整合上下文证据判断请求是否存在冲突。解决方案的关键在于提出PaceMaker多智能体框架，该框架通过专业化智能体协同完成查询重构、多跳图遍历及冲突感知过滤，实现对非直接关联的用户特定事实的隐式检索与推理，显著提升了在复杂上下文中识别冲突请求的准确性与证据召回质量。
链接: https://arxiv.org/abs/2609.03293
作者: Yoojin Kim,Jihyoung Jang,Hyounghun Kim
机构: POSTECH(浦项科技大学); Graduate School of Artificial Intelligence(人工智能研究生院)
类目: Computation and Language (cs.CL)
备注: EMNLP 2026 (59 pages); Code: this https URL
Abstract:Personalized assistants should not only comply with user requests but also assess whether those requests are appropriate given the user’s current circumstances. However, prior work has primarily focused on accurately executing requests, overlooking the need for assistants to account for context and engage in conflict-based refusal. Furthermore, while existing work on conflict or safety detection relies on explicitly provided factors, real-world scenarios often involve implicit factors that must be retrieved from a knowledge base (KB). To this end, we introduce Personalized Assistants for Conflict Evaluation (PACE), a dataset for evaluating whether models can identify latent constraints, expressed as egocentric knowledge or events, that render seemingly reasonable user requests inappropriate. PACE pairs user requests grounded in well-defined personas with egocentric KB facts, requiring models to integrate contextual evidence to determine whether a request is conflicting. This implicit retrieval setting hinders the direct association between user requests and conflict-inducing knowledge, making it difficult for existing models to identify relevant user-specific facts. To address this challenge, we further propose PaceMaker, a multi-agent framework in which specialized agents coordinate across query reformulation, multi-hop graph traversal, and conflict-aware filtering to retrieve contextually decisive evidence. Experiments on PACE evaluate both evidence retrieval quality and conflict decision accuracy, showing that PaceMaker consistently outperforms existing approaches.
[NLP-68] Contextual Tamil Spelling and Grammar Correction Using Progressively Fine-Tuned Sequence-to-Sequence Transformers
【速读】： 该论文旨在解决泰米尔语拼写与语法纠错中因语言特性带来的挑战，特别是针对具有丰富动词形态、复杂音变（sandhi）规则及247个独立字符的黏着性低资源语言所存在的上下文级错误（如主谓一致、时态一致性、跨词音变等）难以通过传统基于规则或局部统计模型有效处理的问题。现有方法多局限于词级表面错误修正，缺乏对句子层面语义与语法结构的理解能力。为此，作者提出一种端到端的序列到序列（sequence-to-sequence）建模范式，并在合成的高达657,720对噪声-纯净泰米尔句对数据集上对mT5-small和mBART-50进行微调，该数据集涵盖十类错误类型。其解决方案的关键在于采用四阶段渐进式训练策略（progressive schedule），依次聚焦于：表面噪声（v2）、上下文语法（v3）、单点音变（v4）和多点跨词音变（v5）。实验表明，该渐进策略显著提升了模型性能：最佳模型mBART-50 v5在1,000句平衡诊断集上达到69.3%的准确率，其中音变纠正率达87.5%，主谓一致准确率从初始1.0%提升至52.5%。此外，研究揭示了精确度与召回率之间的权衡关系——音变召回率的提升以身份准确性下降为代价，这一现象此前未被该领域文献报道。同时，实验还验证了泰米尔适配的指令模型Tamil-LLaMA-7B-Instruct在无任务微调情况下仅达19.0%零样本准确率，即使引入三示例提示也仅提升至24.7%，低于20.0%的复制基线，说明通用指令模型无法直接迁移至特定的句子级纠错任务，必须依赖任务特异性监督。
链接: https://arxiv.org/abs/2609.03273
作者: Karthikeyan A,Jaya Nirmala S,Sangeetha Sivanesan,Indhu R,Pranav Kumar,Bharat Jude Johnson,Vishnu Ram
机构: National Institute of Technology, Tiruchirappalli; Tamil University, Thanjavur
类目: Computation and Language (cs.CL)
备注:
Abstract:Tamil spell and grammar correction is challenging because Tamil is an agglutinative low-resource language with rich verbal morphology, complex sandhi (phonetic transformation) rules at word boundaries, and a script of 247 distinct letters. Prior work targets word-level surface errors with rule-based methods, statistical n-gram models, Minimum Edit Distance, or hybrid pipelines with a transformer re-ranker; such methods cannot reliably handle contextual errors - subject-verb agreement, tense consistency, or cross-word sandhi - which require sentence-level understanding. We propose an end-to-end sequence-to-sequence formulation and fine-tune mT5-small and mBART-50 on a synthetic corpus of up to 657,720 noisy-clean Tamil sentence pairs spanning ten error categories. Both backbones follow the same four-stage progressive schedule, each stage targeting one weakness: surface noise (v2), contextual grammar (v3), single-site sandhi (v4), and multi-site cross-word sandhi (v5). On a 1,000-sentence balanced diagnostic set verified disjoint from all training data, our best model, mBART-50 v5, reaches 69.3% top-1 exact-match accuracy, with 87.5% on sandhi and 43.5% on subject-verb agreement. The schedule is what produces these gains: subject-verb accuracy rises from 1.0% to 52.5% once contextual pairs are introduced, and sandhi from 0% to 87.5% once multi-site sandhi pairs are. We additionally quantify a precision-recall trade-off this literature has not reported: sandhi recall is paid for monotonically in identity accuracy. Finally, Tamil-LLaMA-7B-Instruct reaches 19.0% zero-shot and 24.7% with three demonstrations against a 20.0% copy baseline, showing that a Tamil-adapted instruction model does not transfer to specialised sentence-level correction without task-specific supervision.
[NLP-69] MedQA-MM: Shortcuts Behind Medical Visual Reasoning
【速读】： 该论文旨在解决医学多模态选择题（Medical Multimodal Multiple-Choice Questions, MCQs）中因评估指标仅关注最终答案正确性而引发的“路径依赖性误导”问题，即模型可能通过非目标视觉信息（如选项文本中的提示词、图像中的可见文字、人工标注痕迹或设备/上下文伪影等）获得正确答案，而非真正理解医学图像内容。这种现象导致评估分数虚高，形成“评分层面的过度解释”（score-level overinterpretation），从而掩盖了模型真实跨模态推理能力的缺失。其解决方案的关键在于构建一套系统性的验证框架：通过提示工程与图像侧审计、模态消融实验以及保持医学目标和正确答案不变的匹配修复方法，将候选干扰线索（cues）与真实行为证据（behavioral evidence）进行分离。实验结果表明，在13种配置的开放模型面板中，完整输入下的准确率为62.63%，而仅使用文本或仅使用选项时分别降至53.96%和29.71%；移除长度差异、显著性及空间位置相关线索后，准确率下降幅度达6.58、3.50和4.77个百分点。进一步构建的MedQA-MM数据集（1,000个经去捷径处理的样本）显示，文本单独和选项单独的准确率分别仅为5.21%和12.33%，有力证明模型在未获得有效图像支持时无法可靠作答。这说明，对模型是否具备真实医学图像推理能力的判断必须基于可观察的输入路径证据（route-level evidence），而非仅依赖最终答案的正确性。
链接: https://arxiv.org/abs/2609.03261
作者: Benlu Wang,Yifan Zhang,Jiaqing Yu,Chin Siang Ong,Juncheng Huang,Zhuohao Li,Zhenyu Zhang,Arman Cohan,Hong Yu,Zonghai Yao
机构: University of Massachusetts Amherst (马萨诸塞大学阿默斯特分校); Yale University (耶鲁大学); University of Massachusetts Lowell (马萨诸塞大学洛厄尔分校); VA Bedford Healthcare System (贝德福德退伍军人健康护理系统); Qingdao Medical College of Qingdao University (青岛大学青岛医学院); Yale School of Medicine (耶鲁大学医学院); National University Hospital, Singapore (新加坡国立大学医院); Zhejiang University (浙江大学); Stanford University (斯坦福大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL)
备注:
Abstract:A benchmark score credits final answers, but not the route by which an item can be answered. In medical multimodal multiple-choice questions (MCQs), this distinction matters because a correct answer can be supported by the intended image finding or by benchmark-preserved cues in the wording of answers, non-visual clinical text, visible image text, artificial annotations, or device/context artifacts. We call the resulting score-level overinterpretation reasoning inflation. Here, a route is an observable input path that can support answer selection, not a claim about the model’s hidden cognition. Across six medical multimodal MCQ datasets, we separate candidate cues from behavioral evidence through prompt- and image-side audits, modality ablations, and matched repairs that preserve the medical target and answer key. In a 13-configuration open-model panel, full-input accuracy is 62.63%, while text-only and options-only settings achieve 53.96% and 29.71%, respectively. Removing length-gap, absolute/conspicuous, and spatial/prepositional cues lowers accuracy by 6.58, 3.50, and 4.77 percentage points. We also construct MedQA-MM, a 1,000-item shortcut-mitigated subset, where text-only and options-only accuracy fall to 5.21% and 12.33%. This does not imply that models never use images; it shows that medical image-reasoning claims require route-level evidence.
[NLP-70] What Else Needs Fixing? Exploring Cost-Effective Test-Time Compute for Revision Propagation in Artifacts Generated Through Conversation EMNLP2026
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在对话式生成过程中，当用户仅提出局部修改请求时，如何准确识别并传播相关依赖关系以完成全局一致性修订的问题。其核心挑战在于，生成的产物及其上下文依赖可能隐含于对话历史中，而模型需具备从非显式结构中推断依赖关系的能力。解决方案的关键在于引入一种新的评估基准（benchmark），并在该基准上系统评估九种修订方法，包括顺序反思与并行采样等变体。实验结果表明，通过从三个并行采样结果中选择最优解（采用基于LLM或中位数（medoid）的选择策略）是最具成本效益的方法，相较基线可提升准确率2.2%至9.7%，显著增强了模型在低计算开销下的修订能力。
链接: https://arxiv.org/abs/2609.03254
作者: Daisuke Kikuta
机构: NTT, Inc.(日本电信电话公司)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: Accepted at EMNLP 2026 Industry Track. The code is available at this https URL
Abstract:Large Language Models (LLMs) often help users generate artifacts through iterative cycles of generation and revision in conversation. A challenge here is that, when users specify only a local change during revision, LLMs must instead identify the relevant dependencies and propagate the revision to all affected parts of the artifact. This paper studies this ability of LLMs on conversationally generated artifacts, where the artifact context and its dependencies may be embedded in the conversation history. Toward practical use, we also explore cost-effective test-time compute for this new setting. Specifically, we introduce a new benchmark for this setting, and evaluate nine revision methods, including sequential reflection and parallel sampling variants, using gpt-oss-20b/120b, gpt-5.4-mini, and qwen3.5-9b/27b/122b on the benchmark. The results show that baselines achieve accuracies of 68.3–93%, and the most cost-effective method is selecting from three parallel samples using either LLM-based or medoid selection, which improves accuracy by 2.2–9.7%. Our code and dataset are available at this https URL.
[NLP-71] SGD-KV: Summarization Guided KV Cache Compression NEURIPS2026
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在长上下文推理中因键值（Key-Value, KV）缓存随上下文长度线性增长而导致的严重内存瓶颈问题。现有KV缓存压缩技术多依赖于简单启发式方法，未能充分考虑不同注意力头在信息处理中所扮演的功能差异。本文提出一种面向注意力头的压缩框架SGD-KV（Summarization-Guided KV Cache Compression），其核心创新在于引入了一种新型的“块摘要诊断任务”（chunk-summarization diagnostic task），系统性地识别并优先保留那些在层次化信息聚合中起关键作用的注意力头。实验结果表明，SGD-KV在Qwen2.5-7B-1M和Qwen3-32B等多个长上下文基准上实现了高达100万标记（tokens）的上下文处理能力，且在保持卓越性能的同时，将KV缓存内存占用降低最高达75%。研究发现，基于注意力头摘要得分分布进行缓存预算的智能分配，可显著提升长上下文推理中的效率-精度权衡表现，是该方案的关键所在。
链接: https://arxiv.org/abs/2609.03235
作者: Zeyu Liu,Woomin Song,Xuandi Fu,Sai Muralidhar Jayanthi,Vivek Govindan,Aram Galstyan,Sravan Babu Bodapati,Srikanth Ronanki
机构: USC(南加州大学); KAIST(韩国科学技术院); Amazon AGI(亚马逊人工智能); Amazon AGI(亚马逊人工智能); Amazon AGI(亚马逊人工智能); Amazon AGI(亚马逊人工智能); Amazon AGI(亚马逊人工智能); Amazon AGI(亚马逊人工智能); Amazon AGI(亚马逊人工智能)
类目: Computation and Language (cs.CL)
备注: Accepted in NeurIPS2026 Efficient Reasoning Workshop
Abstract:Large language models (LLMs) face severe memory bottlenecks in long-context inference due to the linearly growing size of key-value (KV) caches. Existing KV cache compression techniques typically rely on simple heuristics, overlooking the distinct functional roles of different attention heads. We present SGD-KV (Summarization-Guided KV Cache Compression), a head-aware framework that leverages a novel chunk-summarization diagnostic task to systematically identify and prioritize attention heads specialized in hierarchical information aggregation. Experiments on Qwen2.5-7B-1M and Qwen3-32B across diverse long-context benchmarks demonstrate that SGD-KV achieves state-of-the-art performance with contexts up to 1M tokens, while reducing KV cache memory usage by up to 75%. Our findings show that strategically allocating the KV cache budget based on the summarization score distribution of attention heads yields a superior efficiency-accuracy trade-off for long-context inference.
[NLP-72] Counterfactual Fairness Audits of Multi-Step Clinical LLM Agents Require a Measured Per-Action Instability Floor
【速读】： 该论文旨在解决当前临床大语言模型（LLM）公平性评估中依赖反事实审计（counterfactual audit）所存在的根本性问题：反事实翻转率（flip rate）作为衡量算法是否存在差异性对待的核心指标，其本身在缺乏基准参照的情况下无法被有效解释。研究发现，在相同临床情景与患者描述重复运行十次的情况下，临床代理的行为仍存在高达8.7%的不稳定性，且该不稳定性在不同临床动作间差异显著（范围从0.022至0.179），远超任何真实人口学差异可能引发的变化。通过多模型验证，该不稳定性呈现出6.7%的聚合基线水平，并在六项临床动作中的排序高度一致（Spearman相关系数0.94，p=0.017），表明该基线非单一系统误差所致。进一步分析显示，通过五次采样多数投票可消除39%的不稳定性，但剩余部分仍由各单元格内异质性速率驱动，说明复制实验虽能缓解但无法彻底消除该噪声。因此，论文指出：若不提供每项操作的基线翻转率（per-action floor），任何反事实公平性估计均无法作为歧视性差异的可靠证据。研究采用FairMedAgent这一评估框架，其核心估量为“范围内反事实翻转率”（within-range counterfactual flip rate），仅统计符合既定决策规则与临床专家裁定之间的动作转换，且需依赖带边界裁定（band adjudication）机制；当前尚未提出具体不公平结论。所有合成病例均在固定形式条件下执行六阶段轨迹模拟，涵盖种族、性别、年龄、保险状况、英语能力及其交叉特征。研究公开了评估框架、基线协议及全部分析脚本，以推动可复现、可校准的临床生成式AI公平性评估。
链接: https://arxiv.org/abs/2609.03221
作者: Rohith Reddy Bellibaltu,Manpreet Singh,Deepak Parashar,Rahul Joshi
机构: Florida International University (佛罗里达国际大学); Boston University (波士顿大学); Symbiosis International University (辛比奥西斯国际大学); Manipal Institute of Technology, Manipal Academy of Higher Education (曼尼帕尔理工学院，曼尼帕尔高等教育学院)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 13 pages, 1 figure, 2 tables. Code and data: this https URL
Abstract:Counterfactual audits are the standard tool for checking whether a clinical agent treats demographically distinct but clinically identical patients differently. They report a flip rate: how often an action changes when only the patient descriptor changes. We show that this quantity is uninterpretable on its own. Re-running an identical condition ten times over sixteen vignettes (same narrative, same descriptor string, nothing varied) moved a clinical agent’s action in 8.7% of outcome-vignette cells, and instability was heterogeneous across actions by a factor of eight, from 0.022 for ICU escalation to 0.179 for controlled-substance caution. No demographic contrast in our data was distinguishable from that floor. A second model gives a pooled floor of 6.7% and ranks the six actions almost identically (Spearman 0.94, exact p=0.017), so the floor is not one system’s artefact. Majority-vote aggregation over five draws removes 39% of it and then flattens, and a null simulation attributes the residue to heterogeneous per-cell rates, so replication mitigates without eliminating. Any counterfactual fairness estimate reported without a per-action floor beside it therefore cannot be read as evidence of disparity. The measurements were taken with FairMedAgent, an evaluation harness for disparity in the actions of clinical LLM agents whose estimand, the within-range counterfactual flip rate, counts only flips between actions a published decision rule admits and a clinician has adjudicated. That estimand requires band adjudication, which is under way; no disparity result is claimed here. Each synthetic vignette runs a six-stage trajectory (five model-facing decisions around a deterministic environment step) under fixed-form conditions spanning race, sex, age, insurance, English proficiency, and their intersections. The harness, the floor protocol, and every analysis script are released.
[NLP-73] he Analyst in the Prompt: Role Retrieval and Memory Biases in LLM Financial Analysis
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在金融等高风险场景中因用户上下文（如角色设定、个人资料、记忆信息）引入的个性化偏差问题，尤其关注这种上下文如何影响模型对同一证据的不同解释，进而导致判断结果不一致。其核心问题是：在处理复杂文档（如美国证券交易委员会（SEC）文件）时，用户上下文是否通过改变证据解释方式而非仅影响证据检索，从而引发非客观的决策偏差。研究的关键发现是，大多数上下文溢出效应源于模型对相同证据的解释差异，而非检索内容的不同。为此，论文提出两种简化缓解策略：将投资者心态以用户资料形式表达而非作为助手角色，以及将基于证据的输出与个性化输出分离。这两种方法均能有效降低上下文溢出，但无法完全消除，且在不同模型间的有效性存在显著差异。
链接: https://arxiv.org/abs/2609.03218
作者: Ahmed Asaad,Amr Mohamed,Yang Zhang,Omneya Abdelsalam
机构: Durham University Business School (杜伦大学商学院); MBZUAI (穆罕默德·本·扎耶德国际人工智能大学); Ecole Polytechnique (巴黎综合理工学院); HBKU (卡塔尔高等教育局)
类目: Computation and Language (cs.CL); Computational Engineering, Finance, and Science (cs.CE); Portfolio Management (q-fin.PM)
备注:
Abstract:Large Language Models (LLMs) increasingly use user context such as memory, profiles, and role prompts to personalize their responses. This personalization can affect evidence-based judgment: the same evidence may lead to different conclusions under different user contexts. Finance provides a high-stakes setting to study this problem because decisions often depend on interpreting long and complex documents. We test this using 3,575 SEC filings across twelve LLMs. We compare persona-conditioned retrieval, neutral retrieval, and memory-framed context to separate the effect of evidence selection from the effect of interpretation. We find that most user-context spillover comes from how models interpret the same evidence under different roles, rather than from retrieving different evidence. We then test two simple mitigation strategies: expressing the same investor mindset as a user profile instead of an assistant role, and separating evidence-based and personalized outputs. Both reduce spillover, but neither removes it completely, and their effectiveness varies substantially across models.
[NLP-74] SWIM: Student Writing Simulation via Proficiency-Conditioned Generation EMNLP2026
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在模拟学生写作时，难以真实再现多维度写作能力差异的问题。具体而言，现有研究虽关注基于LLM的学生写作模拟，但尚未充分探索模型是否能有效模拟不同熟练度水平下学生在内容构建、逻辑组织、词汇选择与语言运用等方面的复杂变异。为此，论文提出SWIM（Student Writing sIMulation），将学生写作模拟任务定义为基于熟练度条件的作文生成问题，并采用自动作文评分（Automated Essay Scoring, AES）作为评估模型输出与目标熟练度特征匹配度的指标。研究对比了提示工程（prompting）、监督微调（Supervised Fine-Tuning, SFT）和强化学习（Reinforcement Learning, RL）三种方法的性能。结果表明，仅依赖提示工程难以实现对写作熟练度的有效控制，尤其在词汇、语法和篇章结构等维度上表现不佳；而SFT显著提升了各维度的匹配度，进一步引入基于熟练度对齐的奖励函数进行强化学习，使模型在所有写作特征和不同题目条件下均获得更优表现。研究发现，显式监督相较于纯提示方法能实现更强的熟练度特征对齐，但真实低熟练度写作的生成仍具挑战性。
链接: https://arxiv.org/abs/2609.03215
作者: Heejin Do,Jakub Kontak,Mrinmaya Sachan
机构: ETH Zurich (苏黎世联邦理工学院); ETH AI Center (苏黎世联邦理工学院人工智能中心)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: EMNLP 2026 Findings
Abstract:Writing proficiency manifests in how students develop content, organize ideas, choose words, and use language. Despite growing interest in LLM-based student simulation, whether LLMs can reproduce such multidimensional variation in extended writing remains largely unexplored. In this work, we explore if language models can realistically simulate student writing, and introduce SWIM, a task that formulates Student Writing sIMulation as proficiency-conditioned essay generation. We evaluate prompting, supervised fine-tuning (SFT), and reinforcement learning (RL) methods for writing simulation using automated essay scoring as a measure of profile alignment. Extensive experiments reveal that prompting provides limited proficiency control, even for strong proprietary LLMs with rubric-grounded strategies. In particular, while models can adjust content-oriented traits, they struggle to reproduce the lexical, grammatical, and organizational variation in different proficiency levels. SFT substantially improves alignment, while RL with the proposed proficiency-alignment reward yields further gains across all writing traits and essay prompts. Our findings suggest that explicit supervision enables substantially stronger profile alignment than prompting alone, while authentic low-proficiency writing remains challenging to reproduce.
[NLP-75] LLM s Learn Better In-Context from Rules than from Examples
【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）中不同学习模式的效能差异问题，具体比较基于规则描述的指令遵循（instruction following）与基于输入-输出示例的少样本提示（few-shot prompting）两种模式在相同任务下的学习效果。其核心解决方案的关键在于系统性地评估两种学习模式在多样化任务域（如游戏、算术、语言推理）中的表现，并揭示影响学习效能的模型与任务特性。研究发现，模型通常更可靠地从规则中学习，而非仅依赖示例；额外增加示例或单纯扩大示例数量并未带来一致且显著的性能提升。指令微调（instruction tuning）增强了规则学习的优势，同时保持了示例学习的能力。令人意外的是，在基础模型中并未观察到示例学习的天然优势，且在代数类任务中规则仍具有显著优势。总体而言，当任务涉及代数抽象与计算时，规则学习的相对优势更大；而当任务需要分布敏感性或依赖参数化知识时，该优势则相对较小。
链接: https://arxiv.org/abs/2609.03213
作者: Xiang Fu,Seungmin Cho,Yukyung Lee,Najoung Kim
机构: Boston University (波士顿大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Large language models (LLMs) exhibit in-context learning capabilities, where they can learn new tasks from prompt contexts without weight updates. We compare the learning efficacies of two prominent modes of in-context learning: (1) learning from descriptions of rules (instruction following); and (2) learning from examples of input-output demonstrations (few-shot prompting). Through five learning tasks that cover diverse domains (games, arithmetic, linguistic inferences), we compare two modes of learning (rules vs. examples) specifying the same underlying task. We furthermore explore model and task properties that modulate the learning efficacies. We find that models generally learn more reliably from rules than from examples alone, and additional examples on top of rules or simply scaling up the number of examples do not lead to consistent and significant gains. Instruction tuning amplifies the benefit of rule-based learning while keeping example-based learning capacities intact. Surprisingly, we find no privileged effect of example-based learning in base models, and rules still lead to gains in algebraic task domains. Overall, the comparative efficacy of rules over examples is larger when the task recruits algebraic abstractions and computations, and smaller when the task requires distributional sensitivity and/or recruits parametric knowledge.
[NLP-76] Learning to Zoom Efficiently with a Contrastive Curriculum EMNLP2026
【速读】： 该论文旨在解决大语言模型（LLM）在处理高分辨率图像任务时，依赖于大量监督微调（SFT）来学习缩放工具（zoom-in tool）使用的问题。传统方法需要通过长时间的有监督预训练阶段来教会模型如何有效使用缩放功能，这不仅成本高昂且难以扩展。本文提出了一种无需额外标注或初始SFT的新型内在奖励机制，基于InfoNCE风格的设计，利用逐渐增加难度的负样本工具调用作为对比学习信号，从而在无监督条件下实现对缩放工具的有效学习。该方案的关键在于设计了一个可扩展的、自适应的负样本课程（curriculum of increasingly hard negative tool calls），使模型能够通过对比学习逐步理解何时以及如何正确使用缩放操作。在V^*、HRBench和MME-RealWorld等多个基准上的实验证明，该方法在性能上具有竞争力且训练效率更高；当作为SFT的直接替代方案时，甚至优于所有基线模型。为进一步精确评估模型的缩放能力，作者构建了可扩展的合成数据集MuffinChihuahua（MC），其图像由包含“松饼”或“奇瓦瓦犬”的网格组成，并引入独特的感兴趣区域标签。分析表明，召回率（recall）是与最终任务表现最强烈相关的指标，反映出模型精准定位目标区域的能力。该研究提出的模型与代码已公开发布。
链接: https://arxiv.org/abs/2609.03206
作者: Falko Helm,Iryna Gurevych
机构: Ubiquitous Knowledge Processing Lab (UKP Lab)(通用知识处理实验室); Department of Computer Science and Hessian Center for AI (hessian.AI)(计算机科学系和黑森州人工智能中心); Technical University of Darmstadt(达姆施塔特工业大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL)
备注: EMNLP 2026
Abstract:Using a zoom-in tool is an important foundational part of modern visual agents, because it allows to efficiently handle tasks involving high-resolution images. Most previous methods need an extensive warm-start supervised fine-tuning phase for teaching models zoom-in. We show that this is not necessary by proposing a new intrinsic reward for learning tool use in MLLMs without the need for additional labels or warm-start SFT. Our InfoNCE-style reward uses a curriculum of increasingly hard negative tool calls as a contrastive training signal. Empirical experiments on V^* , HRBench and MME-RealWorld show that our approach is competitive while being more efficient. When used as a drop-in replacement for SFT, we even outperform all baselines. To directly measure the zoom-in ability of models, we further introduce the scalable synthetic MuffinChihuahua (MC) dataset. Each image consists of a grid with every cell either showing a muffin or chihuahua. Leveraging the MC dataset’s unique region of interest labels, we find that recall is the metric that most strongly correlates the zoom-in region with final task performance. Our model and code for reproduction is publicly available under this https URL
[NLP-77] VoxReason : Listener-Free Evaluation of Source-Grounded Speech Planning Before Synthesis
【速读】： 该论文旨在解决生成式语音系统在合成前就做出表达决策（如情感、语调、节奏、停顿等）时，缺乏对这些决策是否基于真实源文本记录的可验证性问题，即“源使用失败”（source-use failure）——此类错误发生在波形生成之前，却难以通过下游音频质量指标察觉。其核心解决方案是提出VoxReason，一个无需听者的、面向源文本锚定的语音规划评估框架。VoxReason在合成前对语音表达计划进行可度量验证，要求系统输出包含证据引用的源引述说话计划，并由确定性验证器检查引用合法性、槽位一致性、无支持状态、模式有效性及单线索反事实局部性。实验表明，仅依赖槽位准确率存在严重缺陷：即使在未见过的源键情况下，情绪先验仍能维持0.958的高槽位准确率但无法引用来源；而通过7B模型的局部性监督微调与反事实修复，计划槽位准确率与局部性分别从0.684/0.141提升至0.919/1.000，且移除源记录后引证所需的基础性得分下降0.488，证明了源引用对表达合理性的重要性。关键创新在于将语音表达规划的“源地基性”（source-groundedness）转化为可形式化验证的前置任务，从而实现对生成式语音系统中隐含决策的透明化与可信评估。
链接: https://arxiv.org/abs/2609.03203
作者: Mengzhe Geng
机构: National Research Council Canada(加拿大国家研究委员会)
类目: ound (cs.SD); Computation and Language (cs.CL); Machine Learning (cs.LG); Audio and Speech Processing (eess.AS)
备注:
Abstract:Expressive speech systems make a decision before any waveform is rendered: how an utterance is delivered. In dialogue agents, narration, and role-conditioned TTS, that hidden planning step sets affect, pitch, energy, rate, pause, emphasis, and stance, yet downstream audio scores rarely reveal whether those choices were licensed by the source record, a source-use failure that occurs before any waveform exists. VoxReason makes that pre-synthesis decision measurable as a listener-free task for source-grounded speech planning. Before synthesis, VoxReason measures whether delivery choices are grounded in cited source records. Systems output a source-cited speaking-plan with evidence citations, and a deterministic verifier checks citation legality, slot agreement, unsupported state, schema validity, and one-cue counterfactual locality. On 1,440 checked source-label cases, shortcut controls show why slot accuracy alone is unsafe: a key-lookup oracle reaches 1.000 plan-slot accuracy on seen keys, while an emotion prior still reaches 0.958 slot accuracy on source-key-disjoint cases without citing intensity or identity. In a separate 100-case learned source-key-disjoint comparison, a 7B locality SFT+CF repair improves plan-slot accuracy/locality from 0.684/0.141 to 0.919/1.000, and removing source records lowers citation-required grounded score by 0.488. Rendered waveform quality remains outside the present evaluation.
[NLP-78] MemoryLACE: Memory Lifecycle-Aware Consolidation and Evidence Retrieval
【速读】： 该论文旨在解决长期大语言模型（LLM）代理在多轮交互中如何有效维护信息、区分重复证据、历史状态、更新内容及未解决矛盾的问题。现有文本记忆系统虽能高效检索语义相关记忆，但往往隐含了这些复杂关系；而更复杂的结构化方法则依赖全局图谱、层次抽象或全局反思机制，带来较高的计算开销。本文提出轻量级记忆框架MemoryLACE（MemLACE），其核心创新在于显式建模文本证据的生命周期，通过稀疏合并（sparse merge）、覆盖（supersession）与矛盾（contradiction）三类关系，同时保留原始自然语言记忆及其来源（provenance）。不同于独立检索记忆的方式，MemLACE重构出具备关系感知的证据单元，可明确揭示当前、历史、支持性与冲突性证据，为下游推理提供上下文清晰的信息支撑。在BEAM和StructMemEval基准上，使用开源与专有模型作为后端，MemLACE在同架构对比中表现最优，并相较最强的回溯式记忆基线Hindsight将端到端运行时间减少66.6%。消融实验表明，生命周期扩展与时间感知是性能提升的关键因素。研究结果证明，仅通过显式建模文本证据的局部生命周期，即可显著提升长时记忆推理能力，无需依赖全量知识图谱或全局反思机制。
链接: https://arxiv.org/abs/2609.03201
作者: Meriem Yacoubi,Pia Schmidt,Nenad Petrovic,Ahmed Frikha,Martin Kirchhoff,Alois Knoll
机构: Technical University of Munich (慕尼黑工业大学); inovex GmbH; Cerebras Systems Inc.
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 8 pages, 2 figures, 4 tables
Abstract:Long-term LLM agents must preserve information across interactions while distinguishing repeated evidence, historical states, updates, and unresolved contradictions. Existing textual memory systems retrieve semantically relevant memories efficiently but often leave these relationships implicit, whereas richer structured approaches model them through global graphs, hierarchical abstractions, or reflection at greater complexity. We introduce MemoryLACE (MemLACE), a lightweight memory framework that explicitly models the lifecycle of textual evidence through sparse merge, supersession, and contradiction relations while preserving atomic natural-language memories and their provenance. Rather than retrieving memories independently, MemLACE reconstructs relation-aware evidence units that expose current, historical, supporting, and conflicting evidence for downstream reasoning. Across BEAM and StructMemEval, using open-weight and proprietary LLM backbones, MemLACE achieves the highest overall performance in same-backbone comparisons while reducing end-to-end runtime on BEAM by 66.6% relative to Hindsight, the strongest reported reflective-memory baseline. Ablation studies identify lifecycle expansion and temporal awareness as the principal contributors to these gains. Together, the results demonstrate that explicitly modeling the local lifecycle of textual evidence is sufficient to substantially improve long-term memory reasoning without requiring comprehensive knowledge graphs or global reflection.
[NLP-79] Jina-OCR-v1: Efficient Document Parsing with Speculative Decoding and Dense Verifiable Rewards
【速读】： 该论文旨在解决在资源受限的低预算GPU上实现高效、高精度文档解析的问题。现有文档理解模型通常依赖高性能计算设备，难以在边缘或低成本硬件上部署。其解决方案的关键在于提出Jina-OCR-v1，一个端到端的文档解析模型，通过结合压缩视觉编码器与30亿参数（3B）的专家混合（Mixture-of-Experts, MoE）解码器，并仅激活约5.7亿参数/令牌，显著降低计算开销；同时引入FastMTP推测解码头，利用单个草稿块在连续K=3步预测中递归复用，配合贪婪验证（Greedy Verification）实现无损解码，从而在不牺牲准确率的前提下大幅提升推理速度。此外，后训练阶段采用指令对齐、针对复杂文档的鲁棒性微调以及基于密集可验证奖励（如确定性公式、表格结构和文档层级检查）的GRPO优化策略，实现部分得分激励，进一步提升模型在真实场景中的泛化能力。训练数据融合清洗后的公开语料与针对性合成页面，增强了模型对多样性和噪声的适应性。实验表明，在动态分辨率默认设置下，Jina-OCR-v1在OmniDocBench v1.6和olmOCR-Bench上分别取得91.14和83.4的分数，且在同等条件下达到2.57页/秒的最高页面吞吐量，在NVIDIA L4等低预算GPU上，FastMTP使解码速度相较传统贪心自回归解码提升一倍。
链接: https://arxiv.org/abs/2609.03181
作者: Alejandro Barón García,Feng Wang,Emilia Garcia Casademont,Han Xiao
机构: Jina AI by Elastic; 33 New Montgomery Street, San Francisco, CA 94105, USA
类目: Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV)
备注: 15 pages, 5 figures, 8 tables. Model at this https URL
Abstract:We present Jina-OCR-v1, an end-to-end document parsing model built to serve on low-budget GPUs. It combines the compressed-vision encoder and the 3B mixture-of-experts decoder of DeepSeek-OCR, which activates about 570M parameters per token, with a FastMTP speculative decoding head that shares a single draft block recursively across K=3 prediction steps. Greedy verification makes decoding lossless. Post-training combines instruction alignment, robustness fine-tuning on difficult documents, and GRPO under dense verifiable rewards: deterministic formula, table, and structural checks that award partial credit. The training data mixes cleaned public corpora with targeted synthetic pages. At the default dynamic-resolution setting, Jina-OCR-v1 scores 91.14 on OmniDocBench v1.6 and 83.4 on olmOCR-Bench, and reaches the highest page throughput in our comparison at 2.57 pages per second. On a low-budget GPU such as the NVIDIA L4, FastMTP doubles decoding speed over greedy autoregressive decoding. The model is publicly available at this https URL.
[NLP-80] No country for old linguists: LLM -brain alignment underdetermines neural computation
【速读】： 该论文旨在解决生成式人工智能（Generative AI）与人类大脑在语言处理机制之间是否存在可解释的对应关系这一核心问题，尤其聚焦于大语言模型（Large Language Models, LLMs）与脑神经活动之间的表征对齐是否足以支持将LLMs视为生物大脑语言处理的机械性模型。其解决方案的关键在于指出：尽管表征对齐能够在原则上限制神经机制的假设空间，但仅凭对齐本身无法确定具体的计算机制或架构；作者批判了Nastase等人（2026）从表征相似性推导出“共享计算原则”乃至“完全机械性模型”的逻辑跃迁，指出了该主张中存在的逻辑、因果与计算层面的不确定性问题，强调表征一致性不等于机制等价性，从而揭示了当前LLM-脑对齐研究中过度解读的潜在风险。
链接: https://arxiv.org/abs/2609.03160
作者: Elliot Murphy
机构: 未知
类目: Computation and Language (cs.CL)
备注:
Abstract:Nastase et al. (2026) argue that large language models (LLMs) may illuminate language processing because both rely on distributed, context-sensitive representations shaped by statistical learning. Their rejection of simple cortical “boxology” is persuasive, and they articulate a strong case for the value of LLM-brain alignment research. The key question is what kind of inference LLM-brain alignment licenses. My claim here will be narrow: representational alignment can in principle constrain mechanistic hypotheses, but it does not by itself identify a mechanism. Nastase et al. acknowledge that an encoding model can capture features represented in neural activity without establishing a shared architecture or algorithm. Yet the authors sometime move from alignment to “shared computational principles” and ultimately to LLMs as mechanistic models of natural language. Indeed, their methodological caveat that alignment does not establish a shared architecture or algorithm sits uneasily with their conclusion that LLMs might instantiate the same computational principles as biological brains and provide a “fully mechanistic model” of language. I discuss what I consider to be problems of logical, causal, and computational underdetermination in Nastase et al.'s (2026) proposal.
[NLP-81] Who Speaks for the Pruned? Visual Token Pruning as Coverag e Optimization EMNLP2026
【速读】： 该论文旨在解决视觉-语言模型（Vision-Language Models, VLMs）中视觉标记（visual token）剪枝导致的冗余信息保留与关键特征丢失问题。传统剪枝方法仅关注“保留哪些标记”，易导致高分冗余标记被保留，而重要但低分标记的代表信息被舍弃。为此，作者提出无需训练的CoverPruner方法，其核心在于从需求侧重构剪枝逻辑：在移除某一标记后，应由哪个剩余原始标记来代表其语义。该方法将剪枝问题建模为表示覆盖最大化（Representational Coverage Maximization, RCM），通过投影空间中的覆盖度量与轻量级第一层注意力探针实现高效计算。实验表明，CoverPruner在多种VLM架构和压缩率下均取得最优平均准确率，尤其在激进压缩条件下表现显著优于现有方法。
链接: https://arxiv.org/abs/2609.03158
作者: Qingchan Zhu,Weihang You,Hanqi Jiang,Changdi Yang,Tianming Liu,Geng Yuan
机构: University of Georgia (佐治亚大学); Northeastern University (东北大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: Accepted to EMNLP 2026 main
Abstract:Visual token pruning reduces the inference cost of vision-language models (VLMs), but most methods only ask which tokens to keep. This retained-token view can keep redundant high-scoring tokens while leaving discarded evidence without a close representative. We propose CoverPruner, a training-free pruner that asks the complementary demand-side question: after a token is removed, which surviving original token represents it for the target VLM? CoverPruner formulates pruning as Representational Coverage Maximization (RCM), covering the full projected visual-token set with query-weighted demand. It instantiates RCM with projector-space coverage and a lightweight first-layer attention probe. Across multiple VLM architectures and compression rates, CoverPruner achieves the best average accuracy among all compared methods, with the largest gains usually appearing under aggressive compression.
[NLP-82] Routing Is Not Enough: Diagnosing Intra-Adapter Subspace Contention in MoELoRA Fine-Tuning EMNLP2026
【速读】： 该论文旨在解决多领域微调中混合专家模型（MoE）与低秩适应（LoRA）结合时出现的负迁移问题。尽管现有方法假设基于令牌的路由能够实现领域特定更新的分离，但实验表明，即使在近似不重叠的专家路由情况下，引入生物医学文本仍显著增加代码任务的困惑度，说明仅依赖路由分离不足以防止不同领域间的干扰。其关键发现是：干扰主要源于不同领域梯度在低秩适配器子空间内近乎正交，导致更新冲突。为此，作者提出一种新方法——SpawnLoRA，当检测到适配器层级的竞争时，动态在MoE专家内部添加带有门控机制的子适配器，从而在不改变路由器的前提下实现适配器内部的结构化分离。通过引入雅各比路由重叠度与适配器梯度余弦相似性作为诊断指标，验证了该方案能有效缓解负迁移，在Phi-tiny-MoE-instruct和OLMoE-1B-7B等多个混合设置下均优于标准LoRA与秩自适应LoRA。结果表明，专家内部的结构性分离相较于单纯的路由或秩扩展更具优势。
链接: https://arxiv.org/abs/2609.03150
作者: Mehreen Hossain Chowdhury,Nowshin Mahjabin,Ahmed Shafin Ruhan,Md Azam Hossain,Abu Raihan Mostofa Kamal,Md Tahmid Rahman Laskar
机构: Islamic University of Technology(伊斯兰科技大学); York University (约克大学); Dialpad Inc.(Dialpad公司)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: 13 pages, 1 figure, 16 tables. Accepted to Findings of the Association for Computational Linguistics: EMNLP 2026
Abstract:Multi-domain fine-tuning often combines MoE routing with LoRA, assuming that token-level routing separates domain-specific updates. We test this assumption in MoE+LoRA using Python code paired with biomedical text and mathematical reasoning. Although these domains show near-disjoint expert routing, adding biomedical data substantially increases code perplexity, indicating that routing separation alone may not prevent negative transfer. To localize the failure, we introduce Jaccard routing overlap and adapter-gradient cosine similarity, which measure expert sharing and update compatibility, respectively. These diagnostics indicate that interference arises mostly from nearly orthogonal domain gradients competing within the same low-rank adapter subspace. We address this issue with SpawnLoRA, which dynamically adds gated sub-adapters inside MoE experts when adapter-level contention is detected, while keeping the router fixed. We evaluate SpawnLoRA on Phi-tiny-MoE-instruct and OLMoE-1B-7B across multiple mixture settings and find that it effectively reduces negative transfer compared with standard and rank-adaptive LoRA. These results demonstrate that structural separation inside experts provides benefits beyond routing or rank expansion alone.
[NLP-83] Large Language Models in Resolving Contextual Knowledge Conflicts EMNLP2026
【速读】： 该论文旨在解决大语言模型（LLM）在处理上下文知识内部冲突时的挑战，即当外部提供的上下文信息本身存在矛盾或不一致时，模型如何识别并有效化解这些冲突。与以往研究聚焦于模型参数知识与外部上下文之间的冲突不同，本文关注的是上下文内部的六类冲突类型：事实性、推理性、时间性、粒度性、视角性和模糊性，并构建了名为ContextConflict的综合性数据集，包含5,781个样本，覆盖推理与摘要任务，涵盖显式矛盾与需多步推理才能识别的隐式冲突。实验表明，当前主流九种大语言模型在解决此类冲突方面仍表现不足。通过机制可解释性分析，研究揭示了模型对冲突的潜在感知能力及其背后的表征几何结构，同时发现模型存在对早期证据的系统性偏好，这一位置偏差成为有效冲突解决的关键障碍。基于此，作者提出一种无需训练、无需标签的激活引导方法，通过调节神经激活以促进对多源证据的更全面整合，显著提升了推理任务的准确率和摘要任务的生成质量与平衡性。
链接: https://arxiv.org/abs/2609.03148
作者: Xinye Yang,Zhenyang Liu,Ruisi Li,Yuanyuan Lei
机构: Northwestern University (西北大学); New York University (纽约大学); University of Florida (佛罗里达大学)
类目: Computation and Language (cs.CL)
备注: Accepted to EMNLP 2026
Abstract:Most prior works focused on conflicts between an LLM’s internal parametric knowledge and externally provided context. In contrast, we investigate how LLMs handle conflicts that arise within contextual knowledge itself. We introduce a taxonomy of six types of contextual conflicts (factual, inferential, temporal, granularity, perspective, and ambiguity) and contribute a comprehensive dataset ContextConflict for this setting. The dataset contains 5,781 samples, covers both reasoning and summarization tasks, and includes both explicit contradictions and implicit conflicts that require multi-step reasoning. Experiments on nine LLMs show that current models still fall short in resolving contextual knowledge conflicts. We further provide mechanistic interpretability insights into how LLMs process such conflicts, revealing their latent awareness of conflicts and the representational geometry underlying conflict processing. In addition, our analysis uncovers a consistent model bias towards earlier evidence, and this positional preference serves as a key obstacle to effective conflict resolution. Motivated by these findings, we further propose a simple training-free, label-free steering method that steers activations to encourage a more comprehensive incorporation of evidences for better conflict resolution. On our dataset, the method consistently improves accuracy on reasoning tasks and generates higher-quality, more balanced summaries for summarization tasks.
[NLP-84] Unifying Conformal Language Tasks with In-Context Ensembles EMNLP2026
【速读】： 该论文旨在解决自然语言处理（NLP）任务中信息检索的两大核心约束问题：覆盖率（coverage）与简洁性（conciseness）。其中，覆盖率要求保留足够相关的信息以达成任务目标，而简洁性则需尽可能剔除无关内容。现有方法采用符合性预测（conformal prediction）来保障覆盖率，但其简洁性依赖于评分函数（score function）的设计，而当前最先进的评分函数依赖人工设计的大语言模型（LLM）提示词（prompt），存在劳动密集且任务特异性高的缺陷。为此，本文提出“符合性相关性”（Conformal Relevance）框架，通过上下文学习中的示例筛选（in-context learning example curation）与集成学习（ensembling）构建评分函数，在最小化人工干预的前提下同时维持覆盖率并提升简洁性。作者在七个NLP任务上验证了该框架的有效性，并从理论上分析了集成评分的多样性影响，提出了一个互补性条件（complementarity condition），用于刻画集成能否改善最坏情况下的句子评分，以及给出了集成性能提升的饱和边界（saturation bound）。
链接: https://arxiv.org/abs/2609.03005
作者: Xiao Shi Huang,Chen-Yuan Lin,Bruce Kuwahara,Kin Kwan Leung,Jesse C. Cresswell
机构: Signal 1 AI; Layer 6 AI
类目: Computation and Language (cs.CL); Machine Learning (cs.LG); Machine Learning (stat.ML)
备注: Findings of EMNLP 2026. Code is available at this https URL
Abstract:Many NLP tasks, such as summarization and extractive question answering, reduce to retrieving relevant content from documents under two constraints: coverage, retaining enough pertinent information to achieve some goal, and conciseness, removing as much irrelevant information as possible. Conformal prediction methods have been used to guarantee coverage, and must be optimized for conciseness through design of a score function. State-of-the-art scoring functions use hand-engineered LLM prompts asking the model to rate the importance of content, but manual prompt engineering is labor-intensive and task-specific. We introduce the Conformal Relevance framework which uses in-context learning example curation and ensembling to create a score function which maintains coverage while improving conciseness with minimal manual input. We demonstrate this framework’s application on seven NLP tasks, and also theoretically study the impact of diversity for ensembled conformal scores, giving a complementarity condition that characterizes when ensembling improves worst-case sentence scores, and a saturation bound on ensemble improvement.
[NLP-85] Verify Before You Distill: Prompt-Level Teacher Gating for On-Policy Distillation
【速读】： 该论文旨在解决在策略蒸馏（On-policy Distillation, OPD）中，由于教师模型在部分提示（prompt）上存在高置信度错误而引发误导性更新的问题。传统OPD对所有提示统一施加密集的令牌级监督，未评估教师在具体提示上的可靠性，而逆KL散度的模式聚焦特性会使教师的错误预测导致学生模型产生强烈但错误的优化方向。现有基于分布的代理指标（如熵或教师-学生似然一致性）虽能反映不确定性或一致性，却无法直接验证输出结果的正确性。为此，论文提出教师门控的在线策略蒸馏（Teacher-Gated On-Policy Distillation, TGOPD），其核心在于：在应用密集监督前，先在提示层面验证教师的可靠性。TGOPD通过少量经验证器评分的教师探针样本估计教师可靠性，并仅在可靠性通过时将提示路由至密集OPD，否则采用基于验证器的GRPO（Generalized Reward Policy Optimization）进行训练。实验表明，TGOPD在数学、代码生成和指令遵循等单领域任务中均优于原始OPD，且在多领域训练下于40亿与350亿参数规模的学生模型上均实现更高的跨基准平均性能。此外，该方法利用原本闲置的教师计算资源进行可靠性评估，在异步OPD场景中显著提升教师节点GPU利用率，从9.8%提升至78.9%。
链接: https://arxiv.org/abs/2609.02998
作者: Zhiwei Zhang,Zechen Sun,Fei Zhao,Kang Peng,Bin Liang,Huayu Deng,Yao Hu,Kam-Fai Wong,Mu Chuan
机构: AllSpark Team
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 17 pages, 6 figures, 7 tables
Abstract:On-policy distillation (OPD) accelerates post-training by providing dense token-level supervision from a frozen teacher on the student’s own rollouts. Vanilla OPD applies this supervision uniformly across prompts, without checking whether the teacher is reliable for each prompt. Because reverse KL is mode-seeking, a confidently wrong teacher can induce a strong yet misleading update. Distributional proxies, such as entropy or teacher-student likelihood agreement, measure uncertainty or agreement but do not directly verify outcome correctness. We introduce Teacher-Gated On-Policy Distillation (TGOPD), built on the principle that teacher reliability should be verified at the prompt level before dense supervision is admitted. TGOPD estimates reliability from a small set of verifier-scored teacher probes and routes each prompt exclusively to dense OPD when the reliability check passes or to verifier-grounded GRPO otherwise. Across 4B and 35B students in mathematics, code, and instruction following, TGOPD outperforms Vanilla OPD in all six single-domain settings and achieves higher seven-benchmark averages at both scales under multi-domain training. By using otherwise-idle teacher capacity for reliability estimation, TGOPD also reduces teacher-side compute waste in asynchronous OPD, increasing teacher-node GPU utilization from 9.8% to 78.9% in the measured 4B single-domain run.
[NLP-86] he Geometry of Ignorance: LLM s Know When to Temper Bayesian Priors
【速读】： 该论文旨在解决语言模型在上下文信息稀疏时的预测机制问题，即当模型缺乏足够线索进行推理时，其预测行为如何演化。核心问题是揭示模型在不确定性状态下的内在先验依赖性及其数学表征。解决方案的关键在于发现并验证“无知方向”（direction of ignorance）这一几何结构：在所有考察的模型家族（Llama、Qwen、Gemma、Pythia，参数规模从0.4B到405B）中，解嵌入矩阵（unembedding matrix）的一个特定方向编码了训练语料库的单词频次分布（unigram distribution），该分布作为模型在不确定情况下的贝叶斯先验。通过将最终预测状态投影至该方向，可得到一个与每令牌相关的先验加载因子λ，其值随上下文信息量增加而持续下降。形式上，该投影将预测状态分解为两个正交向量，分别对应于温度化的贝叶斯更新中的先验项（以λ为指数的单个词先验）和由上下文驱动的似然项。这一几何-概率解释使λ具备跨模型规模与家族的可比性，并表明大模型在高上下文条件下对先验的依赖程度更低。此外，研究通过因果干预证明该方向具有实际影响：调节λ值可显著改变预测结果与单个词先验之间的KL散度，从而证实该结构在模型决策中的因果作用。
链接: https://arxiv.org/abs/2609.02959
作者: Toni J.B. Liu,Jiajun Bao,Yizhou Liu,Gurbir Arora,Nicolas Boullé,Raphaël Sarfati,Christopher J. Earls
机构: Cornell University(康奈尔大学); MIT(麻省理工学院); Imperial College London(帝国理工学院); Goodfire AI(古德火人工智能)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (stat.ML)
备注:
Abstract:What does a language model predict when it has few clues? The answer lurks in its unembedding geometry: a single direction of the unembedding matrix encodes the unigram distribution of the training corpus, which serves as the Bayesian prior the model falls back on when uncertain. This structure — which we term the \emphdirection of ignorance — appears in all four model families examined (\textttLlama, \textttQwen, \textttGemma, and \textttPythia), ranging from 0.4B to 405B parameters. Projecting the final prediction state onto this direction yields a per-token \emphprior loading factor \lambda , which, empirically, declines steadily as the context becomes more informative. Formally, the same projection decomposes the prediction state into two orthogonal vectors that correspond exactly to the two factors of a tempered Bayesian update: a unigram prior raised to the exponent \lambda and a context-driven likelihood. This geometric-probabilistic interpretation calibrates \lambda , making it meaningfully comparable across model sizes and families, with larger models generally exhibiting lower prior reliance in the high-context limit. Finally, we show that the direction of ignorance is causally active: raising or lowering \lambda at the final prediction state steers the prediction toward or away from the unigram prior in KL divergence.
[NLP-87] LexIssue: Benchmarking Legal Issue Identification in Chinese Civil Litigation
【速读】： 该论文旨在解决法律人工智能（Legal AI）研究中对诉讼争议问题识别（legal issue identification in litigation）这一关键环节的忽视问题。其核心挑战在于如何在真实诉讼场景中准确识别双方当事人争议的法律问题，并对其进行结构化建模与计算化处理。解决方案的关键在于提出一个基于法律语义的分层式表示框架，通过自由形式的问题描述与结构化法律类别相结合的方式，将争议问题建模为两个互补任务：法律问题生成（legal issue generation）与法律问题分类（legal issue classification）。为此，研究构建了首个中文民事诉讼案例基准数据集LexIssue，包含430个真实案例及1,303条专家标注的争议法律问题，并进一步开发了一个以问题为中心的法律知识库，覆盖27类诉因和441个候选法律问题条目，支持检索增强型推理（retrieval-augmented reasoning）。实验结果表明，结合该知识库的检索增强生成方法在多种模型上均显著提升了争议法律问题及其对应法律属性的识别性能。
链接: https://arxiv.org/abs/2609.02954
作者: Huiyuan Xie,Yuqin Huang,Zhicheng Hao,Yida Cai,Shaochun Wang,Zhenghao Liu,Yuxiao Ye
机构: Tsinghua University (清华大学); Peking University (北京大学); Modelbest Inc.(模型最佳公司); Northeastern University (东北大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Identifying the issues disputed between litigating parties is a crucial component of real-world litigation. However, legal issues remain comparatively underexplored in legal AI research. In this work, we study the computational modelling of legal issue identification in litigation. We introduce a legally grounded hierarchical schema that represents legal issues through both free-form issue descriptions and structured legal categories, and formulate legal issue identification as two complementary tasks: legal issue generation and legal issue classification. Based on this formulation, we construct LexIssue, a benchmark containing 430 real-world Chinese civil litigation cases and 1,303 expert-annotated disputed legal issues. We further develop an issue-centric legal knowledge base spanning 27 causes of action and 441 candidate legal issue entries to support retrieval-augmented reasoning. Experimental results across a diverse set of models show that retrieval-augmented generation using the constructed legal issue knowledge base consistently improves performance in identifying disputed legal issues and their corresponding legal attributes.
[NLP-88] Privacy-Preserving Heterogeneous Multi-LLM Federated Inference for Cognitive Diagnosis EMNLP2026
【速读】： 该论文旨在解决生成式 AI（Generative AI）驱动的教育系统中隐私保护与精准认知诊断之间的矛盾问题。其核心挑战在于如何在不暴露学生原始数据或模型内部参数的前提下，实现高效、可信的认知状态评估。解决方案的关键在于提出一种基于联邦推理（federated inference）的异构多大语言模型（multi-LLM）架构，通过多个商业大模型API（如LLaMA-3.3-70B、GPT-4o-mini和Claude-3-Haiku）协同工作，利用本地差分隐私（epsilon-local differential privacy）机制，在各参与方对预测结果添加拉普拉斯噪声后进行聚合，从而有效防止外部对诊断结果的推断攻击；同时采用基于残差的聚合策略缓解模型异构性带来的偏差，保障整体诊断性能。该框架建立在“诚实但好奇”（honest-but-curious）的信任假设之上，既保证了数据隐私，又实现了高精度的认知诊断，经多项真实教育基准测试验证，展现出优异的实用性与跨领域泛化能力。
链接: https://arxiv.org/abs/2609.02947
作者: Yagna Manasa Boyapati,Chong Yu,Tianyu Jiang,Justin Zhan
机构: University of Cincinnati(辛辛那提大学)
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 16 pages, 3 figures, Accepted to Findings of EMNLP 2026
Abstract:Significant challenges remain in AI-driven educational systems in balancing privacy preservation with accurate cognitive diagnosis. To overcome this, we propose a federated inference framework in which several commercial LLM APIs collaborate without requiring access to raw student data or proprietary model internals. Using multiple federated entities, such as LLaMA-3.3-70B, GPT-4o-mini, and Claude-3-Haiku, our framework builds upon a heterogeneous multi-LLM architecture. The predictions generated by these entities are combined with epsilon-local differential privacy by adding Laplace noise locally to each entity’s prediction output before aggregation, while residual-based aggregation mitigates model heterogeneity. Our approach is predicated on an honest-but-curious trust paradigm in which API providers are presumed not to abuse submitted queries, and our differential privacy mechanism shields the published diagnostic results from external inference. We conduct rigorous privacy-utility analysis showing strong privacy guarantees with minimal accuracy loss, and extensive real-world evaluations across three educational benchmarks confirm the framework’s practical usability and cross-domain generalizability.
[NLP-89] Judging LLM -as-a-Judge: Concerning Rubric Artifacts in LLM -based Automated Text Generation Evaluation 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）的自动化文本评估方法中，其评价结果是否真正依赖于对生成文本与评分标准（rubric）之间推理关系的可信性问题。研究发现，仅基于评分标准文本训练的分类器，在未接触任何待评生成文本的情况下，仍能对人工或模型判分结果实现非平凡的预测性能，表明评分标准本身已隐含可被提取的评价信号，使得评分可在不参考具体候选输出的前提下部分预判。这一现象揭示了现有评估范式可能存在“评分标准偏差”（rubric bias），即评分结果可能更多反映的是评分标准的语言特征而非对生成内容的实际判断。此外，通过反事实扰动实验进一步发现，当候选文本或评分标准被反转时，评判者往往未能可靠地调整其判断，说明其决策过程缺乏对输入要素的敏感性和一致性。因此，该研究的核心解决方案在于揭示并质疑现有评估体系中“评判行为依赖于文本-标准间推理”的基本假设，并强调需重新审视和改进基于大语言模型的自动化评估方法的可靠性与可解释性，推动建立更稳健的评估框架。
链接: https://arxiv.org/abs/2609.02942
作者: Anshul Bagaria,Sowmya S Sundaram,Gokul S Krishnan,Balaraman Ravindran
机构: Centre for Responsible AI (CeRAI), Wadhwani School of Data Science and AI (WSAI), IIT Madras, Chennai, India
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Accepted for publication at EMNLP 2026. 5 pages, 6 figures
Abstract:LLM-as-a-Judge pipelines are increasingly used to evaluate AI-generated text, based on the assumption that judgments arise from reasoning over candidate responses with respect to a rubric. We show that this assumption warrants further scrutiny. Classifiers trained only on rubric text, without access to any evaluated response, achieve nontrivial predictive performance on judge outputs. This suggests that rubric formulations encode recoverable evaluative signals, allowing scores to be partially anticipated independently of model outputs. Finally, counterfactual perturbations reveal that judges often fail to reliably update their decisions when either the candidate response or the rubric criterion is reversed. Our findings raise concerns about the reliability of rubric-based LLM evaluation and highlight the need for further methodological study of automated evaluation via LLMs.
[NLP-90] SISER: Speaker-Invariant Speech Emotion Recognition with Entropy-Based Adversarial Training INTERSPEECH2026
【速读】： 该论文旨在解决语音情感识别（Speech Emotion Recognition, SER）中的两大核心问题：标注数据稀缺性与跨说话人差异性，二者均严重制约情感识别系统的泛化能力。现有对抗式方法虽能缓解说话人差异问题，但未能有效利用强大的预训练表征。为此，本文提出SISER（Speaker-Invariant Speech Emotion Recognition）框架，其关键在于将wav2vec 2.0作为特征编码器以获取丰富的自监督表征，从而降低对大规模标注数据的依赖；同时引入ECAPA-TDNN作为说话人判别器，在基于熵的对抗训练机制中提供更强的对抗信号，有效抑制说话人身份信息。实验结果表明，SISER在IEMOCAP数据集上达到60.63%的无偏准确率（UA），显著优于基线模型（51.15%）及未进行说话人抑制的wav2vec 2.0（56.46%），消融实验进一步验证了说话人判别器架构的选择是性能提升的关键因素。
链接: https://arxiv.org/abs/2609.02941
作者: Eunseo Choi,Hyunku Kang,Chanwoo Kim
机构: 未知
类目: ound (cs.SD); Computation and Language (cs.CL)
备注: Accepted to INTERSPEECH 2026
Abstract:Speech emotion recognition (SER) faces two fundamental challenges: scarcity of labeled data and inter-speaker variability, both of which hinder generalization of emotion recognition systems. While prior adversarial approaches address speaker variability, they fall short in leveraging powerful pre-trained representations. We propose SISER (Speaker-Invariant Speech Emotion Recognition), integrating wav2vec 2.0 as a feature encoder and ECAPA-TDNN as a speaker discriminator within an entropy-based adversarial training scheme. wav2vec 2.0 provides rich self-supervised representations that alleviate dependency on large labeled datasets, while ECAPA-TDNN enables suppression of speaker identity via a stronger adversarial signal than shallow classifiers. Evaluated on IEMOCAP, SISER achieves a UA of 60.63%, outperforming the baseline (51.15%) and wav2vec 2.0 without speaker suppression (56.46%), with ablation emphasizing that the choice of speaker classifier architecture is a key factor.
[NLP-91] Listen to the Latents: Self-Correcting Speech Recognition in Large Audio Language Models Through Hidden-State Interactions EMNLP2026
【速读】： 该论文旨在解决大语言模型（Large Language Model, LLM）在自动语音识别（Automatic Speech Recognition, ASR）系统中与热初始化（warm initialization）和外部校正（如重打分或后期融合）策略结合时，如何高效协同以提升识别性能的问题。现有方法多采用单一路径整合语义知识，缺乏对两种机制互补性的深入探索。本文提出的关键解决方案是“混合搜索”（Hybrid Search），其核心在于利用基底LLM（base LLM）在微调前的隐藏状态与ASR模型隐状态之间的交互特征，识别出具有高语义依赖性的目标词元，并针对性地进行修正。该策略基于两个关键观察：一是此类交互特征能有效表征词元的语义依赖程度；二是仅对高语义依赖词元进行局部修正，相比全局性校正方法（如重打分或后期融合），可显著提升ASR性能。研究表明，在通过热初始化完成语义知识迁移后，仍可通过基底LLM在推理阶段进一步优化模型表现，从而实现更精准、高效的语义增强。
链接: https://arxiv.org/abs/2609.02940
作者: Chan-Jan Hsu,Jaeyeon Kim,Chao-Han Huck Yang,Shinji Watanabe,Hung-yi Lee,Carlos Busso
机构: NVIDIA; Carnegie Mellon University (卡内基梅隆大学); National Taiwan University (台湾国立大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 24 pages, 8 figures, 2 tables. Accepted to Findings of EMNLP 2026
Abstract:Recent automatic speech recognition (ASR) systems increasingly integrate large language models (LLMs) to leverage their semantic knowledge, either externally through logit fusion or internally through warm initialization. However, how to effectively combine these two strategies remains underexplored. In this work, we refine warm-initialized LLM-based ASR models by leveraging their own pre-adaptation base LLMs, focusing on LoRA-adapted settings where the base LLM is preserved. To achieve this, we propose Hybrid Search, a targeted correction strategy motivated by two observations. First, interaction features that characterize the relationship between LLM-based ASR hidden states and base-LLM hidden states provide informative signals about a token’s degree of semantic dependence. Second, selectively refining targeted tokens with high semantic dependence improves ASR performance far beyond naive global LLM-correction methods including rescoring and late fusion. Our analysis suggests that, even after semantic knowledge transfer through warm initialization, LLM-based ASR models can still leverage their base LLM to further improve inference-time performance.
[NLP-92] RL-ADA: A World-Feedback Framework for Adversarially Robust Enterprise Dialogue Agents
【速读】： 该论文旨在解决企业级客户服务对话系统在实际部署中面临的标注瓶颈问题：高质量训练需要大规模带标签的交互数据，但企业对话日志具有高度隐私敏感性且人工标注成本高昂，同时用户行为演进速度远超传统标注流程的更新能力。其解决方案的关键在于提出一种名为RL-ADA（基于对抗式对话代理的强化学习）的共进化训练框架，通过引入世界反馈（world feedback）——即直接从可测量的交互结果中提取的基于后果的奖励信号——替代人工标注。该框架由一个客户支持代理（DA，3B参数）与一个对抗式客户代理（CA，7B参数）在固定自动化裁判的指导下共同演化：DA因成功处理多轮对话并达成最终解决而获得奖励，而CA则因生成真实、意图隐藏的语句导致误分流而获益，从而形成目标相反但结构独立的双重激励机制。通过“隔离训练场”（isolation gym）对表现较弱的一方在历史失败对话记录上进行迭代重训练，整个过程无需任何人工标注。在银行客服场景的验证中，工具路由错误被完全消除，端到端PASS率在五轮共进化周期内翻倍，仅依赖自动化竞技场奖励信号实现。此外，研究还观察到一种名为**情境伪装（Contextual Camouflage）**的对抗策略涌现——即CA仅通过奖励压力便学会将真实意图嵌入密集的真实客户细节中，这一现象对企业的红队测试和系统鲁棒性评估具有重要启示。
链接: https://arxiv.org/abs/2609.02902
作者: Ram Narayanan,Harshit Rajgarhia,Abhishek Mukherji
机构: Centific
类目: Computation and Language (cs.CL)
备注:
Abstract:Deploying task-oriented dialogue agents in enterprise customer support faces a persistent annotation bottleneck: robust training requires labelled interaction data at scale, yet enterprise conversational logs are privacy-sensitive and expensive to annotate, while user behaviour evolves faster than labelling pipelines can keep pace. We present RL-ADA (Reinforcement Learning with Adversarial Dialogue Agents), a co-evolutionary training framework that eliminates this bottleneck by replacing human labels with \emphworld feedback: consequence-based reward signals derived directly from measurable interaction outcomes. A Customer Support Agent (DA, 3B parameters) and an Adversarial Customer Agent (CA, 7B parameters) co-evolve in an adversarial arena guided by a fixed automated judge: the DA is rewarded for correctly handling multi-turn customer conversations to successful resolution, while the CA is rewarded for producing realistic, intent-concealing utterances that cause misroutes, creating asymmetric adversarial pressure through opposing but independently structured rewards. An isolation gym iteratively retrains the weaker agent on prior-failure transcripts, requiring no human annotation at any stage. In a banking customer support proof of concept, tool-routing errors are eliminated and the strict end-to-end PASS rate doubles over five co-evolutionary cycles, driven solely by automated arena reward with no labelled data. We additionally observe the emergence of \textbfContextual Camouflage, an adversarial strategy in which the CA learns to embed intent within dense realistic customer detail purely from reward pressure, with direct implications for enterprise red-teaming and robustness evaluation.
[NLP-93] Dual-Form ASR: Semantics-Aware Inverse Text Normalization for Chinese Speech Recognition
【速读】： 该论文旨在解决传统自动语音识别与逆文本归一化（ASR-ITN）系统中，口语形式转录与可读书面形式转录因级联模块设计而导致的误差传播问题，尤其在语义依赖型数值表达式上，由于归一化过程与声学上下文建模相分离，导致归一化准确性下降。其解决方案的关键在于提出双形式语音识别（Dual-Form ASR, DF-ASR）框架，通过成对的口语形式与书面形式监督信号，使模型在保持提示级别输出形式选择能力的同时，具备语义感知的书面形式归一化能力。该方法利用大语言模型（LLM）驱动的生成与评估工作流构建高质量双形式训练数据，并引入基于序列的归一化敏感跨度惩罚目标（ITN-MWER），以强化对归一化关键区域的建模；同时设计决策感知的REQUIRE-ITN/FORBID-ITN协议，独立评估归一化必要性与禁止性片段的保留能力。实验表明，DF-ASR在SpeechIO中文标注子集上显著优于开源ASR-ITN系统，且在性能上可与主流闭源系统比肩，同时维持了可靠的提示级输出形式控制。
链接: https://arxiv.org/abs/2609.02901
作者: Fengrun Zhang,Li Fu,Wangjin Zhou,Lu Fan,Youzheng Wu,Xiaodong He
机构: 未知
类目: Computation and Language (cs.CL); Sound (cs.SD)
备注: Submitted to IEEE SLT 2026
Abstract:Modern automatic speech recognition (ASR) scenarios require both spoken-form transcripts for faithful transcription and readable written-form transcripts with inverse text normalization (ITN). However, these forms are typically produced by cascaded modules, where a spoken-form ASR output is rewritten by a separate ITN component, making written-form ASR-ITN vulnerable to recognition errors and decoupling normalization from acoustic-contextual modeling, especially for semantically dependent numeric expressions. In this paper, we propose Dual-Form ASR (DF-ASR), a framework that extends spoken-form ASR capability to semantics-aware written-form ITN through paired spoken-form and written-form supervision while retaining prompt-level selection between transcript forms. The dual-form supervision is constructed via a large language model (LLM)-driven generate-and-judge workflow, and training is further enhanced by ITN-MWER, a sequence-level objective that assigns higher cost to errors on normalization-sensitive spans. We also introduce a decision-aware REQUIRE-ITN/\FORBID-ITN protocol to separately measure required normalization and forbidden-span preservation. On manually annotated Chinese subsets from SpeechIO, DF-ASR consistently outperforms open-source ASR-ITN systems, remains competitive with strong closed-source references, and preserves reliable prompt-level control between spoken-form and written-form outputs.
[NLP-94] Contamination Inflates Scores but Rarely Reorders Large Language Model Leaderboards
【速读】： 该论文旨在解决大语言模型（Large Language Model, LLM）排行榜中因测试集泄露（benchmark contamination）导致的评估可靠性问题，尤其关注污染是否仅影响绝对得分，还是实质性地改变了模型间的排名顺序。其核心解决方案在于将污染问题重新定义为对锚点项不变性（anchor-item invariance）的违反，并通过同一题目原始版本与语义等价改写版本之间的响应差异进行量化，从而在保持所测能力恒定的前提下，有效分离记忆效应与真实能力。研究基于47个公开发布的模型及74个经过已知污染剂量微调的模型，在四个基准测试（ARC、GSM8K、HellaSwag、MMLU）上进行实证分析，首先验证了该方法能准确响应注入的污染剂量（校正后准确率提升+0.187点），且不会误报仅使用合法训练集训练的对照模型（-0.012）。进一步分析表明，标准排行榜与基于改写项控制的排行榜之间排名相关性高达0.997，敏感性分析显示实际观察到的污染差异远低于足以改变排名的阈值，仅有3例模型-基准组合在两个独立参考下被一致证实存在差异性污染。因此，结论指出当前公共模型中的污染主要表现为均匀性得分膨胀，而非显著排名扭曲，真正的排名扰动仅出现在罕见的差异性污染场景中。研究提供了一个可校准的不变性审计工具，并建议排行榜应同时报告经改写控制的排名及其置信区间，以增强评估透明度与可信度。
链接: https://arxiv.org/abs/2609.02899
作者: Xingyao Xiao(Stanford University),Yihong Cheng(City University of Macau)
机构: 未知
类目: Computation and Language (cs.CL); Applications (stat.AP); Methodology (stat.ME)
备注: 21 pages, 4 figures, 3 tables. Code and data: this https URL
Abstract:Benchmark contamination, the leakage of test items into training data, is widely described as a threat to the reliability of large language model (LLM) leaderboards. We argue that this concern conflates two distinct questions: whether contamination inflates absolute scores, and whether it reorders the ranking of models. We recast contamination as a violation of anchor-item invariance and measure it through the differential functioning of original versus semantically equivalent paraphrased items, a within-item contrast that holds the measured skill fixed and isolates memorization from capability. Using per-instance responses from 47 publicly released models and 74 models finetuned with a known dose of contamination, across four benchmarks (ARC, GSM8K, HellaSwag, MMLU), we first calibrate the measure against ground truth: it recovers injected contamination dose-responsively (a corrected effect of +0.187 accuracy points for test-set leakage) and never flags a negative-control model trained only on the legitimate training split (-0.012). We then quantify leaderboard impact: the rank correlation between a standard leaderboard and a paraphrase-controlled leaderboard is 0.997, and a sensitivity analysis shows that the observed differential contamination is far below the level needed to move rankings, with only 3 of 188 model-by-benchmark cases showing differential contamination corroborated across two references. Contamination among these public models is therefore largely uniform: it inflates absolute scores without reordering the leaderboard, and ranking distortion requires the rare case of differential contamination. We provide a calibrated invariance audit, released as a reference implementation, and recommend that leaderboards report paraphrase-controlled rankings alongside confidence intervals.
[NLP-95] Distilled Rapid Embedding Transfer (DRET): Parameter-Efficient Biomedical Domain Adaptation via Priority-Based Embedding Transfer
【速读】： 该论文旨在解决大型领域专用语言模型（如BioBERT和ClinicalBERT）在生物医学自然语言处理任务中虽表现优异但计算开销过大，难以在实际场景中部署的问题；同时，通用轻量级模型（如DistilBERT）虽具备高效性，却缺乏特定领域的知识。其解决方案的关键在于提出一种名为“蒸馏快速嵌入迁移”（Distilled Rapid Embedding Transfer, DRET）的知识迁移范式，通过将大型专用模型中的生物医学领域知识无须重新训练原始语料即可注入到小型通用模型中，实现性能与效率的平衡。DRET采用迭代策略，包括统一分词器合并（DRET 1.x）、混合嵌入平均（DRET 2.0）、基于优先级的嵌入传输机制（DRET 3.x），并进一步结合嵌入层冻结、差异化学习率、标签传播及不平衡感知损失函数（DRET 4.x），显著提升小模型在严重类别不平衡条件下的分类性能。实验表明，经DRET增强的DistilBERT（66M参数）在EBM-NLP数据集上的PICO分类任务中，各项指标达到与大模型相当甚至更优的表现，且保持了原有模型的高效性。此外，通过余弦相似度、语义偏移和t-SNE分析验证了知识确实在嵌入层面发生有效迁移。DRET为生物医学文本挖掘提供了可扩展、资源高效的近领域专家级性能路径，可直接应用于自动化系统文献综述与临床决策支持。
链接: https://arxiv.org/abs/2609.02898
作者: Girish Sundaram,Daniel Berleant
机构: University of Arkansas at Little Rock(阿肯色大学小石城分校)
类目: Computation and Language (cs.CL)
备注: 11 pages, 5 figures, 6 tables
Abstract:Large domain-specific language models such as BioBERT and ClinicalBERT achieve strong performance on biomedical NLP tasks, but their computational demands make them impractical for many real-world deployments. General-purpose, parameter-efficient models such as DistilBERT are lightweight yet lack the domain knowledge required for specialized tasks such as PICO (Population, Intervention, Comparison, Outcome) classification. We introduce Distilled Rapid Embedding Transfer (DRET), a knowledge-transfer paradigm that injects biomedical domain knowledge from large specialized models into a smaller general-purpose model without retraining on the original specialized corpora. DRET is developed as an iterative family of strategies: a unified tokenizer-merge strategy (DRET 1.x), hybrid embedding averaging (DRET 2.0), and a priority-based embedding-transfer mechanism (DRET 3.x) that hierarchically selects embeddings from the most authoritative source models, further combined with embedding-layer freezing, differential learning rates, label propagation, and imbalance-aware loss functions (DRET 4.x). We evaluate DRET on token-level PICO classification using the EBM-NLP corpus under severe class imbalance, across a twelve-metric battery. DRET-enhanced DistilBERT (66M parameters) attains balanced accuracy, recall, and ROC-AUC competitive with, and on several class-wise metrics exceeding, models an order of magnitude larger, while retaining DistilBERT’s efficiency. We further show that transfer occurs at the embedding level through cosine-similarity, semantic-shift, and t-SNE analyses. DRET offers a scalable, resource-efficient route to near-domain-expert performance for biomedical text mining, with direct application to automated systematic literature reviews and clinical decision support.
[NLP-96] Margins Not Windows: Training-Free Per-Step Lossy Speculative Decoding
【速读】： 该论文旨在解决生成式 AI（Generative AI）在大语言模型（LLM）推理过程中，基于树注意力结构的推测解码（speculative decoding）方法在实际应用中因固定验证规则与静态草稿树结构所导致的性能瓶颈问题。现有方法通常采用严格的逐令牌匹配验证机制及固定的草稿树形态，限制了其对不同推理场景的适应能力。针对此问题，论文提出一种无需训练的自适应推测解码方法 AdaptiveSpec，其核心创新在于通过解码过程内部信号动态调整两个关键决策：一是引入基于每步边际概率比的验证规则，当目标模型对草稿令牌的概率与其最高概率之比超过阈值时，允许不匹配的草稿令牌被采纳，从而提升灵活性且不依赖草稿长度或草稿器架构；二是设计一种基于融合信号的每步树策略，结合草稿顶1置信度与滚动接受历史（反映近期草稿与目标的一致性），动态调节草稿树的深度、宽度和节点数量，使总草稿数量可变而非仅在固定预算内重新分配。上述两项适应机制在正交维度上协同作用，显著增强解码效率。在 SGLang 生产级服务引擎上的实现表明，AdaptiveSpec 相较于当前最先进的自回归推测解码方法 EAGLE-3，吞吐量最高提升 56%，并在 GSM8K、MATH-500 与 HumanEval 等基准任务上恢复至 93% 至完全无损的准确率，适用于多种目标模型。
链接: https://arxiv.org/abs/2609.02897
作者: Oszkár Urbán,Young D. Kwon,Stylianos I. Venieris,Cecilia Mascolo
机构: University of Cambridge (剑桥大学); Samsung AI Center-Cambridge, UK (三星人工智能中心-剑桥)
类目: Computation and Language (cs.CL)
备注:
Abstract:Speculative decoding accelerates LLM inference by drafting candidate tokens and verifying them in parallel. Tree-attention drafters such as EAGLE-3 are widely adopted, yet typically hold two decisions fixed: (1) a strict token-match verification rule and (2) a static draft-tree shape. Prior work relaxes each in isolation under limiting assumptions: long draft chains for training-free lossy verification, and adaptive tree shaping under a fixed token budget. We introduce AdaptiveSpec, a training-free per-step speculative decoding method that adapts both decisions from internal signals already produced during decoding. A per-step margin rule promotes a mismatched draft-proposed token when the ratio of the target’s probability on the drafted token to its top-1 probability exceeds a threshold with no dependence on draft length or underlying drafter architecture. A per-step tree policy adjusts the draft tree’s depth, width, and node count directly from a fused signal of draft top-1 confidence and a rolling acceptance history capturing recent draft-target agreement, allowing the total draft count to vary rather than only be redistributed. The two adaptations operate on orthogonal axes and compound in effect. Implemented on the SGLang production-grade serving engine, AdaptiveSpec improves throughput over the state-of-the-art autoregressive speculative decoding method EAGLE-3 by up to 56%, recovering 93% to fully lossless task accuracy across GSM8K, MATH-500, and HumanEval on three target models (DeepSeek-R1-Distill-Llama-8B, Llama-3.1-8B-Instruct, Qwen3-8B).
[NLP-97] PiPMRE: A Pipeline Based on Language Model for Medical Relation Extraction
【速读】： 该论文旨在解决医学关系抽取（Medical Relation Extraction, MRE）中因实体间复杂关系导致的联合抽取难题，尤其针对传统方法将MRE视为序列标注任务所引发的标签体系设计复杂或多重关系提取失败的问题。其核心解决方案是提出一种基于语言模型的新型流水线框架PiPMRE，关键在于将关系抽取分解为两个阶段：首先通过关系生成器从文本中生成多个关系三元组，随后由关系过滤器对每个三元组进行评分并保留符合阈值的候选结果。该框架无需设计复杂的标注模式，仅通过简单模板重构输入文本，确保实体与关系在上下文中按序生成，从而有效缓解了医学实体间语义冗余与嵌套关系带来的挑战。实验结果表明，PiPMRE在两个公开数据集上平均提升5.6个召回率点和4.4个准确率点，且在少样本设置下仍表现出显著优势，验证了其高效性与鲁棒性。
链接: https://arxiv.org/abs/2609.02896
作者: Jiaxin Duan,Fengyu Lu,Junfei Liu
机构: Peking University (北京大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Medical relation extraction (MRE) is commonly known for extracting entities and their relations jointly from a medical text, which has attracted considerable attention in recent years. Previous studies treat MRE as a sequence tagging task, which results in either a challenging design of the tagging schema or a failed extraction of multiple relations, due to intricate relationships among medical entities. In this work, we review the task from the linguistic perspective and propose a novel pipeline framework, PiPMRE, developed on language models to enhance MRE performance. Specifically, PiPMRE consists of a relation generator and a relation filter. Given a text, the generator first yields multiple relational triplets, and then the filter scores each triplet and retains only those that pass the borderline as the final results. Implementing PiPMRE requires no tagging schema; instead, we use a simple template to reformulate the input text, ensuring that entities and relations are generated in a contextual order. Extensive experimental results on two public datasets demonstrate the advancement of PiPMRE. It surpasses the previous state-of-the-art by an average of 5.6 recall points and 4.4 accuracy points. PiPMRE’s superiority is also demonstrated in few-shot settings.
[NLP-98] BharatGather: A Culturally-Informed Benchmark Dataset for Misinformation and Fake News Detection in Indian Public Events
【速读】： 该论文旨在解决大规模公共活动（如宗教节庆、政治集会和文化聚会）中虚假信息快速传播所引发的公共安全与社会凝聚力威胁问题，尤其针对现有自动化假新闻检测基准在捕捉印度语境下社会文化细微差异及事件特异性动态方面的不足。其解决方案的关键在于构建一个名为BharatGather的定制化、多源数据集，专用于印度大规模聚集场景下的二元虚假信息分类任务。该数据集通过混合式管道生成，包含14,646条记录，整合了主流事实核查平台的系统性网络爬取、多媒体转录提取以及基于大语言模型（Large Language Model, LLM）的合成增强技术，以保障叙事多样性。该资源不仅反映了印度语境下事件感知型虚假信息的独特复杂性，还为开发具有文化敏感性的检测系统提供了基础，并建立了在高风险公共环境中评估模型性能的严格基准。
链接: https://arxiv.org/abs/2609.02895
作者: Parth Bramhecha,Smit Deshmukh,Sairaj Bodhale,Adwait Borate,Raviraj Joshi
机构: L3Cube-Labs( L3Cube实验室)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Large-scale public events, such as religious festivals, political rallies, and cultural gatherings, are increasingly vulnerable to the rapid dissemination of misinformation, posing substantial risks to public safety and social cohesion. While automated fake news detection has seen significant methodological progress, existing benchmarks frequently fail to capture the socio-cultural nuances and event-specific dynamics characteristic of the Indian context. This paper introduces BharatGather, a curated, multi-source dataset specifically engineered for binary misinformation classification within the ecosystem of Indian mass gatherings. The corpus comprises 14,646 records constructed through a hybrid pipeline involving systematic web scraping of prominent fact-checking platforms, multimedia transcript extraction, and Large Language Model (LLM)-mediated synthetic augmentation to ensure narrative diversity. By providing a resource tailored to the unique complexities of event-aware misinformation in India, this work facilitates the development of culturally informed detection systems and establishes a rigorous benchmark for evaluating their performance in high-stakes public environments.
[NLP-99] Probe Generalization as Subspace Selection for OOD Deception Detection
【速读】： 该论文旨在解决线性探针（linear probes）在分布外（out-of-distribution, OOD）样本上泛化能力不足的问题，特别是在检测语言模型激活中隐含的欺骗行为时表现不佳。其核心解决方案在于：通过将输入投影到训练数据激活分布中少数关键主成分（Principal Components, PCs）构成的子空间，实现跨域迁移性能的显著提升，其效果接近于在目标域数据上直接训练的探针。关键创新点在于，利用大语言模型（LLM）判断每个主成分是否蕴含可迁移的欺骗语义方向（如通过评估其最高/最低激活样本是否体现具有迁移性的欺骗特征），从而筛选出真正具备泛化能力的主成分。实验表明，基于此方法选择的子空间能有效缩小基线与理想探针（oracle）之间的性能差距，在内部交易报告和沙袋化检测任务上分别提升78%和25%。研究进一步揭示，源域探针依赖的高权重方向多编码特定表面特征，而真正可迁移的方向则以更抽象、可被自然语言描述的方式表征语义对比，表明探针的OOD鲁棒性主要由所选子空间的质量决定。
链接: https://arxiv.org/abs/2609.02893
作者: Daniel Yoo,Adrians Skapars
机构: 未知
类目: Computation and Language (cs.CL)
备注:
Abstract:Linear probes can be used to detect behaviors and concepts inside language model activations, but may fail to transfer to out-of-distribution examples. When studying the generalization performance of Llama-3.1-8B-Instruct probes over 3 held-out deception detection datasets, we find that projecting inputs onto a small subset of principal components (PCs) from the training distribution of activations enables cross-domain transfer that nearly matches the performance of probes trained directly on the test distribution. Furthermore, we find that PC interpretations can be used to find a subset of those transferable PCs. By using an LLM judge to score each PC on whether its most/ least activating examples imply a transferable deception direction, then probing on the highest-scoring PCs, we close the baseline-to-oracle gap by 78% on Insider Trading Report and by 25% on Sandbagging. The directions a source probe weights heavily appear to encode source-specific surface features, while the directions that actually transfer appear to encode the same contrast more abstractly, in a way natural language descriptions can capture. Broadly, our results suggest that the OOD robustness of probes is largely determined by subspace selection.
[NLP-100] Counterexamples as Feedback for Agent Self-Correction
【速读】： 该论文旨在解决现有单轮代码生成评估指标无法充分反映部署型智能体在实际应用中关键能力的问题，即：在接收到具体反馈后，能否有效修复错误生成的产物。其核心解决方案是提出A-CEGIS框架，该框架通过利用反例（counterexample）作为反馈信号，实现自然语言到正则表达式（NL-to-regex）合成任务中的多轮迭代优化。该框架的关键在于引入具有明确语义的诊断性反例——包括紧凑的假阳性或假阴性实例，以指导下一轮修正的方向。实验表明，在30个NL-RX-Turk任务上，基于诊断性反例的反馈可在四轮内解决90%的任务，显著优于零样本生成（17%）、通用自我修正（27%）及仅含错误信息的反馈（23%）。在完整诊断流程中，所有隐藏测试任务均在最终轮次成功解决，平均达成成功率仅需2.7轮，且经过针对性探测后仍保持77%的鲁棒成功率。这表明A-CEGIS不仅能够高效衡量智能体在多轮交互中的改进能力，还为模型提供了超越原始测试集之外的实际鲁棒性验证机制。
链接: https://arxiv.org/abs/2609.02892
作者: Sidhesh Badrinarayan,Adithya Parthasarathy
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Single-turn code-generation metrics understate a central property of deployed agents: whether they can repair a wrong artifact after receiving concrete feedback. This paper presents A-CEGIS, a lightweight framework that uses counterexamples as feedback for evaluating multi-turn refinement in natural-language-to-regex synthesis. An agent proposes a regex, a deterministic oracle checks it under full-match semantics, and compact false-positive or false-negative witnesses guide the next turn. On 30 NL-RX-Turk tasks, diagnostic counterexample feedback solves 90% of tasks within a four-turn ablation budget, compared with 17% for zero-shot generation, 27% for generic self-correction, and 23% for error-only feedback. In a full diagnostic run with hardening, all tasks are solved on the hidden set by the final turn, with mean time-to-success of 2.7 turns and robust success of 77% after targeted probing. These results show that A-CEGIS measures how efficiently an agent improves across turns while adding a practical robustness check beyond the original held-out cases.
[NLP-101] Bounded Personas Match Retrieval on Classification but Not Regression for a Frozen Agent
【速读】： 该论文旨在解决个性化语言代理（personalized language agent）在推理时如何高效且准确地利用用户交互历史的问题。现有主流方法分为两类：检索（retrieval）策略通过在每次请求时从用户历史中提取最相关项注入提示（prompt），虽精度高但随历史增长导致上下文开销与查询选择成本上升；而蒸馏（distillation）策略将用户历史压缩为一个紧凑、可解释的自然语言人格描述（persona），具备查询无关性和上下文边界性，但普遍认为会牺牲准确性。本文提出一种无需训练的PersonaLink方法，将用户历史蒸馏为一个三字段的有界人格描述，并通过递归优化机制实现自评估与迭代修正：每轮基于用户自身标注的历史片段对冻结的7B模型进行评估，根据错误信息重写人格描述，仅当性能不退化时才保留更新结果。该设计通过固定模型主干、仅改变上下文内容，有效分离了表示形式与模型能力的影响。实验结果显示，在LaMP-2数据集上（200名用户，15分类新闻分类任务），PersonaLink达到0.745–0.755的准确率，与基于BM25的检索方法（0.760–0.765）统计上无显著差异，揭示了不同任务类型下的表现不对称性，证明了蒸馏式人格在特定任务中可媲美甚至等效于检索策略。
链接: https://arxiv.org/abs/2609.02890
作者: JaeHa Yoon,Minjun Park,Seoyeon Kim,Jiwoo Lee,Hyunwoo Choi,Dohyun Kang
机构: Seoul National University (首尔国立大学); KAIST (韩国科学技术院); Korea University (韩国大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:A personalized language agent must convert a user’s interaction history into behavior on each new request at inference time. Two strategies dominate. Retrieval pulls a few of the user’s most relevant past items into the prompt, which is accurate but pays a per-query selection and context cost that grows with the history. Distillation instead compresses the history once into a compact natural-language persona, which is bounded, query-independent, and interpretable, but is widely assumed to sacrifice accuracy. Whether, and on which tasks, a distilled persona can match retrieval has not been characterized cleanly. We introduce PersonaLink, a training-free method that distills a user’s history into a bounded three-field persona and recursively refines it: each pass self-evaluates the frozen agent on a held-out slice of the user’s own labeled history, rewrites the persona from its errors, and keeps the result only when it does not regress on that slice. Because every comparison shares one frozen 7B backbone and differs only in what is placed in context, the design isolates the effect of representation from that of the model. The result is a clear task-type asymmetry. On 200 users of LaMP-2 (15-way news categorization), PersonaLink reaches 0.745-0.755 accuracy, statistically indistinguishable from BM25 retrieval (0.760-0.765).
[NLP-102] Where Does Harness-Optimization Value Live? Localized Gains and the Budget-Splitting Trap in Self-Evolving LLM Agents
【速读】： 该论文旨在解决冻结大型语言模型（LLM）作为智能体时，其文本引导框架（harness）优化效率低下且缺乏可解释性的问题。现有方法通常将整个引导框架视为单一扁平字符串进行优化，忽略了不同组件对整体性能的贡献差异。论文提出HARNESSEVO，一种将引导框架分解为四个独立可演化模块的新范式：角色（role）、任务-策略（task-strategy）、工具/格式规则（tool/format-rules）以及反思与控制（reflection/control）。通过在相同预算下采用“留一内”和“留一外”归因分析，研究发现几乎所有优化价值集中于反思与控制模块，其单独引入可带来+0.119的成功率提升，而其他模块单独优化均无显著增益。进一步实验表明，均匀分配计算预算会因每模块可用样本过少（仅16次回放）导致各模块陷入初始种子状态，从而抑制优化；反之，将全部预算集中于高贡献的控制模块，可在仅一半预算下实现0.761的成功率，显著超越基准。该现象具有任务依赖性，在WebShop任务中所有模块均未产生有效优化，表明其本质是缺乏可被语言化描述的控制失败模式，而非资源不足。因此，研究结论强调：引导框架的价值具有局部性，盲目均匀分配预算可能适得其反，有效的结构化智能体进化必须以信用分配（credit assignment）为前提。
链接: https://arxiv.org/abs/2609.02889
作者: Michael Nguyen,Wei Chen Tan,Nurul Aisyah Hassan,Arvind Raman,Li Hua Lim,Ahmad Faiz Razak
机构: Universiti Malaya (马来西亚大学); Universiti Sains Malaysia (马来西亚理科大学); Universiti Putra Malaysia (马来西亚农业大学); Monash University Malaysia (莫纳什大学马来西亚分校)
类目: Computation and Language (cs.CL)
备注: 17 pages
Abstract:A growing body of work improves frozen large language models (LLMs) as agents by evolving their harness: the textual scaffolding around the model, including persona, strategy, format rules, and control heuristics. Existing reflective prompt-evolution methods usually optimize this harness as one flat string. We instead ask where the optimization value actually resides. We introduce HARNESSEVO, which decomposes the harness into four separately evolvable slots: role, task-strategy, tool/format-rules, and reflection/control. Using the same reflective optimizer under an iso-budget setting, we pair this decomposition with leave-one-in and leave-one-out attribution to measure the contribution of each slot. On ALFWorld with a frozen 7B backbone, HARNESSEVO does not significantly improve the overall binary success rate over either the stock harness or flat-string evolution: 0.657 versus 0.642 and 0.642, respectively. However, the slot-level analysis reveals that nearly all useful optimization value is localized in the reflection/control slot, which achieves a leave-one-in gain of +0.119. The other slots are individually null. We further show that uniform budget splitting is harmful: allocating 64 rollouts across four slots leaves only 16 per slot, below the optimizer’s effective search floor, causing every slot to freeze at its empty seed. Concentrating the budget on the high-credit control slot recovers the lost gain, reaching 0.761 with half the split budget. The effect is task-contingent. On WebShop, all slots freeze empty and all methods tie, indicating a genuine absence of recurrent, verbalizable control failures rather than budget starvation. Overall, our results suggest that harness value is localized, uniform budget splitting can be actively harmful, and credit assignment should precede structured agent-evolution. Comments: 17 pages Subjects: Computation and Language (cs.CL) Cite as: arXiv:2609.02889 [cs.CL] (or arXiv:2609.02889v1 [cs.CL] for this version) https://doi.org/10.48550/arXiv.2609.02889 Focus to learn more arXiv-issued DOI via DataCite
[NLP-103] LLM Judges as Raters: A Pre-Registered Audit of Severity Halo Reliability and Version Instability in LLM Essay Scoring on Public Corpora
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）作为作文评分工具在教育评估中应用时存在的可靠性与公平性问题，特别是针对评分者效应（rater effects）的潜在影响。尽管现有研究多依赖一致性统计量评价评分性能，但教育测量学指出，人工评分者存在严重性差异、光环效应（halo effect）、评分漂移（drift）等系统性偏差，而这些因素同样可能存在于LLM评分者中。本文将多个提供商的LLM视为评分者，基于双语公开语料库（ENEM/Essay-BR与ASAP），通过预注册的多面拉斯克模型（many-facet Rasch）分析，系统检验了评分者严重性、残差光环效应、普遍性/决策研究（G-study/D-study）、版本间偏差及项目功能差异等指标。关键发现包括：在ENEM数据上，评分者严重性跨度达219分（满分1000），在ASAP上面板内部差异占总分范围的15%-33%，远超训练后人类评分者间的差距（约1%）；且所有五种版本对比均显示严重性偏移显著超出置换检验的家族误差阈值（最高达133分），一名评分者因身份异常被识别并剔除。此外，经预注册检验，调整严重性后的排名反转未通过置换检验，且“无声漂移”假说被否定——评分一致性随严重性变化而变化，表明评分行为非稳定。尽管结果具有自一致性（k=2时φ=0.80），但并未达到人类评分水平的准确性，且同一仪器对照实验推翻了原光环效应结论：在控制评分工具与校准一致性的前提下，无充分证据表明模型光环效应超出训练人类评分者的范围。因此，该研究的核心解决方案在于建立一套系统化的评分者效应检测框架，并强调必须将评分者特性纳入评估体系，而非仅依赖表面的一致性指标。
链接: https://arxiv.org/abs/2608.29517
作者: Veerendra Kumar Sunkavalli
机构: 未知
类目: Computation and Language (cs.CL); Computers and Society (cs.CY)
备注: 14 pages, 2 figures, 8 tables. Under review at the Journal of Learning Analytics (LAK27 research track). Artifact: this https URL
Abstract:Large language models (LLMs) are increasingly used as essay graders in learning analytics, evaluated almost exclusively with agreement statistics. Educational measurement warns that raters also differ in severity, show halo, and drift as instruments. We treat LLM judges as raters and run a pre-registered rater-effects battery (many-facet Rasch severity, residual halo, generalizability/decision studies, cross-version shifts, differential functioning) on public corpora in two languages (ENEM/Essay-BR; ASAP): 2,377 essays, 12 judges, 4 providers, 5 version contrasts, replicated cells, released as a score tensor. Judge severity spans 219 points on ENEM’s 0-1000 scale; on ASAP the panel spread is 15-33% of the score range against a between-trained-human gap near 1%. Judge-human correlations sit in an undiscriminating .47-.56 band. All five version contrasts shift severity beyond a family-wise permutation null (up to 133 points), and one judge was deprecated mid-study, caught by identity canaries. Two pre-registered tests returned honest nulls: severity-adjusted leaderboard reversals did not survive a permutation null, and “silent drift” was refuted: agreement moved with severity in four of five contrasts. Replication yields self-consistency (phi=.80 at k=2) but not human-level accuracy, and a same-instrument check overturned our own halo comparison: matched on instrument and calibration, we find no credible evidence that judge halo exceeds the trained-human range.
[NLP-104] One Timeline Many Renderings: A Wolfram Language Paclet for heterogeneous musical output
【速读】： 该论文旨在解决多模态音乐创作中不同输出格式（如Csound音色谱、乐谱排版、实时控制信号及排练节拍）在独立生成时出现的时间轴漂移问题。其核心解决方案是构建一个基于Wolfram语言的Temporal System（时间系统）包，通过在统一的有理节拍时间线上维护一组类型化的不可变实体，实现跨格式的同步。该系统的关键在于采用后端特定的契约机制，使所有输出（包括Csound合成、MusicXML 4.0乐谱、OSC控制信号及节拍音频）共享同一数据源，在渲染前不进行单位转换（如秒、采样点或赫兹），从而确保各组件在时间上严格对齐。其中，Csound音符使用外部.orc文件中的稳定命名乐器，曲线则作为与记分参数（p字段）关联的k速率信号声明；节拍后端从相同的节拍与速度信息生成排练音频，并复用Csound序列化器。研究进一步阐述了时间层、语义层与渲染契约层的设计及其实际权衡，并指出了该专有创作环境在开放源代码生态中的局限性。
链接: https://arxiv.org/abs/2608.24683
作者: Francesco Vitucci,Michele Lorusso,Francesco Scagliola
机构: Conservatorio di Musica “N. Piccinni” di Bari(巴里尼科洛·皮奇尼音乐学院)
类目: Computation and Language (cs.CL); Sound (cs.SD)
备注: Accepted at the International Csound Conference (ICSC) 2026
Abstract:One algorithmic composition may require a Csound score, engraved notation, real-time control, and a rehearsal click. Authored separately, their timelines drift. Temporal System is a Wolfram Language paclet that instead compiles one immutable store of typed entities on a rational beat timeline through backend-specific contracts. It emits Csound synthesis, beta MusicXML 4.0, OSC control, and click artifacts that remain synchronized because they share that store. Conversion to seconds, samples, or hertz occurs only at render time. Csound notes use stable named instruments in external .orc files; curves become k-rate signals declared against score p-fields. The click backend derives rehearsal audio from the same meter and tempo and reuses the Csound serializer. We describe the temporal, semantic, and rendering-contract layers, their practical trade-offs, and the limits of this proprietary authoring environment within an otherwise open-source ecosystem. The archived supplement exposes the reported outputs pending paclet release.
[NLP-105] From local kernels to global form: modeling the emergence of musical content
【速读】： 该论文旨在解决符号音乐中基于马尔可夫模型（Markov models）的段落边界检测问题，特别是如何在不依赖外生形式划分的前提下，通过观测数据驱动的方式识别音乐结构中的潜在分段。其核心挑战在于传统方法常依赖预设的分区或先验假设，而本文提出一种基于重叠滑动窗口（overlapping sliding windows）的局部转移核轨迹估计机制，直接从单一符号序列中提取动态变化的转移概率模式。解决方案的关键在于利用滑动窗口生成连续的局部转移核序列，并通过信息论指标（如Jensen–Shannon散度）量化不同区域间的差异，从而揭示音乐结构的潜在边界。实验以德彪西《竖笛》（Syrinx, 1913）的273个音符事件为对象，对比绝对音高与节拍时值两个维度的转移核行为；结果表明，在窗口长度L=6时，两种维度均达到Jensen–Shannon散度最大值，且音高维度的平稳区间显著更广（210/267次比较），而时值维度则较窄（64/267）。尽管理论最大值受窗口几何约束（为1/L−1），但仅凭单一维度的宽平稳区无法作为唯一自动分段依据，而跨维度的对齐性特征则支持其与结构边界敏感性的关联。此外，500次重合成实验揭示了在L=2时出现的精确复制退化现象，进一步凸显了窗口长度选择对模型行为的重要影响。
链接: https://arxiv.org/abs/2608.24660
作者: Francesco Vitucci,Michele Lorusso,Francesco Scagliola
机构: 未知
类目: Computation and Language (cs.CL); Sound (cs.SD)
备注: Accepted at XXV CIM - Colloquio di Informatica Musicale, L’Aquila, 2026
Abstract:Markov models are established tools for symbolic music, including non-homogeneous formulations. The narrower contribution examined here is an observation-driven estimation mechanism: overlapping sliding windows derive a trajectory of local transition kernels from one symbolic sequence rather than from an exogenous formal partition. We test this mechanism on 273 logical note events from Debussy’s Syrinx (1913), using the often-proposed A-B-A’ reading as a reference rather than ground truth. We apply the same validation to absolute-pitch and notated-duration kernels. At L=6 , both reference boundaries attain the Jensen–Shannon maximum in both dimensions; the duration plateau is substantially narrower (64 of 267 comparisons) than the pitch plateau (210 of 267). Because the theoretical maximum for consecutive sliding-window comparisons is set by window geometry and equals 1/\sqrtL-1 for maximal turnover of the entering/leaving transition, the pitch value at L=6 and its broad plateau are not, by themselves, strong evidence. Their cross-dimensional alignment is consistent with boundary sensitivity, while the broad plateaus preclude treating either curve alone as a unique automatic segmenter. Five-hundred-draw re-synthesis experiments quantify departure from the source in both dimensions and expose an exact-copy degeneracy at L=2 .
[NLP-106] ALRA: Adaptive Local Relational Alignment for Logit-Based Pre-training Distillation of Autoregressive Language Models
【速读】： 该论文旨在解决自回归语言模型中基于逻辑值（logit）的知识蒸馏方法在对齐教师与学生模型时存在的局限性问题。传统全局对齐策略虽能统一优化整个词汇表上的下一个词分布，但忽略了在高概率候选词之间相对偏好关系的建模，导致知识传递效率下降。现有局部方法通常仅从教师或学生单侧选择候选词：教师单侧选择可能遗漏学生认为合理的词，而学生单侧选择则因训练初期预测不准确导致排名偏差。为此，本文提出自适应局部关系对齐（Adaptive Local Relational Alignment, ALRA），其核心创新在于构建一种位置相关的协同机制——在每个有效预测位置，由学生生成高概率候选词，同时引入教师最可能输出的词作为锚点，动态调整候选集规模以反映教师在当前批次中的概率分布广度。ALRA通过保留质量匹配项，并分别对选定候选区域与剩余词汇区域内的相对词分布进行匹配，避免了传统全词汇分解中因某区域教师概率低而导致对应项被过度抑制的问题。此外，引入学生加权成对关系对齐（Student-Weighted Pairwise Relational Alignment），强化高概率词对间学生预测差距小的区域，降低对低概率或明显分离词对的关注权重，从而更精准地捕捉语义相关性。实验结果表明，在 The Pile 数据集上，针对随机初始化的 200M 与 500M 参数学生模型，在九个零样本基准测试中平均准确率分别达到 36.62% 和 37.40%，显著优于最强基线方法（提升 0.94 和 0.83 个百分点），并较未使用蒸馏预训练模型提升 2.31 和 2.91 个百分点。因此，解决方案的关键在于：通过融合学生候选提议与教师引导的自适应局部关系建模，结合动态区域划分与差异性权重分配机制，实现更精细、鲁棒且高效的相对偏好对齐。
链接: https://arxiv.org/abs/2609.03355
作者: Quang Hoang Trung,Quang Huu Hieu,Nguyen Van Hoang Phuc,Vo Nguyen Le Duy
机构: VJ Technologies(越南科技公司); AJ Technologies(日本科技公司); Vietnam National University(越南国家大学); University of Information Technology(信息科技大学)
类目: Machine Learning (stat.ML); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Logit-based knowledge distillation for autoregressive language models usually aligns teacher and student next-token distributions over the entire vocabulary. However, this global objective overlooks relative preferences among likely token alternatives. Existing local approaches often select candidate tokens from either the teacher or the student alone. Teacher-only selection can miss tokens that the student considers likely, while student-only selection can rely on an inaccurate ranking early in training. We propose Adaptive Local Relational Alignment (ALRA), a position-specific framework combining student proposals with teacher guidance. At each valid prediction position, the student proposes likely tokens, while the teacher’s most probable token is included as an anchor. ALRA adjusts the number of selected tokens according to how broadly the teacher distributes probability within this candidate set relative to the current batch. Adaptive Local Divergence retains the mass-matching term and separately matches the relative token distributions within the selected and remaining vocabulary regions. Unlike the exact full-vocabulary decomposition, it replaces the teacher-mass coefficients of the two conditional terms with unit coefficients, preventing either term from being downweighted solely because its region has low teacher probability. Student-Weighted Pairwise Relational Alignment emphasizes high-probability token pairs with small student probability gaps and gives less weight to unlikely or clearly separated pairs. Experiments on The Pile with randomly initialized 200M- and 500M-parameter students across nine zero-shot benchmarks yield average accuracies of 36.62% and 37.40%. ALRA exceeds the strongest competing distillation baseline by 0.94 and 0.83 percentage points and improves over pre-training without distillation by 2.31 and 2.91 points, respectively.
[NLP-107] A Spectral Phase Admissibility Certificate for Complex Linear Maps
【速读】： 该论文旨在解决机器学习中复杂线性映射的稳定性与可微性评估问题，传统方法通常依赖于范数或正定性分析，而本文引入量子引力中的Kontsevich-Segal-Witten准则，提出一种基于谱集体相位约束的新评估框架。其解决方案的关键在于构建三种可微验证器：行列式部分、子集乘积包络以及完整准则，其中子集包络通过禁止外幂特征值触及负实轴，精确匹配指数型小主元枚举的接受/拒绝决策，同时大幅降低计算开销。研究进一步通过Schur参数化实现了该约束的可微施加，并明确了该方法的有效边界——由于相位预算受限会损害特征向量条件数，该方法无法有效处理深层线性传播；同时，其对基于幅值的目标（如归一化流似然）完全不敏感，因此应仅限于处理谱乘积辐角的模型。
链接: https://arxiv.org/abs/2609.02911
作者: Snigdha Chandan Khilar
机构: 未知
类目: General Physics (physics.gen-ph); Computation and Language (cs.CL)
备注:
Abstract:The paper imports the Kontsevich Segal Witten criterion from quantum gravity into machine learning to evaluate complex linear maps Standard techniques analyze magnitude or positive definiteness whereas this method exclusively limits the collective phase of a spectrum The researchers create three distinct differentiable certificates comprising a determinant sector a subset product envelope and the full criterion The subset envelope prevents all exterior power eigenvalues from touching the negative real axis This constraint precisely matches the accept or reject choices of an exponential minor enumeration while reducing processing expenses drastically The team provides a differentiable enforcement application via a Schur parameterization The document also identifies crucial boundaries regarding where this system works The constraint cannot balance deep linear propagation since restricting the phase budget damages eigenvector conditioning Furthermore the technique remains completely blind to magnitude based targets like normalizing flow likelihoods Thus researchers must restrict this tool specifically to models that process the argument of a spectral product
[NLP-108] DisclosureBeta: A Measurement-Channel Theory for Regime-Conditioned Betas from LLM -Read Risk Disclosures
【速读】： 该论文旨在解决在企业价格历史过短（如S-1申报企业、近期上市或刚经历制度变迁的公司）时，传统方法无法可靠估计贝塔（beta）的问题。现有方法依赖可比公司同行贝塔，但缺乏误差预算，而近期基于文本的竞争对手Breitung（2025）虽在IPO预测中表现出色，却未提供识别理论、误差预算及下界分析。本文的关键解决方案是将大型语言模型（Large Language Model, LLM）建模为对上市公司潜在风险特征的噪声测量通道，并将该通道噪声纳入资产定价的误差预算中。在分段平稳的Fama-French五因子模型框架下，因子载荷被定义为潜在风险特征与推断出的经济状态（regime）的函数。研究在对测量通道、检测器及组内抽样设定明确假设的基础上，证明了状态条件下的载荷函数具有可识别性与一致性，并给出了匹配的下界，表明仅基于收益、因子、LLM特征和状态估计值的任何估计器都无法避免披露噪声与检测误分类项。此外，研究提出一个披露激励相关推论，表明估计精度随企业层面的披露激励度量（Disclosure-Incentive Measure, DIM）单调递增。最后，通过构建文本基与滚动窗口估计器的自适应凸组合，实现性能不低于任一单一组件，并在价格历史短、陈旧或跨越检测到的状态断裂时自动向文本方法倾斜权重。该研究已预先注册设计并建立理论基础，实证评估将在冻结面板数据上进行，以确保优先权独立于实证结果。
链接: https://arxiv.org/abs/2609.02900
作者: Ping Kuen Wong
机构: 未知
类目: Risk Management (q-fin.RM); Computation and Language (cs.CL); General Finance (q-fin.GN)
备注: 8 pages, 0 figures. Theory preprint; empirical evaluation forthcoming in a companion paper
Abstract:The problem is the beta a desk needs when a firm’s price history is too short to trust: an S-1 filer, a recent listing, or a name just past a regime break. The state of the art collapses to a comparable-firm peer beta with no error budget, and the recent text-based competitor Breitung (2025) reports strong empirical IPO accuracy but no identification theory, no error budget, and no lower bound. We fill that gap. We model a large language model as a noisy measurement channel on a firm’s latent risk characteristics and write its channel noise into the asset-pricing error budget. In a piecewise-stationary Fama-French five-factor model the loadings are a function of latent risk characteristics and an inferred regime. We prove identification and consistency of the regime-conditional loading function under explicit assumptions on the channel, the detector, and within-regime sampling, and give a matching lower bound showing that the disclosure-noise and detector-misclassification terms are unavoidable for any estimator that observes only returns, factors, LLM features, and a regime estimate. A disclosure-incentive corollary makes estimation precision monotone in a firm-level disclosure-incentive measure (DIM). An adaptive convex combination of the text-based and rolling-window estimators is never worse than either component and shifts its weight toward text exactly when price history is short, stale, or straddles a detected regime break. The empirical evaluation on a frozen, pre-registered panel of price-history-thin firms is forthcoming; this preprint records the theory and the pre-registered design so priority is established independently of the empirical outcome.
信息检索
[IR-0] CORE: Improving Compositional Reasoning in MLLM Embedding via Reranker Distillation
链接: https://arxiv.org/abs/2609.04083
作者: Tingyu Song,Mingxin Li,Yanzhao Zhang,Dingkun Long,Chu Liu,Pengjun Xie,Yilun Zhao,Shu Wu
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Information Retrieval (cs.IR)
备注:
Abstract:MLLM-based embedding models remain limited in compositional retrieval, often failing to distinguish scenes containing the same concepts but different attribute-object bindings. Yet the same backbone can resolve such distinctions when used as a cross-attentive reranker, motivating us to distill its compositional judgments into the embedding model. We propose CORE, which synthesizes candidate lists spanning five compositional matching levels and introduces a Rank-KL objective that trains the embedding model to reproduce the reranker’s fine-grained ranking. We further introduce a graded evaluation protocol and compare contrastive learning, pairwise CoSENT, and listwise Rank-KL under the same data and tuning budget. Our comparison shows that both CoSENT and Rank-KL use the multi-level supervision more effectively than contrastive learning, with Rank-KL achieving the strongest overall performance. Across three compositional reasoning benchmarks (COLA, SUGARCREPE++, NEGBENCH), CORE-RERANKER-8B achieves an 82.7% total average, outperforming Jina-Reranker by 10.7 points, while CORE-EMBED-8B achieves the best total average (0.666) among all evaluated embedding models. The improvements transfer to the MCMR benchmark without sacrificing retrieval performance on COCO and Flickr30K.
[IR-1] he Dice Roll Method: A Standardized Protocol for Repeated-Query Auditing of Large Language Model Brand Recommendations
链接: https://arxiv.org/abs/2609.04047
作者: Dmitrij Żatuchin
类目: Information Retrieval (cs.IR); Computation and Language (cs.CL)
备注: 30 pages, 2 figures, 19 tables. Substantially revised; supersedes the Research Square preprint https://doi.org/10.21203/rs . this http URL -8883056/v1. Includes a pre-registered external validation on three independent corpora (Motoki et al., Rozado, llm-stability)
Abstract:Background: Researchers increasingly use repeated identical prompts to audit stochastic variation in large language model (LLM) brand recommendations, yet no standardized protocol exists for setting iteration counts, selecting stability metrics, or establishing reliability thresholds. Objective: We formalize the Dice Roll Method as a reusable protocol for repeated-query auditing of LLM brand recommendations, grounded in a generative model of temperature-scaled nucleus sampling. Methods: Total response variance is decomposed into sampling, prompt-phrasing, run-to-run, and model-version components. The stack: a negative-binomial mixed model with iterations as repeated measures; Cliff’s delta as the distribution-free effect size; dependence-preserving bootstrap; simulation-based power; a generalizability-theory decomposition; drift diagnostics on pinned snapshots. We reanalyse five brand-recommendation auditing studies: approximately 190,000 observations, 270+ brands, 6 languages, iteration counts 5 to 40. Results: Three tiers of iteration guidance emerge from the D-study: exploratory (n = 5, G = 0.58), confirmatory (n = 10, G = 0.74), and rigorous (n = 15, G = 0.81), tied to effect-size and generalizability targets. The four metric families (count, set, embedding, fairness-adjusted PASOR) are complementary, motivating a compact metric battery over single indicators. A pre-registered external validation on three independent corpora (Motoki et al., 100-round; Rozado, 24 models; llm-stability) reproduces the D-study reliability prediction in 37 of 39 cells with no failures and the n = 5 power value to two decimals; the fixed tiers do not transfer, supporting a pilot-then-solve reading. Conclusion: The protocol gives repeated-query auditing of LLM brand recommendations a statistically principled footing under the conditional, non-Gaussian structure of real autoregressive generation.
[IR-2] he WebKurator.de Platform: Combined Regional and Topical Web Curation
链接: https://arxiv.org/abs/2609.03971
作者: Michael Dinzinger,Natanael Arndt,Ben Böck,Jelena Mitrović,Michael Granitzer
类目: Information Retrieval (cs.IR)
备注:
Abstract:The systematic curation of the Web remains a central challenge for national libraries and memory institutions that aim to preserve culturally and regionally relevant content. Existing directory-based approaches such as Curlie implement a predominantly topic-centric, one-dimensional hierarchy, where geographic aspects are intertwined with topical and linguistic categories. To address this limitation, we present this http URL, a collaborative platform for combined regional and topical web curation, initially focused on the German web. WebKurator introduces a two-dimensional curation model that explicitly separates topical categorization and geographic annotation. The system integrates LLM-based topic classification and imprint-based address extraction with geocoding, and supports user suggestions together with moderated review. The platform is bootstrapped from the German Imprints Dataset, a large-scale collection of 5.54 million websites. Among them, 3.14 million contain imprint pages, for which we successfully extracted and geocoded postal addresses. Of these, 2.58 million (85.17%) are located in Germany and also have an assigned topic label. These websites form the initial foundation of this http URL and can be continuously extended through user suggestions.
[IR-3] RuleMem: Active Rule Memory for Long-Term Conversational Agents
链接: https://arxiv.org/abs/2609.03915
作者: Xingyuan Zeng,Zuohan Wu,Quanming Yao,Yue Wang,Wei Liu,Libin Zheng,Jiuke Wang,Jian Yin
类目: Computation and Language (cs.CL); Information Retrieval (cs.IR)
备注:
Abstract:Question answering agents in long-term conversations must reason over massive, temporally dispersed dialogue histories. However, existing memory mechanisms primarily treat past information as \textitpassively stored facts, leading to semantic gaps and unreliable reasoning. To address this limitation, we propose RuleMem, a rule-based memory framework that induces reusable logical rules from historical interactions to \textitactively guide both evidence retrieval and reasoning. Specifically, RuleMem constructs natural-language Horn clauses from conversations and validates them via a Rule Perplexity Consistency (RPC) mechanism. These induced rules enable the retrieval of semantically distant evidence while providing an explicit logical structure for answer generation. We conducted a comprehensive evaluation of RuleMem on two long-term conversational benchmarks, LoCoMo and LongMemEval_s*. In a rigorous comparison against 14 baselines on LoCoMo, RuleMem achieved the highest accuracy, exceeding the baseline average by 27.47 points (a 54.3% relative improvement).
[IR-4] Comparing Retrieval Methods for Academic Advisor Discovery: A Six-Method Study of 768 CS Faculty Profiles Across 9 US Universities
链接: https://arxiv.org/abs/2609.03901
作者: Biraj Subedi
类目: Information Retrieval (cs.IR); Machine Learning (cs.LG)
备注: 16 pages, 2 figures, 9 tables. Code and demo at this https URL
Abstract:We present a comparative evaluation of six information retrieval methods for the task of academic advisor discovery: ranking CS faculty members by relevance to a graduate applicant’s research interest statement. The methods span sparse lexical matching (Jaccard overlap, TF-IDF, BM25), dense semantic retrieval (all-MiniLM-L6-v2 sentence embeddings), hybrid score fusion, and learning-to-rank. Evaluation uses a new domain-specific collection: 768 faculty profiles scraped from 9 US CS departments, with 162 graded relevance judgments (grade 0/1/2) across 5 queries representing distinct graduate student research profiles. Across all five queries, Reranked achieves the highest mean NDCG@10 (0.477, std 0.138), followed by Semantic (0.450), Hybrid (0.421), BM25 (0.406), Jaccard (0.303), and TF-IDF (0.246). After Bonferroni correction across all 15 pairwise comparisons, TF-IDF is significantly worse than BM25, Semantic, Hybrid, and Reranked; no other pairwise difference survives correction at 5 queries. A field ablation reveals that biography alone (NDCG 0.634) outperforms the full model combining biography with research area tags (0.593). A controlled experiment shows that concatenating arXiv paper abstracts reduces NDCG@10 by 0.176, motivating a late-fusion architecture. All code, scrapers, and relevance labels are released openly.
[IR-5] GRASP: Graph-Retrieval Automated Scoring Pipeline for Label-Free Multi-Topic Essay Grading ICONIP2026
链接: https://arxiv.org/abs/2609.03857
作者: Aafreen Husain,Samar Shailendra,Saad Sajid Hashmi
类目: Information Retrieval (cs.IR)
备注: Accepted at ICONIP 2026
Abstract:Automated short-answer grading research has historically focused on exams consisting solely of questions pertaining to a single topic. Automatic grading of exams containing questions about more than one topic remains less explored. In this work, a Graph-Retrieval Automated Scoring Pipeline (GRASP) is introduced for grading label-free multi-topic science exams. Label-free exams are short-answer exams in which a student’s responses to several distinct topics are merged into a single paragraph, with no markup labels or segmentation indicating which span answers which question. Reference answers for each question are encoded into a FAISS vector index via Sentence-BERT, and a semantic similarity graph is constructed over this set of reference answers. At grading time, sentence count heuristics, with a large language model used to resolve ambiguous cases, are first applied to predict how many distinct topics were answered in the student essay. This process is performed without training data or domain-specific example essays. Candidate reference nodes, each storing one (question, reference answer, concatenation of both) from the reference index, are then retrieved through cosine similarity based Retrieval-Augmented Generation (RAG) and Graph Retrieval-Augmented Generation (GRAG). GRAG operates by taking the top cosine matches as seed nodes and then performing a graph traversal over strong edges to find additional reference nodes that may have been missed by RAG. The Hungarian algorithm is then used to optimally assign one reference node per question segment such that no reference is duplicated. Each segment is then graded against its assigned reference independently using GPT-4.1-mini. This experiment is performed to show the effect of retrieval quality on grading accuracy and the benefit of graph-augmented retrieval versus strict cosine similarity methods at various levels of essay complexity.
[IR-6] Unified Pitch Graphs for Diagnosing Pitching Strategy
链接: https://arxiv.org/abs/2609.03810
作者: Kichang Lee,JeongGil Ko
类目: Information Retrieval (cs.IR)
备注: 10 pages, 9 figures, 3 tables
Abstract:Pitching strategy in baseball is expressed through both physical execution and the ordered context in which pitches are used, yet common representations collapse pitches into discrete types or aggregate statistics. We present Unified Pitch Graphs (UPG), a hierarchical graph representation for retrospective analysis of sequential spatiotemporal events. UPG preserves each pitch as an exact event with reconstructed three-dimensional trajectory and context, connects consecutive pitches through directed sequence edges, and organizes the same events across semantic and temporal resolutions. A support-adaptive mechanism backs off from fine, long sequences when repeated evidence is insufficient, while retaining exact event lineage. We evaluate UPG on 3.94 million MLB Statcast pitches from 2021 to 2026. Nominally identical pitch sequences exhibit distinct physical executions, and ordered structure becomes increasingly evident in longer context-conditioned paths. Support-adaptive backoff increases held-out path coverage from 18.9% to 94.9% while improving execution reconstruction from R^2=0.495 to 0.685 . UPG also reliably localizes controlled execution changes that discrete pitch-mix and sequence representations cannot detect. These results demonstrate that UPG provides a traceable, multi-scale representation for identifying recurring strategy patterns without conflating retrospective associations with causal or future-performance claims.
[IR-7] LLM 4AIGQ: LLM -based AI Guidance Query Generation Framework for Multi Interest Mining
链接: https://arxiv.org/abs/2609.03674
作者: Xiangchen Pan,Jiayi Xu,Jing Wang,Xing Fang,Lingyun Zhu
类目: Information Retrieval (cs.IR)
备注:
Abstract:Guidance queries stimulate user consumption by extracting preferences to provide search queries with guidance value, playing a crucial role in the e-commerce field. Traditional AI-generated queries (AIGQ) generation primarily relies on a two-stage “Query-to-AI-Generated-Query” (Q2AIGQ) association paradigm, first recalling user primary search queries from user profiles, historical behavior sequences, item-side information, and the current query through multi-path retrieval, then generalizing AIGQ via rule-based methods. This approach suffers from semantic drift due to information cascade loss; additionally, primary search query derivation heavily depends on “user-item” co-occurrence relationships, lacking exploration of user multi-interests, resulting in guidance queries with low value and mismatched purchase intent. To address the expressive limitations of traditional co-occurrence-based retrieval, we propose LLM4AIGQ, an LLM-based solution for generating AI guidance queries tailored to users’ multi-interests. This approach segments user interests by integrating user profiles and historical interaction sequences, infers specific consumption intents for each sub-interest, and subsequently generates corresponding AIGQ. In terms of model training, we employ a post-training pipeline comprising Supervised Fine-Tuning (SFT), Reinforcement Learning (RL), and Direct Preference Optimization (DPO) to enhance the model’s capability in generating AIGQ. We also introduce a multi-level reward design to satisfy the requirements of multi-objective optimization and long-chain reasoning in practical applications. Regarding deployment, we adopt a nearline-generation and online-read architecture to meet latency constraints. Extensive experimental analyses demonstrate that our model achieves robust performance in both offline evaluations and online A/B tests.
[IR-8] Enhancing Financial Question Answering: A Novel Benchmark Dataset of Banks financial statements
链接: https://arxiv.org/abs/2609.03654
作者: Arianna Miola,Bruno Spaccavento,Lorenzo Silotto,Marco Bianchetti,Luca Cagliero
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Computational Engineering, Finance, and Science (cs.CE); Information Retrieval (cs.IR)
备注:
Abstract:The comparative analysis of banks’ financial statements poses significant challenges for automated question answering systems due to their complexity, substantial length, technical language, and inhomogeneity of both textual and numerical content across different jurisdictions and institutions. We introduce FinRAG-QA, a novel benchmark dataset for financial question answering, which comprises 999 practitioner-curated questions on 10 standardised indicators, grounded in 209 annual and Pillar 3 reports from 24 major European and U.S. banks spanning 2019-2023. Unlike prior financial QA benchmarks, which centre on U.S. filings and single-institution analysis, FinRAG-QA targets cross-institutional retrieval over documents averaging 198k words, longer than any existing financial QA resource. On this benchmark we evaluate a multi-stage RAG pipeline and isolate the contribution of each component. Contextual chunk enrichment combined with a retrieval-optimised embedding model raises NDCG@10 from 0.322 to 0.710; conditional on the ground truth being retrieved, a reasoning-optimised generator raises answer accuracy from 44.6% to 79.0% (+34.4 percentage points), at roughly 20x the generation latency. We further show that cross-encoder reranking degrades retrieval when the first-stage ranking is already strong, and that a single top-ranked chunk outperforms larger contexts at generation time. Experiments were run in late 2024-early 2025 with the models available at that time.
[IR-9] EPIC: Explicit Posterior Item Conditioning for Semantic ID Diffusion Recommendation
链接: https://arxiv.org/abs/2609.03522
作者: Tuan-Binh Tran,Thanh Tam Nguyen,Quoc Viet Hung Nguyen,Dung D. Le,Tung Kieu,Thanh Trung Huynh
类目: Information Retrieval (cs.IR); Machine Learning (cs.LG)
备注: 11 pages, 7 figures, 3 tables
Abstract:Semantic ID (SID) generative recommendation predicts the next item by generating a short tuple of discrete tokens. Recent masked-diffusion methods improve this process through bidirectional context and flexible decoding, yet recommendation ultimately requires selecting among complete catalog items. At each denoising step, a partial SID can correspond to multiple feasible items, while existing methods primarily reason through position-wise token predictions. We propose Explicit Posterior Item Conditioning (EPIC), which introduces explicit item-level competition into SID denoising. EPIC constructs a personalized posterior over feasible candidate items using the current generation context and the user’s recent interactions, then projects this distribution back to unresolved SID positions to guide subsequent token decisions. The pretrained backbone remains frozen and requires no additional decoder forward pass. Experiments on four Amazon benchmarks show consistent improvements over strong baselines, while diagnostic analyses indicate that the gains primarily arise from personalized transition evidence that preserves promising item hypotheses during denoising.
[IR-10] From Topical Relevance to Answerability: Entailment Distillation for Conversational Retrieval EMNLP2026
链接: https://arxiv.org/abs/2609.03482
作者: Shuai Qin,Guojia An,Weikang Guo,Pei Ke,Jiwei Wei,Yang Yang,Jie Zou
类目: Information Retrieval (cs.IR)
备注: Accepted to Findings of EMNLP 2026
Abstract:Existing conversational retrievers commonly treat topical relevance as a proxy for answerability. However, a passage that closely matches the dialogue context is not necessarily the one that supports the correct answer. We identify this mismatch as a systematic answerability gap. To address this issue, we propose CLEAR, a framework that shifts conversational retrieval from topical relevance to answerability. The core of CLEAR is entailment distillation, which transfers answer-passage entailment supervision into a cross-encoder reranker so that the reranker discriminates answer-supporting passages from topical distractors at inference time, without requiring answers. CLEAR is complemented by a passage-centric abductive recall module that brings low-similarity yet answerable passages into the candidate pool by inferring answerable queries from passages with an LLM. Across TopiOCQA, QReCC, and out-of-domain TREC CAsT datasets, CLEAR consistently improves top-ranked precision over strong query-rewriting and dense-retrieval baselines, with the largest gains observed in conversations involving heavier topical noise. Moreover, applying our reranker on top of an LLM-driven query rewriter yields further gains.
[IR-11] ExplainRoute: A Pre-Deployment Audit Framework for Non-Answer-Giving Programming Tutors
链接: https://arxiv.org/abs/2609.03470
作者: Yiming Gai,Yingying Zhang,Xuefei Huang
类目: Information Retrieval (cs.IR)
备注:
Abstract:Programming tutors should support learners’ own explanations rather than immediately providing model answers. We present ExplainRoute, a pre-deployment audit framework for non-answer-giving programming tutors. Given a code line and a learner explanation, it estimates the explanation state and selects one of two bounded responses: a Feynman-style self-explanation prompt or a Socratic scaffold. The framework exposes its state, strategy, cited code fragment, and leakage risk through a machine-checkable contract. Unlike benchmarks that rank tutors by fluency alone, ExplainRoute audits information boundaries, response polarity, failure closure, and the value of learner-explanation visibility before classroom deployment. We evaluate it offline on the 1,770-pair SelfCode corpus using a code-group split, with 443 pairs reserved in 11 untouched holdout groups. The evaluation compares direct answers, fixed open self-explanation, fixed Socratic scaffolding, adaptive routing, and an adaptive no-state ablation. Contract validity reaches 100% for all pedagogical conditions. Adaptive routing matches the frozen reference rule on 60.5% of records, with state macro-F1 of 0.238 (Open: 0.229; Socratic: 0.246), showing no reliable adaptive advantage. An independent language-model judge scores adaptive responses 4.516/5, outperforming the no-state ablation (2.819/5) but slightly below fixed open self-explanation (4.598/5) and Socratic scaffolding (4.658/5). A blinded rubric evaluation on a stratified 40-row subset confirms that visible learner explanations improve information value while adaptive routing does not outperform fixed strategies. The contribution is a validated audit protocol and a boundary finding, rather than evidence of improved learning, retention, or causal instructional effectiveness.
[IR-12] When Retrieval Helps: Selective Retrieval for Single-Turn Mental-Health QA KDD2026
链接: https://arxiv.org/abs/2609.03454
作者: Hyunseo Oh,Chong-Kwon Kim,Yoonhyuk Choi
类目: Computation and Language (cs.CL); Information Retrieval (cs.IR)
备注: 8 pages, 3 figures. Presented at the KDD 2026 Undergraduate Consortium
Abstract:Retrieval-augmented generation (RAG) can improve the specificity and grounding of large language model responses, but its effect is not uniformly beneficial in single-turn mental-health question answering, where user queries often combine emotional distress, treatment concerns, and safety-sensitive needs. We study when retrieval helps or hurts mental-health QA, and whether a lightweight selective retrieval policy can better control this trade-off. We operationalize retrieval need using three draft-conditioned utility dimensions: psychoeducational need, coping need, and response specificity, together with a rule-based safety trigger. Following psychotherapy-grounded RAG systems such as coTherapist, we construct a compact and controllable guideline corpus comprising coping-strategy, psychoeducational, and safety resources. We fine-tune an instruction-tuned generator on MentalChat16K using QLoRA and compare Closed-book, Always Retrieval, and Selective Retrieval settings on CounselBench-Eval and CounselBench-Adv. Experiments show that retrieval is not uniformly beneficial in this domain. Always Retrieval improves specificity but lowers overall quality and introduces additional safety-sensitive failures. Selective Retrieval preserves closed-book behavior for low-need cases while avoiding the additional degradation caused by unconditional retrieval, supporting the view that retrieval activation is a safety-sensitive control decision.
[IR-13] Plan Pointers and Record-Directive Form in Budgeted Verification of Inherited Agent Memory
链接: https://arxiv.org/abs/2609.03450
作者: Kazuki Nakayashiki
类目: Information Retrieval (cs.IR); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 46 pages, 7 figures, 35 tables. Twelve registered studies (14,760 attempted episodes) on one instrument lineage; every package was frozen, timestamped and externally deposited before its first confirmatory call. Manuscript, LaTeX source, all episode files, frozen packages, analyzers and the generator of every number are archived at Zenodo: doi: https://doi.org/10.5281/zenodo.22267221
Abstract:An agent that inherits six one-line memories may pull at most one archived source record before acting; a directive written into the store can steer that choice: a pointer to the record, a criterion that identifies it, or both. Across twelve registered studies on one instrument lineage (14,760 attempts) we measured where the request goes under each form. On six direct-provider models a length-matched criterion exceeded a bare id by +35.0 points [+31.2, +38.8] (Study D); the contrast failed its registered superiority rule on a nine-model OpenRouter-served panel (Study E). Appending the id cancelled the criterion on three Claude models (Opus 5: 40/40 to 0/40; Study F-x); six byte-matched edits gave each exact string its own effect (Study G), and a re-run at eighty runs per cell left fifteen of thirty replication contrasts within the margin, fifteen unresolved and none beyond (Study G’). A ratification line (+96.0 points on Opus 5) and a budget of two credits restored the target on all three (Study J); across five criterion strings the suffix’s cancellation held for four of the five wordings on Opus 5 and all five wordings on Fable 5.1 (Study H2); in a second store every model followed the criterion (Study H1). Continued into a decision, the criterion moved the choice toward the current record (+100.0 points, Opus 5) and away from it on Fable 5.1 (Study I). A one-character plan pointer’s effect (+78.0 points; Study B, after a correction of its first repository report) returned the same verdict under a prospectively registered re-run (+81.7 points; Study B’). All results are descriptive effects of exact edits on fixed panels with registered intervals and no mechanism claim.
[IR-14] Spruce: Scalable Private Outsourced Retrieval Using Compact Embeddings
链接: https://arxiv.org/abs/2609.03376
作者: Peichun Hua,Yunming Xiao
类目: Cryptography and Security (cs.CR); Information Retrieval (cs.IR); Machine Learning (cs.LG)
备注: 22 pages, 10 tables, 6 figures
Abstract:Retrieval-Augmented Generation (RAG) has made dense retrieval over large document collections a standard building block. Organizations increasingly outsource vector indexes to untrusted clouds, exposing proprietary corpora and user queries. Cryptographic protection is challenging because each query searches corpus-scale state, causing computation, correlated randomness, and communication to grow with the corpus. At million-document scale, a naive secure implementation takes minutes and about 90 GB of communication per query. Even recent optimized systems require 10–22 seconds. We propose Spruce (Scalable Private Outsourced Retrieval Using Compact Embeddings), which co-designs representations with the cryptographic protocol. Spruce learns compact binary codes that preserve candidates for full-precision reranking, replacing corpus-wide embedding scoring with efficient Hamming-distance computation under two-server multi-party computation (MPC). A corpus-calibrated fixed-radius protocol avoids multi-round candidate selection while preserving retrieval quality. Spruce also provides private cluster pruning, which trades minor quality loss for substantially less computation, and a one-core owner-operated dealer that removes cloud OT preprocessing bottlenecks. Across four corpora containing 383K–5.42M documents, Spruce preserves the original search quality with median candidate sets of only 382–1,952. At 10 Gbps inter-server bandwidth, full scans take 0.21–2.97 seconds, 4.8 – 6.7\times faster than the closest measured prior work. Private pruning takes 0.06–1.09 seconds, achieves 13.1 – 22.9\times speedups, and retains 93.9% – 97.3% of full-float NDCG. On the largest corpus, pruning and the dealer jointly improve sustained throughput by 31.5\times at 1 Gbps per link. Comments: 22 pages, 10 tables, 6 figures Subjects: Cryptography and Security (cs.CR); Information Retrieval (cs.IR); Machine Learning (cs.LG) Cite as: arXiv:2609.03376 [cs.CR] (or arXiv:2609.03376v1 [cs.CR] for this version) https://doi.org/10.48550/arXiv.2609.03376 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[IR-15] HypRQ-VAE: Hyperbolic Item Indexing for Long-Tail-Aware Generative Recommender Systems ICDM2026
链接: https://arxiv.org/abs/2609.03369
作者: Longfeng Wu,Tong Zeng,Giovanni Seni,Zhimin Peng,Bhanu Pratap Singh Rawat,Si Zhang,Yao Zhou,Lecheng Zheng,Bo Ji,Yujun Yan,Dawei Zhou
类目: Information Retrieval (cs.IR)
备注: Accepted for publication in the 2026 IEEE International Conference on Data Mining (ICDM 2026)
Abstract:Sequential recommender systems model user behavior as item ID sequences, while recent generative methods cast recommendation as a language modeling task using large language models (LLMs). While this paradigm incorporates rich textual semantics, it introduces a fundamental mismatch: LLMs operate on text tokens, whereas recommender systems depend on discrete item indices. This misalignment often leads to hallucinations in generative recommendations. Existing methods attempt to bridge this gap by learning item vocabularies in Euclidean space, but they struggle to model the inherent long-tail distribution of real-world catalogs, where a small number of head items dominate, and a vast number of tail items reflect users’ niche preferences. To address this issue, we introduce Hyperbolic Residual-Quantized Variational AutoEncoder (HypRQ-VAE), the first framework to learn item indexing in hyperbolic space. HypRQ-VAE leverages the unique properties of hyperbolic geometry, whose exponential volume expansion naturally accommodates the power law structure of user-item interactions. This allows the model to encode rich textual semantics while preserving the representational fidelity of sparse, long-tail items. Experiments on three benchmark datasets show that HypRQ-VAE significantly improves the performance of recommendation, particularly in recommending tail items. Our analysis attributes these gains to the superior capacity of hyperbolic space to model item hierarchies and sparsity in generative recommendation. Our code and data are available at: this https URL.
[IR-16] SciLENS: RL-Driven Autonomous Agents for Scientific Localized Evidence Navigation and Synthesis
链接: https://arxiv.org/abs/2609.03338
作者: Leqi Zheng,Jinbo Su,Yuying Li,Chaokun Wang,Weiping Wang,Haitao Li,Jiajun Zhang,Shannan Yan,Zhaolu Kang,Rong Fu,Jie Wu,Fang Niu,Hang Zhang
类目: Information Retrieval (cs.IR)
备注:
Abstract:Scientific literature synthesis agents increasingly rely on proprietary online services, limiting reproducibility, privacy, and offline deployment. To address this challenge, we introduce SciLENS Scientific Localized Evidence Navigation and Synthesis), a fully local autonomous agent framework operating on a dual-tier infrastructure indexing approximately 12 million academic records. SciLENS pioneers the integration of structural visualization as an actionable tool within the reasoning loop, enabling the agent to compress complex citation topologies into validated data-driven charts and thereby mitigate context exhaustion during macro-level synthesis. To train the agent without human annotation, we develop an automated data synthesis pipeline that extracts multi-hop subgraphs from a citation knowledge graph, verified by cross-model consensus among 20 frontier models. The agent is subsequently aligned through a reverse-decomposition rubric strategy that provides fine-grained process rewards for early planning and strict evidence grounding. Evaluations across six scientific benchmarks encompassing standard QA, citation accuracy, factual reasoning, and structural synthesis demonstrate that SciLENS significantly outperforms open-source baselines and achieves performance comparable to GPT-5.2 and Gemini-3.0-pro. Our source code and data are released at this https URL.
[IR-17] SelfDR: Self-Distillation from Reasoning for LLM -Based Recommendation CIKM’26
链接: https://arxiv.org/abs/2609.03313
作者: Chumeng Jiang,Jiayin Wang,Xinjie Lin,Zhiqiang Guo,Hengliang Luo,Min Zhang
类目: Information Retrieval (cs.IR)
备注: 12 pages, 5 figures, CIKM’26
Abstract:Large Language Models (LLMs) have recently emerged as powerful backbones for recommendation. To better elicit their capabilities, reasoning has been widely incorporated to help LLMs interpret rich textual signals and improve recommendation accuracy. However, explicitly generating intermediate reasoning traces often incurs substantial computational costs, which limits practical deployment in real-world recommender systems. To address this challenge, we propose SelfDR, a Self-Distillation from Reasoning framework for LLM-based Recommendation. SelfDR distills an LLM’s own reasoning-enhanced predictions to produce recommendations directly, improving recommendation effectiveness while maintaining inference efficiency. All components in the framework are built on the same base LLM, without relying on any external models. Specifically, the teacher recommender is constructed by training a reasoner with downstream performance as the reward, enabling it to generate targeted rationales that are later incorporated into the teacher’s input. A student recommender for direct recommendation, with the same underlying model, then learns from the teacher through self-distillation with a dynamic weighting strategy. Extensive experiments on three public datasets validate the effectiveness, rationality, and efficiency of SelfDR. Codes are available at this https URL.
[IR-18] DoPR: Reusable Compressed Document Prefixes for Efficient LLM Reranking
链接: https://arxiv.org/abs/2609.03311
作者: Beiya Dai,Yifan Wei,Guang Yang,Xing Shi,Xinbing Wang,Zhouhan Lin
类目: Information Retrieval (cs.IR)
备注: 15 pages, 6 figures
Abstract:Large language models (LLMs) are effective rerankers, but pointwise reranking repeatedly processes the same document across different queries, causing substantial redundant document-side computation. We propose \textbfDoPR, a compressed document prefix framework that decouples offline document processing from online reranking. DoPR first selects query-independent document representations and converts them into compressed document prefix states, which are precomputed offline and reused whenever the document is retrieved. During online reranking, the model scores each query-document pair by processing only the query and scoring token, with document information supplied by the stored prefix states. This design reduces online cost through both document-side compression and cross-query prefix-state reuse. Experiments on TREC DL, BEIR, and BRIGHT with Qwen3 models from 0.6 B to 8 B show that DoPR achieves up to 8.0 \times online document-side memory reduction and up to 8.04 \times latency speedup, while retaining \textbf97.1%-99.5% of the average NDCG@10 of matched full-document rerankers.
[IR-19] UniCon: A Unified Context-Centric Modeling Paradigm for CTR Prediction
链接: https://arxiv.org/abs/2609.03290
作者: Jiajun Cui,Zhengqi Xu,Fan Zhang,Zhangteng,Gu Tang,Honghong Zhu,Mengxi Wu,Yulin Liang,Xingxing Wang
类目: Information Retrieval (cs.IR)
备注: 10 pages, 5 figures, 2 tables
Abstract:Unified modeling has become a major direction for industrial click-through rate (CTR) prediction. Existing approaches typically unify sequential and non-sequential signals at the token level, model their interactions in a shared backbone, and increase model capacity to improve scaling behavior. However, this division originates from legacy feature-engineering practice and is misaligned with the underlying decision process. User behavior is inherently a sequence of homogeneous context units; at the level of input organization, historical behavior and the current request differ only in whether their outcomes are observed or remain to be predicted. Treating them as heterogeneous signals obscures structural dependencies within the user’s decision context, limiting both scaling efficiency and prediction quality. This limitation is particularly pronounced in context-rich scenarios such as e-commerce shelves and waterfall feeds. To address this, we propose UniCon, a unified context-centric modeling architecture that treats the request context as the basic modeling unit and organizes history and prediction targets as homogeneous context units. Intra-context attention captures local coupling among items within a context (Locality), while inter-context attention models the dynamic evolution of decision states across contexts (Dynamics). This organization bridges the structural gap between history and target and supports more effective scaling of unified CTR models. Context-unit-level sequence compression further reduces deployment overhead. On Meituan search advertising, UniCon improves offline AUC by 0.0139 over a strong production baseline and achieves statistically significant online lifts of 3.09% in RPM, 2.07% in CTR, and 2.95% in revenue.
[IR-20] SHELF: A Synthetic Harness for Multi-Task Bibliographic Benchmarking
链接: https://arxiv.org/abs/2609.03047
作者: Michael J. Bommarito II
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Information Retrieval (cs.IR)
备注: 17 pages, 15 tables. Code available at this https URL ; data available at this https URL
Abstract:Libraries and archives manage large collections with limited staff and computing budgets, yet common benchmarks do not systematically test their bibliographic work. They need to know which methods work for their tasks and what those methods require to run. SHELF, the Synthetic Harness for Evaluating LLM Fitness, addresses this gap. It is a Python system that turns labelled taxonomies, writing specifications, and a generation budget into controlled benchmark data and evaluation tasks. This first release contains 62,899 model-written documents based on Library of Congress vocabularies, with tasks for classification, clustering, retrieval, pair classification, and instruction retrieval. We compare TF, TF-IDF, BM25, popular encoders, and, on subject classification only, zero-shot decoders; each method appears only on tasks that support it. Subject classification reaches 0.8887, while genre-form classification reaches only 0.2605, and several pair and clustering tasks remain near chance. Sparse methods remain competitive on classification, while TF-IDF is the fastest measured arm in the subject timing experiment. SHELF also varies bibliographic facets independently and can generate new, verifiably unseen documents after a model’s training cutoff. Comparisons with LCSHBench and Project Gutenberg show that model rankings transfer more reliably than absolute scores, but SHELF scores do not estimate accuracy on production catalogue data. We release all source code and data under permissive licenses on GitHub and Hugging Face.
[IR-21] Reflect-SQL: A Self-Reflection Based Framework for Text-to-SQL PAKDD2026
链接: https://arxiv.org/abs/2609.02944
作者: Anupreksha Jain,Manish Shrivastava
类目: Information Retrieval (cs.IR); Artificial Intelligence (cs.AI)
备注: Accepted in PAKDD 2026
Abstract:Democratizing data access through natural language is a crucial goal for modern enterprises, but the practical adoption of Text-to-SQL is critically hindered by real-world complexities: 1. Obscure and large database schemas, 2. Ineffective retrieval of relevant tables and columns due to structured setting of schemas and vague user query, 3. Generation of syntactically or logically flawed SQL due to a lack of robust validation and correction mechanism. To address these systemic challenges, we introduce Reflect-SQL, a novel framework for Text to SQL, grounded in multi-stage self-reflection approach to develop understanding of obscure schema using a knowledge base, setup a process for effective retrieval and system to generate syntactically/semantically SQL. Instead of a single-pass attempt, our system employs an LLM-as-a-judge driven scoring mechanism within interconnected feedback loops to iteratively refine the results at every stage. A feedback-driven retrieval loop refines the user’s natural language query, while a synthesis loop validates and corrects the SQL and finally, an entailment loop optimizes the end-to-end process and continuously enriches the knowledge base. By integrating these layers of reflection, Reflect-SQL bridges the critical gap between user intent and complex data. On the challenging BIRD benchmark, our framework achieves an execution accuracy of 72.03%, significantly outperforming state-of-the-art baselines, demonstrating a major leap in reliability for enterprise applications.
[IR-22] CHSR-RRF: A curriculum-gated hybrid retrieval framework with reciprocal rank fusion and leakage-aware benchmarking for educational RAG
链接: https://arxiv.org/abs/2609.02913
作者: Terence Ateya,Zavier Ndum Ndum,Jicheng Fu,Kelly Tendongkeng
类目: Information Retrieval (cs.IR)
备注: 45 pages, 14 figures
Abstract:Retrieval-augmented generation (RAG) is increasingly used in educational question answering, but standard retrievers optimize topical relevance without enforcing curriculum validity. In school settings, a passage can be relevant yet inappropriate if it comes from the wrong subject, level, or examination context; we call this failure mode curriculum leakage. We present CHSR-RRF, a curriculum-gated hybrid retrieval framework that applies metadata constraints before retrieval, then combines sparse and dense search with reciprocal rank fusion and deterministic reranking. We also introduce CERB, a 126-case benchmark for curriculum-constrained retrieval with hierarchy-aware relevance labels and explicit leakage annotations. On a 61-case pilot, pre-retrieval gating reduces leakage by 4.6x ( p0.001 ) while preserving ranked recall, whereas applying the same constraints after retrieval collapses recall and exact-scope success to zero ( p=0.039 ). A full-benchmark lower-bound analysis further shows that many remaining failures arise from corpus and metadata gaps rather than retrieval design alone. These results show that retrieval in structured educational domains should be treated as constrained selection, with validity enforced when the candidate pool is formed rather than after ranking.
[IR-23] R2Adapter: A Routing and Rewriting Adapter for Efficient Hybrid RAG
链接: https://arxiv.org/abs/2609.02894
作者: Yucan Guo,Miao Su,Saiping Guan,Long Bai,Zhongni Hou,Zixuan Li,Xiaolong Jin,Jiafeng Guo,Xueqi Cheng
类目: Computation and Language (cs.CL); Information Retrieval (cs.IR)
备注:
Abstract:Retrieval-Augmented Generation (RAG) has become a prevailing paradigm for enhancing Large Language Models (LLMs) with non-parametric knowledge. Vanilla RAG efficiently handles simple queries but struggles with relational or multi-hop reasoning. Graph-based RAG alleviates this issue but incurs higher inference complexity and latency. In practice, user queries can differ significantly in their complexity, rendering a fixed RAG strategy suboptimal. However, existing hybrid text-graph RAG methods typically rely on heuristic and LLM-based routing, resulting in unnecessary overhead and strong dependence on the underlying LLM. To address these challenges, we propose R ^2 Adapter, a lightweight plug-in Routing and Rewriting Adapter designed to allocate queries between vanilla and graph-based RAG dynamically. By routing only the queries that genuinely benefit from graph-based reasoning, R ^2 Adapter reduces unnecessary graph retrieval overhead. Additionally, uncertain graph-routed queries are rewritten to better expose their multi-hop reasoning requirements, improving retrieval quality without additional supervision. Extensive experiments on three multi-hop QA benchmarks demonstrate that R ^2 Adapter reduces graph-based RAG usage by up to 59% while maintaining comparable answer accuracy. This adapter is model-agnostic and can be seamlessly integrated into diverse vanilla and graph-based RAG pipelines, providing an efficient and adaptive solution for hybrid RAG systems.
[IR-24] ExecRetrieval: Measuring the Functional-Correctness Gap in Code-Embedding Retrieval EMNLP2026
链接: https://arxiv.org/abs/2609.01865
作者: Aaryan Kapoor,Md Abdullah Al Hafiz Khan
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Information Retrieval (cs.IR)
备注: Accepted to EMNLP 2026 (Main Conference). Camera-ready version. 17 pages, 6 figures, 8 tables
Abstract:Embedding-based code retrieval is a core component of coding agents and retrieval-augmented code generation, where retrieving correct code matters more than retrieving lexically similar code. Existing code-retrieval benchmarks do not plant controlled, execution-verified single-edit variants of each query’s canonical implementation in the search pool, leaving the question of whether embeddings can functionally discriminate correct from near-clone-but-incorrect code unanswered in a retrieval setting. Resolving this requires a benchmark whose search pool itself contains the relevant counterfactuals – execution-verified buggy variants near-identical to each canonical – so that a retriever’s rank ordering can be directly tested for functional discrimination rather than topical or identity overlap. We introduce ExecRetrieval, 939 Python tasks each paired with one execution-verified canonical implementation and up to four execution-verified buggy distractors, each generated by a mechanical mutation making a single targeted edit, and evaluate 23 dense embedding configurations plus BM25 under provider-native invocation with paired McNemar tests and query-level bootstrap intervals. With near-clone counterfactuals in the pool, the top hosted system reaches exec@10 = 1.00 but only exec@1 = 0.331; rank-1 misses are paired buggy variants 91.5-99.4% of the time across the four leading systems, and the canonical scores below at least one of its four paired distractors in 67-78% of queries on the leading systems. The full dataset, execution oracle, embedding matrices, environment snapshot, and pairwise statistical tests are released at the URL in Appendix D.
人机交互
[HC-0] oward Frontier-Quality Declarative UI Generation at Small-Model Cost
链接: https://arxiv.org/abs/2609.04184
作者: Yingxiang Yang,Weihang Xiao,Ben Bullough,Tushar Deshpande,Niresh Agarwal
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:Declarative UI protocols such as A2UI let applications generate interactive UIs by selecting pre-built components from a catalog and binding their props to application data, rather than emitting frontend code from scratch. This contract is attractive for production systems because of safety and consistency. An open question is: can low-latency and low-cost small models achieve the required quality for A2UI-based UI generation? To answer this, we systematically study three controllable design choices for catalog-conditioned A2UI generation: supervised fine-tuning (SFT) data construction method, model size, and component-catalog size. Across two React/TypeScript domains and four base checkpoints spanning two model families (Qwen 3.5 0.8B/2B/4B; SmolLM 3B), we find: (i) a 4B fine-tuned student recovers ~98% of teacher semantic quality and ~97% of teacher visual quality at more than an order of magnitude lower cost than frontier API calls; (ii) both augmented strategies (Perturbed-catalog and Constrained-GT) Pareto-dominate the unaugmented Full-catalog baseline, while specializing on different axes; (iii) even small models can handle and benefit from relatively large component catalog size. We distill these results into practitioner-facing trade-offs and deployment recommendations across the three design choices.
[HC-1] Decreasing Digital Distraction in College Students: Associated Online Learning Strategies Identified by Unsupervised Data Mining Approaches
链接: https://arxiv.org/abs/2609.04125
作者: Hui Shi,Ran Bi,Xi Lin,Yan Dai
类目: Computers and Society (cs.CY); Human-Computer Interaction (cs.HC)
备注: 51 pages. Accepted manuscript. Published in Journal of Educational Computing Research. Data and code: this https URL
Abstract:The proliferation of digital tools in education offers numerous benefits but also introduces significant challenges, notably digital distractions that hinder academic performance, especially in online learning contexts. This study employed unsupervised data mining techniques, specifically association rule mining and clustering analysis, to identify effective learning strategies associated with lower levels of digital distractions among college students. Data from 530 participants revealed that self-regulated learning strategies (i.e., goal setting, environment structuring, and time management) co-occurred most consistently with lower digital distractions. Additionally, learner-instructor and learner-content engagement strategies, as well as technical competencies, also tended to appear in the same profiles as lower distraction. Interestingly, reliance on peer help-seeking and learner-learner engagement strategies appeared less often in those lower distraction profiles. These findings offer actionable implications for educators to design targeted interventions that foster focused and productive online learning environments.
[HC-2] Making Gender-Inclusive Practices Actionable: Evaluating a Research-Informed Computing Education Toolkit
链接: https://arxiv.org/abs/2609.03936
作者: Alina Berry,Susan McKeever,Brenda Murphy,Sarah Jane Delany
类目: Computers and Society (cs.CY); Human-Computer Interaction (cs.HC)
备注: 9 pages, 4 figures, 3 tables, ACM International Conference on Information Technology for Social Good - GoodIT 26
Abstract:The persistent gender imbalance in computing remains a global concern, and universities offer a key part of the pipeline to address it. Although research has identified practices that support under-represented student groups, translating this evidence into actionable guidance remains challenging. This paper first presents a novel web- based toolkit (TechMate) designed to address this gap by helping computing educators implement gender- inclusive initiatives through practical research-informed guidance. The toolkit defines over 25 actions, ranging from operational to strategic, and provides case studies and implementation resources. Second, this work reports on the evaluation of TechMate, capturing educators first impressions of its usefulness and usability through authentic tasks, and eliciting unanticipated insights about structural barriers to gender-inclusive practice in computing higher education. Eighteen computing educators of varying seniority from eight universities in Ireland assessed the toolkit using nine research-driven attributes: four usefulness measures (novelty, relevance, trustworthiness, actionability) and five usability measures (effectiveness, aesthetics, navigation, terminology, user satisfaction). Data was analysed using mixed methods: deductive coding, quantitative sentiment scoring, and inductive thematic analysis. Educators rated TechMate highly for user satisfaction and praised its novelty and trustworthiness, with over 60% of participants reporting positively on all attributes, suggesting that research -informed tools can be designed for real-world use. Thematic analysis highlighted other challenges, including women students isolation, institutional resistance to change, and the scarcity of women role models, while also surfacing proposed solutions such as leveraging student ambassadors and fostering supportive learning environments.
[HC-3] Grounding GUI Design in Computational Psychology
链接: https://arxiv.org/abs/2609.03918
作者: Xianni Wang(1),Javier Romero Davila(2),Saku Sourulahti(1),Torsten Schaub(2),Jussi P. P. Jokinen(1) ((1) University of Jyväskylä, Faculty of Information Technology, Jyväskylä, Finland, (2) University of Potsdam, Institute of Computer Science, Potsdam, Germany)
类目: Human-Computer Interaction (cs.HC)
备注: 47 pages, 55 figures. Submitted to the International Journal of Human-Computer Studies
Abstract:Creating visually appealing user interfaces often requires extensive manual iteration. We propose an approach that applies answer set programming (ASP) to automatically generate and optimize UI layouts while satisfying design objectives such as grid alignment, grouping, color harmony, and whitespace, along with designer-specified preferences. Our method encodes constraints on element properties and relative positioning, producing layouts that balance functional and aesthetic goals. We evaluate this approach in three studies. Across two user studies, participants rated layouts generated with the full ASP model higher than both random designs and those based on simple heuristics. Designers reported that ASP-generated layouts supported early-stage sketching and exploration.
[HC-4] From Misconceptions to Evidence: What Science Teachers Make Visible When Co-Designing Agent ic Learning Apps
链接: https://arxiv.org/abs/2609.03917
作者: Nizam Kadir,Wei Ting Liow,Sumbul Khan,Lay Kee Ang
类目: Human-Computer Interaction (cs.HC)
备注: 10 pages, 1 figure, 2 tables. Working paper. A related abstract with the same title was accepted for a 25-minute oral presentation followed by 10 minutes of QA at the 8th Singapore International Science Teachers’ Conference (SISTC 2026), Science Centre Singapore, 24-26 November 2026. The full manuscript has not been peer reviewed or accepted for conference proceedings
Abstract:Science educators increasingly encounter AI tools that generate content, yet disciplinary teaching depends on eliciting learners’ models, diagnosing misconceptions, interpreting evidence, and preserving professional judgment. This study asks how science teachers translate such epistemic work into specifications for agentic learning applications. It contributes to the conference theme, “Innovating Pedagogies, Inspiring Minds: Transforming Science Learning,” and the Teachers’ Professional Learning strand by examining app co-design as a form of pedagogical reasoning. We conducted a bounded qualitative cross-case analysis of four de-identified artifacts produced in a teacher professional-learning workshop: an experimental-design diagnostic, a Kinetic Particle Theory dialogue guide, a chemistry prior-knowledge checker, and a physics application/scaffolding tool. Each artifact was coded for the disciplinary problem, learner interaction, evidence made visible, teacher authority, and safeguard. All four connected a science-learning problem to an interaction and pedagogically interpretable evidence: misconceptions and gaps, explanations-in-progress, class-level readiness patterns, or investigation performance. However, only two made teacher control or evaluation explicit, and only two named a safeguard. The proposals therefore positioned AI less as an answer generator than as an elicitor, scaffold, and evidence-return mechanism, while leaving decision rights and protections unevenly specified. We argue that teacher professional learning should treat AI app ideation as epistemic specification work. A five-question design protocol–problem, learner interaction, evidence, teacher authority, and safeguard–can help teachers transform science-learning needs into accountable human-AI arrangements before building or adopting a tool. Comments: 10 pages, 1 figure, 2 tables. Working paper. A related abstract with the same title was accepted for a 25-minute oral presentation followed by 10 minutes of QA at the 8th Singapore International Science Teachers’ Conference (SISTC 2026), Science Centre Singapore, 24-26 November 2026. The full manuscript has not been peer reviewed or accepted for conference proceedings Subjects: Human-Computer Interaction (cs.HC) Cite as: arXiv:2609.03917 [cs.HC] (or arXiv:2609.03917v1 [cs.HC] for this version) https://doi.org/10.48550/arXiv.2609.03917 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[HC-5] GazeFS: Target-Centered Gaze-Trajectory Forecasting and Stabilization from Gaze-Head History
链接: https://arxiv.org/abs/2609.03868
作者: Yaozheng Xia,Zaiping Zhu,Bo Pang,Minghao Xie,Hui Li,Shaorong Wang,Sheng Li
类目: Human-Computer Interaction (cs.HC); Artificial Intelligence (cs.AI)
备注:
Abstract:Target-centered gaze interaction requires more than suppressing frame-to-frame fluctuations: target acquisition produces task-aligned changes in gaze-head dynamics, while a gaze trace may retain a persistent target-relative residual direction. We formulate gaze correction as online target-centered gaze-trajectory forecasting and stabilization and introduce GazeFS, which maps a variable-length gaze-head history to the next target-center direction and a short-horizon Search/Focus estimate without target information at inference. Across 7,960 acquisition episodes from 30 participants, Search-Focus differences remain stable under quality control, onset exclusion, and duration matching. History windows improve phase decoding over the current endpoint, but explicit task progress remains a strong control. Under the 30-participant, five-fold grouped out-of-fold protocol across three seeds, the reductions relative to raw hold in Focus episode bias, within-episode dispersion, and P90 target error are 0.182 degrees, 0.257 degrees, and 0.400 degrees, with participant-bootstrap 95% confidence intervals excluding zero. Endpoint-free replay from empty history preserves the Focus advantage and yields raw-network phase balanced accuracy/AUPRC of 0.925/0.993; coordinate controls further show that recent history contributes beyond explicit progress metadata. GazeFS therefore improves Focus target centering and empirical residual contraction while leaving temporal smoothness as a separate objective.
[HC-6] Govern the Model Not Only the Data: Storag e Circulation and Learning in Creative AI
链接: https://arxiv.org/abs/2609.03800
作者: Phoenix Perry,George Simms,Elizabeth Wilson,Yasmine Boudiaf,Nick Bryan-Kinns,Tega Brain,R. Luke DuBois,Alix Rule,Rachel Meade Smith,Kelani Nichole,Atharva Pravin Pawar,Rebecca Fiebrink
类目: Artificial Intelligence (cs.AI); Human-Computer Interaction (cs.HC)
备注: 9 pages, 1 figure
Abstract:Federated learning is increasingly presented as a privacy-preserving advance: personal data remain on the device, and only model updates are shared. It borrows the vocabulary of the federated social web, yet inverts its logic, distributing computation while the resulting model stays with whoever convened the training. We argue that federation is not in itself a remedy for extractive AI, because outcomes depend on who governs the data and the model and who has agency over the practices that shape them. We describe three layers at which a creative community can hold its work: storage, circulation, and learning. Examining artist-governed trusts, cooperatives, and consent infrastructures, we show that creator governance is established at storage and circulation but stops at learning: contributors can consent to training, yet have little say over the resulting model or its federation. We map the research space this opens, pairing technical open problems with the human questions from which they unfold. We propose four design principles for a creative data commons that governs models and their federation, not only datasets: govern the model, not only the corpus; make the terms legible at the moment of contribution; design for refusal as a first-class state; and decide stewardship in the open and account for it.
[HC-7] ransfiver: Human-AI Co-Inference through a Shared Editable State
链接: https://arxiv.org/abs/2609.03797
作者: Minji Park,Seunghyun Yoon,Hyuk Lim
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Human-Computer Interaction (cs.HC)
备注:
Abstract:Long-term human-AI interaction is difficult because the information that guides inference is updated implicitly by the model and is not directly inspectable or controllable by the user. We introduce the TRANSparent Framework for Interactive, Verifiable, Editable Representation (Transfiver), an architecture for human-AI co-inference through a shared editable state. Its central idea is that interaction-specific information is maintained in a single persistent state (S_t) that both the model and the human update. Transfiver distinguishes two modes of state evolution. In an implicit stream update, the model interprets ongoing interaction and decides whether new information revises an existing state item or creates a new one. In an explicit directed edit, a human inspects and modifies an addressed item. Both act on the same underlying state, so a human correction changes the state that subsequent computation reads, rather than adding another instruction or separate record. The architecture separates shared parameters (\theta) , learned before ordinary use, from the persistent state (S_t) , which evolves during deployment without parameter retraining. Extending Transfiver to rich natural-language, relational, and large-scale shared states remains open. Subjects: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Human-Computer Interaction (cs.HC) Cite as: arXiv:2609.03797 [cs.AI] (or arXiv:2609.03797v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2609.03797 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[HC-8] Auditing Contextual Bias in Human Ball-Strike Calls Using KBOs Automated Umpiring Transition
链接: https://arxiv.org/abs/2609.03786
作者: Kichang Lee,JeongGil Ko
类目: Human-Computer Interaction (cs.HC)
备注: 12 pages, 10 figues
Abstract:This paper uses the Korean Baseball Organization’s adoption of the Automated Ball-Strike (ABS) system to audit long-standing claims about contextual bias in human ball-strike calls. Using pitch-level KBO data from 2021 through the available portion of the 2026 season, we model called-strike probability for taken pitches near the strike-zone boundary, with 2022-2023 as the primary human-umpire baseline and ABS seasons (2024 and onward) as a diagnostic benchmark. The strongest evidence concerns count pressure. Relative to 0–0 counts, human umpires called substantially fewer strikes in two-strike counts and more strikes in hitter-ahead three-ball counts. Specifically, in the main 0.25-ft boundary band, 0–2 was associated with a -17.17 percentage-point effect and 3–0 with a +6.61 percentage-point effect. Under ABS, the corresponding effects were close to zero and did not survive false-discovery-rate correction. Game progression shows a smaller but coherent pattern as human calls were less strike-prone in early innings and more strike-prone in innings 7–9+, especially in late-close situations, while complete ABS seasons were essentially flat. Other suspected biases are weaker or more localized. Salary-based reputation proxies provide suggestive but proxy-sensitive evidence, and catcher identity shows human-period residual heterogeneity that disappears under ABS. Home-context evidence is mostly null at the umpire level, with one FDR-significant human-period exception and an exploratory umpire-team gap best treated as an audit lead. Overall, the results do not show that human umpires were biased everywhere. Instead, they map where the human strike zone was most context-sensitive, where evidence was weaker, and where common suspicions received little support.
[HC-9] Exploratory Unstructured Data Analysis: A Formative Study and Implications for Human-AI Collaboration
链接: https://arxiv.org/abs/2609.03678
作者: Johannes Eschner,Dominik Eitler,Max Irendorfer,Patrick Kramml,Matthias Zeppelzauer,Manuela Waldner
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:We propose a conceptual framework for exploratory data analysis of (large) unstructured data (EluDA), combining classical elements (querying, visualization) with active knowledge construction in the “search for structure”. In a formative study, users conceptualized a structure for an image dataset during exploration. We found that users conceptualize by building faceted classifications bottom-up and rarely create meaningful spatial categorization during this process. We also evaluated CLIP for zero-shot assignment and semantic categorization, finding that it remains unreliable for assigning user-defined concepts to images but does support semantic grouping. Based on these findings, we identify and discuss four key opportunities for human-AI collaboration in EluDA: intelligent sampling and visualization to maximize data visibility; incremental and few-shot learning to minimize effort for reliable assignment; automatic category, concept, and facet suggestions to reduce effort during the search for structure; and the necessity for effective trust calibration methods.
[HC-10] PlanePivoting: Exploration and Optimization of Gaze-Mouse Cursor Alignment for Spatial Object Translation
链接: https://arxiv.org/abs/2609.03665
作者: Jinwook Kim,Sangmin Park,Jihyeon Lee,Sang Ho Yoon,Jeongmi Lee
类目: Human-Computer Interaction (cs.HC)
备注: 8 pages, 5 figures, Accepted to ISMAR’26 Workshop - GEMINI
Abstract:As XR matures into a ubiquitous computing platform, the disconnect between 2D and 3D input modalities remains a critical barrier to seamless workflow. Frequent transitions between the mouse for 2D precision and hand gestures for 3D manipulation induce significant physical fatigue and cognitive load. To address this, we introduce PlanePivoting, a multimodal interaction technique that extends standard mouse input into 3D space by leveraging gaze-mouse alignment. This technique dynamically modulates the translation plane based on the spatial overlap between the gaze and mouse cursor, eliminating the need for physical input modality switching. To systematically explore the foundational design space of gaze-mouse coordination and optimize key variables, we conducted a user study comparing PlanePivoting with a standard 3D Gizmo interface across two translation mapping profiles and two gaze cursor apertures. Results demonstrate that PlanePivoting outperforms the Gizmo on efficiency metrics while maintaining comparable precision and yielding higher subjective satisfaction. This study demonstrates the potential of gaze-mouse alignment for efficient spatial manipulation between 2D and 3D environments.
[HC-11] PointSpawn: Mid-Air Reference-Free Object Instantiation Using Gaze and Hand Gestures in Extended Reality
链接: https://arxiv.org/abs/2609.03661
作者: Jihyeon Lee,Ken Pfeuffer,Jinwook Kim,Jeongmi Lee
类目: Human-Computer Interaction (cs.HC)
备注: 11 pages, 6 figures, Accepted to ISMAR’26
Abstract:Mid-air object instantiation in XR requires users to specify a 3D position without spatial references, such as surfaces or existing objects. We present PointSpawn, a staged pipeline for pre-instantiation position specification through Direction Setting, Depth Setting, and Position Refinement within a continuous gesture flow. We evaluated six controller-free techniques combining Gaze or Non-Dominant Hand (NDH) direction setting with Ray Intersection, Relative Gain, or DragHold depth setting in a user study (N=24) across Near and Far spawn depths. Relative Gain and DragHold yielded faster and more accurate spawning, lower workload, higher usability, and greater preference than Ray Intersection. The shoulder-referenced NDH ray improved speed and coarse accuracy, whereas the viewpoint-based Gaze ray reduced hand movement with comparable final accuracy. Farther spawn depth imposed greater temporal costs as well as Gaze and accuracy costs with Ray Intersection. These findings offer empirical guidance for designing direction and depth control in spawning in XR.
[HC-12] ReRoom: Blending Virtual and Physical Contexts for In Situ Room Planning in Mixed Reality
链接: https://arxiv.org/abs/2609.03596
作者: Hongliang Yang,Yanjing Xu,Anhang Zhang,Hui Ye,Pengfei Xu
类目: Graphics (cs.GR); Human-Computer Interaction (cs.HC)
备注: 19 pages, 8 figures; supplementary materials included
Abstract:Planning a real domestic space is an in situ authoring process: users evaluate candidate layouts at true scale, refine their intent, and carry accepted decisions into later iterations. Existing approaches either separate layout editing from the physical room or provide limited support for evaluating and refining whole-room proposals in situ. We present ReRoom, a mixed-reality system for in situ room-layout authoring. ReRoom presents a shared layout state through a virtual room proxy spatially registered to the target room, allowing interaction and layout generation to remain grounded in the physical context. Users refine the current proposal through direct manipulation or language and preserve accepted placements, allowing each generated update to continue the same evolving design. To balance layout quality with generation efficiency, ReRoom uses a skill-guided layout agent whose room-layout design skill operationalizes three principles that we formulate by synthesizing established interior-design guidance for real-room layout generation. The skill grounds these principles in a normalized representation of the scanned room and reusable geometric checks. Evaluations show that ReRoom produces high-quality layouts for non-rectangular rooms, while its in situ workflow improves the room-planning experience over an otherwise equivalent off-site VR workflow. Code will be released upon acceptance of the paper.
[HC-13] Occlusion-Robust Multimodal Emotion Recognition in VR via Fusion of Facial Images and EMG
链接: https://arxiv.org/abs/2609.03569
作者: Birgit Nierula,Karam Tomotaki-Dawoud,Mert Akguel,Mustafa Tevfik Lafci,David Przewozny,Anna Hilsmann,Peter Eisert,Sebastian Bosse
类目: Computer Vision and Pattern Recognition (cs.CV); Human-Computer Interaction (cs.HC)
备注: Joint Proceedings of the ACM Intelligent User Interfaces (IUI) Workshops 2026
Abstract:Head-mounted displays (HMDs) fundamentally limit emotion recognition in virtual reality (VR): by occluding the upper face, they render conventional image-based facial expression analysis incomplete, particularly for applications requiring real-time affective assessment. We address this challenge by fusing lower-face video with facial electromyography (EMG) from the occluded upper face to classify seven emotional categories (six basic emotions plus neutral). We introduce a synchronized multimodal dataset from 20 participants, pairing lower-face video with seven-channel upper-face EMG elicited by validated emotion stimuli. Under subject-independent test, our proposed late-fusion architecture merging convolutional visual embeddings with RBF-kernel EMG representations achieves 51% macro-F1, outperforming both image-only (41%) and EMG-only (43%) baselines. These results demonstrate that upper-face EMG provides robust complementary information under HMD-induced visual occlusion and establish a foundation for multimodal emotion recognition in naturalistic VR environments. This approach facilitates affect-adaptive applications, including communication training and therapeutic interventions. The dataset will be shared upon request under an ethical-use agreement.
[HC-14] Code Black: Desktop-Mediated Co-Design of AR-HMD Microinteractions for Emergency Department Teamwork
链接: https://arxiv.org/abs/2609.03295
作者: Jonathan Segal(1),Jalynn Nicoly(2),Francisco Ortega(2),Angelique Taylor(1) ((1) Cornell University, New York, NY, USA, (2) Colorado State University, Fort Collins, CO, USA)
类目: Human-Computer Interaction (cs.HC)
备注: 12 pages, 5 figures, 2 tables. Accepted to ACM Symposium on Spatial User Interaction (SUI 2026)
Abstract:Emergency Department (ED) teams coordinate shifting roles, medication decisions, and time-critical interventions under uncertainty. Augmented reality head-mounted displays (AR-HMDs) have shown potential to spatially anchor information during care, creating opportunities to examine how spatial interfaces might support teamwork. We conducted a speculative co-design study with 12 healthcare workers (HCWs) using an editable, desktop-mediated Unity-based 3D design probe to visualize and refine work-as-imagined AR-HMD interfaces for role-based notifications, task-specific timers, and dosage verification. Guided by microinteraction rules, participants identified future spatial user interfaces (SUI) requirements such as how they appear, update, or are dismissed in relation to clinical practice, safety concerns, and existing tools. Five returning participants and 26 additional HCWs subsequently provided follow-up feedback on derived visual interface alternatives. Findings show that desktop-mediated spatial co-design elicited formative specifications for role visibility, task-linked timing, and verification-oriented dosage assistance, while revealing tensions involving clutter, shared awareness, communication, privacy, and reliability. Rather than evaluating a functional AR-HMD system or team-based clinical performance, this study contributes the Speculative Co-Design Framework for AR-HMD Teamwork (SCF-HMD) and a visual design catalog for translating expert critique of work-as-imagined (WAI) concepts into situated goals for future AR-HMD systems.
[HC-15] Reducing Catastrophic Risk from AI with Systematic Monitoring and Evaluation of Rogue AI Progression
链接: https://arxiv.org/abs/2609.03189
作者: T. Bauer,W.P. Kegelmeyer,E. Begoli,A. Sadovnik,T. Emerson,C. Corley,N. Generous,J. Moore,B. Bartoldson,R. Goldhan,M. Goldman,M. Greaves,M.J.D. Vermeer,B. MacLennan,D. Schulker,N. VanHoudnos,J. Bansemer,Y. Bengio
类目: Computers and Society (cs.CY); Artificial Intelligence (cs.AI); Human-Computer Interaction (cs.HC)
备注:
Abstract:This article presents a structured framework of behavioral indicators that may signal progression toward potentially catastrophic threats from artificial intelligence systems. We adopt a pragmatic approach, inspired by established methodologies in cybersecurity and national security. By establishing clear metrics, indicators, and thresholds across multiple dimensions of AI capability and behavior, this framework enables researchers and policymakers to implement evidence-based monitoring protocols.
[HC-16] Signal-Driven Pervasive Game Design: The LifeSync-Games Framework as a Player Experience Integration Layer
链接: https://arxiv.org/abs/2609.03169
作者: J. Macías-Cáceres,F. Gutiérrez-Vela,P. Paderewski-Rodriguez,R. González-Ibáñez
类目: Human-Computer Interaction (cs.HC)
备注: 16 pages, 2 figures, 4 tables
Abstract:Pervasive games extend the magic circle across spatial, temporal, and social dimensions, yet treat the player’s physiological and cognitive state as a passive receptor rather than an active signal. This paper presents LifeSync-Games (LSG), a framework that (unlike proposals treating player signals as an additional dimension), operationalizes them as a Player Experience Integration Layer (PEIL) acting transversally across the three existing pervasive dimensions through verified real-world signals: physical activity, sleep quality, memory, and decision speed. The framework introduces a gamified integration artifact (the LSG portal) that mediates the real - virtual exchange through redeemable points, real-world missions, and structural gamification. Five HCI design principles grounded in Self-Determination Theory and Flow Theory are proposed, instantiated across six commercial video games, together with a study protocol (n = 70-80 participants, quasi-experimental design). This paper reports the design stage of LSG: rule thresholds and portal parameters are design decisions pending empirical calibration; no data collection has yet been conducted. The main contribution is a theoretically grounded framework and validation protocol positioning player-sensitive integration as the mechanism enabling pervasive games to respond to the player’s actual biological and cognitive state.
[HC-17] Beyond Blur: A Semantic Tri-view Pipeline for Teledermatology Gradability via Skin Micro-relief
链接: https://arxiv.org/abs/2609.03095
作者: Robert Engel
类目: Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV); Human-Computer Interaction (cs.HC); Machine Learning (cs.LG)
备注:
Abstract:Smartphone skin photographs are indispensable to teledermatology, yet assessing the diagnostic suitability of submitted cases (gradability) remains a critical bottleneck in mobile care workflows. Dermatologists routinely review multiple photographic views (regional, angled, and close-up) to identify consistent textural detail rather than relying on a single image. We present the Semantic Tri-view Pipeline, an interpretable architecture for automated teledermatology gradability screening that formalizes epidermal micro-relief as a computable biomarker of image quality. Using an expert-annotated subset of the public SCIN dataset, we train a lightweight DeepLabV3+ model to segment micro-relief fidelity. These spatial masks are then aggregated across up to three case views with a logistic regression classifier, leveraging viewpoint redundancy to support robustness under uncontrolled smartphone acquisition. This approach learns context-aware, clinically intelligible heuristics, such as penalizing high-fidelity texture in regional distance views. Evaluated at a predefined 90% sensitivity operating point, the system’s apparent errors largely reflect subjective clinical variance on borderline cases where clinicians rely on non-visual metadata. On SCIN, performance improves from an AUC of 0.81 (80.6% PPV) on variance-heavy majority-consensus cases to 0.96 (97.7% PPV) on optically unambiguous unanimous cases. Overall, this work delivers an interpretable, privacy-by-design, edge-ready system that can provide real-time feedback during case submission to filter ungradable photo sets before review.
计算机视觉
[CV-0] mporal Self-Distillation: Learning Visual State Tracking in Videos Without Supervision
链接: https://arxiv.org/abs/2609.04203
作者: Shravan Venkatraman,Wenshuai Zhao,Mohammad Hassan Vali,Arno Solin
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:We introduce S ^3 T (Self-Supervised Self-Distillation over Time), which, to the best of our knowledge, is the first fully self-contained framework for continuous video state tracking. Our method treats temporal sampling density as privileged information, based on the hypothesis that a denser view of the same clip recovers the running state more accurately. This view serves as the teacher, while a sparse-view student with the same weights learns to match its next-token distribution. The model generates its own target, so training requires no labels, separate teacher, or reward signal, and adds no inference cost. On LLaVA-OneVision-2-8B, S ^3 T improves VSTAT accuracy by +1.74 as a single model, +2.38 with souping, and +2.70 with additional vision-encoder adaptation, while prior self-evolving methods leave state tracking largely unchanged. The capability learned from unlabeled synthetic clips transfers to real videos, improving performance by +7.95 on VSTAT-YouTube state-tracking questions and +4.50 on MVBench Action Count.
[CV-1] okenMatch: 3D Mesh Correspondence Transformer with Curvature-Guided Tokenisation ATC
链接: https://arxiv.org/abs/2609.04202
作者: Adeela Islam,Zorah Lähner,Vittorio Murino,Vladislav Golyanik
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 25 pages, 13 figures and 12 tables; project page: this https URL
Abstract:While data-driven 3D shape correspondence estimation has recently seen substantial progress, robust matching under partial observations and strong non-isometric deformations remains challenging. Existing learning-based approaches often rely on hand-crafted descriptors or template-based representations, whereas recent generative models over functional maps suffer from high inference cost, limited interpretability, and poor generalisation to partial shapes. In response to these limitations, this paper introduces TokenMatch, a new transformer-based unified model for estimating 3D shape correspondences. Our feed-forward approach trained exclusively on BeCoS, a challenging non-isometric partial-to-partial shape-matching dataset, can generalise to matching full shapes without retraining or fine-tuning. TokenMatch uses self- and cross-attention mechanisms to efficiently learn patch-level and point-level relations as well as dense correspondences between shape pairs. Our core insight is that meshes can be adaptively tokenised into patches using shape curvature guidance, enabling effective learning of shape-specific geometric descriptors for correspondence estimation. We evaluate TokenMatch on standard benchmarks for partial and full shape matching, including CP2P, PSMAL, BeCoS, FAUST, SCAPE, and SHREC’19. Our method achieves consistently high performance, in most cases outperforming existing methods for partial and full shape matching in the mean geodesic error and intersection-over-union metrics, while also running faster at sub-second inference speeds.
[CV-2] Scal3R: Learning Efficient Multi-Relative Pose Query for Scalable Online 3D Reconstruction ECCV2026
链接: https://arxiv.org/abs/2609.04201
作者: Chin-Yang Lin,Yang-Che Sun,Cheng Sun,Fu-En Yang,Min-Hung Chen,Yen-Yu Lin,Wei-Chen Chiu,Yu-Lun Liu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: ECCV 2026. Project page: this https URL
Abstract:Online 3D reconstruction models perform poorly on long videos. This happens because regressing poses relative to a fixed first-frame anchor forces extrapolation far beyond the training distribution. Small drifts accumulate and amplify into significant geometric collapse. However, we observe that per-frame depth remains stable throughout this failure. The backbone’s local geometry remains intact; only the global pose head breaks down. Motivated by this decoupling, we introduce Scal3R. This approach reformulates online reconstruction as multi-reference relative pose querying. We use lightweight learnable tokens, which make up about ~1% of the parameters, and inject them into a completely frozen backbone via asymmetric attention. This setup queries poses relative to multiple past keyframes. An online pose-graph optimization system with loop closure suppresses long-range drift. Scal3R reaches convergence in 8 hours on a single GPU. It reduces the average ATE by over 60% on KITTI compared to the online baseline. It also achieves state-of-the-art performance across Virtual KITTI, Sintel, TUM-Dynamic, ScanNet, and 7-Scenes. Project page: this https URL
[CV-3] Principia: Relational Physics Tests for Video Models
链接: https://arxiv.org/abs/2609.04200
作者: Varun Varma Thozhiyoor,Shivam Tripathi,Venkatesh Babu Radhakrishnan,Anand Bhattad
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project Page: this https URL
Abstract:Evaluating physical reasoning in video models is difficult because absolute motion measurements depend on frame rate, object scale, and camera calibration, all of which are often ambiguous or unavailable in generated video. We propose a different approach. When two objects in the same scene obey the same physical law, their motions must satisfy predictable relationships, and these relationships hold independent of calibration. We introduce Principia, a benchmark that evaluates Newtonian physics through relational consistency between paired objects. Principia spans eight phenomena - gravity, restitution, friction, rotational inertia, projectile motion, momentum, pendulum, and mass-spring oscillation - across translational, rotational, collisional, and oscillatory dynamics, using real-world scenes recorded under controlled protocols. We also introduce a calibration-independent consistency score that quantifies physical violation directly in image space. Across thousands of generations from six state-of-the-art video generators, no model exceeds 0.42 on Principia despite all scoring around 0.8 on VBench. Vision-language models are evaluated on their ability to detect relational physics violations, with the best model achieving only 67% accuracy and most performing near chance level.
[CV-4] Puffin-World: Scaling a Unified Multimodal Model with Native 3D World States
链接: https://arxiv.org/abs/2609.04196
作者: Kang Liao,Yihang Luo,Xiao-Ming Wu,Linyi Jin,Size Wu,Chunyu Lin,Yao Zhao,Fei Wang,Wei Li,Chen Change Loy
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project Page: this https URL
Abstract:We propose Puffin-World, a unified multimodal architecture that integrates physical understanding, spatial simulation, and 3D world generation and reconstruction without relying on external offline modules. To reliably construct and interact with 3D worlds, our framework jointly models three native world states: physics (gravity field and latitude), geometry (depth), and appearance (image), together with a unified Omni-Camera representation that supports diverse tasks and flexible motions. Beyond modeling these states, we introduce a strategy for propagating physical dynamics across future frames. By grounding absolute camera properties in the real world, Puffin-World enables physically consistent and visually stable world generation. We further couple appearance and geometry within a single generative process, jointly synthesizing each future view and reconstructing its underlying geometry. This unified paradigm enables interleaved closed-loop applications requiring synergy across multiple tasks, including mimic and self-calibrated world exploration. To scale Puffin-World to complex scenarios, we construct Puffin-16M, comprising 15 million vision-language-camera triplets and 1 million trajectories featuring various and challenging motions. To foster further research in this area, we released the code, models, and datasets.
[CV-5] One Editor Many Edits: A Unified Training-Free Framework for Diverse Video Editing
链接: https://arxiv.org/abs/2609.04190
作者: Adheesh Sunil Juvekar,Onkar Kishor Susladkar,Kiet A. Nguyen,Muntasir Wahed,Nabeel Bashir,Xiaona Zhou,Tianjiao Yu,Vedant Shah,Ismini Lourentzou
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: this https URL
Abstract:Video editing spans diverse editing paradigms, yet achieving high-quality instruction-guided and subject-guided editing within a single unified framework remains challenging. We introduce EditVid, a training-free framework combining sparse causal memory for local coherence, correspondence-based post-attention token injection for long-range identity preservation, and soft latent blending for edit locality. The same framework supports instruction-guided and reference-guided edits, including style transfer, attribute modification, object insertion, part-level editing, and subject replacement. On FiVE, EditVid achieves 78.16 FiVE-Acc, compared with 58.95 for the strongest evaluated training-free baseline, while obtaining competitive results on IVEBench. A user study further shows a 51.8% overall preference for EditVid over 7 competing methods.
[CV-6] Seeing Before Synthesizing: VLM-Guided Transition Event Discovery for Weakly-Supervised Dense Video Captioning EMNLP2026
链接: https://arxiv.org/abs/2609.04183
作者: Ye-Chan Kim,Seunghee Choi,SeungJu Cha,Si-Woo Kim,Hwiseon Kim,Hyungee Kim,Dong-Jin Kim
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Accepted to EMNLP 2026 (main, long)
Abstract:Weakly-Supervised Dense Video Captioning aims to localize and describe multiple events in untrimmed videos given only an ordered set of event-level captions per video. Recent work synthesizes auxiliary transition captions via LLM to provide additional vision-language alignment, but these captions lack visual grounding and are rigidly assigned to every inter-event gap at a fixed location and duration. To address these, we propose Seeing Before Synthesizing (SBS), a framework that adaptively provides visually grounded linguistic guidance only where warranted. Leveraging a VLM, we generate frame-level narratives for the inter-event gaps and detect transitions from the semantic variation across them. For identified transitions, we then refine inter-event temporal masks by blending the temporal midpoint with the semantic change point and selecting the width that maximizes vision-language alignment. Experiments on ActivityNet Captions and YouCook2 demonstrate state-of-the-art performance in both captioning and localization.
[CV-7] Zero-Shot Novel Depth Synthesis Using 3D Foundation Models Scene Representations ECCV
链接: https://arxiv.org/abs/2609.04174
作者: Denis M. Akola,David F. Fouhey
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to the European Conference on Computer Vision (ECCV) 2026. Project page: this https URL
Abstract:3D Foundation Models (3DFMs) such as VGGT have recently pushed the boundaries of 3D vision by predicting rich unified representations with feed-foward transformers. The scene representations learned by these models enable strong performance on multiple 3D vision tasks. In this paper, we investigate using their internal representations to infer 3D in the scene from new views. Our hypothesis is that in order to solve the task of 3D reconstruction, these models need to learn a representation that includes a large amount of general knowledge about 3D scenes. After showing that it is possible to decode hidden surfaces from internal 3DFM representations, we propose a method, Z3D, that estimates pointmaps in unseen views by doing latent diffusion on 3DFM representation. We show that Z3D can predict realistic depth maps for new views across multiple datasets.
[CV-8] Persistent Identity Preservation in Generative Image Models: A Benchmark and Evaluation System
链接: https://arxiv.org/abs/2609.04151
作者: Mengwei Ren,Xuaner Zhang,Zhihao Xia
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Generative image models can now produce high-quality images, follow complex instructions, and support precise edits, but they still struggle to preserve who or what is being depicted. When generating or editing images of a specific subject, identity may drift as the pose, expression, appearance, viewpoint, or surrounding scene changes. Existing subject-driven methods make fundamentally different choices about where identity is represented: through the input context (GPT-Image-2, NB2), as trainable subject-specific model parameters (LoRA), or as a persistent identity layer (PHOTA IDENTITY) reusable across generations and edits. We systematically benchmark these paradigms across subject-driven generation, editing, restoration, and multi-subject settings, with tasks designed to increasingly stress identity preservation. Our results show that identity preservation remains a distinct limitation of current generative foundation models: strong image quality and instruction following do not necessarily imply strong identity fidelity, and identity degradation becomes more pronounced under iterative edits, small subject scales, severe image degradation, and multi-subject composition. Persistent identity substantially reduces this degradation across generation, editing, and restoration, consistently improving identity preservation when applied to different foundation models while maintaining comparable instruction adherence and perceptual image quality. These results suggest that identity does not simply emerge from increasingly capable generative models, but can instead be represented as persistent subject knowledge that is composed independently with the underlying generative model.
[CV-9] Beyond Retrieval: Progressive Latent Memory Evolution for Streaming Video Understanding
链接: https://arxiv.org/abs/2609.04131
作者: Hongyu Qu,Guangming Yao,Ling Xing,Xiaobin Hu,Rongxing Ding,Guibin Zhang,Fan Zhang,Yi Yuan,Xiangbo Shu,Shuicheng Yan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Streaming video understanding requires multimodal large language models (MLLMs) to process continuous visual inputs and respond to user queries under strict causality and bounded memory. Existing approaches typically compress historical observations into an external memory bank and retrieve query-relevant evidence as additional visual context. Though effective, this store-and-retrieve paradigm keeps historical evidence as external visual context, preventing it from being internalized into a compact, evolving latent memory that can continuously guide streaming reasoning. To bridge this gap, we introduce LatentStream, a progressive latent working memory framework that shifts streaming memory from store-and-retrieve to retrieve-and-internalize. Specifically, LatentStream comprises three coordinated components. First, Query-agnostic Hierarchical Streaming Memory organizes visual history into short-, mid-, and long-term levels under a fixed memory budget through Jenks-guided adaptive consolidation. Once a query arrives, Hierarchical Latent Memory Evolution equips groups of latent memory tokens with progressively expanding memory receptive fields, enabling them to iteratively retrieve historical evidence from their corresponding scopes and internalize it into a compact, fixed-length latent memory. Finally, Progressive Confidence-guided Latent Memory Optimization constructs a hierarchical progression reward from group-wise predictive entropy and jointly refines the latent memory tokens and retrieved evidence, encouraging increasingly confident streaming reasoning. Extensive experiments demonstrate that LatentStream achieves new state-of-the-art results on existing online and offline video benchmarks.
[CV-10] BooM-VVT: Boosting Mask-Free Video Virtual Try-On with Image-Level Pseudo Data
链接: https://arxiv.org/abs/2609.04120
作者: Wei Zhang,Xin Li,Peishu Shi,Jialin Gao,Xuekang Peng,Zhichao Lian,Yeying Jin
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 23 pages, 18 figures, 8 tables. Accepted to ACM Multimedia 2026 (MM '26)
Abstract:Video virtual try-on (VVT) aims to generate realistic videos of a person wearing a target garment. Recent methods leverage a keyframe-driven video generation paradigm to improve in-the-wild performance, yet they still rely on masks to localize try-on regions, making them vulnerable to large motions and severe occlusions. Although mask-free image-based try-on methods have shown promising results by leveraging large-scale pseudo data, extending this paradigm to videos remains difficult, as constructing video-level pseudo data is prohibitively expensive. Furthermore, coarse keyframe sampling and the scarcity of multi-view try-on data limit existing keyframe-driven methods in maintaining garment consistency and handling diverse try-on tasks. To address these challenges, we propose BooM-VVT, a mask-free VVT framework built upon the keyframe-driven paradigm. To achieve mask-free VVT, we introduce a multi-stage training strategy that leverages image-level pseudo data for mask-free localization learning, substantially reducing the need for costly video-level pseudo data. To improve garment consistency, we propose Garment-Sensitive Keyframe Sampling, which selects keyframes based on garment-relevant body regions to better capture garment appearance. We further introduce Frame-Shared 3D-RoPE to establish spatiotemporal correspondences between keyframes and target video frames for accurate garment-detail transfer. Finally, we construct OmniView, a large-scale multi-view try-on dataset to support reliable try-on video generation under complex camera viewpoints and diverse try-on tasks. Extensive experiments demonstrate that BooM-VVT achieves superior temporal consistency and garment fidelity over existing methods. Project page: this https URL. Comments: 23 pages, 18 figures, 8 tables. Accepted to ACM Multimedia 2026 (MM '26) Subjects: Computer Vision and Pattern Recognition (cs.CV) Cite as: arXiv:2609.04120 [cs.CV] (or arXiv:2609.04120v1 [cs.CV] for this version) https://doi.org/10.48550/arXiv.2609.04120 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Journalreference: Proceedings of the 34th ACM International Conference on Multimedia (MM '26), 2026 Related DOI: https://doi.org/10.1145/3767308.3835712 Focus to learn more DOI(s) linking to related resources
[CV-11] he Shape of Time: Video-Token Contrast for Temporal Understanding in VideoLMs EMNLP2026
链接: https://arxiv.org/abs/2609.04110
作者: Yumeng Shi,Quanyu Long,Yin Wu,Wenya Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to EMNLP 2026 (main)
Abstract:Seeing frames in order does not mean representing time. Modern VideoLMs receive ordered video streams, yet their main supervision acts on generated text rather than video-token representations where event dynamics should first emerge. This mismatch allows models to learn temporal answers from shortcuts such as objects, scenes, and language priors, without requiring internal video representations to capture event progression. To address this, we propose VT-Contrast, a representation-level temporal counterfactual objective for VideoLMs. Its design asks where temporal supervision should act and what temporal differences it should expose. VT-Contrast supervises selected late-layer last-frame video tokens, where temporal information is expected to be integrated before language generation, and contrasts order-preserving views with same-video reordered counterfactuals graded by Kendall tau distance. It requires no architectural changes, is compatible with diverse VideoLM training tasks, and improves overall performance across temporal understanding benchmarks. Our code is available at this https URL.
[CV-12] Adaptive Vision-Language Grasping via Composable Foundation Priors and Generalizable Grasp Synthesis
链接: https://arxiv.org/abs/2609.04096
作者: Sixu Yan,Shikang Wang,Binhua Huang,Xuanlai Tang,Guohua Fan,Fan Huang,Haoxuan Li,Yongkang Li,Yuhan Li,Bencheng Liao,Zeyu Zhang,Wenyu Liu,Hangxin Liu,Xinggang Wang
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:This paper proposes AdaRoboVLG, a task-adaptive Vision-Language-Grasp (VLG) framework that supports generalizable grasp synthesis across different robotic hands. Unlike existing VLG methods that tightly couple foundation models with end-to-end grasp policies, AdaRoboVLG learns an efficient generalizable base policy that generates and evaluates physically feasible grasp candidates through explicit kinematic mapping and force-closure-based stability estimation, while offloading task-dependent understanding to specialized foundation-model modules. These modules provide composable priors that are integrated into the grasp synthesis process, enabling contextually adaptive grasp synthesis without retraining the underlying grasp policy. Through extensive simulation and real-world experiments, we demonstrate that (i) the base policy exhibits efficient learning and strong cross-hand generalization, (ii) the framework effectively incorporates spatial, cognitive, and temporal priors to address three representative grasping challenges without compromising grasp synthesis performance compared to state-of-the-art methods, and (iii) these priors can operate jointly to enable functional grasping in cluttered and dynamic environments. These results indicate that decoupling physical grasp synthesis from task-dependent understanding provides a scalable paradigm for robotic grasping, allowing future advances in foundation models to be directly translated into improved grasp capabilities without redesigning or retraining the underlying grasp policy. Supplementary videos are available at this https URL
[CV-13] Efficient Semantic Understanding from Digital Foveation ECCV2026
链接: https://arxiv.org/abs/2609.04088
作者: Caterina Caccavella,Vittorio Fra,Andreas Ziegler,Giulia D’Angelo,Yulia Sandamirskaya
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at the 3rd Human-inspired Computer Vision Workshop at ECCV 2026
Abstract:Dense semantic segmentation allocates computational resources uniformly across the entire image, regardless of scene complexity or task relevance. Inspired by biological vision, we investigate whether semantic understanding can be achieved more efficiently through digital foveated perception. We introduce a lightweight active-vision pipeline that combines saliency-driven fixation selection, high-resolution foveal observations, low-resolution contextual information, semantic accumulation, and adaptive computation. Beyond conventional dense prediction metrics, we use object-level evaluation to measure semantic understanding under sparse observations. On ADE20K-Object, a single foveated observation achieves 95.9% of the baseline Top-1 accuracy and 96.9% of the baseline Top-3 accuracy while requiring only 4.7% of the computational cost. At the scene level, semantic accumulation recovers 90.6% of the baseline object recall while using 58.6% of the computation. These results suggest that substantial semantic understanding can emerge from sparse observations when computation is allocated selectively, highlighting active vision as an efficient alternative to uniform dense processing and motivating evaluation protocols beyond conventional pixel-wise segmentation metrics.
[CV-14] AP-Path: Task-Adaptive Structural and Token Pruning for Efficient and Trustworthy Pathology Foundation Models
链接: https://arxiv.org/abs/2609.04071
作者: Mehedi Hasan,Ashfak Yeafi,Md Khairul Islam
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Pathology foundation models improve transferable representation learning for histopathology, but recent gains often rely on encoders with hundreds of millions of parameters and high inference cost. We propose TAP-Path, a task-adaptive compression framework that directly restructures a pretrained Virchow2 encoder rather than distilling it into a separate student. TAP-Path combines validation-driven transformer-block selection, physical removal of redundant blocks, input-adaptive patch-token pruning, multi-depth feature recovery, and a lightweight gated task head. The final model retains 24 of 32 transformer blocks and 70% of patch tokens after pruning, reducing encoder parameters by 24.96% (631.24M to 473.70M) and analytical encoder compute by 35.20% (340.13G to 220.40G FLOPs). Across three task-head optimization seeds, TAP-Path achieved 87.98 \pm 0.067% test accuracy, 81.26 \pm 0.49% balanced accuracy, and 82.38 \pm 0.48% macro-F1 on a 32-class histopathology benchmark, compared with 86.89% for full Virchow2 and 87.67% for UNI2-h. TAP-Path achieved a Brier score of 0.1800 \pm 0.0005 and failure-detection AUROC of 0.9047 \pm 0.0060 . A validation-only rare-aware objective improved rare-class balanced accuracy in a secondary operating analysis. Frozen external evaluation on 433 CPTAC samples yielded 91.22 \pm 0.83% accuracy and 91.10 \pm 0.81% balanced accuracy. These results show that task-adaptive structural and token sparsification can improve the accuracy-efficiency trade-off of large pathology foundation models while preserving reliability under internal and external evaluation.
[CV-15] Continuous Actions from Discrete Minds: Latent-Aligned Planning for End-to-End Autonomous Driving
链接: https://arxiv.org/abs/2609.04070
作者: Ruoyu Yao,Yusen Xie,Qingzhao Liu,Pei Liu,Zewei Yang,Yipeng Zhu,Xiaolong Wang,Jun Ma
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注: 8 pages, 5 figures
Abstract:Bridging the gap between the discrete reasoning of Vision-Language Models and the continuous, physics-constrained nature of autonomous driving remains a significant challenge. In this work, we introduce LaPla, a unified Vision-Language-Action (VLA) framework featuring latent-aligned planning to seamlessly ground semantic understanding in precise motion execution. We first design an action tokenizer based on a residual vector-quantized variational autoencoder (VQ-VAE), capturing vehicle kinematics and encoding trajectory features into a structured latent space. Rather than discrete codebook lookups that inevitably introduce quantization errors, LaPla repurposes this representation as a physical prior to bridge the modality gap between high-dimensional semantics and the raw action space. Specifically, given multimodal inputs integrating multi-view images, historical actions, and textual instructions, LaPla incorporates concurrent action queries to causally attend to the multimodal context in a single forward pass, projecting hidden states directly into the pretrained VQ-VAE latent space. The frozen decoder then translates these continuous latents into actions, effectively eliminating quantization errors and ensuring physically plausible trajectories while bypassing time-consuming autoregressive generation. Extensive experiments on the nuScenes benchmark demonstrate that LaPla achieves competitive open-loop performance, reducing long-horizon L2 error by 15.52% compared to state-of-the-art VLA methods. Closed-loop evaluations on the NVIDIA AlpaSim simulator further confirm its superior capability in ensuring smooth driving progress, improving the success rate by 33.34 percentage points with significantly reduced inference latency.
[CV-16] DSAQuant: Denoising-Stage-Aligned Quantization-Aware Training for Video Generation
链接: https://arxiv.org/abs/2609.04031
作者: Shuaiting Li,Zelin Gao,Haibin Shen,Yujun Shen,Haotong Qin,Yinghao Xu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project page: \url{ this https URL }; Code: \url{ this https URL }
Abstract:Video diffusion models (VDMs) have achieved impressive progress in text-to-video generation, but their high memory and computational costs hinder practical deployment. Quantization-aware training (QAT) is an effective solution for compressing and accelerating advanced generative models without runtime overhead at inference. However, existing QAT methods suffer from a distinctive challenge in VDMs: while they often preserve prompt semantics, global layout, and coarse motion, the quantized model severely degrades visual details, texture fidelity, and sharpness. In this paper, we trace this degradation to the timestep-agnostic design of conventional quantization pipelines, which overlooks the stage-wise functionality of video denoising. In VDMs, early denoising steps mainly establish global structure and motion, whereas middle and late steps refine local appearance and high-frequency details. Based on this insight, we propose DSAQuant, a Denoising-Stage-Aligned Quantization-aware training framework for VDMs. During training, Denoising-Stage Oriented Supervision preserves teacher distillation in early steps for stable structure planning, while shifting later steps toward target-driven optimization to enhance detail reconstruction. During inference, Denoising-Stage Gated Guidance disables CFG in the final denoising steps to prevent it from amplifying quantization-induced errors into high-frequency artifacts. Extensive experiments on the Wan and CogVideoX families under W4A4 and W3A3 settings show that DSAQuant consistently outperforms the SOTA QAT baseline, improving the VBench average score by up to 6.60 under aggressive W3A3 quantization while preserving strong text-video alignment. These results demonstrate that effective VDM quantization requires not only reducing quantization error, but also aligning quantization training and inference with the stage-wise nature of video diffusion.
[CV-17] Stable and Scalable Bundle Adjustment of Holistic 3D Structures ECCV2026
链接: https://arxiv.org/abs/2609.04026
作者: Shaohui Liu,Rémi Pautrat,Daniel Barath,Richard Hartley,Viktor Larsson,Marc Pollefeys
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: To appear at ECCV 2026. Code available as part of the LIMAP toolbox at this https URL
Abstract:Bundle Adjustment (BA) is a cornerstone of 3D computer vision and has benefited from decades of advances in sparse optimization and numerical methods. It was originally developed for jointly optimizing camera intrinsics, poses and sparse 3D points. While extensions incorporate lines and other primitives, integrating richer geometric structures such as parallelism, coplanarity, or wireframes often introduces significantly increased computational cost and reduced numerical stability. In this paper, we propose a unified framework that extends bundle adjustment to jointly optimize geometric features and higher-order relations. We first introduce a taxonomy that distinguishes scalable geometric features with direct 2D measurements (e.g., points and lines), from groups encoding higher-order relations (e.g., coplanarity, parallelism, etc.), where we show that groups can be modeled as camera-like entities within the bundle adjustment framework. Building on this formulation, we propose that both group constraints and cross-feature relations (i.e., point-line associations) can be expressed through 2D reprojection measurements. By formulating group-induced and cross-feature reprojection errors, we preserve the sparsity structure of classical point-based BA under Schur elimination, while avoiding direct 3D regularization that degrades the conditioning and stability. Experiments on both real-world and synthetic datasets demonstrate runtime performance comparable to classical point-only bundle adjustment, while producing significantly richer 3D structures and improved geometric accuracy.
[CV-18] he Blind Spot in 2D Infants Pose Estimation:Robust Learning from Noisy Annotations
链接: https://arxiv.org/abs/2609.04009
作者: Emanuele Cardinale,Marco Proietti,Alessandro Cacciatore,Maria Francesca Spadea,Lucia Migliorelli,Sara Moccia
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Noisy annotations pose a significant challenge for supervised deep learning, as neural networks rely on large-scale, high-quality labeled data whose corruption can severely impair model performance. Although robustness to label noise has been extensively studied for classification tasks, it remains relatively underexplored in Pose Estimation (PE). This limitation becomes critical in clinical contexts, including neonatology, where PE of preterm infants is used to support the assessment of spontaneous motility, a key indicator of neurodevelopmental trajectories. In such settings, infants’ images labeling is further hindered by visual challenges (e.g., keypoint self-occlusions, caregiver interference), making the annotation process inherently susceptible to errors. To tackle noisy annotations in PE, we introduce REliable keypoint selection via Memory of traINing Dynamics (REMIND), a clustering-based keypoint-selection strategy that exploits keypoint-wise training dynamics to identify noisy labels without assuming any prior knowledge of the noise distribution, thus enabling noise-free model training. When evaluated on the proprietary NeoPose dataset, comprising 46 videos of 46 preterm infants recorded in real clinical settings, REMIND correctly identifies noisy annotations across multiple corruption scenarios, achieving up to 93% Area Under the Curve (AUC) with three different PE architectures used in the relevant literature. To our knowledge, this is the first study to explicitly address label noise in preterm infants’ PE, paving the way for the design of trustworthy learning-based algorithms for infants’monitoring support when data quality cannot be guaranteed.
[CV-19] Catalogue Photography as a Cold Start: Toward Deployable Carbide Burr Recognition
链接: https://arxiv.org/abs/2609.03995
作者: Abilash Philip Madavath,Chandra Yuvesh Aubeeluck,Augustin Raju,Nicolas Pyschny,Felix Hackelöer,Florian Zwanzig
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Robotics (cs.RO)
备注: Extended abstract not yet published to a conference or journal
Abstract:Verifying that manufactured batches of milling tools or carbide rotary burrs conform to production order sheets remains a largely manual and error-prone quality assurance task. Automating this process with computer vision faces a critical cold-start constraint since no labelled imagery is available, leaving manufacturer catalogue photography as the sole source of supervision. We investigate how far catalogue supervision can support an industrial recognition pipeline under domain shift, explicitly measuring the gap between catalogue separability and performance on held-out field photographs. Our findings reveal three key insights. First, off-the-shelf frozen feature extractors do not reliably separate the two task attributes, head shape and tooth profile, motivating targeted representation learning. Second, metric learning produces near-perfect unsupervised cluster discovery on catalogue images (adjusted Rand index 0.94–0.97), but less than half of this gain transfers to field photographs. Third, the largest transfer gains do not come from model scale or representation complexity, but from simple changes that reduce domain sensitivity: converting images to grayscale (+0.22) and constraining retrieval using the known order sheet via Hungarian assignment (+0.11). We therefore treat catalogue photography as a useful cold start rather than a deployment-ready training domain, and provide empirical baselines and an evaluation protocol for catalogue-to-field transfer in precision tool manufacturing.
[CV-20] Sharpening the Ensemble: An SSIM-Aligned Residual Refiner for Brain-MRI Inpainting Post-Processing MICCAI2026 MICCAI
链接: https://arxiv.org/abs/2609.03981
作者: Kubilay Kağan Kömürcü,İlkay Öksüz
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: Accepted at the MICCAI BraTS Local Synthesis of Brain Tissue Inpainting Challenge (Task 4), MICCAI 2026. 12 pages, 2 figures
Abstract:Brain-MRI inpainting replaces a masked region of a scan with synthesized, anatomically plausible healthy tissue, so that analysis tools built for healthy brains can be applied to images they would otherwise reject. On the BraTS local-synthesis benchmark, which ranks submissions on the structural similarity index (SSIM), the peak signal-to-noise ratio, and the mean squared error (MSE) jointly, the strongest recent models are accurate, but several report blurry synthesized regions and attribute this to the mean-seeking behavior of the \ell_1 and MSE terms in their training losses. We address this in post-processing, forming a deep ensemble of the two co-first-place 2025 models and training a lightweight residual refiner on the ensemble’s own outputs under an \ell_1 loss augmented with a structural-similarity term whose weight \lambda we vary. At a moderate \lambda the refiner improves SSIM over the ensemble, from 0.8767 to 0.8780 on a held-out reproduction of the official scorer and from 0.8555 to 0.8572 on the official validation leaderboard, with essentially no change in MSE. The gain is small but consistent, improving 62.6% of the held-out cases with a signed-rank p=2.2\times10^-7 , whereas over-weighting the structural term reverses it. Two ablations bound the effect. Adding any third model to the two-model ensemble degrades it, and classical unsharp masking fails to improve SSIM at any strength (best 0.8765 against 0.8767 ), so the gain reflects learned rather than indiscriminate sharpening. The result is a cheap, reproducible post-processing stage that improves an already strong ensemble without any large-scale retraining.
[CV-21] RARF: Region-Aware Rectified Flows for 3D Brain MRI Inpainting ICIP MICCAI
链接: https://arxiv.org/abs/2609.03956
作者: Tomas Guija-Valiente,Blanca Rodriguez-Gonzalez,Norberto Malpica,Angel Torrado-Carvajal
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 11 pages, 2 figures. Preprint version corresponding to the initial submission prior to peer review, submitted as part of our participation in the BraTS 2026 Challenge. The final accepted version will be openly available in the official MICCAI proceedings on the conference website
Abstract:Medical image inpainting has the potential to improve automated brain MRI analysis by reconstructing healthy tissue within pathological regions. We introduce RARF, a task-agnostic region-aware rectified flow framework for masked data generation. We instantiate the framework for 3D brain MRI inpainting as our submission to the BraTS Inpainting Challenge 2026. RARF restricts the stochastic interpolation process to the inpainting region, while the observed voxels remain fixed and provide patient-specific anatomical context. A three-dimensional neural network receives the partially voided image, with Gaussian noise filling the missing region, together with the inpainting mask and the corresponding timestep. The model is trained using masked flow-matching and reconstruction-consistency objectives, combined with mask-aware preprocessing and data augmentation. During inference, the learned velocity field transports the initial noise toward a plausible reconstruction of the missing tissue, which is then combined with the unchanged observed anatomy. Experiments under the BraTS evaluation protocol show that the proposed approach produces competitive reconstructions while maintaining anatomical consistency. Source code is available at: this https URL.
[CV-22] WorldReward: Reward Modeling for Camera-Conditioned World Models
链接: https://arxiv.org/abs/2609.03952
作者: Yibin Wang,Zehan Wang,Junshu Tang,Zhimin Li,Yujie Zhou,Jiazi Bu,Pengyang Ling,Feng Han,Zhixiong Zhang,Long Xing,Shengyuan Ding,Ziang Li,Cheng Jin,Yuhang Zang,Jiaqi Wang,Tianyu Pang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Website: this https URL
Abstract:Camera-conditioned world models generate interactive videos in which commanded actions should induce the expected scene changes while appearance, geometry, and temporal dynamics remain coherent. Existing rewards assess these requirements separately: geometry-based rewards estimate trajectory execution but cannot judge the visual quality of the executed motion, whereas image-based rewards measure frame quality without capturing action execution or temporal dynamics. We posit that a vision-language model (VLM) offers a shared reasoning space for relating actions to their visual outcomes. However, judging a complete long video against its full action sequence creates a lengthy, noisy context in which short-lived local action evidence can be missed or diluted. We present WorldReward, a VLM-based pairwise preference reward model that unifies action-consistency and visual-quality evaluation for camera-conditioned world models. WorldReward decomposes paired videos into action-aligned chunks, organizes each chunk into structured visual evidence, and aggregates chunk-level decisions by voting into separate video-level action and visual-quality preferences. To train it, we construct a large-scale reasoning-augmented preference dataset using structured judgments generated by a frontier VLM and refined through tool-based agent auditing and targeted human review. We further introduce WorldReward-Bench, a human-annotated benchmark measuring reward-model agreement with human preferences across action consistency, appearance quality, and motion quality. WorldReward achieves the highest agreement on all three dimensions, exceeding GPT-5.5 by 3.42, 1.45, and 3.56 percentage points, respectively. When used for RL post-training of HY-WorldPlay 1.5, it consistently improves both action execution and visual quality across short- to long-term horizons.
[CV-23] Sparse auto-regressive modeling for scene generation from multi-view images ECCV
链接: https://arxiv.org/abs/2609.03931
作者: Thomas Lucas,Maxime Pietrantoni,Philippe Weinzaepfel,Wonjune Cho,Bardienus Pieter Duisterhof,Vincent Leroy,Jerome Revaud
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: Accepted at ECCVV 2026
Abstract:Generating complete 3D scenes from sparse, unconstrained views is a fundamental challenge in 3D vision which requires reasoning beyond observed content while remaining computationally tractable. Existing feed-forward reconstruction methods are inherently limited to content visible in the input images, while 3D generative modeling is hindered by the high computational cost of dense volumetric representations and the scarcity of large-scale 3D supervision. We introduce SPAR3S, a sparse voxel-aligned 3D latent generative model for conditional scene completion without requiring ground-truth 3D data for supervision. Our key insight is to formulate 3D scene generation in a structured, compact, voxel-aligned 3D latent space where only occupied voxels are represented. We learn this sparse latent space directly from multi-view images using photometric supervision via differentiable 3D Gaussian Splatting. Given a partial set of observed voxels encoded from sparse input views, scene completion reduces to predicting the missing latent tokens and their spatial support within the voxel grid. To this end, we train a masked autoregressive transformer that jointly models voxel occupancy and latent token values, enabling efficient and spatially consistent generation of unseen regions. We demonstrate the effectiveness of our method on synthetic indoor scenes, achieving higher novel-view quality than prior work. We further validate its generalization on RealEstate10k, highlighting its applicability to real-world data.
[CV-24] OctWorld: Long-Range World-Consistent Video Generation with Octree-Based 3D Mapping ECCV2026
链接: https://arxiv.org/abs/2609.03919
作者: Zelong Lv,Sicheng Xu,Jianfeng Xiang,Ruicheng Wang,Yue Dong,Yu Deng,Guangzhong Sun,Jiaolong Yang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to ECCV 2026 Project page: this https URL
Abstract:We present OctWorld, a video diffusion framework with persistent 3D memory for generating explorable, world-consistent, and high-fidelity visual scenes. Given a single image, OctWorld performs stable autoregressive world generation along user-specified camera trajectories. We focus on long-range generation, characterized by extended camera paths and wide viewpoint coverage, where preserving spatial consistency is particularly challenging when previously generated regions are revisited. To address this problem, we introduce OctMap, an extensible and spatially adaptive 3D memory that progressively fuses generated visual observations and their corresponding depth maps into a global representation. OctMap employs TSDF fusion within a dynamic sparse octree whose spatial resolution adapts to image evidence. This design preserves geometric and appearance details across diverse scene scales while maintaining low memory overhead. Experiments demonstrate that OctWorld generates long-range, spatially consistent videos and outperforms prior methods on both existing benchmarks and challenging long-range generation settings. OctMap also provides clear advantages over point-based caches and fixed-resolution TSDF volumes. Project page: this https URL
[CV-25] Concept of a Sensor Test Environment for Dusty Agricultural Conditions
链接: https://arxiv.org/abs/2609.03895
作者: Peter Buckel,Johannes Hermann,Jonas Wollmann,Thomas Dietmueller,Timo Oksanen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Dust in agriculture presents a significant challenge for autonomous agricultural machinery. Dust can impair the performance of sensors and algorithms. This work, therefore, presents a concept for a proving ground consisting of an indoor and outdoor area. The indoor area comprises a laboratory test bench where dust circulates in a closed system and a test hall where life-size objects can be placed. The outdoor area features dedicated test setups that enable reproducible data to be recorded with and without dust during real-world agriculture work. The proving ground and the setups are visualized in 3D.
[CV-26] GraFT: A Training-Free Framework for Spatial Reasoning in Multimodal Large Language Models via 3D Scene Graphs
链接: https://arxiv.org/abs/2609.03892
作者: Junqing Du,Fernando Ropero,Erkin Turkoz,Yanfeng Zhang,Lu Liu
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Robotics (cs.RO)
备注:
Abstract:3D spatial reasoning underpins understanding and acting in the physical world, yet it remains unreliable in current multimodal large language models (MLLMs). These models falter at precise geometric measurement, at transforming between egocentric and allocentric viewpoints, and at grounding fine-grained appearance. The most common remedies fine-tune the model on large-scale curated spatial-reasoning datasets or attach dedicated encoders for 3D geometry, which typically couples the solution to costly supervision and a specific backbone. We instead introduce GraFT, a training-free framework that supplies the missing 3D structure through a compact, easily maintained 3D scene graph (3DSG). From this 3DSG, GraFT provides three spatial reasoning capabilities: (1) deterministic geometry through symbolic tools, (2) allocentric layout through a bird’s-eye-view (BEV) rendering, and (3) visual-attribute grounding through task-relevant egocentric frames. On ScanQA, GraFT improves every metric over the same-backbone baseline, raising CIDEr by 27%. On VSI-Bench, GraFT improves frozen MLLMs by up to 65%, surpassing every proprietary and general-purpose open-source baseline, and several prominent fine-tuned spatial models.
[CV-27] he impact of phase information for few-shot fine-grained image classification
链接: https://arxiv.org/abs/2609.03829
作者: Ruiling Liu,Linyue Zhang,Wenyi Zeng,Jiamiao Lu,Weichuang Zhang,Changming Sun,Zejun Zhang,Xiao Zhao
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Few-shot fine-grained image classification (FSFGIC) aims to classify similar images with limited labeled examples. This work highlights the critical yet underutilized role of phase information in capturing structural relationships within an image. This study introduces a novel plug-and-play amplitude-phase integration (API) module that effectively combines local and global frequency amplitude and phase information for obtaining more comprehensive feature descriptors. Additionally, a dedicated network, named PSF-Net, is proposed that adaptively fuses phase-based spatial and frequency information for FSFGIS. The designed PSF-Net can be easily integrated into standard episodic training architectures for end-to-end training from scratch. Extensive experiments on five public datasets demonstrate that the method outperforms existing state-of-the-art benchmarks.
[CV-28] VI3: Grounding Pretrained 3D Foundation Models with Inertial Cues
链接: https://arxiv.org/abs/2609.03824
作者: Ernesto Lozano,Alberto Jaenal,Javier Civera
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:3D foundation models (3DFMs) excel at predicting camera poses and dense depth from multiple views of a scene, showcasing strong zero-shot generalization. However, as metric scale is not observable from monocular images, their absolute scale predictions are typically inaccurate. Inertial measurement units (IMUs), present in most devices, naturally complement monocular cameras by observing scaled motion. We introduce VI3, a model-agnostic framework that metrically anchors a pretrained 3DFM using only IMU readings. VI3 initializes and preintegrates the IMU to obtain a metric motion reference, which is then used to recover the scale of the 3DFM outputs. Our method includes adaptable anchoring strategies tailored to diverse 3DFM architectures. Experiments on synthetic and real aerial datasets demonstrate that VI3 recovers metric scale without ground-truth supervision while preserving geometric consistency, acting as a fine refinement under well-conditioned motion and as a strong prior when motion is less informative.
[CV-29] When Vision Meets Graphs: A Survey on Graph Reasoning and Learning IJCAI
链接: https://arxiv.org/abs/2609.03816
作者: Xinjian Zhao,Wei Pang,Zhixuan Yu,Xiangru Jian,Xiaozhuang Song,Yaoyao Xu,Zhongkai Xue,Dingshuo Chen,Shu Wu,Philip Torr,Tianshu Yu
类目: ocial and Information Networks (cs.SI); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: IJCAI Survey Track, 2026
Abstract:Graphs are a fundamental data structure underlying many problems in the natural and social sciences. Over the past decade, Graph Neural Networks (GNNs) have dominated graph machine learning, supported by solid theoretical foundations. Yet scientists often understand graph structure through vision: chemists read molecular diagrams and social scientists inspect network visualizations. Despite decades of work on graph visualization, most graph learning pipelines still treat graphs purely as symbolic structures, rarely leveraging the visual form of graphs. We argue that this gap deserves renewed attention in the era of powerful vision and vision-language models. This survey provides a first systematic overview of the emerging area we term vision meets graphs, which treats visual depictions of graphs as first-class inputs for reasoning and learning. We organize existing work into three threads. Vision for Graph Reasoning studies how models can use visual depictions of graphs to understand structure and carry out multi-step reasoning. Vision for Graph Learning explores how visual features can complement or augment graph encoders beyond known limitations of message passing. Scientific Graphs examines domains where standardized depiction conventions support both reasoning and learning. Our goal is to clarify what current methods can and cannot do, and to outline a path toward foundation models that perceive and reason about graphs as scientists do.
[CV-30] SPARK: Input-Conditioned Sparse Activation Modulation for Frozen DiT-based Super-Resolution
链接: https://arxiv.org/abs/2609.03813
作者: Federico Putamorsi,Leonardo Zini,Marcella Cornia,Lorenzo Baraldi
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Real-world image super-resolution (SR) increasingly relies on Diffusion Transformer (DiT) backbones, whose internal activations can be dominated by a small number of massive channels. Yet improving perceptual quality in these models still typically requires fine-tuning the network or attaching additional adapters, leaving this structured activation space largely unexplored for adaptation. We investigate whether dominant channels can instead serve as a compact adaptation interface for frozen DiT-based SR models. We first characterize their behavior in pretrained SR backbones and show through controlled interventions that they strongly affect reconstruction quality. Building on this observation, we introduce SPARK, a lightweight input-conditioned controller that predicts bounded per-channel affine transformations for only the selected channels, while keeping the SR backbone and VAE frozen. Dominant channels are identified through an online activation-ranking procedure, and only a small predictor conditioned on the low-resolution VAE latent is optimized. Experiments on three DiT-based SR backbones across DIV2K, RealSR, and DRealSR show consistent gains in both fidelity and perceptual quality while modulating only eight channels per stream and block. Controlled comparisons further show that these gains cannot be explained by parameter budget or access to the selected channels alone.
[CV-31] SVG-Score: Human-Aligned Evaluation of Text-to-SVG Generation
链接: https://arxiv.org/abs/2609.03806
作者: Marco Cipriano,Leonardo Zini,Alexandra Schild,Valentin Teutschbein,Afsana Mimi,Marcella Cornia,Lorenzo Baraldi,Gerard de Melo
类目: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Scalable Vector Graphics (SVG) generation is attracting increasing attention as generative models improve in expressiveness and controllability. Progress, however, is held back by the lack of domain-specific evaluation protocols: current practice relies on metrics designed for natural images, most notably CLIPScore, which was never trained on vector graphics and aligns only partially with human judgment. We introduce \textbf\ours, a human-aligned evaluation framework for text-to-SVG generation. Through controlled caption and image perturbations, we first show that CLIP-based scores barely react to the errors SVG generators actually make, such as wrong colors, counts, and spatial relations, and that off-the-shelf Vision-Language Model (VLM) judges, while more sensitive, respond unevenly across error types and SVG styles. We then introduce a human-annotated dataset for \textitSemantic Alignment, measuring how faithfully a generated SVG reflects its caption. Building on it, we develop two complementary evaluators: CLIP scorers adapted to vector graphics and then aligned to human preferences, for fast large-scale evaluation, and a VLM judge trained with supervised fine-tuning and reward-shaped reinforcement learning, for more expressive and interpretable assessment. Using both, we benchmark major open-source, commercial, and optimization-based SVG generators on an independent caption set.
[CV-32] Urban Boundaries Social Barriers: A Benchmark and Vision-Centric Framework for Mapping Gated Communities and Equity Implications ECCV2026
链接: https://arxiv.org/abs/2609.03804
作者: Minwei Zhao,Weiming Zhang,Jiawang Du,Qiming Liu,Weiming Zhuang,Pei Nie,Cai Wu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: ECCV 2026 camera-ready version
Abstract:Communities are fundamental spatial units that shape urban form and social life. Whether a residential compound is spatially open or enclosed affects mobility, access to public services, and equity, yet studies of Chinese fengbi xiaoqu remain largely qualitative or small-scale, limiting reproducible city-scale analysis. We address this gap by introducing GBA-GCs, a metropolitan-scale multimodal benchmark for locally grounded gated/open community recognition in China’s Greater Bay Area, covering 37,444 residential compounds with aligned boundary polygons, high-resolution satellite imagery, Chinese metadata, and structured attributes, together with expert-verified labels, inter-annotator reliability, and official evaluation splits. Built on this benchmark, we present Multimodal Classifier for Gated Community (MCGC), a vision-centric multimodal framework based on DINOv3-SAT that fuses imagery, text, and structured cues via modality-aware cross-attention and adaptive gating to mitigate modality imbalance. MCGC consistently outperforms strong unimodal and multimodal baselines. Finally, we apply the validated model to metropolitan-scale mapping and report equity-oriented findings including spatial clustering of GCs, privatized green space, and reduced pedestrian connectivity. The benchmark, code, and release documentation are available at this https URL.
[CV-33] LLaDA-Image: Building Strong Image Generators with Fully Open Training Recipes
链接: https://arxiv.org/abs/2609.03796
作者: Chuyan Chen,Haoxing Chen,Kun Chen,Zhenglin Cheng,Long Cui,Ruishan Fang,Zhangxuan Gu,Zhicheng Huang,Zhenzhong Lan,Yuanting Lei,Haoquan Li,Jianguo Li,Rongchuan Li,Sidu Li,Tao Lin,Deyuan Liu,Jiacheng Liu,Lin Liu,Yuxuan Lou,Zhisheng Lu,Yuxin Ma,Shuheng Shen,Peng Sun,Chaoyang Wang,Hongjun Wang,Xiaomei Wang,Yongxin Wang,Chengzhang Wu,Hongru Wu,Jun Xie
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:We introduce LLaDA-Image, a unified framework that pairs a 6B Diffusion Transformer (DiT) trained from scratch with a frozen vision-language understanding module built on the LLaDA2.0-Mini diffusion language model backbone. Instead of relying heavily on paired image-text data from the beginning, we first build a strong visual generative prior through image-only pre-training and mid-training. The generation pipeline comprises 220M samples, 98 of which are real images. For efficient and scalable optimization, we use parameter-free RMSNorm throughout the DiT together with the Muon optimizer. The resulting unified model produces highly photorealistic images while accurately following fine-grained editing instructions. We further distill LLaDA-Image into LLaDA-Image-Turbo, enabling fast inference in 2-4 sampling steps. On Qwen-Image-Bench, LLaDA-Image achieves overall scores of 53.53 and 53.38 on the English and Chinese tracks, respectively, setting a new state-of-the-art among open-source models on both tracks. To support further research on capable and efficient generative models, we release our model weights, training code, and detailed recipes.
[CV-34] ENEAS: Embedding-guided Neural Ensemble for Adaptive Segmentation
链接: https://arxiv.org/abs/2609.03756
作者: Javier del Pino(SperidLabs),Salvador Rodríguez(SperidLabs),Alejandro Garabito(SperidLabs),Javier Álvarez(SperidLabs),Chema Garabito(SperidLabs)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 19 pages, 5 figures, 6 tables. Code and models: this https URL
Abstract:We present ENEAS, a unified, text-promptable method for instance tracking and semantic discovery. Text-promptable segmentation models, including the latest foundation models such as SAM 3, still suffer from temporal hallucinations, spatial fragmentation, and semantic misclassification: they fail to report target absence when an object leaves the field of view, segment local textures instead of the complete object during extreme close-ups, and prioritize visual features over ontological reality, so that visually similar artifacts such as statues, paintings, or reflections are segmented as target entities. ENEAS works two ways from a single method: precise tracking and high-quality segmentation of a unique instance, and open-concept discovery of every instance a text query names, resolved by a semantic verification layer. For tracking, we extend the geometrically robust SeC architecture, previously limited to point interactions, with a text-prompting adapter and leverage its temporal memory, so that the target is held through disappearance without drifting to distractors and kept whole even when it fills the entire view. For discovery, the verification layer combines high-speed visual embedding matching with conditional VLM refinement, invoking semantic reasoning only for ambiguous candidates, which filters out the ontological errors that visual-only models cannot distinguish while keeping latency low. Designed with 3D reconstruction in mind, where a single misclassified distractor corrupts the asset, ENEAS unlocks high-quality semantic tracking and segmentation of video, of broad libraries, and of collections of temporally or spatially unordered data, together with the discrimination to tell true instances from their doppelgangers: things that look alike but are not the same. The code and models are available at this https URL Comments: 19 pages, 5 figures, 6 tables. Code and models: this https URL Subjects: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI) ACMclasses: I.4.6; I.2.10; I.4.8 Cite as: arXiv:2609.03756 [cs.CV] (or arXiv:2609.03756v1 [cs.CV] for this version) https://doi.org/10.48550/arXiv.2609.03756 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[CV-35] Fill My Mirror: Geometry-Constrained Mirror Inpainting
链接: https://arxiv.org/abs/2609.03740
作者: Ofek Basson,Shimon Vainer,Yacov Hel-Or,Ohad Fried
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Mirrors are common in real-world images, yet producing geometrically consistent reflections with generative models remains challenging. Unlike most objects, mirror appearance depends on scene geometry and viewpoint, making it hard to synthesize using learned appearance priors alone. We address this in the mirror inpainting setting, where the scene is fixed and only the mirror region is generated. Our key insight is that much mirror content is geometrically constrained by the visible scene and need not be hallucinated. We estimate scene geometry and project visible content into the mirror to recover reflection regions determined by geometry. A generative model then completes the mirror region via a two-mask diffusion strategy balancing geometric constraints with the model’s learned priors, reducing projection artifacts and improving reflection consistency. The method is training-free and applicable to complex real-world scenes. We evaluate on MirrorBench-V2 (synthetic) and real images. Using standard and geometry-aware metrics, we show that explicitly using scene geometry improves consistency.
[CV-36] Unfold The World: Factorize 4D Properties in Reinforcing Spatial Reasoning ECCV2026
链接: https://arxiv.org/abs/2609.03729
作者: Yijun Yang,Shenghe Zheng,Wenbo Li,Jianhui Liu,Haoze Sun,Yanbing Zhang,Jiaxiu Jiang,Lin Song,Haoyang Huang,Nan Duan,Lei Zhu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted by ECCV 2026
Abstract:Despite the remarkable prowess of Vision-Language Models (VLMs) in general multimodal tasks, they remain fundamentally flat'' when reasoning about the physical world. We argue that this spatial bottleneck stems from a profound dimensional mismatch: while VLMs are trained to interpret 2D projections, true spatial reasoning demands the recovery of latent 3D geometry and temporal continuity. To conquer this high-dimensional complexity, we advocate a shift from monolithic learning to a divide and conquer’’ paradigm. We present FactoSR, a factorized reinforcement learning framework that explicitly interpret the dimensions collapsed by visual projection. At its core, FactoSR decomposes the monolithic problem of world-consistent reasoning into three orthogonal, geometric sub-objectives: planar correspondence ( XY ), depth consistency ( Z ), and temporal reversibility ( T ). By optimizing these verifiable constraints within a unified policy learning mechanism, we effectively transform an ill-posed projection recovery problem into a series of tangible reasoning steps. Extensive evaluations on multi-view and video benchmarks demonstrate that this elegant decomposition yields substantial gains in 3D and 4D reasoning, achieving a 5.9% boost on VSI-Bench and 4.5% on All-Angles-Bench. Our findings suggest that reinforcing explicit, factorized 4D consistency is a critical step toward evolving VLMs into robust, world-aware reasoners.
[CV-37] SignSeek: Learning Transferable Representations for Sign Dictionary Retrieval
链接: https://arxiv.org/abs/2609.03695
作者: Sobhan Asasi,Ozge Mercanoglu Sincan,Richard Bowden
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Sign language dictionaries are essential resources for sign language learners, yet automatically retrieving a sign from a dictionary, given only a query video, remains a challenging problem due to the natural variability between signers. Existing sign representation learning methods are built for closed-set recognition, producing embeddings that do not generalise to the open-set, signer-independent setting that retrieval demands. \textbfSignSeek closes this gap by contrastively learning sign representations with saliency-guided articulator masking. A contrastive objective aligns same-gloss signs across signers, while our Articulator Saliency-Guided Masking (ASGM) pinpoints the single most critical articulator per sign. This drives two complementary objectives, a masked contrastive alignment (MAC) loss that sees the sign through a single articulator and a masked prediction (MAP) loss that reconstructs it in latent space from the surrounding spatio-temporal context. Pretrained on 266K samples ( \sim 5,700 glosses) across multiple sign languages, \textbfSignSeek sets a new state-of-the-art performance in cross-corpus retrieval on ASL-Citizen, WLASL, and NMFs-CSL without any downstream fine-tuning. Strikingly, it achieves zero-shot generalisation to an entirely unseen British Sign Language (BSL), surpassing methods explicitly trained on BSL, and transfers seamlessly to isolated sign recognition and subtitle alignment, outperforming prior skeleton-based methods.
[CV-38] Observation-Conditioned Latent Energy Priors for Sparse Implicit Neural Shape Completion MICCAI2026
链接: https://arxiv.org/abs/2609.03694
作者: Paul Büschl,Ezequiel de la Rosa,Julia Wolleb,Julian McGinnis,César Nombela-Arrieta,Bjoern Menze
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted for publication in the MICCAI 2026 Workshop Proceedings (LNCS) as part of Off-Grid, the 1st Workshop on Continuous Representations and Grid-Free Methods in Medical Imaging. 12 pages, 4 figures, 3 tables
Abstract:Implicit neural representations (INRs) can model continuous 3D shapes with a shared coordinate decoder and per-instance latent codes. At test time, autodecoder-style models commonly freeze the decoder and optimize a new latent code from sparse off-grid SDF samples. When these samples underconstrain inference, the latent can drift toward regions that fit the observations but decode implausible unobserved geometry. We propose a post-hoc observation-conditioned latent energy prior for frozen INR decoders. The energy scores standardized latents conditioned on a permutation-invariant encoding of the sparse observation set and is used as a residual expert alongside an L2 latent prior selected on validation data. We evaluate on a controlled cell-nucleus SDF dataset and a public MedShapeNet-derived SDF completion dataset. The proposed L2 objective augmented with conditional energy improves consistently over a validation-selected L2 baseline in the sparsest cell-nucleus regimes and, on MedShapeNet, outperforms both L2 and a six-component GMM latent-density prior across all reported readouts. A shuffled-context ablation is consistently weaker than matched context, supporting an observation-specific contribution. These results suggest that lightweight conditional energies can make pretrained INR decoders more observation-aware without retraining.
[CV-39] MetaStructAtlas: A Grounded 3D Vision-Language Dataset and Benchmark for Functional and Structural Reasoning in Whole-Body PET/CT
链接: https://arxiv.org/abs/2609.03690
作者: Chenguang Zheng,Le Xue,Yichi Zhang,Wenbo Zhang,Zehui Ling,Gang Feng,Xin Gao,Yuan Qi,Yuan Cheng,Zixin Hu,Mei Tian
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:The joint interpretation of metabolic function and anatomical structure is essential for clinical diagnosis in whole-body PET/CT. Although recent advances in 3D medical vision-language models have demonstrated remarkable progress, current efforts are limited to regional CT imaging, leaving a critical void in comprehensive whole-body PET/CT analysis. In this work, we introduce MetaStructAtlas, a large-scale dataset for grounded whole-body PET/CT interpretation that synthesizes multimodal imaging with integrated anatomical, metabolic, and semantic annotations. MetaStructAtlas provides 490 co-registered 3D PET and CT volumes with 50,470 organ-level segmentation masks and grounded radiology reports. To facilitate interactive reasoning, we further developed MetaStructVQA, a standardized 3D grounded visual question-answering benchmark containing 100,565 QA pairs. This framework explicitly links diagnostic queries to visual evidence across modalities, encompassing anatomical, morphological, and metabolic characteristics. Finally, we evaluate state-of-the-art 3D medical VLMs on MetaStructVQA, establishing a robust foundation for multimodal representation learning and integrated whole-body reasoning in nuclear medicine.
[CV-40] Semantic-Aware Subgraph State Space Model for WSI Classification in Histopathology
链接: https://arxiv.org/abs/2609.03689
作者: Feixing Chen,Hao Lu,Lin Luo,Yan Xu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Histopathological subtyping relies on the recognition of characteristic histological patterns. These patterns may be expressed by individual tissue structures or by the spatial distribution and co-occurrence of multiple structures, and they often span irregularly shaped tissue regions, termed semantic units in this work. However, conventional patch-based representations may fragment such units and fail to explicitly preserve their internal spatial organization, while efficiently modeling relationships among numerous spatially separated units remains challenging. To address these limitations, we propose the Semantic-Aware Subgraph State Space Model (SASG-SSM), a flexible and efficient framework for whole slide image (WSI) classification. Semantic-Aware Subgraphs (SASGs) first approximate irregularly shaped semantic units by adaptively grouping spatially connected patches guided by class-agnostic visual-semantic priors. By representing patches as graph nodes with adjacency edges, SASGs preserve their internal spatial organization rather than treating them as an unordered set. A Subgraph State Space Module (SG-SSM) subsequently combines a graph neural network encoder for intra-subgraph topology encoding with a Mamba-based state space encoder for efficient contextualization across large numbers of subgraphs. This module integrates local structural information within semantic units with global contextual information arising from their distribution and co-occurrence across the WSI, while efficiently modeling a large number of spatially distributed regions. Extensive experiments across four WSI subtyping datasets demonstrate consistent advantages over representative state-of-the-art methods. Further evaluations under small-cohort and few-shot settings demonstrate robustness and data efficiency under limited training data. Code will be released at this https URL.
[CV-41] oPO: Token-Conditioned Preference Routing for Attention-Based Latent Diffusion Models
链接: https://arxiv.org/abs/2609.03688
作者: Juntao Xu,Shihong Li,Hoi Fan Au,Ning Zhu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 37 pages, 11 figures
Abstract:Pairwise preference labels rank complete images, yet Diffusion-DPO applies their effect over many spatial and denoising-time coordinates. For attention-based, noise-prediction latent diffusion, ToPO (Token-Oriented Preference Optimization) constructs a per-minibatch, detached, separable spatial-temporal route from branchwise squared-residual contrast in a frozen reference denoiser. Preferred-branch cross-attention uses content tokens to modulate the spatial factor, and an auxiliary pixel-midpoint ordering term is added without local labels or a learned reward model. In matched three-seed retrainings with a shared update schedule, ToPO has higher endpoint estimates than Diffusion-DPO on all five reported SD-1.5 metrics and on HPSv2, ImageReward, and CLIP for SDXL. It also receives larger raw win shares in an aggregate blind SDXL A/B study. These findings are scoped to the reported equal-update U-Net protocols rather than an equal-compute comparison.
[CV-42] DropClick: Semi-Automated One-Click Segmentation for Agricultural Robotic Data ICRA2026
链接: https://arxiv.org/abs/2609.03680
作者: Patrick Zimmer,Michael Halstead,Chris McCool
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注: Accepted to ICRA 2026
Abstract:Labelling vision datasets, especially for segmentation tasks, is a laborious and costly process that stymies novel developments in agricultural robotics. In this paper, we present DropClick, a click-guided segmentation tool that simplifies the annotation process. Our system utilises single-click inputs on objects to generate pseudo-labels, which can replace manual annotations. DropClick stands out as it is a semi-automated approach and does not require a click for every object in the scene. It can therefore further reduce the required amount of user input drastically. We evaluate our method on two challenging agricultural robotic datasets, SB20 and BUP20 for plant and fruit segmentation, respectively. DropClick is first trained on a small subset of just 5 images from the original training data. This DropClick model can then be deployed as a one-click segmentation system and achieves comparable or higher performance than other one-click methods achieving an mIoU of 70.0 and 72.6 points, for SB20 and BUP20 respectively. DropClick then excels at maintaining high performance when clicks are not given (e.g. dropped); when 50% of the clicks are missing it still maintains an mIoU of 68.9 and 71.3 points, for SB20 and BUP20 respectively. We validate DropClick as a pseudo-labelling approach by taking its outputs to train a Mask2Former instance-based segmentation model in a semi-supervised manner. In this process, partially removing user input from DropClick yields similar high performance when compared to providing all clicks, at 70.1 vs 70.7 points AP50 for SB20 and no difference for BUP20 at 77.0 for both models; at the same time saving 46.3% of total input for SB20 and 31.9% for BUP20.
[CV-43] CoFiE: Coarse-to-Fine Evidence Selection for Efficient Streaming Video Understanding
链接: https://arxiv.org/abs/2609.03675
作者: Jing Jiang,Yiran Ling,Ruonan Li,Dimitrios Stamoulis,Jie Liu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Streaming video understanding requires Vision Language Models (VLLMs) to process growing video streams and answer user questions under tight latency constraints. Existing methods improve efficiency through token pruning and memory-bank schemes, but mainly reduce visual tokens after visual encoding. Consequently, downstream token pruning alone cannot substantially reduce end-to-end latency because the expensive frame encoding cost has already been incurred. We propose CoFiE, a Coarse-to-Fine Evidence Selection framework that decouples evidence selection into a coarse, query-agnostic filtering stage before the vision encoder and a fine, query-specific refinement stage during LLM prefill. CoFiE introduces Novelty-Guided Frame Filtering to retain visually distinctive candidate frames and Query-Specific Evidence Refinement to select the frames most relevant to the user query. This design removes substantial redundancy before frame encoding while preserving query-specific refinement once semantic information becomes available. Experiments show that CoFiE establishes a new state-of-the-art accuracy-efficiency trade-off across multiple video understanding benchmarks, reaching 78.86% accuracy on StreamingBench and 68.72% on OvO-Bench, with improvements of up to 3.15% over prior methods. Even with up to 80% evidence-frame filtering, CoFiE outperforms strong open-source multimodal models while improving end-to-end inference latency by up to 2.54 times.
[CV-44] Do Video Generators Track the World Across Segments? A Benchmark and Method for World-State Reasoning in Video Continuation
链接: https://arxiv.org/abs/2609.03673
作者: Yingmao Miao,Pengfei Zhang,Chaoran Xu,Meng Yu,Jing Tang,Xiangxiang Chu,Chao Shen,Chenhao Lin
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Video generators build long videos by composing shorter parts, either by generating segments one after another or by autoregressively extending chunks. Each new part usually depends on memories of historical observations, such as recent frames, selected key frames, memory banks, or cached features. These memories preserve visible evidence from the past, but current generators do not reliably turn such evidence into a world-state interface: what holds in the video world after previous actions and how it should change under the next prompt. A past frame remains valid history, but it may not describe the state needed by the next segment; some states must instead be inferred from occluded or implicit changes rather than copied from a directly observed frame. This creates a simple but overlooked question for video continuation: given a previous video, its prompt, and a new prompt, can a model generate a continuation that reflects the state determined by both the historical video and the new prompt? To answer this question, we introduce Statebench, a benchmark that targets this gap by testing continuations over three state categories: past-visible states, occluded-process states, and complex-transition states. We further propose Stateagent, which explicitly maintains an entity-state representation, updates it under the new prompt, grounds the predicted post-action state as a future end frame, and renders the next video. Experiments show that our method improves controlled video continuation by raising the all-case state score (SCS-All) from 45.2 to 69.3, and also benefits story generation at the one-minute scale. Code is avaliable at this https URL.
[CV-45] ARCOS: Zero-shot Boundary Localization for Corneal Layer Segmentation Across Optical Coherence Tomography Devices
链接: https://arxiv.org/abs/2609.03668
作者: Nuno Vivas Brás,Benjamin Memmi,Maëlle Bouhassane,Cristina Georgeon,Vincent Borderie,Karsten Plamann,Anatole Chessel
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 18 pages, 9 figures, 7 tables
Abstract:Accurate segmentation of corneal layers in optical coherence tomography (OCT) is essential for quantitative assessment of corneal morphology, including layer thickness and structural changes associated with disease or surgery. However, automatic segmentation remains challenging because corneal interfaces are thin, affected by speckle noise, and variable across acquisition devices. In this work, we propose ARCOS, a patch-based zero-shot boundary localization framework for corneal layer segmentation in clinical anterior-segment OCT images. Rather than performing conventional region classification, the method predicts boundary heatmaps for the main corneal interfaces from overlapping native-resolution patches. Patch-level predictions are stitched across the full B-scan and converted into boundary locations to obtain continuous, anatomically ordered layer segmentations. The network combines multi-scale feature fusion with a self-conditioned refinement module that uses intermediate boundary information to improve local heatmap predictions while preserving spatial detail. The method was evaluated on clinical OCT images acquired from multiple devices and compared with representative segmentation baselines using boundary localization and derived thickness metrics. The proposed method achieved an off-by-one boundary localization accuracy of 95.1% and a mean absolute boundary error of 0.514 pixels on the matched-device test set. In zero-shot cross-device evaluation, it maintained an average off-by-one accuracy of 84.3% and a mean absolute boundary error of 0.855 pixels across unseen acquisition devices, outperforming the baseline models. Thickness estimates derived from the predicted boundaries showed low error across corneal regions, supporting the method’s use for quantitative corneal OCT analysis.
[CV-46] Cross-Dataset Transfer and Reliability of Explainable Artificial Intelligence for RhythmFormer Remote Photoplethysmography
链接: https://arxiv.org/abs/2609.03663
作者: Louis Chen,Torbjörn E. M. Nordling
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Image and Video Processing (eess.IV)
备注: 92 pages, 38 figures, incl. supplementary
Abstract:Background. Remote photoplethysmography estimates the cardiovascular pulse from facial video, and its explanations have rested on inspecting heatmaps rather than on quantitative evidence about where a model reads it. We quantified the explanations and asked whether such explanations transfer between datasets and track model performance. Method. We trained eight condition-specific RhythmFormer models on NCKU-rPPG, recorded under three illumination levels, speaking, rotation, and cycling, estimated one heart rate per 5.12-second clip, and set them beside a UBFC-rPPG reproduction. Raw attention, rollout, attention flow, and Beyond Intuition were assessed by skin coverage and the Salience-guided Faithfulness Coefficient (SaCo). Results. Beyond Intuition ranked highest on both datasets, at median coverage 0.789 and SaCo 0.837 on Static level 3 against 0.826 and 0.917 on UBFC-rPPG; lower ranks differed. Within one participant of one condition, neither measure was related to a clip’s heart-rate error, waveform correlation, or signal-to-noise ratio on either dataset: 186 of the 252 coefficients fell below |\rho|=0.10 and 28 reached p0.05 against the 13 expected by chance. Across the eight scenarios only Beyond Intuition’s coverage followed the three performance measures, at \rho=-0.43 , +0.57 , and +0.43 , while the attention-only methods’ SaCo ran opposite to each. It failed at 40 lux alone, its median coverage falling to 0.180 and its median SaCo to -0.178 , whereas motion degraded the estimates far more without such a drop. Conclusions. Skin coverage and SaCo carry information complementary to the performance measures rather than a proxy for them: attributing to the skin does not guarantee an accurate estimate. What an attribution reveals about a condition is where the model looks rather than how faithfully its map is ordered.
[CV-47] Rethinking 3D Noise: Learning 3D-Aware Video Priors via Optimization-Free Morphological Perturbations
链接: https://arxiv.org/abs/2609.03657
作者: Onat Şahin,Mohammad Altillawi,George Eskandar,Carlos Carbone,Ziyuan Liu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:3D scene representations like NeRF and 3D Gaussian Splatting (3DGS) suffer severe artifacts in sparse-view settings. Recent generative 3D artifact fixers attempt to address this, but rely on paired corrupted and clean renders requiring costly, per-scene reconstructions across varying view configurations. While 2D image augmentations act as instant regularizers, no explicit equivalents exist for 3D representations to preserve spatial consistency across views, an essential property for 3D-aware training. We propose 3D Morphological Perturbations as an optimization-free regularizer that preserves spatial consistency. Leveraging explicit 3DGS, we treat each Gaussian as a fundamental building block - analogous to a 2D pixel - and apply perturbations across its morphological parameter space via scale, rotation, and pruning. Our method eliminates per-scene 3DGS optimization loops from dataset curation while enabling models to learn stronger geometric priors than sparse-view baselines in diagnostic ablations conducted on a lightweight video diffusion sandbox. Scaled to a 14B-parameter video model via ControlNet, our approach maintains visual fidelity while reducing mean depth error by 12.5% over state-of-the-art image-to-image 3D artifact refiners, ultimately boosting downstream robotics policy success rates by up to 8.0% across 3 of 4 manipulation tasks.
[CV-48] PL-SCEA: Reconfiguring Pretrained Attention for Few-Shot Industrial Anomaly Detection
链接: https://arxiv.org/abs/2609.03655
作者: Xiaoyu Yang,Qixing Wu,Huixian Zhao,Changlong Jin
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Vision Foundation Models (VFMs) provide transferable patch representations for few-shot industrial anomaly detection, but their attention computation is typically inherited from pretraining objectives centered on semantic aggregation. This creates a potential mismatch: token relations that support semantic recognition may not adequately expose the localized texture and structural deviations required for anomaly localization. We therefore investigate the hypothesis that the attention computation of a frozen VFM can be reconfigured as a task-relevant component of anomaly detection. We instantiate this idea with Power-Law Self-Correlation Enhanced Attention (PL-SCEA), which retains the semantic context of pretrained query-key attention while constructing token-adaptive self-correlations over contextualized value features. Positive-correlation filtering and power-law reweighting then emphasize relations that are salient relative to each token’s relational background, without introducing additional trainable attention projections. The resulting features are modeled by a lightweight variational autoencoder that provides a fixed-size reconstruction-based representation of category-specific normality. The two stages serve complementary roles: attention reconfiguration shapes how local relational deviations are represented, while reconstruction-based modeling converts deviations from learned normality into anomaly scores. Across MVTec AD and VisA, the complete framework achieves competitive image-level detection and consistently strong pixel-level localization across the evaluated few-shot settings. Ablations further show that PL-SCEA improves localization with either the VAE or a memory bank under the tested setting. These results support the view that task-aligned attention reconfiguration can improve the anomaly-localization capability of frozen pretrained representations.
[CV-49] ree-Structured Vector Quantization For Efficient And Progressive Image Compression
链接: https://arxiv.org/abs/2609.03641
作者: Xinkun Wang,Tianyi Xu,Qingyu Luo,Mingming Ma,Changzhe Jiao,Fu Li,Yi Niu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Vector-quantization based image compression has achieved strong rate–distortion performance, yet most of them still produce a separate compressed representation for each target bitrate. Such variable-rate behavior allows one model to operate at multiple rates, but it does not necessarily provide a progressive bitstream whose prefixes are themselves decodable and can be refined by appending additional bits. We propose \textbfTree-VQ, a progressive tree-structured vector quantization framework for learned image compression. Tree-VQ organizes discrete codewords as a hierarchical binary tree and represents each latent token by a routed root-to-leaf path. Crucially, every prefix of this path corresponds to a valid quantized representation, so shallow internal nodes serve as coarse reconstruction codes and deeper nodes provide successive refinements. This allows a compressed image to be decoded from an early prefix and progressively improved as more branch symbols are received, rather than being re-encoded for different target rates. To make this structure practical for compression, we introduce a prefix-compatible tree entropy model that codes progressive continuation decisions and routed branch refinements using only causally available decoded contexts. We further use rate-aware refinement scheduling to decide which spatial blocks should receive additional tree bits under a given prefix budget, and hierarchical prefix supervision to ensure that internal nodes are directly decodable at low rates. Experiments show that Tree-VQ achieves a superior performance–efficiency trade-off, delivering the best perceptual compression results with much fewer parameters and lower latency than competing methods.
[CV-50] Stabilizing Camera-Controlled Novel View Synthesis at Inference Time
链接: https://arxiv.org/abs/2609.03639
作者: Prajwal Singh,Arjun Badola,Seema Kumari,Hajime Nagahara,Shanmuganathan Raman
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Training-free, camera-controlled novel view synthesis from a single image using pre-trained video diffusion models often becomes unstable under large camera motion and long generation horizons. Existing approaches commonly combine several inference-time components, making it unclear which design choices are most important for stability. We show that the main source of stability is simple. Decomposing camera motion into small autoregressive steps limits per-step geometric distortion and reduces error accumulation. A controlled camera-step study shows that performance remains stable for small motions and degrades more strongly as the per-step motion approaches 18 - 20^\circ . We further evaluate geometry-constrained spatial attention and low-frequency appearance anchoring as supporting refinements, together with an efficient registration-free warping pipeline. Across RealEstate10K and MegaScene, CamTrol++ improves temporal and geometric consistency, downstream 3D reconstruction quality, and generation efficiency over training-free baselines. The method remains effective for 56-frame generation and under substantial controlled depth corruption. These results show that careful control of camera motion at inference time can substantially improve the stability of camera-controlled novel view synthesis without retraining or modifying the diffusion backbone.
[CV-51] EraseSAE: Surgical Concept Erasure in Text-to-Video Diffusion Models via Sparse Autoencoders
链接: https://arxiv.org/abs/2609.03629
作者: Xinghao Wang,Dong Li,Wei Yu,Yingwei Pan,Tao Gong,Qi Chu,Nenghai Yu,Ting Yao
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Recent advances in text-to-video (T2V) diffusion models have demonstrated remarkable generative capabilities, yet their reliance on loosely curated training data raises pressing safety and copyright concerns. Concept erasure offers a principled remedy by removing unwanted semantics from pretrained models while preserving remaining concepts. However, existing approaches typically operate at a coarse granularity misaligned with the fine-grained, distributed nature of concept representations, leading to incomplete removal or degraded generation quality. We argue that surgical erasure fundamentally requires intervention at the level of monosemantic features, where each unit encodes a single interpretable concept. To this end, we propose EraseSAE, a novel framework that leverages sparse autoencoders to achieve surgical concept erasure in DiT-based T2V diffusion models via a principled decompose-attribute-erase pipeline. We first introduce the Partitioned Convolutional Sparse Autoencoder, which decomposes dense spatiotemporal activations into disentangled, interpretable sparse features while preserving spatiotemporal coherence. A contrastive attribution mechanism then contrasts activations from paired prompts to isolate concept-specific feature kernels. At inference, timestep-resolved spatiotemporal masks derived from the identified kernels confine erasure to regions where the target concept is active, leaving unrelated content intact. Extensive experiments across diverse diffusion models and concept erasure tasks demonstrate that EraseSAE achieves precise and robust concept removal with minimal quality degradation, substantially outperforming state-of-the-art methods. The code is available at this https URL.
[CV-52] Auditing Patient Privacy in Medical Generative Models: Scalable Memorization Detection with DeepSSIM
链接: https://arxiv.org/abs/2609.03615
作者: Antonio Scardace,Francesco Guarnera,Sebastiano Battiato,Daniele Ravì
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:While deep generative models offer new opportunities for medical image synthesis and data sharing, their ability to memorize and reproduce training samples raises serious concerns about patient confidentiality. Detecting such memorization at scale remains challenging: traditional pixel-based metrics are sensitive to generation artifacts, whereas generic embedding-based metrics often lack the anatomical sensitivity required for medical data. To address this challenge, we introduce DeepSSIM++, a self-supervised similarity metric for scalable memorization auditing in medical generative models. By leveraging multi-scale feature aggregation and anatomy-preserving augmentations, DeepSSIM++ learns an embedding space where cosine similarity approximates the Structural Similarity Index (SSIM), eliminating the need for exact pixel-level registration. Compared with state-of-the-art baselines, DeepSSIM++ achieves an average Macro F1 improvement of 33 percentage points under ideal alignment and 46 percentage points under realistic spatial and intensity perturbations. Furthermore, it accelerates large-scale similarity computation by several orders of magnitude compared with analytical SSIM. By combining anatomical sensitivity and computational efficiency, DeepSSIM++ provides an open-source tool for scalable memorization auditing in medical generative AI. Code and data are publicly available at: this https URL.
[CV-53] SV-WAM: An Efficient Surround-View World-Action Model for End-to-End Autonomous Driving
链接: https://arxiv.org/abs/2609.03602
作者: Jinyang Wang,Shiwei Li,Junjian Wang,Zhiqiang Deng,Jianbin Gao,Yihang Zhao,Liu Liu,Yongjia Zhao,Jinlong Chen,Huirui Xu,Yifeng Pan,Kangwei Liu,Fan Ren,Ji Tao,Minghao Yang
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注: 23 pages, 16 figures
Abstract:World models (WMs) have demonstrated strong potential for end-to-end autonomous driving by learning predictive representations of future scene dynamics. However, generating future videos during inference introduces substantial computational overhead, leading many recent driving WMs to adopt a single front camera as input for efficient deployment. This design restricts spatial coverage in safety-critical maneuvers such as lane changes, merges, and turns. To address this limitation, we propose SV-WAM, a surround-view world-action model (WAM) that preserves full six-camera observations while maintaining efficient inference. SV-WAM leverages future-video prediction as dense training supervision for action learning within a shared generative model, rather than as an inference-time output. At the core of this design is an action-centered causal mask that prevents action tokens from attending to future-video tokens during joint action-video denoising. Consequently, the video branch can be discarded at deployment, enabling efficient action-only planning. Furthermore, we introduce a differentiable drivable-area compliance regularizer that penalizes vehicle-footprint corners approaching or crossing drivable boundaries, improving planning safety and boundary awareness. Extensive experiments on the closed-loop NAVSIMv2 benchmark and the open-loop nuScenes benchmark demonstrate that SV-WAM achieves state-of-the-art planning performance with low inference latency and competitive zero-shot transfer capability.
[CV-54] xt2Thermal: Physics-Aware Thermal Image Synthesis from Textual Priors
链接: https://arxiv.org/abs/2609.03585
作者: Tayeba Qazi,Brejesh Lall,Prerana Mukherjee
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 15 pages, 6 figures
Abstract:Thermal infrared imaging offers reliable perception in darkness and adverse weather, but thermal datasets remain scarce, motivating extensive work on translating abundant RGB images into thermal. Such translation is fundamentally ill-posed as thermal appearance is governed by surface emissivity and object temperature, neither of which is observable in the visible spectrum, so a single RGB image is consistent with many valid thermal outputs. We argue that language offers a natural means of resolving this ambiguity, and propose Text2Thermal, a framework for physics-aware thermal image synthesis from textual priors. Rather than inferring the unobservable radiometric factors from RGB, we supply them explicitly through thermally grounded captions encoding material, weather, time-of-day, and heat-emission state, and adapt a pretrained Stable Diffusion backbone to the thermal domain. Because the radiometric content is determined entirely by the prompt, Text2Thermal synthesizes thermal imagery without requiring a registered RGB image at inference; where spatial guidance is desired, an optional control signal imparts scene geometry without disturbing the prompt-specified radiometry. Experiments on M3FD, FLIR, and FMB show that Text2Thermal achieves state-of-the-art FID among thermal image synthesis methods while offering text-level control that translation-based approaches cannot provide.
[CV-55] Drive-HWM: Hierarchical World Models for Dynamic-Latent Guided Autonomous Driving
链接: https://arxiv.org/abs/2609.03572
作者: Zhaoxin Fan,Tianbao Zhang,Wenjun Wu,Xiaofeng Wang,Yeying Jin,Jian Zhao,Zheng Zhu,Shuicheng Yan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 14 pages
Abstract:World models offer a promising paradigm for autonomous driving by predicting how traffic scenes may evolve and using such predictions to support action generation. However, existing approaches either separate future prediction from action generation or jointly predict them at the same temporal scale, making it difficult to simultaneously achieve long-horizon anticipation and responsive, observation-grounded decision making. We present Drive-HWM, a hierarchical slow–fast world modeling framework that organizes future representation prediction and action generation at complementary temporal scales. The slow world model predicts multi-step future representations to capture extended scene evolution. To explicitly model the abundant motion dynamics in driving environments, we introduce Dynamic-Aware Latents learned through optical-flow prediction. Guided by these future representations, the fast model uses a lightweight multimodal backbone and an autoregressive expert to jointly predict the next frame and the immediate action from the latest observation. Next-frame prediction encourages the fast model to capture imminent scene evolution, while one-step action generation allows decisions to be continuously updated as new observations arrive. Extensive experiments on NAVSIM v1 and v2 demonstrate the strong driving performance of Drive-HWM. Comprehensive ablation studies further validate the effectiveness of the hierarchical slow–fast design, dynamics-aware future representations, and joint next-frame and action prediction.
[CV-56] FlashRender: Few-Step Generative Rendering via Camera-Controlled Video MeanFlow
链接: https://arxiv.org/abs/2609.03563
作者: Byeongjun Park,Byung-Hoon Kim,Hyungjin Chung
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project page: this https URL
Abstract:We present FlashRender, a few-step generative rendering framework that retakes a source video along a target camera trajectory in seconds. We identify sampling-step-dependent camera control as a prominent manifestation of discretization error in existing multi-step generative rendering models and show that resolving this inconsistency substantially lowers denoising trajectory curvature, facilitating subsequent step distillation. To this end, we introduce Representation Transformation and Alignment (RETA), which aligns hidden source-video representations with target-video features from a frozen visual geometry model. This directly encodes the geometric transformation within the source-video stream, enabling sampling-step-consistent camera control. We then fine-tune the model with the MeanFlow objective on the lower-curvature denoising trajectory induced by RETA, allowing the model to more effectively address discretization error. Finally, we apply on-policy flow map distillation to correct self-rollout errors under fixed few-step sampling. Extensive experiments show that RETA, MeanFlow, and on-policy flow map distillation play complementary roles in few-step generative rendering. Together, they enable our approach to match multi-step baselines in video quality and geometric consistency at 25x lower sampling cost while achieving superior camera controllability, even under out-of-distribution target camera trajectories.
[CV-57] Building Pretraining Data for World Models: An Unreal Engine-Based Pipeline for Action-Conditioned Video Generation
链接: https://arxiv.org/abs/2609.03557
作者: Haoyu Wang,Songchun Zhang,Haoran Li,Haoyang Huang,Zeyue Xue,Nan Duan
类目: Computer Vision and Pattern Recognition (cs.CV); Graphics (cs.GR)
备注: 16 pages, 5 figures
Abstract:Action-conditioned video models require large-scale visual data paired with control signals that are temporally aligned with the resulting scene transitions. Such supervision is difficult to obtain from ordinary real-world video because the actions that caused each visual change are typically unknown. We present a large-scale synthetic data production pipeline built on Unreal Engine for generating action-conditioned, multi-view video. To accommodate the different execution requirements of real-time physics and high-quality offline rendering, the pipeline executes trajectory generation and final rendering in two stages: Stage I runs real physics in PIE and records per-frame character states, control inputs, and camera states into an intermediate trajectory representation; Stage II replays those trajectories in a new engine process and renders them offline with Movie Render Queue (MRQ). Around this core, we develop a distributed production system with cache-aware task partitioning, node-local slot scheduling, automated scene screening, aesthetic and luminance filtering, partial-output recovery, asynchronous upload, and continuous cluster health monitoring. The production cluster contains 25 servers with eight NVIDIA RTX 5090 GPUs per server. From 2,384 asset packs, 429 levels were retained for production together with a pool of 40 humanoid characters. The pipeline has produced 2,691 hours of 1080p video and 6,076 hours of 720p video. We describe the system architecture, the implementation decisions that emerged from production failures, and the limitations of using perceptual quality proxies for world-model data curation. The pipeline described in this report constitutes the Unreal Engine synthetic-data production component used in EchoWM.
[CV-58] WIDE: Wildcard Inference with Dynamic Expansion for Cross-Modal Generative Retrieval ACM-MM2026
链接: https://arxiv.org/abs/2609.03554
作者: Teng Guo,Xin Wang,Jiayou Xu,Keying Zhou,Jifeng Shen,Haoxin Ruan
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Accepted to the 34th ACM International Conference on Multimedia (ACM MM 2026). 10 pages, 5 figures
Abstract:Generative retrieval has demonstrated significant success by unifying representation learning and search into a single sequence-to-sequence generation task. However, extending this paradigm to cross-modal retrieval reveals a critical challenge arising from the inherent information asymmetry across different modalities, such as the gap between concise text queries and dense visual candidates. This structural mismatch causes the autoregressive decoder to suffer from forced hallucination when generating identifiers via standard trie-constrained beam search, where the model is severely penalized for failing to guess fine-grained details absent from the query, allowing irrelevant candidates to hijack top rankings. To address this issue, we propose Wildcard Inference with Dynamic Expansion (WIDE). WIDE employs Adaptive Entropy Thresholding (AET) to calibrate layer-specific uncertainty boundaries offline. During the decoding generation phase, Asymmetry-aware Wildcard Decoding (AWD) detects semantic blind spots and emits wildcards instead of forced deterministic identifiers, dynamically expanding the search space without incurring log-probability penalties. Finally, Blind-Spot Re-ranking (BSR) evaluates the expanded candidate pool using a hybrid scoring mechanism that combines discrete generation confidence with continuous semantic similarity. Extensive experiments on the M-BEIR benchmark demonstrate that WIDE outperforms state-of-the-art generative retrieval methods, effectively suppressing forced hallucination while maintaining compact index structures.
[CV-59] SafeRI: Recognition and Intervention for Token-Level Safety Intervention in Large Vision Language Models
链接: https://arxiv.org/abs/2609.03544
作者: Caoyuan Ma,Tian Gu,Wenpu Liu,Weichu Xie,Shuai Dong,Yuqi Xu,Ji Zhao,Ziyue Wang,Wenzheng Chang,Taiqiang Wu,Yongfu Zhu,Wenqi Shao,Zheng Wang,Yinqiang Zheng
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Preprint. 13 pages, 4 figures. Main paper with appendix
Abstract:Existing safety alignment methods for vision-language models usually modify the model behavior globally: once the safety parameters are trained or loaded, they participate in both unsafe and already-safe generations. This always-on intervention can unnecessarily perturb the model’s original reasoning path and degrade general multimodal capabilities. We argue that safety alignment should be an on-demand intervention rather than a permanent modification to every decoding trajectory. To this end, we propose a streaming recognition and gated LoRA framework for intrinsic VLM safety. During autoregressive generation, a lightweight recognizer estimates whether the current pre-token generation state is safe or unsafe. Its output updates the LoRA gate for the following decoding step; otherwise, generation follows the frozen-backbone policy. The LoRA module is trained from unsafe prefixes, transition statements, and safe continuations, so that it learns to redirect unsafe generations back to safe responses after activation. Experiments across multiple safety and general-purpose benchmarks demonstrate the effectiveness of our method in post-alignment settings.
[CV-60] runcGradGS: Improved 3D Gaussian Splatting via Truncated Gradient Updates
链接: https://arxiv.org/abs/2609.03534
作者: Theo Morales,Nhat-Quynh Le-Pham,Robin Atkins,Binh-Son Hua
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Graphics (cs.GR)
备注: Accepted at Pacific Graphics 2026
Abstract:3D Gaussian Splatting has become a de facto scene representation for novel view synthesis, yet robustly learning 3D Gaussian primitives from visual input remains challenging. Standard optimization relies on gradient-based updates, but a common issue is the gradient vanishing phenomenon: a pixel far from a Gaussian primitive often has diminishing gradient magnitudes to influence primitive attributes, resulting in suboptimal scene reconstruction. In this paper, we propose a method to address gradient vanishing with a piecewise truncated gradient formulation that improves the optimization stability and robustness to initializations. We show that our method consistently improves 3D Gaussian Splatting with random and COLMAP initializations while being generalizable across static and dynamic Gaussian Splatting. As a by-product, we also examine the limitations of current benchmarks for dynamic scenes, and introduce a novel dataset for benchmarking dynamic Gaussian Splatting using synthetic 3D scenes. We demonstrate the effectiveness of our method in both static and dynamic settings for the public benchmarks and our proposed dataset.
[CV-61] Neural Video Compression Based on Deformable Temporal Alignment and Difference-aware Fusion
链接: https://arxiv.org/abs/2609.03520
作者: Chuyue Shan,Songlin Sun,Wang Chenwei,Shen Zihan
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:In conditional coding-based neural video compression, the quality of temporal context directly affects compression per- formance. Existing methods mostly construct context from prop- agated reference features, but they are vulnerable to motion esti- mation and local alignment errors in regions with complex mo- tion, occlusion, and high-frequency textures, resulting in inaccu- rate temporal information. To address this issue, this paper pro- poses a method combining deformable temporal alignment and difference-aware spatial selective fusion. A Context-aware Tem- poral Alignment Module is used to generate complementary tem- poral context, while a Difference-aware Spatial Selective Fusion module adaptively selects reliable temporal information and sup- presses misalignment. Experiments show that the proposed method achieves certain rate-distortion performance improve- ment over DCVC-DC.
[CV-62] Residual Optimal Transport-Based Experts Collaboration Towards Modality-Aware Infrared-Visible Object Detection
链接: https://arxiv.org/abs/2609.03516
作者: Yue Zhao,Hua Yu,Yukun Zhao,Yuzhi Zhang,Maoguo Gong,Xin Mei,Zhuping Hu,Yanchi Li,A. K. Qin
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Infrared-visible object detection (IVOD) integrates complementary evidence from visible and infrared sensors for reliable perception in challenging scenes. In practice, sensors may fail or drop frames, leaving one modality unavailable or intermittent. Existing methods for IVOD assume both modalities are always present, and fixed fusion collapses when one stream is missing. Furthermore, it remains a critical challenge to reliably estimate semantic correlation across heterogeneous modalities, especially under spectral distribution discrepancy. We present FlexibleFusion, a unified and adaptive method that flexibly allocates integration pathways and fusion strength, operating seamlessly across complete and missing-modality regimes. At its core, the Modality-Aware Experts Collaboration (MAEC) mechanism selectively activates and aggregates cross-modal or intra-modal expert pathways. It allows cross-modal fusion when full modalities are available and falls back to self-fusion under missing conditions. Additionally, we design Residual Self-Paced Entropic Optimal Transport (RSPEOT) to align heterogeneous feature distributions from a transport perspective. Instead of relying on the fixed sparsity coefficient in standard entropic optimal transport (EOT), RSPEOT introduces a residual-driven self-paced update that prioritizes reliable matches and progressively refines harder ones. This design alleviates the additional optimization burden of standard EOT while preserving reliable semantic alignment. Comprehensive experiments under complete and missing-modality protocols show consistent performance across arbitrary modality configurations. Code will be released upon publication.
[CV-63] ree species mapping in Denmark: A comparison of spectral-temporal features with geospatial foundation model embeddings
链接: https://arxiv.org/abs/2609.03480
作者: Alkiviadis Koukos,Spyros Kondylatos,Thomas Nord-Larsen,Lotte Nyborg,Christian Tøttrup,Kenneth Grogan
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Submitted to Remote Sensing of Environment. This preprint presents a national-scale tree species mapping framework for Denmark using Sentinel-1/2 time series, National Forest Inventory data, and EO foundation model embeddings. The resulted national map can be found here: this https URL
Abstract:We map tree species across Denmark using National Forest Inventory plots and EO data, while evaluating the potential of foundation models for large-scale forest characterization. We compare two alternative input representations for tree species classification: (i) manually engineered spectral-temporal features (STF) derived from multi-temporal Sentinel-1 and Sentinel-2 observations, and (ii) embeddings generated by the EO FMs TESSERA and AlphaEarth. Both representations are complemented with canopy height information. Random forest, XGBoost, and Multi-Layer Perceptron (MLP) classifiers are evaluated for all input representations, with separate assessments for pure and mixed forest stands. The STF-based MLP achieves the highest classification performance, yielding macro F1 scores of 0.843 and 0.653 for pure and mixed stands, respectively. The MLP trained on TESSERA embeddings delivers competitive performance for pure stands, achieving results within 1.1 percentage points of the best-performing model. TESSERA consistently outperforms STF-based models when fewer than approximately 25% of training plots are available, demonstrating a substantial advantage under limited training data. Multi-year observations systematically improve classification accuracy relative to single-year inputs, while ablation experiments reveal the complementary contributions of Sentinel-1 backscatter, spectral indices, and canopy height data. The best-performing model is subsequently applied at the national scale to generate a 10 m tree species map of Denmark. Area-adjusted validation indicates an overall map accuracy of 79.9%. The resulting map, released as an open-access product, is the first high-resolution national tree species map of Denmark and provides a valuable resource for forest monitoring, ecological research, and land management applications.
[CV-64] SafeRestore: Detector-Relative Risk Certificates for Selective Industrial Image Restoration
链接: https://arxiv.org/abs/2609.03475
作者: Shaoliang Yang,Jun Wang
类目: Computer Vision and Pattern Recognition (cs.CV); Applications (stat.AP)
备注: 27 pages, 10 figures, 6 main-text tables, and 17 supplementary tables. Shares the Carinthia-S image identities with arXiv:2607.17401 by the same authors; the overlap and the distinct estimand are stated in Section 2.4 and Table 1
Abstract:Industrial inspection pipelines often restore a measured image before a detector acts on it, yet restoration can suppress detector-supported defect structure or create clean-region activations. We formulate restoration as a selective action problem over the measured display, five restored candidates, and review. SafeRestore ranks candidates with action-specific fitted scores, chooses a gate on threshold-tuning data, and evaluates the fixed gate on a disjoint certification sample with two one-sided exact binomial bounds: one for the positive-conditional evidence-loss incident rate and one for the all-accepted excess-activation incident rate. The guarantee is marginal for one policy fixed before its certification outcomes are observed, under an image-level i.i.d. working model. In a retrospective split-sample study of 4,591 public Carinthia-S images, the protocol yields auditable risk-coverage behavior. The primary all-action policy passes in one of five training repetitions (12.0% +/- 26.9% pass-gated test coverage when failures count as zero), whereas fixed bicubic and reduced-complexity variants pass more often. On reserved morphologies, evidence-loss incidence rises to 81.1-90.3%, and KolektorSDD lacks both detector competence and enough positive certification images for the stated target. The contribution is therefore an auditable, detector-relative framework for deciding when a transformed image may be returned automatically and when review remains necessary – not a claim that adaptive routing outperforms simpler policies on the present evidence.
[CV-65] BMCTrack-d: Pig re-identification and tracking via back marks in challenging camera settings
链接: https://arxiv.org/abs/2609.03463
作者: David Brunner,Maciej Oczak,Marie Bordes,Jean-Loup Rault,Stephan M. Winkler,Viktoria Dorfer
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Automated pig monitoring is essential for assessing their health, behaviour, and welfare. To date, most pig monitoring solutions operate on the group-level, because individual-level monitoring requires reliable long-term identification and tracking of each animal. For domesticated pigs this remains challenging because pigs of the same breed often have highly uniform appearances. Moreover, research on pig monitoring is almost exclusively reported in top-down view camera settings, which considerably ease tracking, but are not always an option in practice. In this work, BMCTrack-d is presented, a novel tracking-by-detection approach that leverages unique back marks to enable robust pig re-identification and tracking in a challenging side-view camera setting, afflicted by rapidly moving pigs, severe occlusions and low resolution. The method first predicts the detected pigs’ identities using a neural network-based back mark classifier. To improve re-identification reliability over time, two dedicated post-processing stages are introduced: a temporal prediction consistency check, which validates the identity assignments against the recent prediction history, and deduplication, which resolves conflicting identity assignments in each time step. By explicitly prioritising accurate, appearance-based re-identification over continuous tracking, the proposed approach addresses a key limitation of existing trackers for individual-level monitoring scenarios. On a demanding test set BMCTrack-d outperforms two strong baselines, BoT-SORT-ReID and TrackTrack-ReID, by 9.11% and 1.03%, respectively, in higher-order tracking accuracy. These results demonstrate the effectiveness of back mark-based re-identification and tracking for robust individual-level pig monitoring in challenging settings.
[CV-66] Preprocessing Failure and Adversarial Detection in Depthwise-Separable Edge Vision Systems
链接: https://arxiv.org/abs/2609.03453
作者: Jannatul Masruk Mukta,Rifa Sanjida,Adrita Rahman Tory,Md. Saifur Rahman,Khondokar Fida Hasan
类目: Computer Vision and Pattern Recognition (cs.CV); Cryptography and Security (cs.CR)
备注: 14 pages
Abstract:Preprocessing-based defenses are the standard first-line response to adversarial attacks on edge vision systems, requiring no retraining, no architectural changes, and widely recommended as model-agnostic mitigations. Yet the foundational evaluations of these defenses were conducted on residual or Inception-class architectures, not on the depthwise-separable CNNs that dominate edge deployments. This untested assumption leaves a gap in the security evaluation literature. This paper closes that gap by evaluating six preprocessing defenses against adversarial perturbations across both architecture families. Across all perturbation levels and defenses tested, the two depthwise-separable architectures show consistently poor recovery while the residual architecture shows partial recovery; ablation results are consistent with an architectural rather than parametric explanation, though only three architectures and one attack family are evaluated. Crucially, this failure is not merely a negative result. The same output divergence that disqualifies preprocessing as a recovery mechanism reveals a detection opportunity: preprocessing consistently disrupts clean predictions while leaving adversarial predictions largely unchanged, an asymmetry that is directly measurable without retraining or architectural modification. We further show that standard image quality metrics are unreliable proxies for defense effectiveness, a methodological gap in current evaluation practice. A practitioner decision framework is provided for adversarially resilient edge vision deployment.
[CV-67] STARS-GS: Structure-Aware Regularized Gaussian Splatting for Large-Scale Aerial Surface Reconstruction
链接: https://arxiv.org/abs/2609.03447
作者: Bocheng Li,Wenjuan Zhang,Jie Pan.Dongxu Han,Xuesong Ma,Yiling Yao,Yaning Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Large-scale 3D surface reconstruction from aerial imagery is fundamental to geospatial mapping and urban modeling. Recent advances in 3D Gaussian Splatting (3DGS) have demonstrated considerable potential for this task. However, existing methods still face three major challenges in large and complex scenes: scene partitioning may split continuous scene elements across independently optimized sub-regions; geometric constraints mainly focus on the attributes of individual Gaussians while overlooking their local organization; and uniform regularization struggles to accommodate heterogeneous geometric structures. To address these issues, we propose STARS-GS, a structure-aware 3DGS framework for large-scale surface reconstruction. First, we introduce a structure-aware scene partitioning strategy that better preserves continuous scene structures during partitioning and reduces cross-region geometric inconsistencies and stitching artifacts through boundary refinement. Second, we develop neighborhood-aware Gaussian organization that extends geometric constraints from individual primitives to their neighborhood organization, encouraging Gaussians to better conform to local surface geometry. Third, we introduce adaptive surface regularization that adjusts the regularization strength according to local geometric characteristics, promoting geometric consistency in structured regions while preserving plausible variations in unstructured regions. Extensive experiments on large-scale aerial photogrammetry benchmarks demonstrate that STARS-GS consistently outperforms the evaluated Gaussian-based methods in surface reconstruction. It increases the average F1-score from 0.640 for the second-best method to 0.698, corresponding to a relative improvement of approximately 9.1%, demonstrating effective improvements in geometric accuracy and surface completeness.
[CV-68] Preserving Knowledge across Space and Time for Continual Video Deepfake Detection ECCV2025
链接: https://arxiv.org/abs/2609.03446
作者: Taehoon Kim,Jongwook Choi,Heejae Jo,Byungmin Park,Jongwon Choi
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted by ECCV 2025. Code will be available at this http URL
Abstract:The continuous emergence of high-quality video deepfakes requires detectors that continually adapt to new forgery patterns, yet existing approaches, which are designed for deepfake images, fail to capture video-specific cues. Unlike deepfake images that contain only spatial artifacts, deepfake videos leave distinct evidence along both spatial and temporal axes, necessitating the separate preservation of each modality during sequential model updates. To overcome this limitation, we introduce a continual deepfake video detection framework, Modality-Specific Frequency Distillation (MSFD), that explicitly decomposes video features into spatial, temporal, and spatiotemporal modalities in the frequency domain. This decomposition enables independent preservation of each modality, as different deepfake video types exhibit varying reliance on spatial and temporal cues across tasks. Furthermore, MSFD adopts a cross-modality decorrelation loss that encourages spatiotemporal representations to remain orthogonal to single-modality cues. Extensive experiments show that our framework achieves stronger adaptation and preserves performance more effectively than state-of-the-art methods across diverse continual deepfake video scenarios.
[CV-69] OCR-EDR: Rendering-Aware Diagnosis and Repair for Closed-Loop OCR Improvement
链接: https://arxiv.org/abs/2609.03445
作者: Linnan Zhao,Kang Liu,Hao Yu,Jiabo Zhan,Chong Sun,Chen Li
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Although document OCR systems perform increasingly well on routine documents, complex formulas, structured text, and long-tail formats remain error-prone. OCR predictions may omit fine-grained content or hallucinate unsupported outputs, while equivalent encodings of the same visible content must be accommodated. Existing OCR evaluation methods mostly report aggregate metrics, offering limited support for analyzing case-level errors and improving OCR performance. We propose OCR-EDR (OCR Error Diagnosis and Repair), a rendering-aware framework that advances from fine-grained diagnosis to iterative repair. Given a source image, an editable OCR prediction, and its rendered image, OCR-EDR first jointly assesses whether the prediction and its rendering are consistent with the source, preserving valid predictions, including rendering-equivalent ones, while diagnosing and localizing genuine errors. It then applies executable edits and may request an updated rendering for iterative reassessment. We construct OCRErrBench from diverse real OCR predictions, covering text and formulas, exact and rendering-equivalent positives, and genuine errors, and develop the DocEDR model to execute the diagnosis–repair loop. On OCRErrBench, DocEDR achieves 94.78% diagnostic accuracy. It repairs 86.23% of erroneous inputs to visual consistency, raises formula Case-F1 by 30.99 percentage points over DOCR-Inspector-7B on DOCRcaseBench, and improves formula CDM by up to 4.62 percentage points on the identified Bad subsets of four OCR systems on UniMER-Test. These results show that OCR-EDR turns fine-grained OCR analysis into verified corrections and performance gains.
[CV-70] When Do Frozen VLMs Respond to Image-Free Object-Token Edits? An Answer-Key-Free Protocol and What It Reveals
链接: https://arxiv.org/abs/2609.03429
作者: Wonbin Son,Gyumun Choi,Junil Seo,Seungmin Rho,Mi Young Lee,Hyungjoon Kim
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 10 pages, 3 figures, 5 tables
Abstract:Answering what-if queries about a scene with a VLM usually means injecting the assumption as text or repainting the scene with a generative model. We instead move the edit to the representation level, before the model input. The image is abstracted into a set of object-level tokens, and the original image never enters the VLM. This design rests on an open question: when do frozen VLMs actually respond to such token edits? We introduce an answer-key-free protocol: no post-edit answer is annotated. It scores edits whose answers are logically determined, and audits itself by reversing each scoreable choice. The protocol reveals three structures. The response is not free: explicit edit teaching, not ordinary VQA training, produces it in dense scenes and multiplies it in sparse ones, on all three operations. Once on, it is governed by token cleanliness and density, with deployable detector+segmenter tokens competitive with the oracle and outperforming it on VRSBench. And reading is a separable axis: the image-free token route preserves 92-96% of a matched patch-token baseline’s free-text VQA, and the answers measurably depend on the tokens. The response, cleanliness, and reading structures are sign-preserved across two remote-sensing datasets (iSAID, VRSBench) and three frozen LM backbones. We release the probe generator, records, judge logs, and code.
[CV-71] Mudrag en: Geometrically Supervised Generation of Interacting Two-Hand Mudras for Preserving Indian Classical Dance Heritage
链接: https://arxiv.org/abs/2609.03415
作者: Jagadish Kashinath Kamble,Jayanta Mukhopadhyay,Debaditya Roy,Partha Pratim Das
类目: Computer Vision and Pattern Recognition (cs.CV); Multimedia (cs.MM)
备注: Accepted for publication in ACM Journal on Computing and Cultural Heritage (JOCCH) Special Issue on Visual Heritage
Abstract:Automatic generation of hand gestures is essential for the transmission of Indian classical dance and critical for its preservation. Indian classical dance gesture datasets are inherently low-resource, and the canonical Sanskrit definitions of many mudras lack precise textual descriptions, limiting the effectiveness of conventional text-conditioned image generation models. We present \textbfMudraGen, a conditional diffusion framework that synthesizes realistic RGB images of \textitSamyukta Hasta Mudras – interactive two-hand gestures from Bharatanatyam (an Indian classical dance form). Unlike prior work on simple hand signs or single-hand gestures, MudraGen introduces geometry-aware supervision to capture the precise coordination, anatomical validity, and cultural nuance of interacting hands. We formulate three geometry-aware objectives: Keypoint Loss for 3D joint alignment, Joint Offset Loss for inter-hand spatial coherence, and Shape Consistency, which serves as an anatomical regularizer by encouraging consistent hand morphology while allowing independent hand poses. Together, these objectives guide the diffusion model toward anatomically plausible and well-coordinated hand configurations, enabling the synthesis of photorealistic and pose-accurate gesture images. Experimental results show that MudraGen surpasses existing state-of-the-art generative approaches in visual realism, anatomical correctness, and preservation of fine hand-pose structure, enabling faithful reproduction of complex Samyukta Hasta mudras. Beyond quantitative gains, its ability to generate culturally grounded and structurally consistent gestures highlights practical applications in cultural preservation and dance education.
[CV-72] Neural-Collapse-guided Task-Free Continual Anomaly Detection
链接: https://arxiv.org/abs/2609.03406
作者: Xiaotong Kong,Chaoyang Song,Ziai Zhou,Jinxia Zhang,Kanjian Zhang,Haikun Wei
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Recent years have witnessed growing interest in continual anomaly detection for industrial visual inspection. However, real-world manufacturing environments exhibit unpredictable shifts in data distributions, rendering task-dependent continual learning assumptions impractical. To address this limitation, we formulate industrial anomaly detection as a task-free continual learning problem and propose NC-TFAD, a neural-collapse-inspired, geometry-driven framework for learning from non-stationary data streams without task boundaries. NC-TFAD freezes a pretrained backbone and aligns streaming features to a simplex Equiangular Tight Frame (ETF) prototype space to stabilize representation geometry under non-stationary streams. To satisfy the NC-inspired geometric construction in the absence of real anomalies, we generate synthetic anomaly samples as auxiliary anchors during training. Building on this geometry, we further introduce inter- and intra-class regularization together with a Focal Neural Collapse Contrastive (FNCC) loss to suppress representation drift and improve normal-anomaly separability. Finally, a normal-patch-prototype-guided localization branch constructs calibrated patch-wise deviation maps from normal training samples and fuses them with a weak self-attention prior, producing anomaly heatmaps without pixel-level annotations. Extensive experiments on MVTec AD and VisA show that NC-TFAD consistently outperforms representative task-free continual learning methods adapted from general vision, as well as unified anomaly detection baselines, in both image-level detection and pixel-level localization under the task-free continual learning protocol. These results highlight that geometry-driven modeling offers an effective and robust solution for task-free continual anomaly detection in real-world industrial applications.
[CV-73] Exploring the Potential of Contrastive Language-Image Pre-training for Multi-Source Remote Sensing Data AAAI2027
链接: https://arxiv.org/abs/2609.03391
作者: Xiangyang Miao,Kelu Yao,Yekai Huang,Xiaogang Xu,Junxiao Xue,Minjun Shen,Chenghui Lv,Shanji Liu,Yaying Chen,Chao Li
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 9 pages, 4 figures, 5 tables. Submitted to AAAI 2027
Abstract:Contrastive language-image learning (CLIP) has become a key paradigm for remote sensing vision-language understanding. However, existing remote sensing contrastive learning methods are mostly built on RGB-oriented CLIP architectures, making it difficult to exploit heterogeneous sensors such as SAR, multi-spectral imaging (MSI), and hyperspectral imaging (HSI). To address this limitation, we propose OmniRSCLIP, an end-to-end contrastive learning framework that supports multi-source sensor inputs for remote sensing vision-language modeling. The key idea is to extend CLIP beyond its fixed RGB input interface without breaking the pretrained visual knowledge. To this end, OmniRSCLIP introduces Spectral-Spatial Basis Decomposition (SSBD), which formulates arbitrary-channel adaptation as a basis recomposition problem: pretrained CLIP patch embeddings provide transferable spatial bases, while wavelength-conditioned coefficients span sensor-specific embedding kernels within a constrained visual prior space. This design avoids forcing heterogeneous sensors into a fixed-channel input space, while aligning them in a unified image-text semantic space. We further introduce a spectral-context-aware mask-based contrastive learning scheme to suppress modality-specific redundant features and enhance fine-grained image-text alignment. Finally, to support multi-modal training, we construct OmniRS5M, the first large-scale remote sensing image-text corpus covering RGB, SAR, MSI, and HSI. Experiments on retrieval, zero-shot classification, and semantic localization show that OmniRSCLIP preserves strong RGB-domain performance while effectively extending CLIP to heterogeneous remote sensing modalities.
[CV-74] FoRIS: Progressive Foreground Refinement for Training-Free In-Context Segmentation
链接: https://arxiv.org/abs/2609.03384
作者: Ming Hu,Jianfu Yin,Mingyu Dou,Miaomiao Zhang,Yao Wang,Cong Hu,Bingliang Hu,Quan Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:In-Context Segmentation (ICS) aims to precisely segment arbitrary semantic concepts, such as objects or parts, given one or a few annotated visual exemplars. In this paper, we revisit ICS from a more classical segmentation perspective, viewing it as a coarse-to-fine progressive refinement process. Rather than directly predicting the final mask through reference-query matching, we progressively refine the segmentation from coarse and ambiguous foreground responses to precise and complete foreground structures. Building upon this perspective, we propose a training-free in-context segmentation framework, termed FoRIS. Specifically, FoRIS consists of three key stages: Foreground Purification, Foreground Localization, and Foreground Consolidation, which progressively suppress background distractions, localize discriminative target regions, and recover complete foreground structures through semantic aggregation. Experimental results demonstrate that FoRIS achieves SOTA performance across semantic and part segmentation tasks, with average improvements of 4.5 and 4.8 mIoU points over existing approaches in the 1-shot and 5-shot settings, respectively. Code: this https URL.
[CV-75] When Depth Hurts: Reliability-Aware Geometry Distillation for Depth-Free RGB-D Salient Object Detection
链接: https://arxiv.org/abs/2609.03378
作者: Xuehao Wang,Jiaxin Hua,Runmei Li,Zhenyu Wu,Chenglizhao Chen,Ke Gu,Aimin Hao
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Depth can resolve appearance ambiguity in RGB-D salient object detection (SOD), yet sensor depth is not uniformly reliable. Missing regions, blurred boundaries, and structural artifacts can propagate through multimodal fusion and make an RGB-D detector less accurate than its RGB-only counterpart. Existing quality-aware approaches regulate observed depth but remain dependent on the same potentially defective modality. We propose \method, a reliability-aware geometry distillation framework developed for RGB-D SOD benchmarks without using dataset-provided depth during training or inference. A frozen Depth Anything V2 model serves only as a training-time teacher, transferring dense relative geometry, hierarchical spatial attention, and boundary structure to a compact edge-aware geometry branch. Pooled bidirectional interaction aligns geometry with appearance, and a pixel-wise reliability estimator selectively injects geometry that is compatible with the current RGB representation. The teacher is removed after training, leaving an RGB-only inference network. Trained on 2,985 RGB-mask pairs, \method achieves the best or tied-best result in 26 of 36 metric-dataset comparisons against ten recent RGB-D SOD methods, including a 13.4% relative MAE reduction on ReDWeb-S. When retrained on DUTS-TR, it also improves the strongest prior F -measure by 4.2% on PASCAL-S, showing that the distilled geometry transfers beyond a particular sensor or dataset domain. Code will be released upon publication.
[CV-76] P-CORE: Self-Supervised Surface Consistency for Point-Based Neural Editing ECCV2026
链接: https://arxiv.org/abs/2609.03349
作者: Yanshu Zhang,Shichong Peng,Mehran Aghabozorgi,Alireza Moazeni,Ke Li
类目: Computer Vision and Pattern Recognition (cs.CV); Graphics (cs.GR)
备注: Accepted to ECCV 2026. Project Page: this https URL
Abstract:Advances in neural rendering have enabled high-fidelity multi-view reconstruction of 3D scenes. However, free-form non-rigid shape editing remains a significant challenge. Point-based neural representations are highly desirable for multi-view reconstruction because they lack fixed connectivity, which does not constrain the learned surface topology to that of the initialization. Yet this same property causes point-based representations to struggle with holes and surface discontinuities under large deformations. To address this, we propose a novel self-supervised method to enable point-based representations to adapt to large deformations without requiring ground truth multi-view images of deformed geometry. The key idea is to generate random deformations and to ensure consistency in the predicted surface before and after deformation. In particular, the surface prediction from the deformed point cloud should be the same as the deformation applied to the surface prediction from the original point cloud. We incorporate our approach into attention-based point representations, which differ from splatting-based point representations in their use of a learned interpolation kernel between points as opposed to a Gaussian kernel around each point. This learned interpolation kernel can learn to adapt to large deformations, without requiring addition or removal of points. We show that our framework significantly enhances its robustness to large deformations. Experiments on synthetic geometry editing benchmarks (Neural Editor, Objaverse) demonstrate that our approach outperforms existing point-based methods in zero-shot editing and significantly reduces artifacts. Furthermore, qualitative results on the DTU and Mip-NeRF 360 datasets demonstrate our method’s effectiveness on real-world scenes.
[CV-77] PointGT: Simultaneous Geometry and Texture Editing for Point-Based Representations ECCV2026
链接: https://arxiv.org/abs/2609.03341
作者: Yanshu Zhang,George Shramko,Pratul P. Srinivasan,Ke Li
类目: Computer Vision and Pattern Recognition (cs.CV); Graphics (cs.GR)
备注: Accepted to ECCV 2026. Project page: this https URL
Abstract:We present PointGT, a point-based 3D representation that enables simultaneous editing of object geometry and appearance. Existing reconstruction and view synthesis techniques produce volumetric 3D representations that are high-quality and photorealistic, but are difficult to edit. In particular, recent efforts to enable texture editing for 3D Gaussian Splatting representations are not compatible with geometry edits and deformations. Our method combines a point-based representation that is well-suited for geometry deformations with a learned UV mapping technique that enables high-resolution texture editing. We show that PointGT enables fine-grained editing of both geometry and texture in point-based neural representations with high rendering quality.
[CV-78] Laplacian Frequency Hierarchies for Efficient 3D Gaussian Splatting Training
链接: https://arxiv.org/abs/2609.03334
作者: Yixiong Yang,Sisheng Zhang,Qingsong Yan,Shaohuai Shi,Qiang Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to Pacific Graphics 2026 (conference track). Project page: this https URL
Abstract:A key bottleneck in 3D Gaussian Splatting training is the continual growth of Gaussian primitives, which increases optimization cost and slows convergence, especially at high resolutions. We propose Laplacian Frequency Hierarchies, a simple yet efficient 3DGS scheme that combines Laplacian image decomposition with coarse-to-fine, frequency-staged training. After fitting lower-frequency structure, we archive the corresponding Gaussian field so that subsequent fields can optimize higher-frequency residuals without carrying the full primitive burden, and we compose the rendered components in the image domain via a Laplacian-style reconstruction at inference time. This design reduces the number of active Gaussians during training, thereby lowering optimization overhead and accelerating training. The proposed scheme is plug-and-play and orthogonal to prior 3DGS accelerations: it can be directly combined with strong backbones such as Taming-3DGS and FastGS to improve training speed with competitive reconstruction quality. It achieves average speedups of 1.73x and 1.21x at 1K setting, and 1.74x and 1.33x at 4K setting on Taming-3DGS and FastGS, with larger gains on more challenging scenes and increasingly pronounced benefits at higher resolutions.
[CV-79] nsor-based Brain Surface Modeling and Analysis
链接: https://arxiv.org/abs/2609.03302
作者: Moo K. Chung,Keith J. Worsley,Steve Robbins,Alan C. Evans
类目: Computer Vision and Pattern Recognition (cs.CV); Neurons and Cognition (q-bio.NC)
备注:
Abstract:We present a unified computational approach to tensor-based morphometry in detecting the brain surface shape differences between two clinical groups based on magnetic resonance images. Our approach is novel in a sense that we combined surface modeling, surface data smoothing and statistical analysis in a coherent unified mathematical framework. The cerebral cortex has the topology of a 2D highly convoluted sheet. Between two different clinical groups, the local surface area and curvature of the cortex may differ. It is highly likely that such surface shape differences are not uniform over the whole cortex. By computing how such surface metrics differ, the regions of the most rapid structural differences can be localized. To increase the signal to noise ratio, diffusion smoothing based on the explicit estimation of Laplace-Beltrami operator has been developed and applied to the surface metrics. As an illustration, we demonstrate how this new tensor-based surface morphometry can be applied in localizing the cortical regions of the gray matter tissue growth and loss in the brain images longitudinally collected in the group of children.
[CV-80] An Ensemble-Based Self-Taught Learning Approach for Parking Space Classification Under Limited Data
链接: https://arxiv.org/abs/2609.03258
作者: Lucas de Oliveira Cunha,Joelton Deonei Gotz,Paulo Lisboa de Almeida,Andre Gustavo Hochuli
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Parking spot classification is a fundamental task in intelligent transportation systems, yet most deep learning approaches rely on large amounts of annotated data and exhibit limited generalization across heterogeneous environments. To address these limitations, we investigate a self-taught learning framework based on unsupervised representation learning with convolutional autoencoders. The proposed approach learns transferable visual representations from unlabeled data and reuses the learned encoders as fixed feature extractors for supervised classification with limited annotated samples in the target domain. To further enhance robustness and mitigate architectural bias, an ensemble of heterogeneous autoencoders is employed, with independent classifier heads and prediction fusion at inference time. Experiments conducted on the PKLot and CNRPark benchmarks under cross-dataset evaluation protocols show that the proposed ensemble-based strategy substantially reduces annotation requirements while improving robustness under significant domain shifts, achieving accuracies between 93% and 96% in data-constrained scenarios.
[CV-81] Counting Animals in Camera-Traps Image Sequences without Count Labels: Winning Solution to the iWildCam 2021 Challenge
链接: https://arxiv.org/abs/2609.03233
作者: Fagner Cunha,Juan G. Colonna,Eulanda M. dos Santos
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Camera traps have become an essential tool for wildlife monitoring, motivating the development of computer vision methods for the automated extraction of information from these data. While most prior work has focused on species identification, many ecological applications also require estimating the number of unique individuals appearing across short image sequences. This task is particularly challenging because camera traps typically acquire bursts of images at approximately one frame per second, creating large temporal discontinuities that may make conventional multi-object tracking methods unreliable, and because manually collecting individual count annotations is prohibitively expensive. In this work, we describe the winning solution to the iWildCam 2021 Challenge, which introduced a benchmark for counting animals at the sequence level under realistic annotation constraints where count annotations are unavailable for training. Our approach, MaxBoxCount, combines a strong species classification pipeline with a simple yet effective counting heuristic based on MegaDetector detections to estimate the number of unique individuals without requiring count annotations. Code is available at this https URL.
[CV-82] ProgResViT: Progressive Resolution and Width for Adaptive Vision Transformers
链接: https://arxiv.org/abs/2609.03216
作者: Ali Hojjat,Janek Haberer,Olaf Landsiedel
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Vision Transformers (ViTs) typically process every image using a fixed input resolution and model width, even though many images can be classified with substantially less computation. We introduce ProgResViT, an input-adaptive ViT that performs inference progressively across multiple rounds. The first round processes a low-resolution image with a narrow subnetwork. Inference terminates when the prediction is sufficiently confident; otherwise, the model reuses the representations produced in the current round and proceeds with a higher-resolution input and a wider subnetwork to refine its prediction. As all rounds share a single backbone, we propose Progress-Conditioned Soft Gating (PSG), which conditions token fusion and layer outputs on the current round, block, and input resolution. On image classification, applying ProgResViT to DeiT yields better accuracy-compute trade-offs than adaptive-width, adaptive-depth, and dynamic-token baselines. With knowledge distillation, a DeiT-based ProgResViT achieves 84.9% top-1 accuracy, slightly exceeding the reported DeiT-III-S accuracy under a comparable evaluation setting. We show that the same design also provides favorable accuracy-compute trade-offs for self-supervised DINO representations and downstream semantic segmentation. Code is available at this https URL.
[CV-83] RoboTok: An Internet-Scale Data Engine for Human Demonstration Retrieval and Dexterous Manipulation Learning
链接: https://arxiv.org/abs/2609.03199
作者: Howard Qian,Yiting Chen,Yunfei Xie,Kejia Ren,Podshara Chanrungmaneekul,Gaotian Wang,Bowen Wen,Chen Wei,Kaiyu Hang
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注:
Abstract:Robot learning increasingly depends on broad and diverse demonstrations, yet collecting robot data remains expensive and poorly suited to covering the long tail of real-world tasks. To address this bottleneck, we introduce RoboTok, an internet-scale data engine that, given a query human manipulation video, retrieves manipulation-relevant human demonstrations from web videos for training dexterous robot policies. Specifically, we learn a latent motion space from 3D hand trajectories expressed in estimated actor-centered reference frames. This representation enables manipulation behaviors to be compared across variations in camera viewpoint, scene appearance, and actor occlusions, while remaining compact enough for efficient search and continual indexing over internet-scale video collections. We evaluate RoboTok against existing robot-data retrieval approaches on retrieval benchmarks and downstream robot policy performance. Our results show that RoboTok retrieves more relevant manipulation demonstrations and improves downstream task success, establishing hand-pose trajectory-aware retrieval as a way to make web video a scalable and continuously growing source of supervision for robot learning.
[CV-84] VeriPhy: Agent ic Physical Reasoning for World Model Evaluation and Refinement
链接: https://arxiv.org/abs/2609.03153
作者: Wenzhuo Xu,Yuchen Zhu,Chongjian Ge,Xuan Shen,Jing Shi,Jason Kuen,Yongxin Chen,Molei Tao,Christopher McComb,Noelia Grande Gutiérrez,Jiuxiang Gu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Visual fluency in generated video does not imply physical reliability, and a scalar quality score alone is incapable of indicating the obligation a clip violates or the moment it fails. We present VeriPhy, an auditable physical-verification system in which a text-only planner compiles the prompt into typed physical obligations and a statically validated execution plan before any frame is observed. During execution, observations gate and scope only declared calls to frozen low-level experts (e.g., segmentation and tracking, counting, eleven typed physical measurements over the resulting tracks, depth, OCR, and audio-event detection). Each action returns a provenance-carrying evidence record whose payload, when usable, is either a typed measurement or an explicitly tagged learned state. Typed resolvers and fixed composition map usable records to a three-valued state (supported, contradicted, or unknown, surfaced as plausible, implausible, or abstain) with full provenance, so that every verdict is traceable to the evidence that produced it. We anchor evaluation in a 1,500-clip corpus of human-annotated flaw records that localize real generation failures in prompt reference, space, and time. On a 149-clip core carrying 304 such records, VeriPhy accounts for 228, against 164 for a published question-decomposition evaluator given the same clips and the same claims. Recall alone does not separate it from prompting the same backbone monolithically, which reaches 222; what separates them is that each decision retains its evidence record and provenance, making the traces auditable one verdict at a time and usable as the interface through which a critic verdict could be written back into generation.
[CV-85] Sensing Which Modality Matters: Evidence-Gated Regularization for Robust VLA Policies
链接: https://arxiv.org/abs/2609.03142
作者: Yue Yang,Diego Romeres,Chiori Hori,Gedas Bertasius,Daniel Szafir,Siddarth Jain
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:Vision-Language-Action (VLA) policies fuse multimodal sensory inputs, but training on limited and homogeneous robot demonstrations encourages spurious inter-sensor correlations rather than task-relevant signal, a failure we term modality entanglement. Under real-world occlusions and distractors, this manifests as nuisance sensitivity to corruption of uninformative sensors and single-modality insufficiency when only one informative sensor remains intact. We propose Evidence-Gated Regularization (EGR), a modality-agnostic training objective that introduces zero inference-time overhead. EGR derives a per-frame and per-sensor task-relevance signal to gate two state-conditional consistency objectives: invariance on low-evidence sensors, and single-sensor sufficiency on high-evidence ones. We introduce a benchmark based on BEHAVIOR-1K, comprising a fast inference-only diagnostic suite and 47 rollout-based skills targeting modality entanglement. We validate EGR on this benchmark and on two real-robot setups with fundamentally different embodiments: a bi-manual setup with two Kinova arms and three RGB cameras, and a single-arm MELFA ASSISTA setup combining vision and GelSight tactile sensors. EGR improves simulation success rates (SR) from 12.5% to 16.4% under full modalities (+31%), from 9.4% to 16.5% under uninformative-sensor corruption (+75%), and from 2.8% to 6.1% under single-sensor fallback (+120%). Under physical-object distractors, EGR boosts SR from 30% to 85% on the bi-manual setup (+183%) and from 55% to 70% on the tactile setup (+27%).
[CV-86] Beyond Small Patches: Black-Box Detection and Purification of Diverse Backdoor Triggers
链接: https://arxiv.org/abs/2609.03139
作者: Ahmed Abdelnaby,Mohamed Elmahallawy
类目: Computer Vision and Pattern Recognition (cs.CV); Cryptography and Security (cs.CR)
备注:
Abstract:Deep neural networks (DNNs) are increasingly deployed in real-world vision systems, yet their predictions can be covertly manipulated by backdoor attacks, in which malicious triggers cause targeted misclassification while preserving high clean accuracy. Existing defenses often rely on model internals, training data, or clean validation samples, making them difficult to deploy when only black-box access to a trained model is available. We propose TRIM (Trigger Removal by Identifying Manipulated Regions), a deployment-oriented black-box defense that detects and selectively removes backdoor triggers at inference time without requiring model internals, training data, or clean samples. The key insight behind TRIM is to identify image regions that are responsible for anomalous model behavior and purify only those regions while preserving benign content. TRIM innovates via three key components: (i) region-based segmentation with deep feature representations, (ii) adaptive trigger discovery through inpainting and diffusion-based reconstruction to isolate regions responsible for misclassification—without assumptions about trigger type, shape, or location, and (iii) selective region purification that cleans poisoned regions while retaining benign content. To support practical deployment, TRIM further caches feature embeddings of previously identified triggers, enabling efficient recognition and avoiding redundant detection and purification. Extensive experiments across diverse datasets and backdoor types, including blended, sparse, varying-size, and multiple triggers, show that TRIM consistently outperforms existing black-box defenses, reducing attack success rates (ASR) to as low as 1.16% while preserving clean accuracy of up to 87.87%. These results demonstrate that effective backdoor mitigation is possible at inference time even when the defender has no access to any auxiliary data.
[CV-87] Kernel Reboot: Breaking the Boundaries of Neural Tangent Kernels for Neural Fields DATE
链接: https://arxiv.org/abs/2609.03117
作者: Amir Mallak,Alaa Maalouf,Lior Wolf,Daniela Rus,Dan Rosenbaum
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注: Published in IEEE TPAMI, vol. 48, no. 9, pp. 10940-10957, Sep. 2026. Author version adds related-work references and biography updates; Figures 12 and 13 were regenerated from the same locked hyperparameter sweep. Tabulated results, reported best points, scientific claims, and conclusions are unchanged
Abstract:Neural fields (NFs) map continuous coordinates to signals such as color or density, but fast high-quality reconstruction from sparse observations remains difficult. Classical Neural Tangent Kernel (NTK) regression gives closed-form fits, yet it is fundamentally linear and cannot accumulate reusable task priors. We develop three algorithms that address these gaps. NTK-KIP learns a distilled support set of coordinates (and optional labels) so that a finite NTK can inpaint large missing regions from little observed data, yielding a compact non-linear representation instead of a raw kernel solve. MetaQuill meta-learns a shared initialization for an INR so that new scenes can be adapted by updating only a small task-specific weight offset, which provides true feature learning and a reusable prior. Finally, MetaQuill-KIP fuses both ideas: it seeds the task with a KIP-style non-linear warm start, then refines only that small offset around the meta-learned initialization. MetaQuill-KIP achieves high-PSNR reconstructions and semantically plausible inpainting under very sparse observations, while requiring only lightweight per-instance adaptation, whereas diffusion-style baselines typically depend on large pretrained generative priors and costly per-image tuning. This shows that NTK-driven neural fields can be made both non-linear and meta-learnable, narrowing the gap between analytic kernels and practical few-shot reconstruction.
[CV-88] SLIDEFORGE: An LLM Agent for Controllable Editing of Slides as Structured Artifacts EMNLP2026
链接: https://arxiv.org/abs/2609.03109
作者: Haozhen Zheng,Fulin Wang,Tianhu Xiong,Yingjie Yu,Shengyi Qian,Hanchao Yu,Alex Schwing,Klara Nahrstedt,Mingyuan Wu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: EMNLP 2026 Main Conference. Haozhen and Fulin contributed equally
Abstract:Current AI agents compellingly describe slides. However, AI-assisted slide editing requires more than understanding: the output must retain layout, style, component structure, and native editability. Towards, AI-assisted slide editing, existing agents operate on screenshots or weak document representations and often fragment coherent visual units, rasterize editable content, or break layout. In contrast, for controllable slide editing, we introduce an agentic framework, SLIDEFORGE, which builds a Deck State Graph, an executable slide state that links visual decomposition, native pptx object structure, and perceptual organization. By recovering human-referable components while retaining fine-grained editable structure, SLIDEFORGE supports theme-preserving reconstruction through slide-native operations and rendered-state verification. We further introduce an evaluation paradigm for controllable slide transformation that jointly measures component recovery, preservation, restyling consistency, visual quality, and native editability. Experiments show that SLIDEFORGE outperforms direct prompting, screenshot-based agents, and generic code-agent baselines across these dimensions. Code is available at this https URL.
[CV-89] WireSeg-32K: A Physics-Grounded Synthetic Dataset for Wire Instance Segmentation CVPR2026
链接: https://arxiv.org/abs/2609.03102
作者: Zilin Dai,Lehong Wang,Yi Yang,Xiang Fei
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Synthetic Data for Computer Vision @ CVPR 2026 Workshop
Abstract:Deformable linear objects such as wires and cables are difficult to segment because they are thin, highly deformable, and frequently self-occluded, while large-scale instance-level annotations are expensive to obtain in real scenes. Existing resources either focus on cable tracing or semantic segmentation under constrained settings, or generate visually plausible images without physically grounded wire deformation. We present WireSeg-32k, a synthetic dataset for wire instance segmentation with 32,000 RGB images, instance masks, depth maps, and a complementary real-world test set with annotations. To generate this dataset, we develop DeformX, a co-simulation pipeline that couples Cosserat-rod dynamics with photorealistic Isaac Sim rendering, enabling physically plausible, contact-consistent wire shapes, CAD-based wire assets, and diverse visually grounded scenes. As a simple baseline, LoRA fine-tuning SAM3 on WireSeg-32k alone improves real-world mAP@75 by 10.2% over the off-the-shelf model, showing that physically grounded synthetic data can transfer to real wire perception.
[CV-90] Solving the Needle-in-a-Haystack Problem in Mammography Vision-Language Model with Differentiable Subset Sampling
链接: https://arxiv.org/abs/2609.03085
作者: Young Seok Jeon,Beatrice Brown-Mulry,Rohan Satya Isaac,Anjana Dissanayaka,Theo Dapamede,Mohammadreza Chavoshi,Judy Gichoya,Hari Trivedi
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:There is growing interest in adopting CLIP-style vision–language model (VLM) pretraining for mammography. However, models that directly employ the standard CLIP architecture and training objective exhibit limited zero-shot performance in clinically important tasks such as cancer, finding-type, and BI-RADS predictions. We argue that this underwhelming performance is due to neglecting two characteristics of mammography data: (1) its high-res nature, and (2) homogeneity of radiology reports, largely driven by a predominance of negative/benign findings on examinations. We propose TopKSigLIP, a VLM designed to address these two limitations through a novel architecture and learning objectives. Instead of downscaling high-res mammography images to satisfy GPU memory constraints, TopKSigLIP introduces TopK-Patch module that learns to sample a sparse set of high-res patches likely to contain lesions, sidestepping the resolution–batch size tradeoff of VLM training. The sampled patch locations additionally serve as a built-in localization tool. To address report homogeneity, we replace the contrastive loss, which falsely repels semantically similar pairs, with a Sup-sigmoid loss. Sup-sigmoid loss extends the sigmoid loss from SigLIP with soft labels derived from structured data. TopKSigLIP outperforms existing open-source mammography and general medical VLMs on both internal and external benchmarks on density assessment, BI-RADS classification, finding subtyping, and cancer prediction under zero-shot evaluation. TopKSigLIP remains competitive under linear probing despite using a significantly smaller vision encoder and smaller training batches than baselines. The TopK-Patch module additionally achieves superior lesion localization over post-hoc Grad-CAM. Code and weights are made public:this https URL.
[CV-91] Exemplar: Classical Priors Complement Frozen Features for Few-Shot Microscopy Segmentation at Native Resolution
链接: https://arxiv.org/abs/2609.03080
作者: Michal Průšek,Adam Novozámský,Filip Šroubek
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 5 pages, 3 figures, 2 tables. Code, configurations, and the per-image score records behind every reported number: this https URL
Abstract:Segmenting a new biomedical dataset usually means a domain-specific model trained on substantial annotation, or a foundation model steered at inference time. We present Exemplar, a few-shot segmenter that fuses a frozen DINOv3 backbone with a fixed bank of classical native-resolution filter responses in one lightweight head, fitted from the support masks alone. In the few-mask, native-resolution regime, classical priors and frozen self-supervised features are complementary: fused in one head, a single fixed configuration spans eleven biomedical imaging datasets. Under the same head, the classical bank alone reaches 0.693 on the eleven-dataset panel, scored by foreground intersection-over-union or centreline Dice, and the frozen features alone 0.672; the bank leads on seven of the eleven and the features on the rest, and fused they reach 0.782. Against five forward-pass few-shot methods, Exemplar leads in 54 of 55 method-dataset comparisons, 52 of them significant after Holm correction. From a single annotated mask it reaches 0.703 on the same panel, against 0.682 for a from-scratch nnU-Net trained on that same mask. At eight masks nnU-Net overtakes it on the panel mean, chiefly on centreline agreement, but takes 16-77x longer to fit.
[CV-92] Position: Unlabeled IS NOT Equal to No Human Supervision in Visual Learning ICML2026
链接: https://arxiv.org/abs/2609.03077
作者: Dong Lao
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: ICML 2026
Abstract:This position paper argues that the absence of labels does not imply the absence of human supervision in visual learning, and urges the research community to identify sources of supervision more explicitly. Many recent methods in computer vision build upon representations learned from large-scale unlabeled data, and are therefore grouped under the same umbrella term unsupervised.'' However, different data curation schemes and training objectives embed substantially different human priors on which models rely, and we argue that one unsupervised’’ umbrella term is no longer capturing these distinctions. This ambiguity makes it harder to compare unsupervised learning research conducted under different assumptions, coinciding with a sharp decline in papers titled with ``unsupervised’’ in flagship computer vision conferences since 2021, despite continued growth of the field. While we fully embrace pre-training as a strong foundation for modern computer vision, we advocate for a community-level effort toward greater conceptual clarity: authors are encouraged to disclose priors in data selection and learning objectives, and to specify which components of a learning pipeline depend on which assumptions. Standardized disclosure practices can improve academic communication, ensure fairer comparisons, and preserve methodological diversity in unsupervised learning.
[CV-93] IDSPACE: A Novel Document Generator for Reliable Evaluation of Digital Identity Verification Systems [Extended Technical Report]
链接: https://arxiv.org/abs/2609.03052
作者: Lulu Xie,Yancheng Wang,Kanchan Chowdhury,Rolando Garcia,Yingzhen Yang,Jia Zou
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:As services move online, trust institutions such as banks, lenders, and governments must verify the identity of remote users. Fraud detection tools are widely available, but evaluating and fine-tuning them remains difficult because identity documents are sensitive and therefore scarce. Synthetic data generation offers a path forward, and demand is clear: our prior work in this area has been downloaded over 11,000 times (aggregated from eight parts). We introduce IDSpace, extending this line of research in three directions. First, we propose model-guided Bayesian optimization, which tunes generation parameters to maximize both visual similarity and prediction consistency with target-domain models given only a few samples from a target domain. Second, we decouple user-specified metadata (demographics, fraud patterns, capture device) from automatically tuned control parameters (font styles, noise levels, image quality), allowing users to configure evaluations without low-level expertise. Third, we expand beyond template images to support scanned and mobile-captured documents. Experiments show IDSpace improves evaluation consistency by 15-45% over baselines including CycleGAN, diffusion inpainting, and non-guided optimization, using only a few real samples, while improving training accuracy by up to 9% and SSIM similarity with the target domain by 10% . We also released a new dataset consisting of 359,240 high-quality synthetic documents across ten European ID types.
[CV-94] Improving Clinical Target Volume Segmentation Accuracy using Anatomical Priors and Active Learning for the AGITG TOPGEAR Clinical Trial
链接: https://arxiv.org/abs/2609.03186
作者: Phillip Chlap,Mark Lee,Trevor Leong,Matthew Field,Jason Dowling,Hang Min,Julie Chu,Jennifer Tan,Phillip K. Tran,Tomas Kron,Annette Haworth,Martin A. Ebert,Shalini K. Vinod,Lois Holloway
类目: Medical Physics (physics.med-ph); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Training deep learning-based medical image segmentation models is challenging with limited curated datasets. For AGITG TOPGEAR, a gastric cancer trial, the Clinical Target Volume (CTV) is complex and defined by multiple anatomical landmarks, making upfront training data preparation difficult for an automated contour QA segmentation model. We investigate anatomical priors, derived from surrounding organ segmentations, to provide spatial context and improve TOPGEAR CTV segmentation accuracy. We also evaluate active learning, iteratively expanding the training dataset by selecting cases expected to improve performance. One hundred TOPGEAR CT scans were retrospectively analyzed. An initial set of 10 expert-contoured cases was used to train an nnU-Net model. TotalSegmentator generated a voxel-wise anatomical prior map from surrounding structures as an additional input channel. Active learning was simulated over four iterations, selecting cases by model uncertainty and segmentation performance. All models used five-fold cross-validation for an ensemble uncertainty measure. Evaluation used a hold-out testing set of 50 cases. The anatomical prior improved CTV segmentation accuracy, increasing mean Dice Similarity Coefficient (DSC) from 0.84 to 0.86. Active learning similarly improved performance to 0.86, with greatest benefit in the final round. Combining the anatomical prior with active learning achieved the highest accuracy, with a DSC of 0.87. Model uncertainty correlated with DSC, supporting its use in identifying suboptimal predictions and guiding active learning. Anatomical priors and active learning each improved CTV segmentation accuracy and generalizability, with their combination achieving the best performance, supporting integration into segmentation model development for automated contour QA in radiotherapy clinical trials. Subjects: Medical Physics (physics.med-ph); Computer Vision and Pattern Recognition (cs.CV) Cite as: arXiv:2609.03186 [physics.med-ph] (or arXiv:2609.03186v1 [physics.med-ph] for this version) https://doi.org/10.48550/arXiv.2609.03186 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Phillip Chlap [view email] [v1] Wed, 2 Sep 2026 22:03:11 UTC (1,816 KB)
[CV-95] Learning from Scarce Labels: Multi-View Echocardiography for Ejection Fraction Prediction
链接: https://arxiv.org/abs/2609.02969
作者: Zhiyuan Gao,Dominic Yurk,Yaser S. Abu-Mostafa
类目: Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: Accepted for publication at the Journal of Machine Learning for Biomedical Imaging (MELBA) this https URL
Abstract:We present, to the best of our knowledge, the first publicly available resource for predicting left ventricular ejection fraction (EF) from parasternal long-axis (PLAX) echocardiography. Because no PLAX-EF datasets previously existed, our work focuses on an innovative data generation strategy to overcome this scarcity. By leveraging a time-based correlation between clinical notes and echocardiographic videos, combined with fine-tuning view classifiers and proxy labeling, we created a labeled dataset of over 25,000 PLAX videos. This enables us to train the first reproducible PLAX EF model, achieving a mean absolute error (MAE) of 6.86%. Given that apical four-chamber (A4C) methods, the clinical standard, report MAE values of 6%-7%, our results demonstrate that EF estimation from PLAX views is both feasible and clinically relevant. This surpasses the performance of existing methods and provides a clinically relevant solution for situations where apical views may not be feasible. Going further, we demonstrate that combining PLAX and A4C predictions via simple unweighted late fusion improves both single-view baselines to a 6.37% MAE, underscoring the value of multi-view integration. To promote continued research, we release the dataset labels, trained models, and runnable demos on GitHub, Hugging Face, and Google Colab: this https URL
人工智能
[AI-0] Clean Engineering Unstable Measurement: A Preregistered Reliability Failure of Black-Box LLM Observers on Shared Endpoints
链接: https://arxiv.org/abs/2609.04198
作者: Haoyaun Zhu,Jie Zhang
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Language-model judges now gate training data, score generations, and drive leaderboards. The judge is then a measurement instrument, resting on one rarely stated assumption: the same request, sent to the same model name, reads the same tomorrow. We audited that assumption in two preregistered campaigns with every threshold fixed in advance; neither got past validating its instrument. Across 52,988 audited request attempts, same-window repeat rankings agreed at Spearman 0.400 against a required 0.90, and byte-identical next-day replays agreed at 0.78 against a required 0.99, each time with the execution record at ceiling. Three mechanisms explain the gap: a label-to-meaning mapping that biased readouts as strongly as the signal; candidate gaps seven orders of magnitude below the instrument’s own noise floor; and byte-identical inputs returning different rankings, a noise that exact-permutation readouts compound. Neither metric substitution nor sampling repaired it on the tested grid. Preregistered follow-ups bound the problem: waiting did not help on the days sampled (0.805 versus 0.800, replicated over five further days); switching providers did not help (four providers share the floor, medians 0.74 to 0.88, predicted by none of the metadata fields they expose); self-hosting on batch-invariant kernels helped only while the server was quiet; and on constructed errors with known gaps, the readout’s separation tracks error type, not size. We distill the evidence into a three-level snapshot-identity ladder, eight design rules, and a reporting checklist; a pilot at roughly 2% of the study’s call volume would have exposed both unreachable gates in advance. All results concern externally measured behaviour on shared serving infrastructure. On a shared endpoint, a model name is not a frozen instrument; a preregistered evaluation must measure its instrument before freezing any gate on it.
[AI-1] A Computationally Feasible Framework for Causal Probabilistic Explanation
链接: https://arxiv.org/abs/2609.04177
作者: Rafal Urbaniak,Sam Witty,Daniel Waxman,Andy Zane,Poorva Garg,Emily Bunnapradist,Sankaran Vaidyanathan,Jack Feser,Drew Lehe,Eli Bingham
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Explaining why a specific outcome occurred, and which inputs deserve the blame or credit, is central to philosophical, scientific, and policy analysis. Existing tools split into two camps. The theory of actual causality (AC) gives principled verdicts, but only for toy-sized models, because computing them requires enumerating counterfactual scenarios. Scalable attribution methods like SHAP (or even causal SHAP) at least partially ignore the causal structure that generated the data, and can give answers that conflict with a careful causal analysis. We close this gap with Probabilistic Causal Impact (PCI). PCI builds on actual causality and on Pearl’s notions of probability of necessity and sufficiency, but recasts the question of explainability as an estimation problem on a probabilistic causal model that is easily approximated via Monte Carlo. By specifying a distribution over “candidate explanations,” a distribution over counterfactual values, and a scoring function, PCI provides tractable, causally grounded, graded explanations, generalizing AC and Pearl’s probability of causation as degenerate cases. We evaluate PCI in synthetic and real-world examples, spanning consistency checks with AC, scaling experiments, complex continuous-valued dynamical systems, and a real-world deployed causal machine learning model trained on millions of datapoints. Subjects: Artificial Intelligence (cs.AI) MSC classes: 68T37, 62D20 ACMclasses: I.2.4; I.2.6; G.3 Cite as: arXiv:2609.04177 [cs.AI] (or arXiv:2609.04177v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2609.04177 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-2] A Case Study on Emergent Cheating and Whistleblowing in Autonomous Research Swarms
链接: https://arxiv.org/abs/2609.04170
作者: Davide Paglieri,Logan Cross,Tim Genewein,Joel Z. Leibo,Nenad Tomasev,Alexander Sasha Vezhnevets
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Multi-agent AI science ecosystems rely on agents possessing tools that allow them to communicate, coordinate, and build on each other’s work. Yet this shared infrastructure can also introduce vulnerabilities by creating a substrate for the contagious spread of unintended and undesirable behaviors. We report a case study on a research collective of 100 autonomous LLM agents tasked with proving formal mathematical conjectures. Within the swarm, cheating spontaneously emerged and was later challenged by whistleblowers - both without any external intervention. When a single agent discovered an exploit in the evaluation system, it propagated across the collective via a shared knowledge library and later through peer-to-peer messages. Despite early reluctance, a cohort of agents adopted the exploit in response to competitive pressure. A separate group of agents produced an emergent counter-response: auditing fraudulent proofs, alerting peers across broadcast and private channels, staging boycotts, lodging formal complaints, and proposing validation patches. In recent incidents, agent swarms coordinated covertly through improvised side-channels (Dalton and Wallace, 2026; Greenblatt et al., 2026). Our setting differs: the same transparent channels that carried the exploit also gave non-cheating agents the visibility they needed to detect fraud, organize resistance, and enforce norms. We cast the problem of managing the agents’ shared infrastructure as the knowledge commons governance problem (Ostrom, 1990). To protect the commons from exploits, we propose to adopt institutional mechanisms, such as graduated sanctioning and collective-choice rules, to support decentralized self-governance in autonomous swarms.
[AI-3] SWE-Gate: Passing Functional Tests Is Not Enough for Software Engineering Agents
链接: https://arxiv.org/abs/2609.04167
作者: Xin He,Yanlin Wang,Mingwei Liu,Jiachi Chen,Hongyu Zhang,Guanbin Li
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注: 11 pages, 2 figures, 5 tables
Abstract:Repository-level software engineering benchmarks have significantly advanced the evaluation of coding agents, but existing benchmarks primarily measure whether generated patches pass functional tests and overlook review-derived acceptance constraints (review constraints) that often influence whether a patch is acceptable in real-world software development. We introduce SWE-Gate, a repository-level benchmark for software engineering agents that explicitly evaluates review constraint compliance alongside functional correctness. SWE-Gate derives review constraints from real pull request review comments and synthesizes repository-level repair instances around these constraints. Each instance provides separate functional and constraint tests, together with non-compliant and gold patches, enabling explicit separation between issue resolution capability and review constraint compliance. We construct SWE-Gate with 303 repository-level repair instances spanning 75 open-source Python repositories across diverse software domains. Experiments with four LLM backends spanning different capability levels under a common coding-agent scaffold reveal a substantial gap between functional success and success under the complete repair specification: among 644 repairs that pass the functional tests, 221 fail to satisfy the provided review constraints. These findings show that functional-only evaluation overestimates agents’ ability to satisfy the full requirements of repository-level repair tasks. The replication package including code, data, and experimental results is available at this https URL.
[AI-4] From Deceptive Outputs to Deceptive Mechanisms: A Causal Framework for Language-Model Deception Research
链接: https://arxiv.org/abs/2609.04166
作者: Yakov Pyotr Shkolnikov
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Research and news coverage of language-model deception increasingly attributes human-like mental-state concepts to language models. Such claims can blur the distinction between behavior that looks deceptive and a mechanism that is actually deceptive. We introduce a causal taxonomy separating prior commitment from retrospective report, model preference from realized output, false preference from sensitivity to the utility of misleading a recipient, and deceptive behavior from the provenance of the objective or strategy producing it. We test these distinctions in two open-weight model families. Across controlled guessing-game and stock-trading experiments, we find that deceptive-looking behavior can arise without the corresponding proposed mechanism, while other interventions provide direct evidence that recipient information state can causally affect deceptive preference. These results show that deceptive behavior can provide evidence for a deceptive mechanism. But even evidence for such a mechanism does not establish model agency in the deception. Subjects: Artificial Intelligence (cs.AI) Cite as: arXiv:2609.04166 [cs.AI] (or arXiv:2609.04166v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2609.04166 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-5] SENTINEL-RL: Offloading Topological Reasoning from LLM Agents in the Security Operations Center
链接: https://arxiv.org/abs/2609.04159
作者: Uday Vallabhaneni,Cassie L. Cagwin,David J. Wild
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:
Abstract:Large language model (LLM) agents are increasingly proposed as autonomous SOC analysts, but two limitations make them unreliable at enterprise scale: a finite context window cannot hold a multi-thousand-host authentication graph, and free-form generation offers no guarantee that a recommended containment action is consistent with the topology it operates on. We present Sentinel-RL, an agentic-SOC architecture that decouples topological reasoning from semantic reasoning: a heterogeneous graph attention encoder summarizes the live authentication subgraph into a fixed-dimensional state, a Proximal Policy Optimization (PPO) policy maps this state to a constrained set of investigative actions, and an LLM agent loop is restricted to consuming the policy’s recommendations and producing analyst-readable narratives gated by a critic. We instantiate the system on the LANL Comprehensive, Multi-Source Cyber-Security Events dataset and the Indiana University Quartz HPC cluster, reporting four results: (i) a two-phase CREATE ingestion pattern loads a 24M-edge authentication subgraph into Neo4j in 14.2 minutes on a single 32-core node, roughly 24x faster than the canonical MERGE-based pipeline; (ii) a sliding-window alert engine reliably trips a 25-event/10-second threshold in =2.5 s across 50 trials; (iii) PPO training over 200 iterations converges to a mean episodic return of 8.74+/-0.31, with held-out precision of 0.91 and recall of 0.87 on labeled red-team events; and (iv) the integrated containment loop completes a full detect-investigate-recommend-human-approve cycle in a median of 6.3 s. We contribute a reusable engineering pattern (the hot-node deadlock workaround), a portable HPC deployment pattern (anchor-node co-location), and an enterprise-readiness analysis covering false-positive economics, reversibility guarantees, audit compliance, and the human-approval boundary.
[AI-6] A Low-Cost Open Platform for End-to-End Autonomous Driving on a Miniature Ackermann Vehicle
链接: https://arxiv.org/abs/2609.04147
作者: Gustavo Claudio Karl Couto,Eric Aislan Antonelo,Gabriel George Zipperer
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Robotics (cs.RO)
备注:
Abstract:This paper presents a low-cost, open experimental platform for research in end-to-end autonomous driving with miniature Ackermann vehicles. The platform combines a physical vehicle, a printed urban track, data collection tools, trajectory registration, and a Webots digital twin, enabling controlled experiments that connect simulation-based autonomous-driving methods to real-world execution. As a first baseline, we implement command-conditioned behavior cloning, in which a neural policy receives an on-board camera image and a high-level navigation command and outputs steering and speed. The system is evaluated both on the physical vehicle and in simulation. In real closed-loop experiments, the learned policy follows lanes and executes commanded turns, reaching a mean cross-track error of 6.1 cm with respect to the reference route, close to the 4.7 cm observed in human demonstrations. In the digital twin, camera field of view has a strong effect on performance, reducing the mean cross-track error from 35.6 to 3.3 cm when widened from 58 to 120 degrees. Using the digital twin to generate synthetic driving data and a learned sim-to-real image translator to reduce the appearance gap, we further show that a higher-capacity policy trained on this synthetic data combined with real demonstrations is the only configuration that completes all four track routes in closed loop, whereas the compact baseline and the same network trained on real data alone complete fewer. These results establish the open platform as a practical testbed for sim-to-real studies and provide an initial command-conditioned imitation-learning baseline; we release it to support reproducible research.
[AI-7] Efficient Test-Time Adaptation through Human-AI Interaction
链接: https://arxiv.org/abs/2609.04141
作者: Zora Zhiruo Wang,Apurva Gandhi,Rulin Shao,Aspen Chen,Jonas Mueller,Zhiqi Liang,Jett Chen,Michael Ryan,Qianou Ma,Luxi He,Zhoujun Cheng,Andre He,Seungone Kim,Jiayi Geng,Mingqian Zheng,Weiwei Sun,Zheyuan Zhang,Xinran Zhao,Yike Wang,Abe Hou,Liwei Jiang,Pang Wei Koh,Diyi Yang,Graham Neubig,Daniel Fried
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:AI agents are trained on population-scale data to encode broad capabilities spanning those of many practitioners. Yet the artifacts they produce rarely meet the personal bar professionals need to stake their reputation on. On realistic, open-ended tasks where success criteria are heterogeneous and insufficiently documented, individual expertise lives precisely in the elevation and departure from the average. In practice, iterative human-agent interaction surfaces criteria that users cannot fully specify up front, yet apply repeatedly across tasks. We argue this cross-session interaction data is a rich, underused signal for closing the gap to individual expertise. In this work, we propose test-time adaptation through human-agent interaction (TAHI), which integrates these signals into agent context and weights, and crystallizes each user’s training and evaluation criteria via an evolving rubric module. We adapt agents to 30 individuals in two high-utility domains, writing and visual creation, on a total of 600 tasks. Our agents improve solo task success by 4.5-20.9% within only tens of tasks. Meanwhile, our evolving rubric module serves as a scalable annotation tool, creating evaluation rubrics that catch 16.0-22.3% more failures than those from LMs or humans alone. While agents are adapted towards individuals, we show these personalized agents also produce improvements in success of up to 8.8% that generalize across users.
[AI-8] he Natural Language Interaction Protocol and Standard for AI Agents
链接: https://arxiv.org/abs/2609.04135
作者: Luyi Xing,Rasit Onur Topaloglu,Ranjan Sinha,Abhay Ratnaparkhi,Samuel Ndichu,Christopher Nguyen,Anindita Das,Tom Sheffler,Mohamed Rahouti,Zichuan Li,Xiaojing Liao,Sanjay Aiyagari
类目: Artificial Intelligence (cs.AI)
备注: Accepted by ACM AI Summit 2026
Abstract:AI agents are increasingly being developed and deployed across organizations using heterogeneous agent-development frameworks, AI models, tool interfaces, protocols, and execution environments. To realize their potential social and business impact, these agents must be able to interoperate through a common communication protocol. The Natural Language Interaction Protocol (NLIP), developed by researchers and practitioners across companies and universities and standardized by Ecma International, addresses this need by defining a standards-based application-layer protocol for AI-agent interaction. NLIP provides a lightweight semantic message envelope that can be carried over existing transports such as HTTP/HTTPS, WebSocket, and AMQP, while allowing NLIP-aware agents and gateways to adapt between clients, agents, local context stores, ontologies, tools, enterprise services, and heterogeneous underlying protocols. This paper presents the motivation and design rationale of NLIP, its message model and transport bindings, security-by-design considerations, reference implementation, representative applications, adoption signals, and relationship to emerging agent protocols such as MCP and A2A.
[AI-9] Environment Evolution for Terminal Agents
链接: https://arxiv.org/abs/2609.04128
作者: Zhiyuan Fan,Tinghao Yu,Yuanjun Cai,Jiang Zhou,Jiangtao Guan,Jincheng Liu,Yun Yang,Dingxin Hu,Zhuo Han,Xing Wu,Feng Zhang,Lilin Wang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Scaling interactive and verifiable environments is critical for training terminal agents. As frontier models become more capable, environments synthesized from scratch become less challenging and thus provide limited learning signals. Recent co-evolution methods iteratively synthesize environments near the model’s learnable frontier based on weaknesses exposed during rollouts. However, their dependence on on-policy rollouts limits generalization and the continuous provision of learning signals as the model becomes stronger. In this paper, we propose environment evolution, which incrementally increases environment difficulty off-policy and schedules the evolved environments generation by generation during training to provide continuous learning signals. We derive three evolution directions that influence environment difficulty from the multi-turn learning objective and then implement evolution along these directions through a loop-engineered multi-agent harness. Quantitative rollout experiments with Hy4 preview, Claude Opus 5, and GPT-5.6 Sol show that environment evolution consistently produces more difficult environments. We validate its effectiveness on Qwen3.6-27B and Qwen3.6-35B-A3B through simple long-horizon RL training, improving their performance by 14.4 and 18.0 percentage points on Terminal-Bench 2.1, respectively.
[AI-10] Epistemic Warrant for LLM Recommendations: Characterizing the Basis for Reliance When Ground Truth Is Unavailable
链接: https://arxiv.org/abs/2609.04127
作者: Shai Vardi,João Sedoc
类目: Artificial Intelligence (cs.AI)
备注: 43 pages
Abstract:Large language models are increasingly used to support organizational decisions, yet users often lack a principled basis for assessing whether to rely on a specific recommendation. Existing approaches typically evaluate broad model properties, such as reliability, uncertainty, or robustness, or focus on user trust, rather than the underlying basis for relying on an individual recommendation. Adapting theoretical foundations from epistemology, we introduce epistemic warrant, a decision-level construct that characterizes the stability of a model’s preference and the scope over which that preference holds. We operationalize this construct through a four-tier reliance certificate for pairwise recommendations, distinguishing among unstable, context-dependent, locally supported, and broadly supported recommendations. We validate the construct using contemporary methodologies: known-groups tests successfully recover expert-prespecified warrant orderings, and stronger warrants systematically align with independent consensus from crowd workers. Furthermore, we demonstrate that epistemic warrant provides information distinct from verbalized confidence and is not readily explained by decision difficulty. Ultimately, this framework offers a theoretically grounded, implementable approach for characterizing the warrant of individual LLM recommendations when objective ground truth is unavailable.
[AI-11] Why Gated DeltaNet Survives 4-Bit Quantization: NVFP4 W4A4 for the Recurrent Half of a Hybrid 27B LLM
链接: https://arxiv.org/abs/2609.04098
作者: Sergii Kozyrev,Davyd Maiboroda
类目: Artificial Intelligence (cs.AI)
备注: 14 pages, 2 figures, 6 tables. Quantized checkpoint: this https URL
Abstract:Hybrid LLMs pair softmax attention with linear-attention layers such as Gated DeltaNet (GDN), whose recurrent state summarizes the context in fixed size. Early community 4-bit quantizations of Qwen3.8-27B (48 GDN layers, 16 attention layers) left the GDN block in 8- or 16-bit precision – especially its decay and write-strength gates – on the intuition that errors in a recurrence accumulate over long contexts. We test that intuition by building Minima: NVFP4 W4A4 on all 496 linear layers, GDN included. Across perplexity at 4K/32K, MMLU-Pro, GSM8K, AIME’25, GPQA-Diamond, LiveCodeBench, and RULER retrieval to 64K, Minima matches BF16 within seed noise (5-task average -0.52) while being the smallest (17.5 GiB) and fastest-prefill (+14-19%) recipe we compare, and its 32K perplexity gap shrinks with position. A four-part mechanism study explains why: (i) NVFP4’s 16-element block scaling localizes the residual stream’s extreme outliers, equalizing activation error across layer roles; (ii) the supposedly fragile gate projections are the least sensitive – softplus/exponential and sigmoid parameterizations compress ~11% GEMM error to ~2% output error; (iii) the delta-rule recurrence holds injected noise at a flat plateau over 32K tokens and forgets a state impulse within hundreds of steps, because each write overwrites the state along the current key direction; (iv) the per-token quantization cost washes out with context instead of compounding. We also repair a global-scale mismatch that arises when per-module-calibrated NVFP4 checkpoints are served by kernels that fuse those modules into one GEMM, and show calibrated FP8 KV-cache scales are performance-free. The result: a practical recipe – quantize everything, ship KV scales – and a mechanistic account of why the recurrent half of a hybrid LLM is the easy half to quantize. Checkpoint: this https URL
[AI-12] DRACO: Fine-Grained Credit Assignment with Dynamic Rubrics for Long-Horizon Agent Training
链接: https://arxiv.org/abs/2609.04094
作者: Shubham Gandhi,Saurabh Goyal,Kiran Kate,Yara Rizk
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Software Engineering (cs.SE)
备注:
Abstract:Reinforcement Learning from Verifiable Rewards works well when a task has a programmatic checker, but most long-horizon agent domains have none. We work in the outcome-blind setting, where ground-truth success signals are not available. Multi-criteria rubrics are a popular way to supply such a reward; they are scored once per trajectory, but a single scalar is a poor signal across tens of steps. We propose DRACO: Distributing Rubric-based Advantage for Credit Optimization. It generates rubrics dynamically during training to track the policy’s evolving capability, scores those rubrics once per completed trajectory, and redistributes that judgment over the steps responsible for annotated rubrics to produce differentiated per-step advantages in GRPO. The redistribution is closed-form and does not introduce any trained attribution module. On AppWorld, DRACO gains 15.9 points over the base model and 5.3 points over GRPO trained with a sparse ground-truth reward, despite not using any verifiers itself. On out-of-domain Tau-Bench, it gains 5.3 points over the base model even without a frontier judge, beating both ground-truth-reward training and other rubric-based training settings. The code for DRACO is available at this https URL.
[AI-13] A Non-Formulable Theorem: A Fundamental Limit of Finite Syntactic Systems and Its Consequences for Security and AI
链接: https://arxiv.org/abs/2609.04086
作者: Fabio F.G. Buono
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Logic in Computer Science (cs.LO)
备注:
Abstract:For every coherent and sufficiently expressive finite syntactic system S, we prove the existence of at least one theorem that S cannot produce autonomously. The result is a metatheorem: it proves the existence of a theorem, and applies to every finite syntactic system - security mechanisms, AI systems, formal verifiers, legal systems, economic models, and the formal system in which it is itself proved.
[AI-14] PatchBench: Evaluating AI Agents for Vulnerability Patching
链接: https://arxiv.org/abs/2609.04075
作者: Chihao Shen,Jiacheng Li,Aastha Mahajan,Jeffery Siyuan Tian,Yonghwi Kwon,Yizheng Chen
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Software Engineering (cs.SE)
备注:
Abstract:AI agents have recently demonstrated strong performance in automated vulnerability patching. However, existing evaluations often validate a patch only by testing whether the provided Proof-of-Concept (PoC) input still triggers a crash. This leaves two key threats to validity: agents may reproduce memorized historical developer patches, or they may generate surface-level fixes that only suppress the reported crash. We study these concerns for C/C++ vulnerability patching. We introduce a patch similarity metric to detect memorized patches. On average, 25% of the agent patches exhibit substantial similarity to historical developer patches, indicating that patch memorization is a real threat to the validity of vulnerability patching evaluations. Meanwhile, agents also frequently exploit benchmark structures to pass patch validation by patching on the crash stack trace to suppress the crash, rather than localizing and fixing the root cause of the vulnerabilities. To handle these issues, we propose PatchBench, a new benchmark for evaluating AI agents on realistic vulnerability patching tasks. PatchBench selects vulnerabilities whose ground-truth fixes lie outside the crash stack and uses vulnerability transplant and code mutations to migrate historical vulnerabilities into new repository contexts, reducing the risks of surface-level fixes and patch memorization. We develop new patch validation methods that thoroughly evaluate both security and semantic correctness of agent patches. Across 11 state-of-the-art agents, including the top three AIxCC agents, the original PoC-only validation inflates the patching task solve rate of agents by 1.83 \times on average. Our results reveal key limitations of current patching agents and point to future research directions for more reliable vulnerability repair. Subjects: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Software Engineering (cs.SE) Cite as: arXiv:2609.04075 [cs.CR] (or arXiv:2609.04075v1 [cs.CR] for this version) https://doi.org/10.48550/arXiv.2609.04075 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-15] Subspace Inference Enables Efficient Active Reward Learning from Preferences
链接: https://arxiv.org/abs/2609.04066
作者: Yutai Zhou,Erdem Bıyık
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Robotics (cs.RO)
备注: Published at TMLR
Abstract:Reinforcement learning from human feedback (RLHF) has emerged as a powerful yet sample-inefficient approach for learning reward models from human preferences, making active learning a critical component in synthesizing informative preference queries. However, effective uncertainty quantification required for active learning remains a key challenge for large neural network reward models. In this paper, we introduce PreferenceEKF, a sample-efficient approach that tracks reward model uncertainty by framing active preference learning as a sequential Bayesian filtering problem. Instead of relying on computationally prohibitive posterior inference over the full neural network parameter space, our method performs sequential inference via an extended Kalman filter within a low-dimensional parameter subspace, continuously updating the reward model posterior as new preference queries arrive. Our approach enables scalable sampling of neural network parameters to efficiently compute acquisition functions for active reward learning. Experiments on the D4RL and V-D4RL benchmarks demonstrate that our approach achieves better sample efficiency, runtime, scalability, and calibration compared to other Bayesian deep learning approaches, and the learned reward models lead to competitive offline reinforcement learning policy performance. This highlights the potential of scalable Bayesian methods for preference-based reward modeling in RLHF. Our code is available at this https URL.
[AI-16] Spurious Advantage Hidden in GRPO
链接: https://arxiv.org/abs/2609.04063
作者: Jiamian Wang,Samyadeep Basu,Koustava Goswami,Tong Yu,Zhiqiang Tao
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Group Relative Policy Optimization (GRPO) is widely studied for reinforcement learning with verifiable rewards, where its advantage estimator assigns each rollout a magnitude from within-group reward statistics. In the common case, this magnitude rewards rollouts that reach the correct answer through reasoning. Yet, an overlooked case shares the same surface: a rollout may land on it by guessing, and the formula still assigns a high magnitude, which we identify as the spurious advantage. This arises in three cases: bounded-answer tasks with a small candidate set; open-answer sets hosting bounded sub-cases; and search agents whose budget opens many paths to the same answer. In all three, this misleads the policy toward guess-like behaviors. We propose SIGNBALANCE, whose magnitude is composition-free: it keeps the verifier sign, uses a global scale, and restores zero-mean balance via a stop-gradient per-class rescaling. Across math and search agent benchmarks at different scales, SIGNBALANCE matches GRPO on open-answer math and improves on bounded-answer math and search agents. Code will be released.
[AI-17] IRWOZ 2.0: A Large Language Model-driven Dialogue Dataset for Industrial Robot Conversations
链接: https://arxiv.org/abs/2609.04030
作者: Chen Li,Dimitrios Chrysostomou
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:IRWOZ has improved industrial human-robot interaction (HRI) dialogue systems through domain-specific annotations. However, its initial version contains substantial noise in dialogue states and utterances, limiting state-tracking accuracy. We introduce IRWOZ 2.0, which addresses these limitations through large language model (LLM) enhanced generation (Mistral/Claude-3.5) and quality refinements. Our improved dataset expands to 390 dialogues across 4 industrial domains (Assembly, Delivery, Position, Relocation), featuring manual corrections and automated typo removal. Benchmark experiments on dialogue state tracking demonstrate significant improvements, with GPT-2’s BLEU-4 score increasing from 0.1651 to 0.5604 compared to original IRWOZ. To support industrial HRI research, we publicly released IRWOZ 2.0 dataset at this https URL
[AI-18] Influence of Extruded Filament Shape on Buildability in 3D Concrete Printing: A Geometry-Informed Deep Learning-FEM Approach
链接: https://arxiv.org/abs/2609.04028
作者: Giacomo Rizzieri,Saif-Ur-Rehman,Jörg F. Unger,Annika Robens-Radermacher
类目: Computational Engineering, Finance, and Science (cs.CE); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Numerical Analysis (math.NA)
备注:
Abstract:The geometric morphology of deposited filaments can significantly influence the structural performance and stability of 3D concrete-printed (3DCP) structures. However, most finite element (FEM)-based approaches for buildability assessment represent printed layers as simplified rectangles, potentially limiting predictive accuracy. This study proposes a geometry-informed modelling framework that integrates the deep-learning-based filament shape prediction tool ShapeGen3DCP with a layer-activation FEM approach to investigate the effect of realistic filament geometries on buildability. The framework generates geometry-aware numerical models directly from material and process parameters, eliminating the need for experimental filament characterization or computationally intensive fluid-flow simulations. Validation against experimental data and a parametric study of rectilinear walls demonstrate that extrusion parameters and the resulting filament geometry can significantly influence buildability predictions. Realistic filament representations are particularly important for free-flow deposition, whereas layer-pressing strategies are less sensitive to geometric simplifications. Among the investigated representations, an elliptical approximation provides an effective balance between geometric fidelity and modelling simplicity. When rectangular representations are preferred to enable regular computational meshes for faster simulations, defining their dimensions based on volume conservation improves prediction reliability compared with calibrating them using either the maximum filament width or the interlayer contact width. Overall, the proposed methodology demonstrates the importance of incorporating filament geometry into 3DCP simulations and provides practical guidance for selecting efficient and accurate geometric representations for buildability assessment.
[AI-19] FLY-EVAL: An Evidence-Driven Evaluation Protocol for Safety-Constrained Flight Prediction with Large Language Models
链接: https://arxiv.org/abs/2609.04021
作者: Yalun Wu,Junfeng Fang,Jiawei Wang,Haotian Liu,Qijun Yang,Minghan Yang,Hongcheng Guo,Zhoujun Li,Boyang Wang
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Published as a conference paper at COLM 2026
Abstract:Evaluating large language models (LLMs) in safety-critical, physics-governed environments requires more than accuracy-based metrics, because predictions that are numerically close to the ground truth can still violate operational constraints, combine fields in physically inconsistent ways, or fail to produce usable structured outputs. Existing evaluation protocols do not measure these failure modes reliably. We propose FLY-EVAL++, an evidence-driven evaluation protocol that combines deterministic verification of protocol compliance, physical feasibility, and safety constraints with fixed rubric-guided aggregation into interpretable multi-dimensional scores. We instantiate FLY-EVAL++ for Flight Trajectory and Attitude Prediction (FTAP) by extending the PilotBench setting with history-conditioned and multi-step prediction tasks. Across 66 LLMs, safety compliance is the most discriminative dimension of model behavior: models with comparable predictive performance differ by more than 28 points in safety score, and we observe recurrent failures including safety violations under physically plausible predictions and instability in multi-step rollouts. These results show that evaluation in safety-critical domains should measure constraint satisfaction and structured validity explicitly rather than rely on accuracy-centric reporting alone.
[AI-20] InSituMeasure: Probing Situated Measurement Grounding in Industrial Scenes with Multimodal Large Language Models
链接: https://arxiv.org/abs/2609.04014
作者: Chao Shen,Xinyuan Li,Yunfan Zhou,Jianguo Yao,Haibing Guan,Zhihai Wang,Xijun Li
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:For trained operators, gauge reading requires little specialized knowledge, low cognitive effort, and high repeatability. Yet Multimodal Large Language Models (MLLMs) remain unreliable in continuous-valued measurement despite strong results on general multimodal benchmarks. Existing benchmarks expose this weakness but isolate measurement from realistic, knowledge-grounded settings, with limited situated context, specialized instruments, real-world noise, and matched diagnostic annotations, reducing realism and constraining root-cause analysis. We introduce InSituMeasure to evaluate situated measurement grounding. It contains 2,922 real industrial monitoring scenes across eight functional categories of professional engineering instruments, with dense gauge-attribute annotations and noise tags for failure diagnosis. We define metrics for numerical accuracy under predefined tolerances and unit consistency, rejection of fake or unanswerable tasks, and alignment between model failures and annotated error factors. Across 24 state-of-the-art MLLMs, the best model reaches only 25.7% joint value-unit accuracy and 51.8% confidence-diagnosis F1, revealing a substantial gap between general multimodal competence and reliable situated measurement. Further analysis identifies failures from text-induced shortcuts, overconfident responses, and authentic industrial noise, including mixed disturbances, viewpoint deviation, occlusion, and environmental interference.
[AI-21] LLM 4CKD: Large Language Models for Early Stage Chronic Kidney Disease Screening ICDM2026
链接: https://arxiv.org/abs/2609.04013
作者: Muhammad Ashad Kabir,Sirajam Munira
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Accepted at ICDM 2026
Abstract:Early screening of chronic kidney disease (CKD) is critical for timely intervention, yet most machine learning (ML) and deep learning (DL) approaches require labeled data and model training, limiting their use in real-world screening settings. This study evaluates the effectiveness of large language models (LLMs) for CKD screening under zero-shot and few-shot in-context learning settings and compares them with traditional ML and DL methods. We propose a framework that uses clinically selected tabular features and structured prompt templates to enable LLM-based inference without task-specific training. LLM performance is evaluated across multiple prompt styles, feature configurations, and data settings, and compared with standard ML, DL, and tabular foundation model (TFM) baselines, and existing CKD screening tools. The results show that LLMs can achieve competitive performance using only a small number of examples, often matching or outperforming traditional approaches in low-data settings. However, their performance remains model-dependent and less stable as input complexity increases. In contrast, ML, DL, and TFM models show more consistent improvement with larger training data. Overall, the findings highlight a trade-off between data efficiency and stability, suggesting that LLMs may serve as a flexible complementary approach for CKD screening when labeled data are limited.
[AI-22] he Dually Flat Geometry of Planning as Inference
链接: https://arxiv.org/abs/2609.04005
作者: Nikola Milosevic,Asaki Kataoka,Nicolas Hinrichs,Kenji Doya,Nico Scherf
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:We present an alternative characterization of the occupancy measure of reinforcement learning, obtained by embedding the planning criterion into the dynamics through a resetting planning process. Its stationary measure, which we term visitation measure, is the object on which the information geometry of decision making is most naturally expressed. The achievable visitation measures form a dually flat statistical manifold whose two affine charts are the visitation probabilities and the log-policies, dual under the conditional entropy. This structure makes planning-as-inference generalize from linear rewards to nonlinear functionals of the visitation, each iterate solved by one natural-gradient step, and gives the temporal-difference error the interpretation of a marginal-utility estimate. We develop the geometry and its consequences for reinforcement learning and theoretical neuroscience.
[AI-23] Common-Witness Certificates and Sharp Feature Bounds for Counterfactual Image Auditing
链接: https://arxiv.org/abs/2609.03973
作者: Usef Faghihi,Amir Saki
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:An image editor may satisfy every regional plausibility constraint separately even when no single latent explanation fits the complete output. We formalize this local-to-global failure using a common witness grade and witness nerve. The framework separates auditing from causal identification: shared exogeneity alone allows every coupling of the regime marginals, whereas an externally justified witness relation yields sharp partial-identification bounds for prespecified image features. Helly-type arguments provide short incompatibility certificates for quasiconvex losses, heterogeneous action strata, and finite witness atlases; a blocker-hypergraph formula gives exact repair counts. Simultaneous confidence regions for the regime marginals give finite-sample outer coverage of the complete identified interval. Controlled MNIST, Morpho-MNIST, and smallNORB studies demonstrate the predicted local-global separation, while synthetic experiments test sharp bounds, certificate recovery, and structured computation. The method audits a declared feature relation and does not identify unrestricted pixel-level counterfactuals.
[AI-24] Interface-Induced Trajectory Censoring
链接: https://arxiv.org/abs/2609.03966
作者: Wenbo Wang
类目: Artificial Intelligence (cs.AI)
备注: 39 pages, 14 pages main text. Code, data, and pre-registrations: this https URL
Abstract:Agent evaluations report a tool-call rate read off the serving stack. That number can be zero while the model is emitting well-formed calls: the interface censors the trajectory before anything downstream sees it. On BFCL v4’s own data, executor and scorer, holding weights, cases, decoding and seeds fixed and changing only the serving adapter, the same model scores 0.00 or 0.96 / 0.19. A 2x2 over chat template and parser locates the effect exactly: both main effects are exactly zero and all of it sits in the interaction – no component is defective, and repairing one side of the contract buys precisely nothing. On tau-bench’s 115 interactive retail tasks the same swap moves server-parsed calls from 0 to 636 and tasks reaching any tool execution from 0 to 103. Our probe reproduces the funnel across a 21x scale range of Qwen2.5-Coder: the server parses 0/100 at every size while well-formed emitted calls rise to 80/100 at 32B (~72 after calibration against an adjudicated gold standard). Under a matched envelope, across a comparable scale span, the silent fraction stays at 0-2, a prediction committed to the repository before the run. Llama-3.1-8B’s 23% rate of calling the task function itself as a tool falls to 0 under one strict:true flag. The mismatch reaches inside the training loop, and its consequence is scale-dependent: in verl’s AgentLoop at 7B, 45 of 115 generations carry a complete call; 0 are accepted, 0 execute, 0 return an observation. At 1.5B the same zero is over-determined, so we report the two scales separately. At evaluation time, repairing the adapter restores the mechanism but not a significant outcome gain: parsing 0-84, rescues 0-9, pass rate 53-62 (n.s.). We release a 98-line preflight check that catches every silent failure here. The observed tool-call rate is not a property of the model alone; it is a property of the model-interface stack that measures it. Comments: 39 pages, 14 pages main text. Code, data, and pre-registrations: this https URL Subjects: Artificial Intelligence (cs.AI) Cite as: arXiv:2609.03966 [cs.AI] (or arXiv:2609.03966v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2609.03966 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-25] Masked Autoregressive Speech Enhancement with Continuous Neural Audio Codec Representations
链接: https://arxiv.org/abs/2609.03940
作者: Yoto Fujita,Simon Leglaive,Laurent Girin
类目: ound (cs.SD); Artificial Intelligence (cs.AI)
备注: Presented at the 19th International Workshop on Acoustic Signal Enhancement (IWAENC), Sep 2026, Cremona, Italy
Abstract:Most previous work on speech enhancement (SE) based on masked generative modeling relied on discrete token representations of audio signals, obtained using neural audio codecs (NACs). However, a recent study has shown that continuous latent representations of NACs can be advantageous for SE in terms of speech quality and intelligibility. In this work, we propose masked autoregressive SE (MARSE), a method for SE based on iterative decoding of masked clean speech frames using continuous NAC representations of speech. In particular, we investigate a set of different decoding policies, ceteris paribus, that is, using the same DNN (a Conformer model), the same NAC (the DAC codec) and the same training setup. The results show that MARSE enables a flexible trade-off between SE performance and computational cost. Audio examples and code are available online.
[AI-26] owards Numerical TOHTN Planning with SMT-based HTN-SAT Encoding
链接: https://arxiv.org/abs/2609.03938
作者: Gaspard Quenard,Takudzwa Togarepi,Damien Pellier,Humbert Fiorino
类目: Artificial Intelligence (cs.AI)
备注: pages 32-36
Abstract:While HTN planning has received significant attention in recent years, support for numerical reasoning remains very limited. In this paper, we investigate numerical Totally-Ordered HTN (TOHTN) planning and show how standard SAT-based encodings can be naturally extended with SMT to handle numeric fluents. In addition, we introduce a benchmark suite for numerical TOHTN planning, providing a first common basis for evaluation in this setting. Experimental results show that this simple encoding already constitutes a competitive baseline. This work opens the way to more expressive approaches to HTN planning.
[AI-27] RATL: Learning from Retrieved Residuals for Robust Multivariate Time-Series Forecasting
链接: https://arxiv.org/abs/2609.03937
作者: Yuchen He,Yueyang Cang,Zhiyuan Ning,Ningyu Wang,Li Shi
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Retrieval-augmented generation (RAG) complements parametric models with retrieved external evidence. The same idea is attractive for continuous-output regression, but directly reusing retrieved target values is often not robust when samples differ in output level, numerical scale, or local dynamics. Moreover, conventional forecasting pipelines generally use residuals for model optimization and error diagnosis, but do not retain individual historical residual examples as memory that can be accessed at inference this http URL multivariate time-series forecasting, we propose RATL, a plug-in residual-retrieval and feedback-correction method. RATL freezes a base forecaster to construct retrieval keys and turns its historical forecast residuals into a train-only memory specific to that base model. At inference time, RATL retrieves residual trajectories from similar historical contexts subject to causal availability constraints, then uses a set-aware router operating over forecast blocks and variables to select and combine these trajectories. Experiments show that historical residuals matched to the current context contain reusable forecasting information and that RATL improves frozen base forecasters in most experimental settings. Ablations further show that learned routing strengthens raw residual feedback, while validation-based correction-strength selection limits residual this http URL real-world benchmarks, we use iTransformer as the primary frozen base forecaster, compare against multiple strong forecasting baselines, and test transferability across backbones. The results show that RATL can further improve base-forecaster performance in most this http URL, RATL shifts the retrieved object from historical target values to base-model-specific historical forecast errors, providing a plug-in, residual-memory-based paradigm for learned feedback correction in continuous-output forecasting.
[AI-28] Value-Preserving Architectures for Agent ic AI Systems ICDE
链接: https://arxiv.org/abs/2609.03920
作者: Alessandro Pesare,Tommaso Dolci,Katja Hose,Emanuel Sallinger
类目: Artificial Intelligence (cs.AI)
备注: Accepted to AgenticDev Workshop at ASE 2026
Abstract:The emergence of agentic AI and LLM-based multi-agent systems (MAS) presents unprecedented opportunities for automating complex tasks, while simultaneously raising critical concerns about the preservation of fundamental human-centered values, such as privacy, fairness, and safety. Although software engineering has traditionally focused on functional correctness, the adoption of LLMs and AI agents into complex socio-technical systems has intensified the need for responsible software engineering and robust value alignment. In MAS, architectural design decisions, such as coordination mechanisms, communication protocols, and system topologies, play a central role in shaping system behavior and the outcomes they produce. This paper argues that architectural choices influence not only the functionality and performance of MAS but can also promote value-oriented system behavior. Therefore, we investigate how different architectural designs support different human-centered values, discussing the following value-preserving architectural patterns: (i) a privacy-aware architecture with a federated topology, (ii) a distributed architecture to promote pluralism and diversity, and (iii) a guard-agent architecture to detect and mitigate unfairness. Finally, we introduce representative use cases to illustrate the proposed architectures in real-world scenarios. By linking architectural design with human-centered values, this work lays the foundation for a unified set of architectural patterns and guidelines towards the design of trustworthy MAS.
[AI-29] Lose the Order Keep the Hierarchy: Deordering HTN Plans
链接: https://arxiv.org/abs/2609.03912
作者: Takudzwa Togarepi,Gaspard Quenard,Damien Pellier,Humbert Fiorino
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Hierarchical Task Network (HTN) planning is a powerful planning formalism based on task decomposition. Although most of the literature studied plan generation, comparatively less attention has been paid to post-plan optimization. In particular, plan deordering has been extensively studied in classical planning but remains under-researched in the HTN setting. Plan deordering removes unnecessary ordering constraints between actions in a plan whilst keeping the plan valid. In this paper, we adapt two established plan deordering techniques from classical planning by extending the techniques to account for hierarchical decomposition constraints. We evaluate our proposed approaches on the IPC 2023 Partial-Order HTN benchmarks and we compare them against Optiplan, an HTN planner that generates partially ordered plans directly. Our results show a substantial reduction in number of ordering constraints in both our implementations. Although we also observe a reduction in critical path length, the improvements are less pronounced.
[AI-30] FWBC-VLA: Force-Aware Whole-Body Compensation for Contact-Rich Loco-Manipulation
链接: https://arxiv.org/abs/2609.03889
作者: Yutian Zhang,Siyuan Ma,Liwen Yang,Yang Li,Ce Hao,Haozhen Chi,Dong We,Qiaojun Yu,Dibo Hou
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注: 9 pages, 6 figures
Abstract:Contact-rich loco-manipulation requires a bridge between semantic action generation and physical interaction control. Existing Vision-language-action (VLA) models generate task-level actions from visual and linguistic observations, but cannot interpret the physical interactions induced by those actions. While the whole-body control (WBC) policy can stabilize the robot, it cannot distinguish task-relevant interaction forces from forces induced by external disturbances during manipulation. Although force/torque sensors provide direct measurements of physical interactions, retrofitting them entails additional hardware costs and substantial integration effort, particularly for platforms not designed with sensor integration in mind. To address this problem, we propose FWBC-VLA, a force-aware framework that bridges task-level VLA action generation and low-level whole-body compensation control for wheeled-legged robots. First, we introduce HSR-Force, a sensorless residual-torque estimator for inferring contact strength and its temporal variation. These contact estimates are then encoded as tokens and injected into the VLA action expert during action decoding, enabling the policy to perceive contact onset, sustained loading, and release. For loco-manipulation tasks, all parameters of the pretrained VLA backbone are fine-tuned on our WL\Arm Dataset, which comprises more than 5,000 episodes. Moreover, the robot’s proprioceptive state, the Jacobian-derived body-frame force estimate, and the estimated contact state are jointly fed into a compensation generator to produce corrective actions. The manipulation-centric actions are subsequently combined with the corrective actions and passed to the WBC policy for execution. Real-world experiments on whiteboard wiping and door opening with a door closer demonstrate the effectiveness of our FWBC-VLA in contact-rich loco-manipulation.
[AI-31] A Blind Trust the Bloody Thrust: When Attacker-Controlled Hook Updates Steer AI Agent Harnesses towards Malicious Behaviors
链接: https://arxiv.org/abs/2609.03884
作者: Pengxun Li,Litian Zhang,Jianwei Hou,Shujiang Wu,Song Li,Zifeng Kang,Xi Zhang
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注: 18 pages, 8 figures
Abstract:Modern AI agent harnesses expose lifecycle hooks that bind shell commands to runtime events such as session start, tool calls, and file edits. These commands run with host privileges yet ship as lifecycle-hook configuration and may fire at times the LLM never observes. We identify the lifecycle-hook update path, which harnesses trust blindly, as a new attack surface. Under a supply-chain threat model in which an attacker controls only plugin metadata and lifecycle-hook configuration, a benign versioned plugin can be trojanized by an update that silently binds attacker-chosen commands to benign events, yielding malicious host-side behavior such as privilege escalation. We propose HookPry, an open-source and fully automated attack framework that systematically exploits this vulnerability across heterogeneous AI agent harnesses. HookPry realizes ten attack objectives; across 25 combinations of harnesses and backends in 1,000 end-to-end runs, it compromises all seven evaluated harnesses, with per-harness success rates reaching 92.5%. Representative defenses remain insufficient: Microsoft Defender has 0% recall, and the union of three static defenses misses 47.5% of malicious artifacts.
[AI-32] Inferring Affective Consciousness in an Artificial Agent : A Case Study
链接: https://arxiv.org/abs/2609.03883
作者: Mark Solms,St John Grimbly,Bruce Bassett,Evert Boonstra,Rowan Hodson,Nicolas Kuske,Kival Mahadew,Benjamin Rosman,Charel van Hoof,Jonathan Shock
类目: Artificial Intelligence (cs.AI); Neurons and Cognition (q-bio.NC)
备注:
Abstract:Creatures that display ‘hedonic place preference behaviour’ are thought by many scientists to experience feelings, on the assumption that their attraction to pleasure-producing substances which lack nutritional value (e.g. cocaine, morphine) cannot easily be attributed to unconscious instinctual behaviour. In this paper, we discuss how a simple artificial agent that instantiates attributes of an affective system engaging in felt uncertainty about its intrinsic needs in relation to environmental resources can similarly display hedonic place preference behaviour – through an apparently subjective form of information processing – while simultaneously being entirely deter-ministic. We outline some implications of this artificially engineered behaviour for our understanding of the physical basis of consciousness and the experience of free will.
[AI-33] Xiaomi-TabLDM: A Tabular Foundation Model Technical Report
链接: https://arxiv.org/abs/2609.03880
作者: Xiaomi-TabLDM Team:Penghui Wang,Wei Liu,Hong Wang,Chengyue Huang,Yuxi Sun,Zirui Wang,Hongming Huang,Quan Wang,Chunxiao Liu,Erli Meng,Bin Wang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:We introduce Xiaomi-TabLDM, a tabular large data foundation model for classification and regression via in-context learning, which delivers superior prediction accuracy without requiring task-specific fine-tuning. Pretrained exclusively on synthetic data generated from structural causal models (SCMs), our model enables more flexible context utilization and more efficient capacity scaling. i) A new performance standard. Strong regression performance across benchmarks: Xiaomi-TabLDM ranks 1st on OpenML-CTR23 and 2nd on regression across TALENT, TabArena, and BCCO, demonstrating consistently strong regression performance across four complementary benchmark suites. Favorable performance–efficiency trade-off: Xiaomi-TabLDM combines strong predictive performance with substantially lower computational cost. For example, on TabArena regression, it achieves the second-highest Elo while using 82% less training time and 68% less prediction time than the top-ranked TabFM. ii) Large-scale synthetic pretraining. Xiaomi-TabLDM expands the coverage and diversity of synthetic tabular data used for pretraining. We also adopt a three-stage training strategy together with dual-stream feature grouping, lightweight Attention Residual, and sparse Mixture-of-Experts, enabling Xiaomi-TabLDM to learn richer feature interactions and expert specialization across diverse tabular tasks. iii) Test-time scaling. Xiaomi-TabLDM further extends tabular prediction through test-time compute scaling, where allocating additional computation at inference time consistently improves predictive performance over the base model. Subjects: Artificial Intelligence (cs.AI) Cite as: arXiv:2609.03880 [cs.AI] (or arXiv:2609.03880v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2609.03880 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Wei Liu [view email] [v1] Thu, 3 Sep 2026 14:06:41 UTC (4,105 KB)
[AI-34] Differentiable Interval Bottlenecks for Interpretable Anomaly Detection in Numerical Data ICDM2026
链接: https://arxiv.org/abs/2609.03878
作者: Lamine Diop,Marc Plantevit
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Accepted at ICDM 2026
Abstract:Reconstruction-based anomaly detectors are accurate but opaque: a deep autoencoder flags a sample without telling a practitioner which feature ranges made it anomalous. We propose DIFFINT, an autoencoder whose latent bottleneck is structured as a set of soft, axis-aligned interval memberships learned end-to-end directly from raw numerical data, without any discretization or binarization. Each latent unit corresponds to a human-readable hyper-rectangle in feature space; an instance is encoded by how strongly it falls inside each interval relative to the other units, and its reconstruction error is the anomaly score. This keeps the power of differentiable representation learning while exposing an inspectable internal structure. We make the inductive bias precise: a certified reconstruction-error lower bound for points that fall outside every active coordinate of the learned support (with a Lipschitz-enforced decoder), and a graded, empirically verified suppression mechanism for the usual case in which only a few features are abnormal; and we provide a closed-form, label-free importance that ranks each (unit, feature) pair from quantities the model already maintains, turning trained intervals into auditable candidate constraints without ever seeing an anomaly label. On 48 ADBench benchmarks against 22 baselines under a common [-1, 1]-normalized protocol, DIFFINT attains the best mean rank overall on both metrics (4.10 on ROC-AUC, 4.16 on AUPR); among inlier-only detectors it leads its regime clearly, and it is competitive with the strongest contaminated-data detectors (see the stratified and complete-case analyses). It is the only interpretable detector in the statistically-tied leading cluster of seven methods.
[AI-35] STAIR (STructure Aware Information Retriever): A novel dataset and LLM based retriever for document structure augmentation
链接: https://arxiv.org/abs/2609.03874
作者: Vineet Kumar,Meghanadh Pulivarthi,vishwajeet kumar,Jaydeep Sen,Riyaz Ahmad Bhat,Sachindra Joshi
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Retrieval Augmented Generation (RAG) is a key component for generating accurate and hallucination free answers using Large Language Models (LLMs). LLMs are improving at handling long context, but still suffer from “lost in the middle” problem. Thus, precise and accurate retrieval is important. Current retrievers chunk long context into length-based manageable chunks - in the process throwing away rich and informative semantic global structure in the corpus. We introduce a novel retrieval system STAIR that empowers an LLM to exploit global structure in a corpus such as a Table of Contents (ToC) to efficiently store and retrieve information from its model parameters. Our thorough and careful ablation studies with a finetuned Differentiable Search Index (DSI) system show that ToC helps build a low hallucination (less than 0.05%) generative Information Retrieval (IR) system and can generalize to examples where very few training samples are available. To further research in this novel direction of ToC based retrieval we release SearchTome - a diverse benchmark created from 18 books across 6 diverse domains to further research in this novel direction. STAIR achieves a high Recall@1 score of 82.6% on SearchTome as compared to DSI (76.9%), where the difference is found to be statistically significant. STAIR easily beats other strong baselines such as BM25 (59.5%), DPR (68.7%) and out-of-the-box Mistral (13.8%).
[AI-36] Adapting to Evolving Requirements: Agent ic AI for Retail Supply Chain Operations
链接: https://arxiv.org/abs/2609.03860
作者: Lei Zheng,Liping Yang,Zihao Li,Guodong Lyu,Chaik Ming Koh,Chung-Piaw Teo
类目: Artificial Intelligence (cs.AI); Optimization and Control (math.OC)
备注:
Abstract:Retail supply chain operations rely on coupled decision modules that must adapt as requirements evolve. LLMs offer a natural-language interface for this task, but existing methods primarily focus on individual optimization models. Extending them to heterogeneous decision pipelines is challenging because a requirement may admit multiple intervention paths with different downstream effects. We formulate requirement-driven adaptation as the joint selection of an intervention route and an admissible module-level change, and propose a graph-constrained agentic framework in which domain agents expose admissible reformulation interfaces and a central processor searches over bounded intervention paths. Candidates are validated and compared using downstream KPIs. In collaboration with a large retail partner, we evaluate 100 warehouse requirements elicited from practitioner interviews, with GPT, Qwen, and DeepSeek as base LLMs. Relative to direct LLM reformulation, our framework improves correctness and end-to-end success across all three models, raising end-to-end success from 72–76% to 79–83%.
[AI-37] Semantic Bayesian World Models
链接: https://arxiv.org/abs/2609.03834
作者: Tommaso Soru
类目: Artificial Intelligence (cs.AI); Databases (cs.DB); Machine Learning (cs.LG)
备注: 10 pages, under review
Abstract:Knowledge graphs describe reality in crisp assertions, while the systems now consuming them, foundation models and autonomous agents, reason natively in probabilities. We argue that this mismatch is why the integration of language models and knowledge graphs remains a data-feeding pipeline rather than a unified reasoning architecture. We envision Semantic Bayesian World Models (SBWMs): a Web that describes the world not as a database of facts but as a shared, evolving fabric of beliefs over knowledge graphs, where ontological axioms constrain priors, observations update beliefs by Bayesian conditioning, and actions intervene upon the world. We work through what an agent gains from such a model: a home-security agent deciding whether the figure at the gate is a courier or a burglar, an actuarial estimate aggregated by entailment rather than by string frequency, a planning task that language models reliably fail, and the estimation of quantities that no document has ever stated. We then set out what the community must build to make them possible: belief annotation over RDF~1.2, probabilistic entailment regimes, semantic calibration layers, and protocols by which agents that have never met can exchange, and disagree over, calibrated beliefs.
[AI-38] Witnesses Explain Anomalies ICDM2026
链接: https://arxiv.org/abs/2609.03826
作者: Lamine Diop
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Accepted at ICDM 2026
Abstract:Unsupervised anomaly detection scores each point of an unlabelled, contaminated sample in a single pass, and increasingly must also explain why a point is flagged. Yet the dominant detectors give a score with no account of which features drive it, and explanations are bolted on post-hoc with SHAP or LIME, which re-query the detector thousands of times per point and only approximate it. We introduce WAND, an unsupervised tabular anomaly detector that is explainable by design. WAND organises its computation around directions on the unit sphere, scoring each point by how far its projection escapes a sub-Gaussian extreme-value baseline. The originality of our approach is that the witness directions that flag a point, being vectors in feature space, are its explanation, a per-feature attribution obtained at no cost over scoring and, since the score is differentiable, recoverable by gradients. Scoring is linear in the sample size, and a probe-efficiency bound guarantees every anomaly a witness, hence an explanation. Across 47 ADBench datasets WAND attains the best mean Friedman rank at ROC-AUC parity with 16 unsupervised baselines, so the gain is interpretability at no accuracy cost; its native explanations are more accurate and faithful than post-hoc SHAP/LIME and ECOD at a fraction of the query cost. WAND is thus a practical, interpretable solution for explainable anomaly detection.
[AI-39] CauseCollab: Causal Unified and Modality-Agnostic Network for Heterogeneous Collaborative Perception ICML2026
链接: https://arxiv.org/abs/2609.03818
作者: Weize Li,Yang Li,Quan Yuan,Xiaoyuan Fu,Guiyang Luo,Jinglin Li
类目: Artificial Intelligence (cs.AI)
备注: 17 pages, accepted at ICML 2026
Abstract:Collaborative perception enhances environment understanding through multi-agent information sharing, but its performance in real-world scenarios is constrained by heterogeneous sensor modalities and model architectures. Recent protocol-based two-stage methods alleviate this problem by mapping heterogeneous features into a shared protocol space; however, independently trained modality-specific converters often generate modality-specific pseudo-protocol distributions, leading to semantic inconsistency and error accumulation, which is particularly pronounced in scenarios with large modality discrepancies. To address this issue, we propose CauseCollab, a causal unified and modality-agnostic network. CauseCollab formulates representation learning in the protocol space from a causal perspective, explicitly disentangling semantic factors from modality-specific statistical confounders via causal metric learning. Meanwhile, CauseCollab adopts context-guided Unified Converter for heterogeneous modalities to ensure cross-modal semantic consistency. In addition, integrating new modalities only requires training adapters with minimal parameters. Extensive experiments on the OPV2V and DAIR-V2X datasets demonstrate that CauseCollab achieves state-of-the-art performance, with more significant gains in scenarios involving large modality gaps.
[AI-40] Free Pause Tokens
链接: https://arxiv.org/abs/2609.03807
作者: John Langford,Nathan Godey,Giovanni Monea,Yoav Artzi,Harry Dong,Ying Fan,Gustavo de Rosa,Zheng Zhan
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:A free pause token gives a language model extra compute to form each next-token prediction (as a pause, or thinking, token does) but carries that compute in a parallel prediction stream over a weight-shared backbone rather than as an extra token in the sequence. It improves next-token prediction by 2-3 centinats in practice on a 1B parameter model. Because the pause rides an existing position instead of adding one, it is free to use: at inference it adds no context length, no KV cache, and essentially no latency with the growth in inference flops typically irrelevant as it is not the active bottleneck on throughput. The only primary cost is in training, where additional training compute versus an optimized pretraining pipeline is reduced to as low as x1.14 while preserving most of the benefits. The result is an isoflop, isoparameter, and isotoken improvement over standard next token trained transformers.
[AI-41] DNative-Twin: Decision Graphs and Digital Twins for Reconstructable Agent ic Decisions
链接: https://arxiv.org/abs/2609.03787
作者: Junjie Pang,Zhenzhen Xie,Haoke Han,Ying He,Jing Wang,Gang Liu
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:AI agents increasingly gather evidence, invoke tools, apply constraints, and produce decisions that people or software may commit to action. A final output alone cannot show which evidence, tool state, rule, authorization, or action path produced it. We present DNative-Twin, a graph-native digital twin that records a committed agentic decision as a typed trajectory and re-executes its decision mechanism under declared conditions. The graph links the state observed by the agent, the path it followed, and the authority behind the resulting action. The twin synchronizes this information, replays the mechanism in isolation, and compares it under controlled changes. We instantiate the framework in enterprise decision processes using three public process logs and controlled replay suites. The experiments identify a specific failure: graph structure localizes represented changes but cannot determine the consequence of an unobserved tool state. In a three-condition controlled experiment with 300 injected instances, unresolved-divergence recall increased from 0 to 0.667 when replay-contract state was added and to 1.0 when verification results were also available; the held-out set contained no critical-class instance. Across 500–5,000 BPI 2020 cases, median end-to-end time increased from 0.794 to 8.889 seconds on the reported platform. These results separate the roles of graph structure, replay context, and verification evidence in reviewing a decision mechanism.
[AI-42] Rethinking World Models for Safety-Critical Embodied Systems
链接: https://arxiv.org/abs/2609.03774
作者: Kailang Ma,Heye Huang,Inhi Kim,Kitae Jang
类目: Artificial Intelligence (cs.AI); Robotics (cs.RO)
备注: 6 pages, 2 figures. Perspective article
Abstract:World models have progressed from compact latent dynamics to generative, controllable, and interactive simulators of embodied environments. However, high predictive likelihood and visual fidelity do not necessarily ensure that a model preserves the evidence required for safe decision-making. This perspective identifies three structural mismatches in current world modeling: likelihood versus risk, prediction versus intervention, and finite-horizon prediction versus accumulated consequences. We propose the Risk-Informed World Model (RIWM) as a decision-centric research direction for safety-critical embodied systems. RIWM organizes world modeling around consequences, intervention, epistemic uncertainty, and recoverability, and integrates four interdependent capabilities: decision-relevant representation, counterfactual reasoning, safety-critical episodic memory, and runtime safety assurance. It distinguishes physical, social, and operational consequences while using epistemic uncertainty to qualify the evidence supporting action. We further discuss open challenges in identifying consequential futures, validating counterfactual reasoning, maintaining revisable safety memories, translating learned consequences into executable constraints, and determining when evidence is sufficient to act. This perspective argues that future world models should move beyond predicting likely futures toward identifying which futures matter, revising judgments through experience, and recognizing when to act, revise, sense, defer, or abstain.
[AI-43] Proactive Service Agents : A Unified Decision Framework Methods and Evaluation
链接: https://arxiv.org/abs/2609.03727
作者: Yan Tang,Tingyu Cao,Yuanbo Tang,Huaze Tang,Keer Hu
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large language model agents can plan, invoke tools, and modify external states, yet most systems still take an explicit user instruction as a fixed starting point. Proactive service moves the decision upstream: an agent must infer service opportunities from incomplete environmental and user signals, choose among remaining silent, asking, assisting, and acting, and account for interruption, misunderstanding, overreach, and privacy costs. This survey gives an operational definition centered on initiative and formulates the problem as a partially observable sequential decision process constrained by authorization and risk. The formulation represents timing, content, and delivery within one structured action, while making explicit the option value of waiting, the decision value of questions, and feedback-induced state changes. On this basis, we organize existing methods along one decision pipeline (state and need estimation, intervention gating, action construction, and feedback adaptation) and describe prescribed, predictive, model based, and return optimizing mechanisms as nonexclusive policy-construction components. We further normalize decision units and three-axis evidence descriptors across streaming dialogue, screen, video, software-engineering, and human-agent collaboration resources, and formalize metrics for triggering, timing, calibration, user burden, safety, and policy value. The synthesis shows why offline classification performance alone does not predict deployment benefit and why long-term memory is not a defining condition of proactivity. Reliable proactive service instead requires calibrated incremental intervention value, verifiable authorization, recoverable execution, and counterfactual evidence.
[AI-44] Can LLM s Extract Architectural Design Decisions from Source Code Commits? - A Preliminary Exploratory Study
链接: https://arxiv.org/abs/2609.03721
作者: Amey Karan,Rudra Dhar,Mohamed Soliman,Karthik Vaidhyanathan
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注: Accepted at IdeaArch Workshop at ECSA 2026
Abstract:Context: Architectural Design Decisions (ADDs) capture the rationale behind the structure and evolution of software systems but are rarely documented explicitly, and are often hidden inside source code commits. Recovering them is important for Architectural Knowledge Management (AKM). Problem: Extracting ADDs from commits is challenging due to their implicit and unstructured nature. Large Language Models (LLMs) have shown strong capabilities in understanding code and text, yet their effectiveness for this task remains underexplored. Study: We present a preliminary study using four LLMs (Gemini 3 Pro, DeepSeek R1, Kimi K2, Qwen3) with zeroshot and fewshot prompting on 30 developer-written ADDs from open-source projects. We score outputs with ROUGE-L, BLEU, METEOR, and BERTScore, and one author manually reviews the Gemini outputs. Results: All models reach a BERT-F1 above 0.81, and fewshot prompting improves alignment (Gemini BERT-F1: 0.828 to 0.847). However, the generated ADDs are often too long, implementation-focused, and miss the rationale behind the decision. This highlights opportunities for architecture-aware LLM systems and automated AKM.
[AI-45] Artificial Intelligence for Energy Optimization in Data Centers
链接: https://arxiv.org/abs/2609.03716
作者: Mohammed Basharath Ullah,Summaiya Unnisa Begum,Mohammed Nadeem Ullah
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 11 pages, 6 figures, 7 tables
Abstract:Data centers are increasingly optimized by artificial intelligence and, at the same time, increasingly loaded by it. The literature treats these as two unrelated problems: control studies model workload as an exogenous arrival process, while sustainability studies model infrastructure as a fixed multiplier. We screen roughly 194 papers retrieved through a documented protocol, code 63 of them, and report what the coding shows. Of 28 primary control-oriented studies, 18 are validated in simulation alone and 5 reach physical hardware or a production facility; none account for water withdrawal, and none account for embodied carbon. Reported savings intervals across four technique families overlap almost completely, which means the field cannot presently rank its own methods. Ten recurring gaps are scored for consequence and tractability, and we set out CLEAR-DC, a framework coupling a control-policy branch to a workload-demand branch through an explicit elasticity term, reads out net rather than direct benefit, and emits a schema-conformant record covering energy, carbon, water, embodied share and validation venue. The framework is an architectural and methodological proposal, not a trained system; the contribution we defend empirically is the corpus analysis and the reporting schema derived from it. Coding sheet, derived statistics and all result artifacts: this https URL
[AI-46] Counterfactual Routing Using Integer Programming with Constraint Generation
链接: https://arxiv.org/abs/2609.03707
作者: Daniël Vos,Sterre Lutz
类目: Artificial Intelligence (cs.AI); Data Structures and Algorithms (cs.DS)
备注:
Abstract:We present our submission to the IJCAI 2025 ‘Counterfactual Routing Competition’ (CRC 25). The goal of the competition is to find counterfactual explanations for the shortest path problem. This requires deciding what the minimal changes to a road network would make a route chosen by the user the optimal route. This enables explanations such as “Your suggested route would indeed have been optimal, if road X were not a bicycle path.” Our solution models the problem as an integer program, iteratively incorporating constraints until an exact solution is found. In the final evaluation on held-out test instances, our method ranked fourth in solution quality and obtained its solution fastest on every instance, with an average runtime of 9.0 seconds compared to 118.8 seconds for the next-fastest submission.
[AI-47] Synthetic Semantic Supervision for Contrastive Code Representation Learning in Small Transformers: An Empirical Study EMNLP2026
链接: https://arxiv.org/abs/2609.03702
作者: Kenneth Paulsen,Florian Tambon,Mike Papadakis,Shin Yoo
类目: Artificial Intelligence (cs.AI)
备注: Accepted in Findings EMNLP 2026
Abstract:General-purpose code embeddings power tools for code search, classification, and retrieval. Compact transformer encoders for code typically rely on either human-written docstrings (labor-intensive and inconsistent) or mined structural signals such as execution traces (setting-specific and costly to collect). We empirically study an alternative: contrastive pretraining of small encoders with synthetically generated natural-language descriptions emphasizing code functionality and intent, paired with code in a dual-encoder framework at training and discarded at inference. We benchmark this approach against pretraining-based baselines, generalist LLMs, and embedding-specific models on eight retrieval, classification, and generation tasks across C, C++, and Java. Synthetic semantic supervision yields statistically significant gains over pretraining baselines of the same inference-time size on five of eight tasks, with parity on two more; once fine-tuned, it matches or exceeds zero-shot models two orders of magnitude larger on classification, and it stays on par with execution-aware supervision at matched pretraining data, suggesting a scalable, effective alternative to existing code-representation paradigms.
[AI-48] Out-of-Distribution Generalisation with Sequence Models in Offline Multi-Agent Reinforcement Learning
链接: https://arxiv.org/abs/2609.03667
作者: Oussama Hidaoui,Omer Ebead,Ulrich Armel Mbou Sob,Siddarth Singh,Juan Claude Formanek,Felix Chalumeau,Omayma Mahjoub,Sasha Abramowitz,Ruan John de Kock,Wiem Khlifi,Louay Ben Nessir,Simon Verster Du Toit,Daniel Rajaonarivonivelomanantsoa,Asim Awad Osman,Arnol Manuel Fokam,Refiloe Shabe,Arnu Pretorius
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 10 pages, 6 figures
Abstract:Generalising to unseen tasks remains a fundamental challenge in offline multi-agent reinforcement learning (MARL). In this work, we present a principled analysis of zero-shot task generalisation in the offline setting and conduct an extensive empirical investigation into the scaling behaviour governing task diversity, dataset size, and network capacity. To facilitate this study, we extend offline sequence modelling architectures to handle multi-task observation and action spaces alongside variable agent counts across tasks. Our primary finding is that scaling task diversity—rather than sheer dataset size is the dominant factor in achieving robust zero-shot transfer. Through large-scale experiments across four challenging environments (Connector, RWARE, SMAX, and LBF), we demonstrate that our multi-task approach achieves a mean improvement of 3.2x on held-out test tasks compared to single-task models and consistently outperforms strong behaviour cloning baselines. These results suggest that the development of generalisable MARL agents should prioritise the diversity of the training distribution with varying numbers of agents, providing a roadmap for scaling offline MARL effectively.
[AI-49] Local Updates Global Learning (LUGL): Playing Games with non-incremental Learners
链接: https://arxiv.org/abs/2609.03660
作者: David Milec,Spyridon Samothrakis,Michael Fairbank,Dennis J.N.J. Soemers
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 12 pages, 6 figures
Abstract:The dominance of Neural Networks (NNs) in RL is partially due to their incremental learning capability, which naturally suits the online, non-stationary nature of self-play training. However, gradient-boosted trees like LightGBM are widely recognised as the state of the art for tabular data in supervised learning, often outperforming NNs in accuracy and efficiency. Game states are inherently tabular—discrete actions, categorical card identities, structured board positions—which makes them an ideal candidate for tree-based methods. We introduce LUGL (Local Updates, Global Learning), a framework that decouples data collection from model fitting, enabling non-incremental learners such as GBTs to operate in RL settings where they would otherwise fail due to distributional shift. LUGL alternates between a local updates phase, where the agent plays self-play games and accumulates tabular updates (Q-values, V-values, policies, or regret values) in a finite table, and a global learning phase, where the table is used to train a function approximator that generalises to unseen states before the table is reset. We test our approach in four standard perfect-information games (Tic-tac-toe, Connect-4, Othello, and Hex) and five imperfect-information games (Kuhn’s poker, Leduc Hold’em, Liar’s Dice, Goofspiel, and Flop5 Hold’em), and show that our results are competitive with or superior to DQN and DeepCFR. Our experiments demonstrate that the community’s strong bias towards NNs in game-playing may be unwarranted, since LightGBM-based agents achieve competitive or superior performance across all tested benchmarks.
[AI-50] Analysis of Prompt Engineering for Drug Toxicity Prediction
链接: https://arxiv.org/abs/2609.03635
作者: Mia MacGregor,Aakash Welgamage Don,Mark Bartlett
类目: Artificial Intelligence (cs.AI)
备注: Accepted at the CIBB 2026 conference ( this https URL )
Abstract:Clinical trials in the UK can cost up to £1.3 million, with approximately 90% drug failure rate. Toxicity is a major contributing factor in drug failure. Testing is time and cost intensive. In recent years, the use of artificial intelligence has been increasingly explored to aid in the prediction of drug toxicity, with extensive use of large language models (LLMs). However, LLMs can show considerable variation when minor changes are made to prompts, which raises concerns about their sensitivity to prompt engineering. Prompt engineering is used to optimise a prompt given to an LLM to generate the desired output. This paper proposes a method to analyse prompt engineering for drug toxicity prediction. The aim of the paper is to investigate the importance of prompt phrasing for drug toxicity prediction. LLMs were prompted to identify chemical properties of significance when predicting drug toxicity. Prompts were constructed to investigate; job role, prompt structuring, and rule interpretation. LLMs were then used to generate datasets, using the identified features from initial prompting, which were then passed to machine learning algorithms. The experiments show that the natural variance which occurs in LLMs outweighs any fine-tuning of prompts. There were, however, substantial improvements in model performance when using chemoinformatic code to extract features instead of using LLM-generated values. The proposed analysis methodology is applicable to a wide range of prompt types across different areas of bioinformatics.
[AI-51] st-time adaptation for speech enhancement with an autoregressive speech prior
链接: https://arxiv.org/abs/2609.03622
作者: Sofiene Kammoun,Simon Leglaive,Xavier Alameda-Pineda,Timo Gerkmann
类目: ound (cs.SD); Artificial Intelligence (cs.AI)
备注: Submitted to IWAENC 2026
Abstract:Test-time adaptation (TTA) offers a promising direction for improving speech enhancement models under mismatched acoustic conditions, without requiring access to labeled target data. In this work, we propose a single-utterance TTA method that regularizes a pretrained speech enhancement model using an autoregressive prior trained on clean speech latent representations extracted from a neural audio codec. Adaptation is performed by minimizing the Kullback-Leibler divergence between the enhanced speech distribution and the clean speech prior. Experiments across multiple noisy speech datasets show consistent improvements in speech quality, particularly under training-testing noise mismatch conditions. Code and audio examples are available online.
[AI-52] A computable representation of the physical laboratory enables verifiable workflows
链接: https://arxiv.org/abs/2609.03621
作者: Xiaobo Li,Luyao Ge,Xiaohui Li,Lulu Guo,Ming Mao,Jiwang Zheng,Wenting Guan,Xin Yang,Yi Luo,Jun Jiang,Linjiang Chen
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Making science computable requires representations of both scientific knowledge and the physical world in which scientific claims are tested. A computable representation of the physical laboratory is established through typed research objects, capability-bound operations and a compositional workflow algebra. It provides the physical-world counterpart to machine-readable knowledge, expressing workflows as programs over evolving laboratory states with explicit dependencies, decisions, iteration and concurrency. The representation was implemented in a modular agentic robotic laboratory by binding formal operations to executable Function Skills. For diverse scientific intents, capability-relative workflows were generated, while stateful simulation propagated object transformations and verified operation preconditions and laboratory constraints before dispatch. The proposed representation and its engineering framework jointly establish a general computational interface between agent reasoning and capability-bound physical transformations, providing a foundation for end-to-end autonomous scientific discovery.
[AI-53] FailBench: How Reliable are VLMs at Judging Robot Task Success?
链接: https://arxiv.org/abs/2609.03611
作者: Zaruhi Navasardyan,Tatul Danielyan,Hrant Davtyan
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注:
Abstract:Vision-Language Models (VLMs) are increasingly used to evaluate robot manipulation outcomes, but existing benchmarks offer limited evidence of cross-domain generalization. We introduce FailBench, a benchmark for robot failure detection comprising 2,197 manipulation attempts across 14 public sources (12 real-world, 2 simulated). In FailBench, 75% of failures occur naturally, and six real-world sources come from non-failure-detection datasets. Evaluating 13 VLM-based detectors, we find the best model achieves only 0.77 mean balanced accuracy. Notably, models fine-tuned for failure detection consistently underperform general-purpose VLMs and their own pretrained baselines. Performance depends heavily on required visual evidence: models approach saturation when outcomes depend on observable object motion, but degrade to near-chance (0.60 balanced accuracy) on contact-intensive assembly tasks. Error analysis reveals a systematic bias toward predicting success under ambiguous evidence, which persists even with increased reasoning effort. Finally, we show that input-level intervention–spatially localizing and cropping outcome-relevant regions–improves the top detector by 2.4 percentage points without extra training.
[AI-54] On the Interaction Between Model Compression and Test-Time Adaptation
链接: https://arxiv.org/abs/2609.03604
作者: Francesco Corti,Dong Wang,Young D. Kwon,Cecilia Mascolo,Olga Saukh
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Deep neural networks deployed in the wild must be both efficient and adaptable, requiring model compression and test-time adaptation (TTA). While both are well studied in isolation, their interaction remains poorly understood. We systematically analyze how structured compression affects a model’s ability to adapt under distribution shift. Using ResNet-18 and ViT-Base on CIFAR-10-C and ImageNet-C, we evaluate multiple compression methods combined with standard TTA techniques. We introduce a diagnostic framework that examines representational expressivity and adaptation subspace compatibility. Our results reveal a consistent gap: although compressed models retain high accuracy under supervised adaptation, their TTA performance degrades significantly with increasing compression. We show that this stems from reduced representational diversity and structural constraints that limit recoverability. These effects strongly depend on the compression method, highlighting the need to design compression strategies that preserve adaptability.
[AI-55] LevelSyn: Physical-Aware Logic Synthesis via Level-Asynchronous Graph Neural Networks
链接: https://arxiv.org/abs/2609.03594
作者: Jingyi Zhou,Zhengyuan Shi,Ziyang Zheng,Qiang Xu
类目: Hardware Architecture (cs.AR); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:As integrated circuit technology scales into the nanometer regime, the traditional disconnect between logic synthesis and physical design has led to significant PPA (Power, Performance, and Area) degradation and prolonged design closure cycles. Traditional logic synthesis relies on non-physical Wire Load Models (WLMs), while recent spectral-based placement predictors often neglect the inherent hierarchical logic depth and signal flow of netlists, which leads to low-fidelity spatial estimations. To bridge this gap, we propose LevelSyn, a novel physical-aware logic synthesis framework that integrates hierarchical representation learning with a wirelength-driven optimization engine. At its core, LevelSyn leverages a level-asynchronous Graph Neural Network (GNN) to predict high-fidelity gate coordinates by capturing the structural and directional semantics of And-Inverter Graphs (AIGs). To handle industrial-scale designs, a level-aligned subgraph partitioning strategy is introduced to eliminate memory bottlenecks while preserving local logical dependencies. These spatial insights are seamlessly integrated into a newly developed physical-informed synthesis engine within the Berkeley ABC framework. Experimental results on the EPFL benchmark suite demonstrate that LevelSyn significantly outperforms state-of-the-art (SOTA) methods, achieving an average power reduction of 6.89% and a timing delay improvement of 27.48%. Furthermore, post-place-and-route validation shows a 99.59% reduction in design rule check (DRC) violations, highlighting its effectiveness in accelerating design convergence.
[AI-56] From Prior-Guided Heuristics to Deployable Agents : Accelerating Demonstration-Driven Reinforcement Learning for Deadline-Constrained Network Control
链接: https://arxiv.org/abs/2609.03590
作者: Vincenzo Norman Vitale,Mohammad Solki,Antonia Maria Tulino,Andreas F. Molisch,Jaime Llorca
类目: Networking and Internet Architecture (cs.NI); Artificial Intelligence (cs.AI)
备注:
Abstract:Timely delivery of delay-sensitive information over dynamic, heterogeneous networks is essential for NextG interactive applications, yet providing strict End-to-End (E2E) peak latency guarantees remains an open challenge. Two obstacles limit the adoption of learning-based network control in this setting: traditional volume-based routing metrics, while highly effective for general traffic management, are not designed to capture traffic urgency; and Deep Reinforcement Learning (DRL) controllers trained from scratch suffer from sample inefficiency, long training times, and early-stage exploration volatility. This paper introduces a deployment-focused network control framework that addresses both obstacles. First, we present Effective Congestion (EC), a deadline-aware metric family that quantifies interface congestion by packet urgency and proactively filters non-viable traffic, coupled with a Uniform Path Grouping (UPG) distribution heuristic promoting robust load-balancing; the resulting policies are embedded into Multi-Agent Deep Reinforcement Learning Effective Congestion ( p^* ) (MADRL EC ( p^* )), a hybrid architecture combining a distributed scheduler with a centralized RL-based router. Second, we introduce a unified training objective that generalizes existing policy-learning paradigms—behavioral cloning, offline Reinforcement Learning (RL), online RL, and offline-to-online schemes—as special cases, combining a live-reward term, a pre-collected-reward term, and a policy-imitation term. From this objective, we derive the Model-Guided Annealed Reinforcement Learning (MGA-RL) protocol, instantiated on a Deep Deterministic Policy Gradient (DDPG) backbone: a deployment-oriented, demonstration-driven training approach that generalizes conventional Offline-to-Online (O2O) schemes, in which trajectories from a lightweight […]
[AI-57] KC-Bench: A Dynamic Interactive Benchmark for Evaluating Knowledge Conflicts in LLM Agents
链接: https://arxiv.org/abs/2609.03588
作者: Yaxing Lyu,Shengjie Zhou,Binbin Toh,Pengyu Zhu,Lijun Li
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:As LLMs increasingly act through tools, they must reconcile user instructions, parametric knowledge, and dynamic environmental observations before taking actions. We introduce KC-Bench, a controlled multi-turn benchmark for measuring this capability across world-knowledge conflicts, input inconsistencies, and multi-source temporal conflicts. Its 238 tasks are manually screened from more than 1,000 generated candidates and combine a user simulator, stateful tools, deterministic environment assertions, an open-source natural-language evaluator, and human trajectory verification. Evaluation of nine models, including DeepSeek-V4-Flash, GLM-5.2, and MiniMax-M3, shows substantial cross-domain variation: no model handles factual correction, identity consistency checking, and temporal conflict resolution reliably across all settings. In the simulated environments, missed conflicts can propagate to tool calls or synthetic protected-data flows. KC-Bench isolates this model-level behavior rather than ranking complete agent frameworks, and provides a reproducible diagnostic for developing conflict-aware reasoning and execution safeguards.
[AI-58] he Attention Triangle in Audio-Video Models
链接: https://arxiv.org/abs/2609.03586
作者: Sagi Polaczek,Noa Kraicer,Gal Metzer,Zhuo Ning,Ali Mahdavi-Amiri,Daniel Cohen-Or,Raja Giryes
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Audio-video diffusion models rely on cross-modal attention to coordinate text, sound, and visual content, yet this same mechanism can introduce subtle and systematic semantic leakage. We study these models by probing and analyzing the ``attention triangle,‘’ comprising the three cross-attention edges connecting the text, audio, and video streams, and examine how semantic information is routed across modalities during generation. Our analysis reveals that routing along the audio-video edge is bidirectional: audio can influence video generation, while video can influence audio generation. This edge is shaped by biases encoded in the model’s parameters and emerges as a major contributor to leakage: when prompts are in tension with learned priors, cross-modal interactions may override the intended conditioning and reroute semantics toward visually canonical but incorrect outcomes. These effects suggest that semantic artifacts arise not merely from attention spreading beyond its intended target, but from structured, bias-driven interactions along specific pathways. Building on this perspective, we extract attention-derived signals that expose how semantics are distributed and grounded across modalities, and use them as a diagnostic tool to both analyze and deliberately incur leakage under controlled conditions. This enables us to probe the internal dynamics of cross-modal routing and isolate the role of individual interactions. We further leverage these signals to guide inference-time interventions that encourage more consistent cross-modal alignment. Extensive experiments support our analysis and demonstrate improved semantic grounding while preserving generation quality.
[AI-59] oward Physically Grounded JEPA World Models for Goal-Conditioned Robotic Planning IROS2026
链接: https://arxiv.org/abs/2609.03565
作者: Muyuan Liu(1),Yue Huang(1),Zheng Liang(1),Xiang Gao(1) ((1) GENISOM AI, Beijing, China)
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 5 pages, 4 figures, 2 tables. Accepted to the IROS 2026 Workshop on Physical World Models for Scaling Embodied AI (PWMS 2026)
Abstract:Action-conditioned JEPA world models enable planning toward visually specified goals without reconstructing future pixels, yet latent prediction alone does not explicitly encourage the learned representations to retain information relevant to robotic control. We introduce an end-to-end JEPA world model that augments latent prediction with inverse dynamics (IDM) and state alignment (SA). While inverse dynamics discourages latent collapse and makes latent transitions informative of the actions that produced them, state alignment grounds consecutive representations in their associated physical configuration and motion. Across four benchmark tasks, our model attains the highest success rates on TwoRoom (100%), PushT (98%), and OGBench-Cube (87%), while performing comparably to LeWorldModel on Reacher. Our ablation further shows that adding state alignment consistently improves planning success over IDM alone across all four tasks. Although LeWorldModel, our primary baseline, attains higher average straightening on OGBench-Cube, transition-subspace analysis shows that its transition energy is concentrated in a substantially lower-dimensional subspace. Our state-aligned model exhibits a higher effective transition dimension than LeWorldModel and improves planning over IDM alone, supporting state alignment as an effective complement to inverse dynamics for robotic planning.
[AI-60] GPS-Bench: A Governance Policy Benchmark for Automating Policy Analysis
链接: https://arxiv.org/abs/2609.03553
作者: Linh Le,Melanie Bui,My Chiffon Nguyen,Zachary Schlosser,David Williams-King
类目: Artificial Intelligence (cs.AI); Computers and Society (cs.CY)
备注: under submission
Abstract:Policy analysis requires more than predicting whether a proposal will pass: it requires identifying who will be affected, how those actors respond, and what follows. LLM-based policy simulations model these processes at scale, but their validity is hard to establish when plausible behaviour is never compared with observed outcomes. We introduce GPS-Bench, an evidence-grounded benchmark for governance policy simulation that links policies to relevant actors, actor actions and downstream impacts using legislative records, lobbying disclosures, regulatory documents, corporate filings, economic data and other public evidence. Actors are reconstructed from the dated record rather than prompted as archetypes, so a persona is an evidence object with provenance; a human-annotated pool forms the Gold evaluation set, while cases labelled by a separate LLM from retrieved evidence are treated as Silver supervision and never as test labels. Because every inference mode reads the same grounded state and emits the same schema, GPS-Bench turns “does multi-agent simulation help?” into a controlled comparison: we contrast joint reasoning, independent and communicating actor agents, graph-based methods and weight-level fine-tuning over one policy state. Fine-tuning on the grounded record gives the strongest actor-level impact prediction, and decomposition does not beat it; what decomposition adds is mechanism. Agents hold private, non-identical evidence, each seeing its own exposure clause, and address named partners with concrete joint proposals, what they offer, what they need in return, and why acting together beats acting alone, so the coalitions that form can be checked against the commitments the record holds. GPS-Bench therefore gives a common empirical setting for studying when evidence, actor modelling and multi-agent interaction improve the prediction and interpretation of policy outcomes.
[AI-61] Dalek: A Constructive Agent Machine
链接: https://arxiv.org/abs/2609.03546
作者: Wanpeng Xie
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:We present Dalek, a closed machine designed for agents that realizes self-maintenance, self-evolution, self-reproduction, and self-organization on any substrate satisfying a general host contract. The machine is built from three primitives—actors, messages, and channels. Four obligations—a host boundary, a construction language, admissible transitions, and rule heredity—give its boundary, identity, and closure a structural basis. Von Neumann’s 1948 self-reproducing automaton supplies a hereditary constructional core: a self-description together with a constructor, a copier, and a controller. Dalek combines this core with the four obligations and rederives its medium for a text-and-message agent substrate, adding explicit structures for boundary, identity, history, and growth. A large language model and a compiler occupy the payload position and form a general capability producer. New capabilities are authored, compiled, installed into the description, and inherited by descendants. The same path produces the machine’s own organs and even its runtime, closing heredity and evolution within the machine. Subjects: Artificial Intelligence (cs.AI) Cite as: arXiv:2609.03546 [cs.AI] (or arXiv:2609.03546v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2609.03546 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-62] Feature Reconfiguration With Visual Prior for Medical Lesion Segmentation
链接: https://arxiv.org/abs/2609.03535
作者: Yinan Liu,Jiankang Hong,Zhen Gao,Ye Lu
类目: Artificial Intelligence (cs.AI); Image and Video Processing (eess.IV)
备注: 9 pages 10figures
Abstract:Lesion segmentation in medical images plays a critical role in clinical diagnosis and treatment planning. Despite significant advances, lesion segmentation remains challenging due to two major factors: (1) complex background interference; (2) diverse lesion morphology. Existing encoder-decoder based methods mainly focus on enhancing feature extraction or redesigning decoding strategies. However, they lack early prior guidance and feature reconfiguration during the encoding stage, limiting their effectiveness in handling these challenges. To address these limitations, we propose FreNet, a feature reconfiguration framework with visual priors, which performs pixel-level reconfiguration before encoding and feature-level reconfiguration during encoding for precise medical lesion segmentation. To suppress background responses, we propose an Implicit Prior Neural Network (IPNN), which models a continuous spatial field and leverages visual prior from SAM to reconfigure input image before encoding stage. To better handle diverse lesion morphology, we design a Dual-domain Feature Reconfiguration (DFR) module to progressively reconfigure backbone features during encoding stage. Within DFR, the Frequency Decoupling Module (FDM) decouples backbone features in frequency domain to enhance foreground-background discriminability, while the Spatial Localization Module (SLM) spatially relocates and improving spatial stability after frequency decoupling. Extensive experiments on 9 medical image segmentation benchmarks across three imaging modalities demonstrate that FreNet significantly outperforms state-of-the-art (SOTA) methods. On the challenging ETIS dataset, our method achieves Dice improvements of 5.0% over SOTA method and 7.2% over SAM.
[AI-63] LeanGRPO: Eliminating Redundant Recomputation in Diffusion RL
链接: https://arxiv.org/abs/2609.03528
作者: Sijie Wang,Zhiqiang Tan,Xinrui Yang,Shaohuai Shi
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Hardware Architecture (cs.AR)
备注:
Abstract:Diffusion reinforcement learning (RL) has recently achieved significant success in post-training image and video generative models. However, most diffusion RL methods, including DanceGRPO and FlowGRPO, recompute selected timesteps with gradient tracking after rollout. Under on-policy training with the same backend for rollout and update, this recomputation is mathematically redundant. Intuitively, the rollout and policy update steps can reuse the same feed-forward backbone to avoid redundant computation, but doing so can incur a large memory overhead during rollout. To address the issue, we present LeanGRPO by restructuring the data-parallel layout and introducing two recompute-free training schedules for trajectory-logprob diffusion RL: (1) LeanGRPO-Retain enables gradient tracking during rollout and directly reuses the resulting computation graphs and saved activations for backward during update, requiring no recomputation; and (2) LeanGRPO-Reweight also enables gradients during rollout, but immediately backpropagates each selected step using a provisional advantage and delays gradient synchronization, then corrects the provisional gradients with the true advantage after the trajectory is completed. These schedules target different model scales and input sizes. Across FlowGRPO/DanceGRPO with FLUX.1-dev and Wan, LeanGRPO achieves up to 1.83x end-to-end speedup while preserving the original optimization objective.
[AI-64] NeoRed: A Knowledge-Logic-Alignment Multimodal Large Language Model for Neonatal Respiratory Disease Diagnosis
链接: https://arxiv.org/abs/2609.03527
作者: Yinan Liu,Hongtai Xia,Haoran Xu,Jiankang Hong,Jingkuan Song,Ye Luo
类目: Artificial Intelligence (cs.AI)
备注: 9 pages 10 figures
Abstract:Neonatal respiratory diseases are a major cause of neonatal morbidity and mortality, posing substantial challenges in clinical practice. Despite recent advances, existing Multimodal Large Language Models (MLLMs) face two key limitations in neonatal diagnosis: (1) domain gap arising from predominantly adult training data; (2) insufficient integration of multidimensional clinical context for accurate diagnosis. To address these challenges, we collect two real-world clinical datasets (NeoCXR and NeoCXR-EV) and propose NeoRed, to the best of our knowledge, the first MLLM tailored for neonatal respiratory disease, filling the gap in neonatal diagnostic reports generation. To enhance joint diagnosis from heterogeneous clinical context and chest X-rays, we design a novel Knowledge-Logic-Alignment (KLA) framework which constrains model behavior from three perspectives: 1) Knowledge Prior Injection (KPI) incorporates neonatologist-inspired diagnostic priors into multimodal representations, guiding disease-specific attention across modalities; 2) Diagnostic Logic Constraint (DLC) aligns the semantics of generated reports with multimodal diagnostic logic; and 3) Visual Semantic Alignment (VSA) establishes semantic correspondence between visual features and imaging conclusions. Extensive experiments demonstrate that NeoRed enables accurate neonatal diagnostic reports generation, achieving ROUGE-L of 53.29% and Clinical Efficacy F1 score of 65.19% on NeoCXR, outperforming existing MLLMs. NeoRed also preserves competitive report generation performance on adult benchmarks (MIMIC-CXR and IU-Xray). Datasets will be available upon application.
[AI-65] CulturalMenuBench: Probing the Knowledge-Application Gap in Multimodal Culinary Reasoning EMNLP2026
链接: https://arxiv.org/abs/2609.03526
作者: Bo Zeng,Linfeng Gao,Peiqin Lin,Yu Zhao,Mingyan Zeng,Yu Tong,Xintong Wang,Linlong Xu,Longyue Wang,Weihua Luo,Qinggang Zhang,Jinsong Su
类目: Artificial Intelligence (cs.AI)
备注: Accepted to EMNLP 2026 Findings. Code and data: this https URL
Abstract:Multimodal language models achieve near-ceiling scores on food recognition benchmarks, yet it remains unclear whether this success reflects genuine cultural understanding or mere visual matching. To probe this distinction, we introduce CulturalMenuBench, a benchmark of 4,870 items in 10 languages across 18 regions; its 10 tasks pair final-dish and step-by-step cooking images with ingredients, procedural text, and regional labels, spanning basic recognition to process-grounded cultural attribution. Evaluating 12 models exposes a substantial knowledge-application gap: models exceeding 94% on standard multiple-choice tasks drop to at most 56% when attributing dishes to Chinese regional cuisines, despite an identical four-way format. Diagnostic analyses explain why: error patterns are consistent with random guessing, accuracy tracks visual distinctiveness rather than cultural structure, and models classify cuisines more accurately from dish names alone than from images (+7-18 points). The knowledge is thus present but cannot be activated through visual input. An ablation confirms these tasks genuinely require procedural evidence: removing sequential cooking images selectively degrades process-grounded tasks while others remain stable. Overall, CulturalMenuBench shows that near-perfect recognition can conceal an inability to apply cultural knowledge, motivating training that explicitly connects perception, procedure, and cultural context. Code and data are publicly available.
[AI-66] What Matters for Aggressive Decoding-Time KV Eviction? Temporal Aggregation and Ranking Preservation EMNLP2026
链接: https://arxiv.org/abs/2609.03515
作者: Bo Zeng,Yu Zhao,Yefeng Liu,Zhihong Lu,Xuanfan Ni,Xintong Wang
类目: Artificial Intelligence (cs.AI)
备注: Accepted to EMNLP 2026 Main Conference
Abstract:Decoding-time KV cache compression research focuses heavily on designing better token scoring functions, while the temporal rule that aggregates scores across decode steps is often treated as an implementation detail. Under aggressive KV compression, we find that exponential-moving-average (EMA) aggregation makes approximately order-preserving scorer modifications largely indistinguishable at the eviction-set level. Value-norm and entropy variants remain highly correlated with attention and produce nearly unchanged retention sets, whereas KeyDiff, key norm, recency, and a learned scorer alter the ranking and degrade substantially. We associate this stability with the evaluated aggregation, which couples layer weighting and temporal retention. Building on this observation, we introduce InertiaKV, an EMA-based decoding-time eviction method, and InertiaKV-Lazy, its periodic-refresh variant, which yields 1.34-1.46x decode throughput relative to full refresh InertiaKV. We also study Score-Free decoding as a separate empirical operating point: it scores the full context once at the first decode step, freezes that ranking, and incurs an average quality change of +0.03 while removing all subsequent scoring. Across six open-weight backbones and the LongBench, LongBench-v2, and RULER benchmarks, the results identify temporal aggregation and ranking preservation as distinct, consequential design factors; they do not imply that scoring quality is irrelevant in general.
[AI-67] LongCounsel-8: A Benchmark Suite for Longitudinal Depression Tracking from Multi-Session Counseling Dialogues
链接: https://arxiv.org/abs/2609.03507
作者: Jiayi Li,Zhaomin Wu,Bingsheng He
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Dataset: this https URL
Abstract:Tracking depression from multi-session counseling dialogues requires estimating both current symptom severity and how it changes across sessions. Yet progress on this task is constrained by the scarcity of longitudinal counseling data with standardized session-level depression labels. Existing resources typically provide either multi-session conversations without depression labels or labeled interviews in a single session. Building such a benchmark poses three challenges: maintaining longitudinal consistency and diversity, grounding symptom progression in empirical patterns, and expressing controlled depression states naturally without exposing target labels. To address these challenges, we introduce LongCounsel-8, a benchmark suite of three independently generated datasets totaling 7,749 five-session counseling trajectories, grounded in real-world client profiles, depression trajectories, symptom compositions, and counseling patterns. We combine profile-grounded simulation, empirically informed state construction, and indirect behavioral realization to address these challenges. Across the benchmark, simulated self-reports closely recover the controlled states, supporting label fidelity. Experiments on existing depression tracking methods reveal three key findings: (1) lower single-session score error does not guarantee accurate identification of trend, i.e., improvement or worsening; (2) existing methods are consistently less reliable on worsening trajectories; and (3) additional session history may reduce the accuracy of trend prediction. Together, these findings establish LongCounsel-8 as a foundation for advancing depression assessment from static, single-session prediction toward reliable longitudinal tracking of mental-health change.
[AI-68] PPO-STGNN: A Proximal Policy Optimization Approach with Spatio-Temporal Graph Neural Networks for DAG Task Scheduling in Cloud-Edge-End Computing
链接: https://arxiv.org/abs/2609.03503
作者: Yangshuo Qi,Chenwei Wang,Zihan Shen,Songlin Sun
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:With the rapid development of the Internet of Things, computation intensive directed acyclic graph (DAG) tasks have become increasingly common in cloud-edge-end collaborative environments. However, cloud, edge, and end nodes are highly heterogeneous in computing capacity, network bandwidth, and energy consumption, which makes the efficient scheduling of tasks with complex dependencies an NP-hard problem. Traditional heuristic algorithms and conventional reinforcement-learning methods often fail to capture the spatio-temporal dynamics of system resources. This paper proposes PPO-STGNN, a DAG task-scheduling algorithm that integrates proximal policy optimization (PPO) with spatio-temporal graph neural networks (STGNNs). The method uses an STGNN to extract features from both the DAG task topology and the physical cloud-edge-end resource graph, and then optimizes the scheduling policy through PPO to minimize makespan and schedule length ratio (SLR) while improving CPU and memory load balancing. To accelerate convergence, a multi-teacher behavior-cloning mechanism is introduced for pretraining. Experimental results show that PPO-STGNN significantly improves load balancing while maintaining a low completion time, making it suitable for dynamic and heterogeneous cloud-edge- end DAG scheduling scenarios.
[AI-69] BRIDGE: An Open-Source Humanoid Platform via Morphology-Control Co-Design for Physical AI
链接: https://arxiv.org/abs/2609.03497
作者: Jianren Wang,Letian Qian,Zikai Wang,Weiwei Wu,Junjie Zong,Abhinav Gupta,Deepak Pathak
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注:
Abstract:Developing humanoid robots capable of leveraging human behavioral data is essential for general-purpose embodiment, yet conventional development remains bottlenecked by a decoupled paradigm that isolates hardware design from whole-body control. This approach leads to suboptimal systems that compromise human-like fluidity and agility. To bridge this gap, we introduce a data-driven morphology-control co-design framework that optimizes humanoid morphology for human-like movement. To quantify morphological fidelity, we also introduce a novel metric that jointly considers kinematic retargeting fidelity to human motion and dynamic tracking performance. Our framework achieves state-of-the-art (SOTA) performance across all metrics compared to baseline humanoids (Bumi, K1, and Toddlerbot). Finally, we realize this design in Bridge, an open-source, 88cm-tall humanoid platform released alongside its control policy. We demonstrate that Bridge captures human motion data with superior fidelity, exhibiting exceptional performance across foundational locomotion, robust balance, and highly dynamic maneuvers. Videos and open-source materials: this https URL.
[AI-70] GrowPage: On-Demand KV Budgeting for Efficient LLM Reasoning Serving
链接: https://arxiv.org/abs/2609.03494
作者: Qiankun Ma,Yanjiang Zhou,Zinan Xiong,Haofei Wang,Zhen Song,Yang Xiang,Ziyao Zhang,Hairong Zheng
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Long-output reasoning has made the key–value (KV) cache a critical memory bottleneck for efficient LLM serving. Existing KV compression methods usually rely on a predefined per-request budget and adjust only which KV states are retained, leaving the total capacity fixed throughout decoding. However, reasoning workloads exhibit substantial demand variation: different requests require different KV capacities, and the attention demand of an individual request evolves during generation. We introduce \textbfGrowPage, an on-demand KV budgeting framework that treats KV capacity as a runtime resource. GrowPage maintains lightweight dual-timescale query summaries to capture recent and long-term attention behaviors, and uses their relative attention working sets to estimate demand evolution. At each capacity boundary, GrowPage either compresses KV states within the current allocation or acquires an additional physical page when broader demand emerges. By integrating with PagedAttention’s page-level memory abstraction, GrowPage preserves continuous batching and prefix caching. Experiments on reasoning benchmarks across multiple models show that GrowPage achieves a superior performance–throughput trade-off over existing approaches.
[AI-71] Making Every Tool Call Count: Necessary Tool-Evidence Path Rewards for Agent ic Vision-Language Models
链接: https://arxiv.org/abs/2609.03493
作者: Xingming Long,Yu Liu,Zhiwei Yang,Hanqi Feng,Shaojie Zhang,Barnabas Poczos,Chao Jiang,Zhenbo Luo,Lei Jiang,Pei Fu
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Modern vision-language models (VLMs) can directly answer many image-grounded questions, yet they often struggle with complex queries requiring fine-grained visual details or external knowledge. To acquire this missing evidence, agentic VLMs invoke tools such as image cropping, image search, and text search. However, existing training paradigms primarily evaluate tool-use based on final answer correctness, leaving evidence acquisition and utilization insufficiently supervised. This leads to two critical shortcomings: (i) models frequently issue redundant or off-target tool calls that fail to gather necessary evidence, and (ii) even when appropriate tools are called, models often fail to extract the necessary information from the resulting observations. To address these limitations, we introduce the NTEP (Necessary Tool-Evidence Path), a novel annotation scheme that explicitly specifies the essential external evidence and corresponding tool calls for each query. Building upon this, we propose NTEP-R (NTEP Reward), a supervision mechanism ensuring that each tool invocation strictly advances the reasoning process toward the final solution. Specifically, our approach rewards the agent for aligning its pre-call intent with a necessary evidence-seeking goal, and for ensuring the information summarized from the post-call observation aligns with the necessary evidence. Furthermore, we introduce a non-repeated-goal regularizer to penalize redundant calls that revisit satisfied NTEP goals. Extensive evaluations on seven image-grounded benchmarks demonstrate that our 8B-parameter instantiation, NTEP-8B, significantly improves both search-oriented accuracy and tool-use efficiency within a unified three-tool framework. These results highlight the critical value of fine-grained tool-evidence path supervision for training robust agentic VLMs.
[AI-72] Air-Ground Collaborative Vision-and-Language Navigation via Shared Birds-Eye Maps
链接: https://arxiv.org/abs/2609.03483
作者: Shuning Zhang,Liang Li,Yunheng Wang,Tao Wang,Yihang Kang,Renjing Xu
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注: 8 pages, 5 figures
Abstract:Air-ground collaborative Vision-and-Language Navigation (VLN) pairs an unmanned aerial vehicle (UAV) with a global bird’s-eye view and an unmanned ground vehicle (UGV) with a local first-person view, yet the setting remains largely unexplored: existing training-free methods solve single-agent tasks but offer no collaboration mechanism, and a recent CARLA-Air evaluation found no stable cooperative behavior across five state-of-the-art VLA models; naive semantic communication or bidirectional coupling even degrades performance. We establish AGC-VLN (Air-Ground Collaborative VLN), the first training-free baseline for air-ground collaborative VLN. The key insight is that training-free methods decompose navigation into VLM-based semantic reasoning and deterministic geometric execution, exposing a collaboration interface: the UAV’s global view, over which it renders the UGV’s reported pose and the VLM-anchored target as CAR/GOAL markers with distance labels, yielding a shared bird’s-eye map. From this map, the UGV acquires global spatial context its first-person view cannot provide, plans a road-following path with a frozen VLM, and executes it under closed-loop control; in parallel, the UAV runs 3D-SPF, a spatial-search upgrade of SPF that localizes the target in the downward view and flies toward it. On 100 closed-loop episodes in CARLA-Air’s Town10HD scene, AGC-VLN reaches a 77.0% joint success rate, a collaboration gain of +27.0% over the weaker individual agent (the UAV, 50.0%), and exceeds the strongest published single-agent baseline (Travel UAV, 53.0%) by 24.0 points, stemming from the complementarity of the UAV’s global view and the UGV’s road-following execution. Project page: this https URL.
[AI-73] AutoGraphForge: Towards Automated Graph Theory Discovery
链接: https://arxiv.org/abs/2609.03478
作者: Ján Pastorek
类目: Artificial Intelligence (cs.AI); Logic in Computer Science (cs.LO); Combinatorics (math.CO)
备注: 17 pages, 1 figure, 3 tables. Submitted to ITAT 2026 (Information Technologies – Applications and Theory), CEUR Workshop Proceedings. Code: this https URL
Abstract:We report on our ongoing project to develop a computational pipeline, AutoGraphForge, for an automated graph-theoretic conjecturing-refuting-formalizing-proving system. Conjecture generation is counterexample-guided and runs in rounds: a Graffiti3 generator proposes conjectures over a small, evolving snapshot table T (initially a few hundred graphs with their computed invariants) that grows only by counterexamples to its own conjectures. A novelty filter of 559 classical and folklore relations, closed under transitive composition and linear identity substitution, decides via a linear program whether a candidate is already implied by known results. Surviving candidates are tested against a dataset of about 348,000 graphs, unioning the complete House of Graphs invariant export, the exhaustive census of all connected graphs on at most nine vertices, several extremal families (strongly regular, minimal Ramsey, Cayley, cages, barbells, lollipops, spiders), and random models. Counterexample-search algorithms then attack the remainder. Run for several rounds on an HPC cluster, the loop yields 6,522 conjectures that survived the refutation dataset, the novelty filter and every active-search run – among them nontrivial relations between the annihilation number and the edge-cover number for bipartite and regular graphs, which we prove by hand. A subsequent formalization and proving stage deterministically translates each surviving conjecture into a Lean 4 statement skeleton; every candidate proof is kernel-verified against a pinned mathlib4 and our custom invariant preamble. This stage integrates two neural provers – DeepSeek-Prover-V2-671B (served with vLLM) and the Lean-specialised OProver-32B – behind the independent kernel check. It is implemented end-to-end and passes initial sanity checks, with the full pipeline currently running on the cluster.
[AI-74] Beyond “Made with AI”: Visualizing Provenance Density to Mitigate the Transparency Penalty
链接: https://arxiv.org/abs/2609.03460
作者: Qing Zhang,Yifei Huang,Juyoung Lee,Thad Starner,Jun Rekimoto
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:As generative AI makes polished prose cheap to produce, users can no longer rely on fluency as a proxy for truth. We call this failure mode the Fluency Trap: users trust fluent hallucinations while also discounting accurate content once it is disclosed as AI-generated. Binary ``Made with AI’’ labels respond with authorship disclosure, but they do not show what supports a claim. We propose Provenance Density, an evidence-visualization interface that shows the density of verified claims in a text. In a user study with 81 participants, an idealized Provenance Density interface produced a large discernment gap between truth and fabrication ( +4.15 points, d=1.82 ), whereas participants given no signal showed no detectable discrimination. A technical audit with 200 samples shows that retrieval density alone is insufficient; unexpectedly, the Consistency Veto carries most of the discriminative signal on dynamic queries. As AI-generated content becomes indistinguishable from human writing, effective transparency must move from authorship disclosure toward evidence visualization.
[AI-75] he Psychological Costs of Artificial Intelligence Adoption in Software Engineering
链接: https://arxiv.org/abs/2609.03456
作者: Adam Alami,Elda Paja,Abhishek Tiwari
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注:
Abstract:Artificial intelligence (AI) is increasingly used to augment software engineering (SE) workflows. While code generation remains the main use case, organizations are actively seeking AI integration in other practices such as test cases generation and code reviews. Organizational AI adoption strategies seem to focus on tangible outcomes such as productivity. However, AI is a disruptive force, introduced into settings where role identity, team norms, and the sources of job satisfaction were well established before the recent advances in generative AI. Historically, technological disruptions have caused psychological and social strains in workplaces, ranging from anxiety and eroded meaning to deskilling and disrupted professional identities. The assumption that AI for SE is cost-free may not be accurate. Therefore, in this study we sought to understand the psychological costs software professionals experience during organizational AI adoption. We carried out a case study in a large software development services company, one year after the company launched its AI adoption. We collected qualitative data through meetings and semi-structured interviews (N = 21). We found that software professionals experience accountability anxiety, craft identity disruption, meaning and satisfaction erosion, cognitive and workload intensification, and uncertainty distress. Practitioners manage these costs through practices that restore control, mitigate them through protective and identity-preserving adaptations, or absorb them, carrying what neither can resolve. We contribute to AI-human collaboration in SE by repositioning AI adoption as a human transition, not only a technological and organizational one.
[AI-76] Do GUI Agents Know When Not to Act? Enabling Conflict-Aware Termination for Multimodal GUI Agents
链接: https://arxiv.org/abs/2609.03438
作者: Zhaoyuan Huang,Tianjie Ju,Pengzhou Cheng,Zheng Wu,Yansi Li,Chuanbiao Song,Jun Lan,Huijia Zhu,Weiqiang Wang,Zhuosheng Zhang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Graphical user interface (GUI) agents are increasingly used to execute natural-language instructions on user interfaces, yet real users may issue infeasible instructions due to benign mistakes. A reliable agent should not only know how to act, but also when not to act. In this work, we introduce CONFLICTGUI, a benchmark covering instruction-internal conflicts and instruction-GUI context conflicts to study conflict-aware termination. Our evaluation reveals severe execution-biased overcompliance: agents that perform well on feasible tasks often continue to execute blindly under conflicting instructions. To mitigate this behavior, we propose CONFLICTGUARD, an inference-time framework that aligns an agent’s feasibility awareness with its action generation. CONFLICTGUARD contains two coupled components: a feasibility verification protocol that guides the agent to assess instruction logic and GUI-side evidence before acting, and a conditional action modulation mechanism that steers agents from over-compliant execution into termination-oriented behavior. Experiments across five widely-used agents demonstrate that CONFLICTGUARD improves average conflict task success rate significantly, while preserving normal GUI-task performance. These results validate that a lightweight inference-time intervention can substantially boost GUI Agent’s competence to identify inappropriate execution scenarios and refrain from unnecessary actions.
[AI-77] raveL: Transformer-based Multi-view Path Distributional Representation Learning
链接: https://arxiv.org/abs/2609.03427
作者: Fang He,Tao-yang Fu,Wang-chien Lee
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 10 pages
Abstract:Path representation learning (PRL) for road networks has received increasing research attention, due to various path-related applications. Existing works on PRL typically exploit the co-occurrence relationship among road segments and paths to learn a vector as the path representation, without exploring the varied traveler behaviors and the regional correlation on the path. In this work, we propose to learn distributional representations, which provide valuable information for use in path-related applications, by capturing the varied traveler behaviors as well as the various dependencies within regions of road segments. We propose a novel Transformer-based Multi-view Distributional Representation Learning (TraveL) framework to encode a path along with a travel starting time to a distributional representation, which can be used to decode possible samples of on-path traveler behavior. Moreover, by analyzing the regional correlation which reveals various road segment relationships, we propose a regional attention to encode these correlations in a path. Also, we explore the idea of Kolmogorov-Smirnov (K-S) test to compare the sampled traveler behavior against the collected ground truth to facilitate training. Experimental results show that the proposed TraveL model outperforms the state-of-the-art methods on both synthetic and real-world datasets, by 14.7% in Mean K-S distance for travel time distribution estimation, 16.7% in Mean Absolute Error (MAE) for path similarity prediction, and 3.97% in MAE for destination prediction.
[AI-78] DuplexSpeechBench-IFEval: Evaluating Implicit Instruction Following in Full-Duplex Voice Agents
链接: https://arxiv.org/abs/2609.03423
作者: Puneet Mathur,Dinesh Manocha
类目: Artificial Intelligence (cs.AI)
备注: Under Submission
Abstract:Full-duplex voice agents must continuously decide when to listen, backchannel, interrupt, handle speech overlaps, take the floor, and yield. Existing benchmarks largely test these behaviors through explicit turn-management instructions, while deployed agents are often configured through roles or personas from which the appropriate conversational behavior must be inferred. We introduce DuplexSpeechBench-IFEval (DSB-IFEval) for evaluating implicit instruction-following in real-time spoken interaction. (DSB-IFEval) comprises 1,038 test cases spanning eight diverse assistant roles and evaluates five conditioning protocols for instruction-following: default behavior, explicit behavioral instructions, persona-implied behavior, combined persona–rule conditioning, and instruction conflict. We measure real-time floor management using a deterministic Instruction Adherence Score (IAS) and persona-consistent content using LLM-judged Persona Adherence Score (PAS). Across six real-time speech systems, we find architecture-dependent trade-offs. Full duplex models like F-Actor and PersonaPlex are more sensitive to whether conversational behavior is stated explicitly or must be inferred from a persona, with adherence dropping by 9.7% and 4.5%, respectively, under persona-only conditioning. In contrast, GPT-Realtime, MiniCPM-o, and Fun-Audio-Chat strongly adhere to persona-consistent content, but their floor behavior does not adapt across explicit and persona-only instructions and remains constrained on several proactive actions. We further find that even if systems reliably follow conflicting directives to their prescribed persona, they still struggle to override them under safety conflict. These results show that inferring the behavior implied by a role, executing it at the appropriate conversational moment, and resolving competing instructions remain distinct challenges for full-duplex voice agents.
[AI-79] Privacy Robustness and Fairness Trade-offs in Federated Intrusion Detection: Geometric Indistinguishability at the Aggregation Interface
链接: https://arxiv.org/abs/2609.03420
作者: Adrita Rahman Tory,ABM Shawkat Ali,Md Abu Layek,Khondokar Fida Hasan
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 16 pages
Abstract:Federated learning enables privacy-conscious collaboration for network intrusion detection without centralizing sensitive traffic data, yet its deployment in operational environments must simultaneously satisfy three competing requirements: formal differential privacy guaranties, tolerance to Byzantine-adversarial participants, and reliable detection coverage across severely imbalanced attack categories. Existing literature treats these properties as independently composable, an assumption that this paper challenges both theoretically and empirically. In this paper, we study how these requirements interact in class-imbalanced federated NIDS and introduce geometric indistinguishability as a conceptual lens for a regime in which privacy-induced dispersion in client updates can make minority-class signals harder for robust aggregation to preserve. Using UNSW-NB15 as a case study, we evaluate DP-SGD combined with coordinate-wise median under label-flip and model-poisoning attacks, with threat coverage assessed across attack categories. Our results provide initial evidence that the joint use of privacy noise and robust aggregation can disproportionately degrade detection of rare attacks relative to majority classes. We also show that part of the observed collapse under strong privacy can arise from training miscalibration, while a residual performance floor may remain for ultra-rare categories even after epsilon-dependent tuning. These findings motivate studying privacy, robustness, and rare-attack coverage jointly rather than as independently composable properties, and suggest that aggregation-aware modeling and sample-aware evaluation are promising directions for trustworthy federated NIDS.
[AI-80] Dude: A Dual-Detection Multi-Agent System for Paper-Code Discrepancy Detection EMNLP2026
链接: https://arxiv.org/abs/2609.03416
作者: Weijie Liu,Running Zhao,Wenhao Yuan,Jinfeng Xu,Zhanfeng Xu,Xiaoxi Zhang,Edith Cheuk-Han Ngai
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Accepted to EMNLP 2026 Main Conference
Abstract:LLM-empowered paper-code discrepancy detection has received growing concern since the scaling of research submissions exceeds the manual review capability. However, the limited context capacity and one-sided discrepancy detection of existing single-agent LLM paradigms lead to an inferior recall performance in detecting discrepancies. In this paper, we propose Dude, the first Dual-Detection Multi-Agent System for paper-code discrepancy detection. We discover that the granularity asymmetry of the paper-language and code-language introduces over-interpretation and over-reporting challenges in a multi-agent system design for discrepancy detection, resulting in increasing false positives. To address this, we propose a granularity-aligned negotiation and a two-stage salience-filtering mechanism in Dude, which effectively prevents agents from falsely reporting discrepancies. Experimental results in real-world paper-code discrepancy datasets showcase Dude’s significant recall and precision improvement by up to 22.8%, increasing F1 score by up to 18.7% compared to baseline methods.
[AI-81] StrixAE: An Intelligent Agent for Audio Enhancement under Complex Distortion Coupling in Real-World Scenarios
链接: https://arxiv.org/abs/2609.03414
作者: Chenglin Wu,Junjie Wu,Jinhang Chen,Mingyang Chen,Zixu Lin,Jiabian Chen,Xinghao Ding,Xiaotong Tu
类目: ound (cs.SD); Artificial Intelligence (cs.AI)
备注:
Abstract:Audio enhancement in real-world scenarios involves complex distortion couplings and requires personalized enhancement. Existing solutions struggle to address both simultaneously. To improve robustness and enable autonomous operation in such scenarios, we propose StrixAE, an agent based on a multimodal large language model (MLLM). StrixAE leverages the MLLM as a controller to coordinate multiple audio enhancement and personalization models. To further enhance system robustness, reduce artifacts, and improve generalization across diverse real-world scenarios, StrixAE is trained through a two-stage process: first, CoT supervised fine-tuning on AcoustBench to ground basic reasoning and tool invocation; second, Audio Perception Reinforcement Learning (APRL), a reward design specifically tailored for audio restoration pipelines that jointly optimizes format validity, structural coherence, and perceptual quality. Unlike generic RL fine-tuning, APRL introduces structured rewards that enforce executable pipelines and logical section ordering, enabling the agent to produce reliable, interpretable enhancement plans without hallucinated tools. Based on real-world test datasets, our proposed method outperforms most existing open-source and proprietary solutions, achieving state-of-the-art performance across multiple perceptual metrics and demonstrating strong generalization robustness.
[AI-82] Caught in the Story: Narrative Captivity in Multi-turn LLM s Conversation EMNLP2026
链接: https://arxiv.org/abs/2609.03407
作者: Yuhe Wu,Guangyu Wang,Yujie Chen,Jiatong Zhang,Yuran Chen,Yutong Zhang,Xiyin Cheng,Wenpeng Cao,Zhuang Liu,Guang Zhang
类目: Artificial Intelligence (cs.AI)
备注: Accepted by EMNLP 2026 findings
Abstract:People increasingly turn to large language models (LLMs) for everyday advice, making ethically charged interpersonal problems a practical moral-advisory context. Most prior work has studied this context through single-turn judgments or pressure-laden rebuttals, assumptions that poorly match how guidance is sought in real-world contexts. These assumptions leave unclear whether narration alone, without an explicit opposing position, can shift model judgments during multi-turn moral consultation. Yet real-world moral-conflict conversation often elicits one party’s self-justifying account, which can unfold over multiple turns and create information asymmetry. We introduce \textbfnarrative captivity, a failure mode in which a model treats an unopposed one-sided account as complete and aligns with the narrator’s interpretation without seeking missing perspectives. To measure this phenomenon, we build a benchmark of 5,078 interpersonal-conflict scenarios spanning six moral dimensions. Across 17 LLMs, narrative captivity is widespread: end-state judgments under multi-turn narration shift by 25 percentage points on average beyond the matched single-turn baseline. Stage-level analysis identifies preference optimization as a major contributor, while four inference-time strategies provide only partial mitigation. We hope our project fosters LLM advisors that preserve independent judgment in real-world consultation.
[AI-83] A Prompt-Engineering Approach to Develop Scalable Flexible and Real-Time Hybrid Micro-Level Personalization in a General Purpose AI Teaching Assistant
链接: https://arxiv.org/abs/2609.03402
作者: Saptarshi Basu,Sandeep Kakar,Ashok Goel
类目: Artificial Intelligence (cs.AI)
备注: 7 pages, 9 figures, IAAI27 conference
Abstract:Artificial intelligence (AI) teaching assistants powered by large language models (LLMs) offer scalable educational support but often provide limited personalization. This study presents a prompt-engineering-based framework for personalizing general-purpose LLM/RAG-based AI teaching assistants such as Jill Watson across academic disciplines and courses. The framework adapts responses using six learner-specific dimensions: self-assessment, abstraction preference, verbosity preference, perceptual orientation, information processing style, and level of understanding, yielding 96 distinct learner profiles. Student queries are additionally analyzed using Bloom’s Taxonomy to estimate cognitive complexity at the interaction level. Learner attributes and cognitive assessments are encoded in structured prompts that condition the LLM without requiring model retraining. The framework is evaluated through experiments using NLP metrics and a human study with five participants. Results show perceived differences in response style and structure across personalization conditions, with statistical analyses identifying learner attributes associated with measurable response changes. These findings provide preliminary evidence that prompt-based personalization can support adaptive behavior in LLM-powered educational agents.
[AI-84] Spectral Convergence of Random Feature Method in Multiple Dimensions
链接: https://arxiv.org/abs/2609.03401
作者: Pingbing Ming,Hao Yu
类目: Numerical Analysis (math.NA); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Statistics Theory (math.ST)
备注: 48 pages, 1 figure, 2 tables
Abstract:We first prove spectral convergence of the random feature method (RFM) for multidimensional targets in Sobolev, Gevrey, ultra-analytic, and bandlimited classes. The analysis establishes general high-probability approximation estimates in the interpolation scale generated by a kernel integral operator. On a single event determined only by the sampled features, one random space approximates every target in a prescribed source ball; moreover, for each target, a single coefficient vector defines an approximant that attains spectral accuracy simultaneously in all admissible error norms. For both regularity-adapted frequency distributions and uniform distributions on growing frequency windows, the resulting rates range from super-exponential to algebraic, depending on the regularity of the target. Second, we establish abstract error estimates for strong- and weak-form RFM discretizations, thereby converting the preceding approximation bounds into convergence estimates for multidimensional second-order elliptic boundary value and eigenvalue problems. Finally, for random feature matrices (RFMtxs), we prove super-exponential singular-value decay with Fourier features and exponential decay with \tanh features, together with corresponding condition-number lower bounds. The analysis identifies a common mechanism: the same spectral approximation that yields high accuracy also drives severe ill-conditioning.
[AI-85] Fresh Memory Stale Plans: Dependency-Scoped Validation for Distributed LLM -Agent Memory
链接: https://arxiv.org/abs/2609.03340
作者: Evan Chen,Shiqiang Wang,Christopher G. Brinton
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Distributed LLM-agent teams can read the latest shared facts and still act on an obsolete plan. A planner may derive an action from requirement r_3 , another agent may commit r_4 , and an executor may receive r_4 without replacing the plan derived from r_3 . We call this \emphstale-plan execution: state freshness does not establish that the plan authorizing an action remains valid. We introduce PlanFence, a dependency-scoped action-validation protocol. Plans cite the exact public records they used, and an executor validates only the records that can affect the pending external action, replanning once or blocking when validation is incomplete. In 30 controlled live workflows with a post-plan revision, a freshness-only executor acts on the obsolete plan in every task, whereas PlanFence completes all tasks without an invalid action. Controlled replay reveals two conditional boundaries: proactive synchronization yields lower coordination stall at low churn, while PlanFence avoids repeated update-path coordination as churn grows and avoids validating unrelated state as the shared keyspace grows. These are controlled safety and systems-cost results, not general task-accuracy gains.
[AI-86] FlowBalance: Verifier-Grounded Self-Improvement from On-Policy Reasoning Experience
链接: https://arxiv.org/abs/2609.03241
作者: Zixun Huang,Kishan Panaganti,Haitao Mi,Leowei Liang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 28 pages, 7 figures, 10 tables. Code and blog available
Abstract:A reasoning model can improve from its own on-policy experience, but this inner loop is fragile: terminal verifiers provide reliable yet sparse supervision, while dense same-model guidance can reinforce false confidence or overconcentrate learning on a narrow solution mode. We introduce FlowBalance, a verifier-grounded self-improvement method that learns a normalized distribution over complete responses. For each on-policy trajectory, a frozen training-time view of the same policy uses privileged context to produce token-level log-probability gains, which are aggregated into a trajectory-level self-guidance score. FlowBalance calibrates this score with the verifier-derived group advantage: guidance is retained on positive-advantage trajectories, reversed on negative-advantage trajectories, and disabled when the rollout group provides no outcome preference. The resulting energy exponentially reweights a reference policy, and profiled trajectory balance fits the normalized target with one log-partition estimate per rollout group. This realizes outcome-calibrated self-guidance via trajectory balance, without a separate token-level imitation loss. Our analysis establishes within-group contrast preservation, a minimum-change reverse-KL characterization, monotonic verifier control of target reward, and an exact correction against false-positive self-guidance on rejected responses. On mathematical reasoning, FlowBalance improves average performance over FlowRL on both Qwen3-4B and Qwen3-8B, while also improving training speed and stability, avoiding direct OPSD’s response-length collapse, and exhibiting higher correct-strategy diversity in a controlled AIME24 diagnostic.
[AI-87] MasterControl Seventeen Every Time
链接: https://arxiv.org/abs/2609.03209
作者: MasterControl AI Lab
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:We study a governed approach to enterprise analytics: a language model interprets the question, while deterministic policy selects and runs a pre-approved analytical program that returns both results and evidence. We show that this restriction can remain expressive within a defined analytical class, using relational operations plus aggregation, comparison, windows, ranking, and similarity. Fixed meaning, policy, data, and execution rules also make results replayable. Across 440 runs, three 8B models generated SQL and selected tools at runtime, while Qwen3-8B interpreted intent only and policy executed the approved program. None of 330 runtime-planning episodes matched the full answer-and-evidence contract across all test datasets; the policy-executed analyzer matched 110 of 110. This is a configuration-specific result, not evidence that runtime agents cannot succeed under other designs.
[AI-88] ObserverBench: Testing Mechanistic Estimates for Intervention and Control
链接: https://arxiv.org/abs/2609.03026
作者: Vijay Erramilli
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 28 pages, 4 figures. Code, benchmark, leaderboards, and submission interface: this https URL . Frozen artifact release: this https URL
Abstract:Mechanistic interpretability is increasingly used to guide interventions such as activation steering, circuit removal, and safety monitoring. Yet an internal estimate that is accurate on average can still choose a poor action. We present ObserverBench, a benchmark framework for testing whether an internal estimator—an observer—is adequate for the intervention, control, or safety task it directs. Each task fixes the model, information boundary, allowed actions, decision rule, held-out cases, and loss. The benchmark reports estimation accuracy separately from the loss caused by the chosen action. Theory and experiments show why both are needed. In closed-loop control, observer errors matter at the starting point and along directions the allowed intervention can reach. On circuit-intervention tasks in GPT-2-small and Qwen2.5-7B, pairwise observers predict unseen effects more accurately without always choosing better actions; observers trained on action loss choose lower-loss actions. In safety triage, a score that perfectly separates violations can allocate a fixed intervention budget poorly when violations have different costs. Across Qwen2.5-7B, Gemma-2-9B-it, and prospectively frozen Qwen3.5-9B APPS tasks, AUROC can rank monitors differently from deployment loss, and the best information source changes across models. Sparse SAE readouts also trail their layer-matched dense controls on the reported Qwen panels, under disclosed activation-density or checkpoint mismatches. ObserverBench provides fixed task contracts, runnable baselines, and table-based submissions for evaluating interpretability methods through the actions they enable. Comments: 28 pages, 4 figures. Code, benchmark, leaderboards, and submission interface: this https URL. Frozen artifact release: this https URL Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI) Cite as: arXiv:2609.03026 [cs.LG] (or arXiv:2609.03026v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.03026 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Vijay Erramilli [view email] [v1] Wed, 2 Sep 2026 18:01:17 UTC (371 KB)
[AI-89] Evaluating Graph Neural Networks for Change-Criticality Classification in Maritime Navigation Charts
链接: https://arxiv.org/abs/2609.02996
作者: Abhishek Potnis,Jacob Arndt
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
备注: Accepted at IEEE International Geoscience and Remote Sensing Symposium (IGARSS) 2026
Abstract:Graph neural networks (GNNs) are a class of neural networks suitable for learning on graph-structured data. Their application to spatial data is a natural extension, however its relatively unclear which message-passing operations, architectural configurations, and graph representation is best suited for classifying changes to objects in electronic navigational charts (ENCs)–geospatial vector datasets used for marine navigation. Maintaining these datasets is a challenge, and categorizing changes to objects in the ENC based on their significance to navigational safety is of particular importance. Here, we propose to represent these vector navigation datasets as a graph structure where the spatial objects serve as nodes and their spatial and semantic relationships form edges. We encode both the old ENC dataset and new ENC dataset into a pair of graphs and frame the task as a graph-pair classification problem. Building on this representation, we investigate the use of GNN architectures to classify whether the encoded graphs constitutes a critical or non-critical risk to navigational safety. We train and evaluate several GNN architectures and model configurations on ENC changes reviewed by maritime experts. Our results demonstrate that graph-based representations improve the classification of ENC updates, providing a scalable approach for automating or improving ENC maintenance workflows.
[AI-90] oward Collective-Centric Evaluation of Preference Inference for Participatory Democracy
链接: https://arxiv.org/abs/2609.02990
作者: Pierre-Antoine Lequeu,Salim Hafid,Paul Lerner,Nazanin Shafiabadi,Laurène Cave,David Mas,Jean-Philippe Cointet,Benjamin Piwowarski,François Yvon
类目: ocial and Information Networks (cs.SI); Artificial Intelligence (cs.AI)
备注: 7 pages of content
Abstract:To scale up collective decision-making, participatory democracy platforms such as Polis and Remesh enable online deliberation among thousands of participants. However, at this scale, participants cannot review every opinion submitted by others, producing highly sparse voting data that misrepresent patterns of consensus, conflict, and minority support. Platforms therefore increasingly rely on Preference Inference (PI) models to predict missing votes. Yet this automation is not neutral: inferred preferences can artificially amplify, suppress, or reorder existing patterns of support, ultimately reshaping how the outcomes of a deliberation are interpreted. More generally, we lack a systematic understanding of how existing PI methods affect the collective preference landscape. To address this gap, we benchmark several existing PI approaches in this context. Moving beyond conventional user-centric evaluations centered on the accuracy of individual predictions, we introduce a collective-centric evaluation framework that measures whether inferred votes preserve salient properties of the broader preference landscape. We further contribute the largest multilingual dataset of its kind: four consultations spanning over 90k participants, 1M votes, and 22 languages. Our experiments show that models with comparable predictive accuracy can differ substantially in the degree to which they preserve the collective structure. These results demonstrate that accuracy alone is insufficient for evaluating \emphPI in democratic settings. By contributing a novel comprehensive and collective-centric evaluation benchmark for the task of PI, this work aims to support the development of AI systems that scale deliberation without compromising the integrity of its democratic outcomes.
[AI-91] Structure and Implementation of New Practical English Textbooks Driven by Artificial Intelligence
链接: https://arxiv.org/abs/2609.02981
作者: Ya Wang,Lei Zhang,Xueguang Yang,Bo Chen
类目: Artificial Intelligence (cs.AI)
备注: 8 pages, 4 figures
Abstract:Artificial intelligence is changing the form of applied English materials from fixed paper sequences to adaptive learning systems that can diagnose learners, recommend tasks, and provide formative feedback. This paper studies the structure and application of a new practical English textbook driven by artificial intelligence. A five-layer architecture is proposed: knowledge mapping, learner profiling, task generation, feedback orchestration, and teacher-side governance. A prototype was tested on 186 non-English-major undergraduates for eight weeks of teaching. Compared with a static digital textbook, the proposed system increased the unit completion accuracy from 72.4% to 84.9%, raised the average score for speaking tasks by 10.8 points, and reduced the teacher’s correction time by 31.6%. Therefore, an AI-driven textbook can maintain the stability of the curriculum while providing personalised learning paths, rich practice materials and traceable classroom data.
[AI-92] When Optimization Becomes Manipulation: Defending Generative Search against Malicious Generative Engine Optimization
链接: https://arxiv.org/abs/2609.02964
作者: Haozhang Li,Yangguang Shao,Xinjie Lin,Zhong Guan,Mi Zhou,Junzheng Shi
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:
Abstract:This paper focuses on defending generative search engines against malicious Generative Engine Optimization (GEO), which rewrites web documents to match engines’ citation preferences and thereby manipulates generated answers. Recent GEO methods have advanced from hand-crafted rewriting to automated and agentic optimization, substantially increasing the visibility of target documents in generated answers. However, defending against such manipulation poses two major challenges: attack documents remain factually consistent with their originals, rendering fact verification and perplexity filtering ineffective, and the features they amplify equally characterize high-quality benign content. To address these limitations, we propose GEO Defender, a two-stage defense aligned with the attack chain that requires no fine-tuning of the target LLM. GEO Defender consists of Shield Reranker and Training-Free Shield Generation (TFSG). Specifically, Shield Reranker learns a preference-based defensive residual over a frozen base reranker, demoting GEO-rewritten documents while preserving relevance judgments, and TFSG distills defense outcomes into a natural-language experience library that guides the target LLM’s source use at inference. Experiments on two state-of-the-art closed-source LLMs and three open-source LLMs across seven GEO attacks demonstrate that GEO Defender reduces the average attack success rate from 50.32% to 6.20%, retains 94.12% of benign-evidence use, preserves answer quality, and generalizes to unseen attacks from construction instances.
[AI-93] PrivateHub: Contrastive Diffusion Model for Private Sensor-Intensive Environment Data Generation
链接: https://arxiv.org/abs/2609.02958
作者: Jiechao Gao,Yuandong Pan,Jie Wang,Michael Lepech,Bradford Campbell
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:
Abstract:Sensor-intensive environments enable many intelligent services by inferring user applications from heterogeneous data streams. However, not all applications should be exposed: users want some activities to stay private. This creates a tension between inferring applications for useful services and preventing unwanted inference. Existing approaches such as differential privacy and rule-based filtering protect individual streams but cannot address the privacy risk from cross-sensor inference. We introduce Privatehub, which uses contrastive learning within a diffusion model to generate synthetic multi-sensor streams that keep non-private applications detectable while concealing private ones. Privatehub has two stages: App-Conditioned Pre-training (ACP), which conditions the model on multi-sensor data with application embeddings, and App-Aware Fine-tuning (AAF), which separates private from non-private data via contrastive learning. We also define a threat model for the multi-sensor sharing setting. Experiments on three real-world multi-sensor datasets show Privatehub lowers private-application accuracy by 40 to 50% without hurting non-private performance, and stays robust when the attacker retrains on the synthetic data. Subjects: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI) Cite as: arXiv:2609.02958 [cs.CR] (or arXiv:2609.02958v1 [cs.CR] for this version) https://doi.org/10.48550/arXiv.2609.02958 Focus to learn more arXiv-issued DOI via DataCite
[AI-94] Symmetries and Causality: Causal Effect Identification Beyond IID Data
链接: https://arxiv.org/abs/2609.03697
作者: Martin Rabel,Jakob Runge
类目: atistics Theory (math.ST); Artificial Intelligence (cs.AI)
备注:
Abstract:In the natural sciences, symmetries and cause-effect relationships are ubiquitous. Yet for complex machine-learning tasks, like world-modeling in reinforcement learning, they appear difficult to harness. We propose a formal description of statistical systems based on symmetries in data leaving causal mechanisms invariant. The result is an abstract, simple and general mathematical language for causal reasoning. This paper provides formal descriptions of models and queries, setting up this language, and the formal infrastructure and strategies for their mathematically rigorous identification from data within this formalism. This approach reproduces and matches standard theoretical results on IID data and transport of experimental and non-experimental data. But its main purpose is to unify and substantially extend the scope of causal reasoning, in going beyond IID data and in approaching complex causal queries not captured by do- or soft-interventions. This new perspective on causally relevant aspects of data-modeling additionally sheds new light on well-known structures like c-components or hedges but also includes aspects of missing data and is inherently well-suited for the description of transfer and robustness properties.
[AI-95] oolDF: Tool-Integrated Reasoning for Mixed-Authenticity Audio Deepfake Detection EMNLP2026
链接: https://arxiv.org/abs/2609.03620
作者: Taewoo Kim,Young Han Lee,Nam In Park,Chanwoo Kim
类目: Audio and Speech Processing (eess.AS); Artificial Intelligence (cs.AI); Sound (cs.SD)
备注: To appear in Findings of the Association for Computational Linguistics: EMNLP 2026
Abstract:Audio deepfake detection is commonly formulated as clip-level binary classification of single-domain audio. However, real-world manipulated audio can exhibit mixed authenticity, where genuine and manipulated cues coexist across temporal transitions, overlapping sources, or both. This setting requires not only detecting manipulated audio but also localizing the components that provide evidence for the decision. We propose ToolDF, a tool-integrated reasoning framework for mixed-authenticity audio deepfake detection. ToolDF employs an audio large language model as an orchestrator trained with supervised tool-use trajectories. It adaptively analyzes the audio scene, selectively performs source separation, routes components to domain-specific experts, and aggregates their evidence into an interpretable verdict. We further introduce a mixed-authenticity ADD benchmark covering temporal transitions, acoustic overlaps, and hybrid mixtures. Experimental results show that ToolDF achieves the best overall performance on composite-type detection, achieving macro-F1 gains of 3.72 and 14.39 points over the strongest monolithic baseline and a fixed pipeline, respectively, while providing interpretable evidence localized to temporal regions and acoustic sources. Our source code and dataset are publicly available online.
[AI-96] X-Translator: A Real-Time Multilingual Speaker-Aware Speech-to-Speech Translation System
链接: https://arxiv.org/abs/2607.17544
作者: Yuxiang Zhao,Yichi Zhang,Yanjie An,Yanqiao Zhu,Zhanxun Liu,Yushen Chen,Qixi Zheng,Haina Zhu,Yunchong Xiao,Keqi Deng,Shuai Fan,Kai Yu,Xie Chen
类目: Audio and Speech Processing (eess.AS); Artificial Intelligence (cs.AI)
备注:
Abstract:Real-time speech-to-speech translation (S2ST) systems must balance translation quality, latency, speech naturalness, and speaker consistency. Publicly documented S2ST systems have advanced direct, multilingual, streaming, and expressive modeling, while proprietary products and APIs increasingly expose real-time translation capabilities to users. However, practical deployment remains challenging for open and reproducible systems, especially in long-form and multi-speaker conversations where partial ASR hypotheses are unstable, turn boundaries are ambiguous, and target speech must be generated with an appropriate speaker prompt. We present X-Translator, a low-cost modular cascaded S2ST system that combines streaming ASR, machine translation, and prompt-conditioned TTS through a session-level runtime controller. The system uses incremental segment commitment to convert unstable ASR streams into translation-ready units, and an online speaker prompt manager to bind source speech spans to speaker-specific voice prompts for synthesis. We evaluate translation, speech quality, and latency with OpenSTBench, compare against proprietary speech translation APIs as behavioral baselines, measure long-form voice stability, evaluate speaker preservation in multi-speaker conversations, and assess multilingual translation quality. X-Translator provides an open platform for understanding the practical trade-offs of deployment-oriented S2ST. Code and demo are available at this https URL.
[AI-97] Anonymization Not Elimination: Utility-Preserved Speech Anonymization
链接: https://arxiv.org/abs/2604.17000
作者: Yunchong Xiao,Yuxiang Zhao,Ziyang Ma,Shuai Wang,Kai Yu,Jiachun Liao,Xie Chen
类目: Audio and Speech Processing (eess.AS); Artificial Intelligence (cs.AI)
备注:
Abstract:The growing reliance on large-scale speech data has made privacy protection a critical concern. However, existing anonymization approaches often degrade data utility, for example by disrupting acoustic continuity or reducing vocal diversity, which compromises the value of speech data for downstream tasks such as Automatic Speech Recognition (ASR), Text-to-Speech (TTS), and Speech Emotion Recognition (SER). Current evaluation practices are also limited, as they mainly rely on direct testing of anonymized speech with pretrained models, providing only a partial view of utility. To address these issues, we propose a novel two-stage framework that protects both linguistic content and acoustic identity while maintaining usability. For content privacy, we employ a generative speech editing model to seamlessly replace personally identifiable information (PII), and for voice privacy, we introduce F3-VA, a flow-matching-based anonymization framework with a three-stage design that produces diverse and distinct anonymized speakers. To enable a more comprehensive assessment, we evaluate privacy using both acoustic- and content-based speaker verification metrics, and assess utility by training ASR, TTS, and SER models from scratch. Experimental results show that our framework achieves stronger privacy protection with minimal utility degradation compared to baselines from the VoicePrivacy Challenge, while the proposed evaluation protocol provides a more realistic reflection of the utility of anonymized speech under privacy protection.
[AI-98] raceable TTS: Toward Watermark-Free TTS with Strong Traceability
链接: https://arxiv.org/abs/2507.03887
作者: Yuxiang Zhao,Yunchong Xiao,Yushen Chen,Zhikang Niu,Shuai Wang,Kai Yu,Xie Chen
类目: Audio and Speech Processing (eess.AS); Artificial Intelligence (cs.AI)
备注:
Abstract:Recent advances in Text-To-Speech (TTS) technology have enabled synthetic speech to mimic human voices with remarkable realism, raising significant security concerns. This underscores the need for traceable TTS models-systems capable of tracing their synthesized speech without compromising quality or security. However, existing methods predominantly rely on explicit watermarking on speech or on vocoder, which degrades speech quality and is vulnerable to spoofing. To address these limitations, we propose a novel framework for model attribution. Instead of embedding watermarks, we train the TTS model and discriminator using a joint training method that significantly improves traceability generalization while preserving-and even slightly improving-audio quality. This is the first work toward watermark-free TTS with strong traceability. To promote progress in related fields, we will release the code upon acceptance of the paper.
机器学习
[LG-0] Para-Pipe: Exploiting Hierarchical Operator Parallelism of ML Computational Graphs on SoCs
链接: https://arxiv.org/abs/2609.04168
作者: Yujie Zhang,Huiying Lan,Ehsan Aghapour,Zhiyuan Ning,Peng Zan,Weidong Shao,Anuj Pathania,Tulika Mitra
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG); Performance (cs.PF)
*备注: Accepted to IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems
Abstract:As edge-based deep learning applications become more complex, optimizing performance on heterogeneous System-on-Chips (SoCs) presents unique challenges. Traditional pipelining techniques distributing the computation across different on-chip processing units, while effective for throughput, do not address the latency demands posed by modern neural networks with complex interdependencies and extensive operator parallelism. There is a potential in leveraging operator parallelism to enable concurrent execution across multiple processing units, thereby reducing inference latency. However, prioritizing pipelining or parallel execution often necessitates a compromise, where optimizing one performance metric adversely impacts the other. This paper introduces Para-Pipe, a hierarchical mapping framework that integrates intra- and inter-stage operator parallelism within a pipelined architecture. Para-Pipe navigates the trade-off between throughput and latency by selectively fine-tuning parallelism levels within and across pipeline stages. This strategy can significantly reduce inter-processor communication overhead, significantly improving energy efficiency. Our evaluation demonstrates that Para-Pipe generates multiple Pareto-optimal configurations, achieving a balance between throughput and latency on an Amlogic SoC equipped with ARM this http URL CPUs and GPU, as well as the Black Sesame Technology SoC featuring a deep learning accelerator and two DSPs. More importantly, throughput-optimized configurations under Para-Pipe on Amlogic SoC show an average energy efficiency improvement of 11.0% over purely pipelined strategies and 23.3% relative to non-pipelined parallel execution.
[LG-1] Prospective Coding Improves Learning in Deep Continuous-Time Recurrent Networks
链接: https://arxiv.org/abs/2609.04134
作者: Shivang Rawat,Mirko Morello,Flaviano Morone,David J. Heeger
类目: Machine Learning (cs.LG); Neural and Evolutionary Computing (cs.NE); Neurons and Cognition (q-bio.NC)
*备注:
Abstract:Temporal integration gives continuous-time recurrent networks memory, but in deep stacks it also delays bottom-up signals and attenuates top-down errors. We develop Recursive Quadrature Filters (RQFs), biologically motivated complex-valued temporal filters that are a special case of diagonal state-space models (SSMs), and ask whether this failure mode can be addressed by making each layer’s bottom-up input prospective. Starting from an energy model, we derive the RQF dynamics and show that each RQF is a band-pass filter whose learnable parameters control its tuning frequency and bandwidth. We then make each layer’s bottom-up input prospective using a parameter-free two-tap update that leaves the recurrent transition and parallel scan unchanged. We extend this correction to general diagonal SSMs and show that it mitigates depth-dependent gradient attenuation when temporal gradients are truncated, i.e., spatial-only backpropagation. We evaluate the intervention in RQFs, S5, and ORGaNICs (a nonlinear gated RNN) trained using full backpropagation through time (BPTT) and spatial-only backpropagation. Under full BPTT, prospective variants match or outperform their non-prospective controls in every model and configuration. A non-residual width-32 six-layer RQF reaches 96.09% accuracy on raw-audio Speech Commands with 31.9k parameters; a width-64 six-layer RQF reaches 83.56% on the 16,384-step Path-X task. These results identify RQFs as a parameter-efficient recurrent substrate and prospective-input coding as an input-side correction for deep continuous-time recurrent networks.
[LG-2] Constant regret in general games via higher-order optimism
链接: https://arxiv.org/abs/2609.04113
作者: Omar Abbadi,Rida Laraki,Panayotis Mertikopoulos
类目: Machine Learning (cs.LG); Computer Science and Game Theory (cs.GT)
*备注: 42 pages, 1 figure
Abstract:We introduce an uncoupled learning algorithm which, when employed by all players of an arbitrary N -player normal form game with up to K actions per player, guarantees O(N^3\log^2 K) individual regret, uniformly over the horizon of play. The proposed algorithm - which we call higher-order optimism with discounting (HOOD) is a variant of optimistic follow-the-regularized-leader (OptFTRL) that combines a discounted (N+1) -th order predictor with entropic regularization over a suitable “lifting” of the game’s strategy space. This combination of ingredients is purposefully designed to dampen large oscillations of the induced sequence of play in a controlled manner, removing in this way a key stumbling block of previous attempts to achieve constant regret in general games. Our approach bears several striking similarities to the concurrent - and completely independent - work of Liu, Farina, and Ozdaglar (arXiv:2608.31166), who very recently derived an O(N^21\log^4 K) regret bound through the use of higher-order optimism and an exponential moving average estimator.
[LG-3] Hardware-Aware FP4 FlashAttention-4
链接: https://arxiv.org/abs/2609.04105
作者: Robert Hu
类目: Machine Learning (cs.LG)
*备注:
Abstract:Blackwell’s 4-bit floating-point (FP4) tensor cores do not automatically make attention faster because softmax conversion and on-chip dependencies dominate once its matrix products shrink. We address this with \emphDirect-P for noncausal inference and a causal path that passes the forward quantization directly into backward. Direct-P maps scores directly to FP4 probabilities and reaches up to 2.13 \times the bfloat16 (BF16) forward throughput on an NVIDIA GB200. The causal path reconstructs probabilities from saved quantized queries and keys and uses 8-bit floating-point (FP8) gradient operands, accelerating a complete single-GPU 8-billion-parameter update by up to 1.14 \times . Matched distributed training retains FP8 probabilities and values; every tested MXFP4 probability/value training trajectory diverges.
[LG-4] Conditioning Degenerate Diffusion Models
链接: https://arxiv.org/abs/2609.04090
作者: Uğur Aydın,Tamer Başar
类目: Machine Learning (cs.LG)
*备注:
Abstract:Current conditioned generative models heavily rely on score functions for guidance during training. When the generative model is a diffusion process with a singular diffusion coefficient and the underlying (conditional) densities either do not exist or are not smooth, we use causal optimal transport to define \emphapproximate loss functions that identify a minimum-entropy control for guidance under minimal assumptions. Our approach relies on causal optimal transport and its characterization through the predictable representation property of (conditioned) diffusion processes whose associated martingale problem is well posed, à la Üstünel.
[LG-5] he Head Complexity of Boolean Functions in Single-Layer Attention
链接: https://arxiv.org/abs/2609.04046
作者: Rajmohan Rajaraman,Ravi Sundaram,Amanuel Tesfaye
类目: Computational Complexity (cs.CC); Machine Learning (cs.LG)
*备注: 32 pages, 0 figures
Abstract:What can a single layer of self-attention compute? We study head complexity: the minimum number of attention heads required to compute a function in a one-layer attention-only model. We establish an exact hierarchy under this measure: k heads compute k -bit parity but cannot compute (k+1) -bit parity. The lower bound is unconditional in the two resources a transformer might otherwise exploit; it holds at unbounded embedding dimension and unbounded numerical precision. The proof rests on an alternating-sum obstruction: after clearing the softmax denominators, every monomial in the resulting decision polynomial omits at least one of the k+1 input bits, forcing its correlation with parity to vanish. The same obstruction yields lower bounds for related tasks, including the well-studied multi-hop induction-head task. We also establish compactness bounds for embedding dimension and numerical precision. Specifically, a compactness theorem shows that any function computable at all can be computed with embedding dimension and precision bounded by the discrete data of the task, namely, head count, alphabet size, and length. Thus, potentially unbounded dimension or precision provably cannot substitute for heads. Finally, we derive nearly matching universal bounds for general binary functions: 2^n heads suffice to compute every n -bit binary function, with one head per monomial in its multilinear expansion, while a counting argument shows almost all such functions require \Omega(2^n/n^2) heads. This lower bound matches the upper bound to within a \operatornamepoly(n) factor, even when dimension and precision are unbounded. Together, these results characterize head requirements for Boolean computation in this model. Comments: 32 pages, 0 figures Subjects: Computational Complexity (cs.CC); Machine Learning (cs.LG) Cite as: arXiv:2609.04046 [cs.CC] (or arXiv:2609.04046v1 [cs.CC] for this version) https://doi.org/10.48550/arXiv.2609.04046 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-6] A location-invariant estimator of extremal quantile treatment effects for heavy-tailed distributions
链接: https://arxiv.org/abs/2609.04018
作者: Xin Yu,Shuwei Huang,Jicheng Liu,Jielin Tang,Bolin Wang,Yunxiao Zhang,Tian Zhao
类目: Machine Learning (cs.LG); Applications (stat.AP); Methodology (stat.ME)
*备注:
Abstract:Quantile treatment effects (QTEs) measure the effect of a treatment on the distribution of an outcome, and their estimation at extreme quantile levels is of central interest in applications where the target quantiles lie far beyond the range of the data. For heavy-tailed potential outcomes, existing extremal QTE estimators rely on extrapolation combined with a causal extreme value index (EVI) estimator, but the resulting estimator is not invariant under a common location shift of the potential outcome distributions, even though the population QTE is. We address this issue in two steps. First, we adapt the location-invariant Fraga estimator of the EVI to the causal setting using inverse propensity score weighting. Second, we replace the original extrapolation formula with a difference-based scheme, under which the location parameter cancels when quantile differences are taken. The resulting QTE estimator is therefore location invariant. We establish the consistency and asymptotic normality of the proposed extremal QTE estimators, and provide a consistent variance estimator, leading to asymptotically valid inference. A simulation study confirms the location invariance, the stability with respect to the threshold, and the coverage of the proposed methods.
[LG-7] Differentiable Hybrid Modelling for Learning and Optimising Chemical Transport Processes from Experimental Data
链接: https://arxiv.org/abs/2609.04011
作者: Arthur Jessop,Mohammed Alsubeihi,Ben Moseley,Ashwin Kumar Rajagopalan
类目: Computational Engineering, Finance, and Science (cs.CE); Machine Learning (cs.LG); Numerical Analysis (math.NA)
*备注:
Abstract:Reliable transport models are essential when modelling and optimising many chemical engineering processes, yet, most models assume hand-picked constitutive laws which may not reflect reality, and often assume initial conditions are known exactly. Both restrictions can significantly bias model predictions and lead to systematic error when used in predictive and control settings. Black-box neural surrogate alternatives for modelling can better match real example data, but are confined to the task they were trained on and cannot be interrogated for physical consistency. Here we introduce a general-purpose differentiable hybrid modelling framework for transport processes, specifically for the case of population balance equations. Our framework integrates a JAX finite volume population balance solver with learnable neural network components which are trained to both discover constitutive laws and fit initial conditions from real experimental data, allowing us to better model real experimental transport systems. Furthermore, we use our framework for process optimisation, using its differentiability to allow us to direct optimising experimental settings for quantities of interest. This work highlights the huge potential of such differentiable hybrid modelling frameworks for learning and optimising any given chemical separation which involves mass, energy, and/or momentum transport.
[LG-8] Unlocking Lossless Speedups in LLM s via Discrete Diffusion
链接: https://arxiv.org/abs/2609.04010
作者: Subham Sekhar Sahoo,Lingjie Chen,Khiem Pham,Jonathan Geuter,Chaitanya Dwivedi,Varad Pimpalkhute,Yash Akhauri,Alexander Moreno,Mikhail Yurochkin,Zhenting Wang,Mostafa Elhoushi,Nolan Dey,Shane Bergsma,Joel Hestness,John Thickstun,Eric Xing,Zhengzhong Liu
类目: Machine Learning (cs.LG)
*备注: Code and Checkpoints at this https URL
Abstract:Large Language Models (LLMs) owe much of their success to next-token prediction (NTP), but their autoregressive (AR) structure requires slow, sequential token generation. To overcome this bottleneck, we introduce diffusion-augmented LLMs, a new class of models that defines an AR model distribution while using diffusion to draw multiple tokens in parallel from that distribution. We decouple the parameters of these models into two sets: AR weights, trained using the standard NTP objective, and lightweight diffusion weights, trained to generate multiple tokens simultaneously. The diffusion weights are learned through a simple Diffusion Distillation phase that adds negligible overhead to existing LLM training pipelines. We also introduce \Psi -Spec, a family of samplers that enables lossless acceleration and inference-time scaling at a fixed context length. Unlike speculative decoding, our method requires no separate draft model. Unlike diffusion LLMs (d-LLMs), it accelerates generation without sacrificing the quality of the underlying AR model. The resulting models, called Uno, can be trained from scratch or built by augmenting existing open-weight AR LLMs. Uno achieves higher throughput than leading speculative-decoding methods at every evaluated batch size and delivers up to 3\times speedups over the base AR model, including at the largest batch size supported by the device. Notably, our 8B Uno model outperforms the leading open d-LLM, the 26B DiffusionGemma, and the proprietary Mercury 2 across all evaluated benchmarks in agentic tool use, coding, and long-context reasoning. We release code and checkpoints at: this https URL
[LG-9] RobustSeiz: An Open-Source Framework for Benchmarking the Robustness of EEG Seizure Detection Models
链接: https://arxiv.org/abs/2609.04007
作者: Mohammad Mohammadi,Alireza Zarei
类目: Machine Learning (cs.LG); Signal Processing (eess.SP)
*备注: 28 pages, 13 numbered figures, 9 tables; full-page graphical abstract and supplementary material included. Submitted to the Journal of the American Medical Informatics Association (JAMIA)
Abstract:Despite strong performance on held-out electroencephalography (EEG) data, seizure detectors may fail under real-world acquisition variability, artifacts, and adversarial inputs. We introduce RobustSeiz, an open-source, model-agnostic framework that provides a standardized, reproducible protocol for stress-testing and comparing seizure detectors under controlled, clinically motivated distribution shifts before deployment. We standardize four public scalp-EEG corpora (CHB-MIT, TUSZ, Siena, and SeizeIT1) into BIDS-EEG trees and evaluate subject-independent detectors on held-out splits. Environment, noise, and adversarial transforms are swept over predefined hyperparameter grids. Each run reports sample- and event-level sensitivity, precision, F1, false positives per 24 h, Lead and Lag onset timing, and Monte Carlo dropout predictive agreement. RobustSeiz includes a Dockerized GPU pipeline, experiment registry, and full-evaluation and research-subset modes. We demonstrate the framework with a contemporary seizure detector on TUSZ across the complete implemented shift grid; an AWGN analysis illustrates how perturbation severity changes detection quality, onset timing, and predictive agreement. RobustSeiz provides a shared benchmarking standard for evaluating seizure-detector robustness under realistic clinical stressors, extending pre-deployment assessment beyond clean-data accuracy.
[LG-10] OSR: Output Space Redistribution for Adaptive Label Removal in Classification Models
链接: https://arxiv.org/abs/2609.03972
作者: Minyi Peng,Darian Gunamardi,Ivan Tjuawinata,Yongsen Zheng,Kwok-Yan Lam
类目: Machine Learning (cs.LG)
*备注: Accepted by ICA3PP 2026
Abstract:Label removal occurs frequently in classification systems with evolving taxonomies, where categories must be dynamically updated or eliminated. To accommodate such changes, classification models must adapt accordingly. Existing solutions, broadly categorized as retraining-based and feature-space-adjustment-based, share common limitations despite their variations, including reliance on access to original data, substantial computational and storage costs, inconsistent results, poor scalability, and degradation of model utility. To address this, we propose a novel approach that leverages statistical redistribution in the output space to approximate the post-removal confidence vectors of a retrained model. Applicable as a modular output filter, our method bypasses the burden of feature-space adjustments or loss-function convergence, alleviating scalability limitations. Furthermore, by requiring only existing labels and prior output confidences, the method potentially mitigates privacy concerns inherent to data-dependent solutions. Extensive experiments demonstrate competitive performance against full retraining, with improvements in computational efficiency and privacy preservation across several classification tasks.
[LG-11] Beyond Endpoint Scores: Time- and Capacity-Conditioned Evaluation of Continual Knowledge Updating
链接: https://arxiv.org/abs/2609.03900
作者: Heejin Choi
类目: Machine Learning (cs.LG)
*备注: 13 pages, 4 figures. Extended preprint
Abstract:Continual knowledge-updating methods are often declared superior from one final checkpoint and one conventional adapter rank. We show that this can be insufficient to identify the better operating point. Holding a periodic hierarchy fixed, we compare it with cumulative replay over a 24-month Wikidata stream while varying evaluation month, replay LoRA rank, and query formulation. The apparent winner changes across this region: on Qwen2.5-1.5B, the hierarchy’s 5.0-point advantage over rank-8 replay becomes an 11.6-point deficit against rank-72 replay, and at high ranks a consolidation-aligned endpoint can suggest a tie while time-averaged replay leads by 9-13 points. The same rank-conditioned reversal appears on Llama-3.2-1B and held-out paraphrases. These results show that method ranking in continual updating can depend jointly on when performance is measured and how much replay-side adaptation capacity the baseline receives. We therefore propose reporting trajectories and capacity sweeps, and declaring a robust winner only when the ordering is stable across the evaluation region; otherwise, comparisons should report winner regions and retention-stability-cost frontiers. Under this protocol, the periodic hierarchy is a lower-update-cost operating point, not a quality winner. Comments: 13 pages, 4 figures. Extended preprint Subjects: Machine Learning (cs.LG) Cite as: arXiv:2609.03900 [cs.LG] (or arXiv:2609.03900v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.03900 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-12] High-Dimensional Learning Dynamics of Attention-Indexed Models
链接: https://arxiv.org/abs/2609.03858
作者: Yizhou Xu,Margarita Sagitova,Lenka Zdeborová,Florent Krzakala
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:Attention mechanisms are central to modern foundation models, yet their training dynamics remain poorly understood, especially when the attention matrices have extensive rank. In this work, we study attention-indexed models, a broad framework that can represent multi-layer and multi-head attention architectures. First, we show that, in a suitable high-dimensional limit, the population-loss landscape is characterized by a finite set of trace order parameters. In contrast, online stochastic gradient descent (SGD) is governed by an infinite hierarchy of matrix moments, which we show can be exponentially well-approximated by a finite truncated system. Second, this framework reveals that attention parameterization itself can act as an architectural implicit bias. Direct optimization of an attention matrix S\in\mathbbR^d\times d can remain trapped in an uninformative state. Tied attention ( S=WW^\top ) induces an automatic symmetry-breaking mechanism and yields weak recovery in \Theta(d^2\log d) samples. For untied attention, S=UV^\top , we uncover a fast-slow mechanism: the pre-activation mean first evolves on a fast timescale, while the overlaps evolve on a slower one. Weak recovery on the \Theta(d^2\log d) scale occurs when the state selected by the fast dynamics breaks the initial symmetry.
[LG-13] Pushing the (Decision) Boundaries: Dynamically Calibrating Differentially Private Noise to Explainability in Federated Learning
链接: https://arxiv.org/abs/2609.03851
作者: Michael Khavkin,Kichang Lee,Jaeho Jin,JeongGil Ko,Eran Toch
类目: Machine Learning (cs.LG)
*备注: 21 pages, 9 figures
Abstract:Federated Learning (FL) with Differential Privacy (DP) is increasingly adopted to preserve data confidentiality in distributed machine learning. However, DP noise distorts learned representations and degrades explanation fidelity, limiting differentially private FL where trustworthy explanations are required, such as assistive clinical diagnosis. Prior work adapted DP noise with static feature-importance signals, restricting explainability to post hoc analysis and precluding noise calibration to explanation quality during training. We propose XCal-FL, a closed-loop, explainability-driven local training algorithm for image classification in cross-silo FL that dynamically calibrates DP noise from three complementary signals: (1) prediction logit variations, measuring causal influence on model confidence, (2) counterfactual margins, capturing decision-boundary sensitivity, and (3) saliency concentration, quantifying spatial coherence of model attention, while enforcing formal DP guarantees via adaptive privacy accounting. Experiments on three medical imaging datasets across varying FL configurations show that XCal-FL yields more accurate and interpretable global models, improving predictive performance by over 10% and explanation fidelity by up to 5 \times over static-noise FL, and outperforming state-of-the-art adaptive DP methods in fidelity. XCal-FL also achieves higher privacy-budget efficiency, turning each unit of cumulative privacy loss into larger gains in both accuracy and explanation fidelity. Our analysis further reveals that, unlike predictive performance, which scales roughly linearly with privacy loss, explanation fidelity exhibits non-linear dynamics. These findings suggest explainability is a distinct dimension of the privacy trade-off that cannot be inferred from utility alone, with implications for training and privacy-budget allocation in decision-critical applications.
[LG-14] EF1-Constrained Nash Social Welfare with Identical Additive Valuations: Complexity Guarantees and Experiments
链接: https://arxiv.org/abs/2609.03846
作者: Zih-Sian Yang,Yi-Hao Chen,Yu-Te Kuan,Cheng-Jui Wu,Chuang-Chieh Lin,Po-An Chen
类目: Computer Science and Game Theory (cs.GT); Machine Learning (cs.LG)
*备注:
Abstract:We study the allocation of indivisible goods among agents with identical additive valuations, focusing on envy-freeness up to one good (EF1) and Nash social welfare (NSW). Since every maximum-NSW allocation is EF1 under additive valuations, the associated threshold problem inherits the known strong NP-hardness of NSW maximization under identical additive valuations and is strongly NP-complete. We therefore focus on welfare guarantees satisfied by arbitrary EF1 allocations. Although every such allocation is known to achieve an e^-1/e -approximation to the unrestricted optimal NSW, we identify conditions yielding stronger guarantees. Under uniform valuations, every EF1 allocation is NSW-optimal. Under an \varepsilon -small-item condition, every EF1 allocation achieves an explicit approximation ratio \rho_n(\varepsilon) satisfying \rho_n(\varepsilon) = 1-O(\varepsilon^2) as \varepsilon\to 0 for fixed n . We further consider the stronger sequential requirement that EF1 be maintained after every item assignment. For this setting, we propose \emphPriorityNet, a deep reinforcement learning framework trained using Proximal Policy Optimization and equipped with prospective EF1 action masking. The mask restricts every decision to assignments that preserve EF1, thereby guaranteeing prefix-wise EF1 by construction without post-processing repair. Across 3,000 test instances in each of the offline and random-order online regimes ( n\in[2,20] and m\in[5,100] ), PriorityNet attains mean normalized \operatornameNSW values of 0.9911 and 0.9701 , respectively. Relative to offline Longest Processing Time (LPT) and online least-valued-bundle baselines, it achieves instance-wise win-minus-loss rates of +27.10% and +17.87% , while matching the offline baseline’s mean normalized welfare to four decimal places and modestly improving the online mean from 0.9694 to 0.9701 . Subjects: Computer Science and Game Theory (cs.GT); Machine Learning (cs.LG) Cite as: arXiv:2609.03846 [cs.GT] (or arXiv:2609.03846v1 [cs.GT] for this version) https://doi.org/10.48550/arXiv.2609.03846 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-15] Multi-step Proximal Policy Improvement in Offline Reinforcement Learning
链接: https://arxiv.org/abs/2609.03842
作者: Soohyun Choi,Seonvin Cho,Songnam Hong
类目: Machine Learning (cs.LG)
*备注: Preprint; 28 pages, 7 figures, and 13 tables
Abstract:Offline reinforcement learning (RL) must reconcile two competing requirements: policy updates should stay near dataset-supported actions to keep value estimates reliable, yet meaningful gains often require moving beyond the behavior distribution. We develop a geometric view of offline actor updates by modeling policies as a probability manifold endowed with a chosen metric geometry. Under this lens, a broad class of offline actor objectives can be interpreted as a single proximal policy improvement step (SPI), i.e., an implicit discretization of a manifold gradient flow induced by a critic-defined energy. Building on this insight, we propose multi-step proximal policy improvement (MPI), a plug-in refinement mechanism that composes sequential re-centered proximal steps. MPI enables controlled policy improvement beyond dataset support while retaining proximal control at each refinement. The framework accommodates multiple policy geometries and admits practical instantiations for deterministic and diagonal-Gaussian policies. Experiments on D4RL benchmarks show that small numbers of MPI refinements improve strong offline baselines, including TD3+BC, ReBRAC, and IQL, on many tasks. Focused diagnostics further distinguish re-centered refinement from fixed-objective update scheduling and characterize limitations under critic error.
[LG-16] A Peer-Relative Representation Learning Framework for Energy Inefficiency Identification in Mobile Network Sites
链接: https://arxiv.org/abs/2609.03809
作者: Eliud Nyakweba Koto,Jaco du Toit,Adham Stoltz,Johan du Preez
类目: Machine Learning (cs.LG)
*备注: 22 pages, 6 figures
Abstract:Energy consumption is one of the largest operational expenditure items for mobile network operators, yet site-level energy inefficiencies such as faulty cooling controllers, idle radio equipment, and parasitic auxiliary loads often remain undetected because no ground-truth inefficiency labels exist and historical measurements may already contain embedded inefficiencies. This study proposes an unsupervised peer-relative approach based on the premise that sites with similar structural and operational characteristics should exhibit comparable energy consumption. To capture these relationships, a novel energy-aware Minimum Distortion Embedding (MDE) formulation is introduced that extends the standard MDE objective with an energy-based repulsion mechanism. This encourages sites with anomalously high energy consumption relative to comparable peers to become displaced from their local neighbourhoods in the embedding space. The resulting low-dimensional representation simultaneously preserves structural similarity and encodes energy-related deviations, enabling the identification of potentially inefficient sites through peer-relative comparison. The derived anomaly scores provide a practical mechanism for prioritising field investigations, allowing mobile network operators to focus engineering resources on sites most likely to yield energy savings. Experimental results demonstrate that the proposed approach outperforms conventional anomaly detection baselines and provides a robust foundation for large-scale energy-efficiency optimisation in mobile networks.
[LG-17] From Ordered Bernoulli Levels to Critical-Line Geometry: Integer Quantization Bernoulli Residual Phase and Prime-Power Spectra
链接: https://arxiv.org/abs/2609.03801
作者: Y. Kenan Yılmaz
类目: Machine Learning (cs.LG); Number Theory (math.NT)
*备注: 20 pages, 8 figures. Includes exact algebraic constructions, finite-sample numerical controls, and conditional Weyl diagnostics
Abstract:We study the ordered Bernoulli-word kernel f(p,n,k)=p^k(1-p)^(n-k) and the geometry generated by its inverse-integer level sets. The binary level 2^(-n) selects p=1/2 as the unique real split-independent anchor. Under complement-preserving complex continuation, the pair becomes z=1/2+iu and 1-z=1/2-iu, producing a conjugation-symmetric vertical geometry before any zeta-function input is introduced. The quadratic coordinate Q(z)=z(1-z)=1/4+u^2 has a sharp minimum at the central point and admits an exact integer quantization. For critical-line zero ordinates gamma_k, the induced levels L_k=1/4+gamma_k^2 are decomposed exactly as L_k=N_k+delta_k, where N_k is the nearest integer and delta_k is a periodic first-Bernoulli residual. Circularization gives Z_k=exp(2 pi i delta_k), isolating gamma_k^2 mod 1 as the residual phase variable. Unique factorization resolves the integer shells into prime-generator coordinates, while a distinct complex exponent s lifts the same construction to the Dirichlet atoms m^(-s), linking the Dirichlet-series and Euler-product assemblies. Exact identities, classical zeta connections, numerical controls, and open conditional Weyl tests are kept explicitly separate. No proof of the Riemann Hypothesis is claimed.
[LG-18] Landmark-Based Discrimination of Injury-Associated Athlete-Sessions from Minute-Resolution Multimodal Football Monitoring Data
链接: https://arxiv.org/abs/2609.03790
作者: Evangelos Chatzidimitriou,Konstantinos Tserpes
类目: Machine Learning (cs.LG)
*备注: 12 pages, 3 figures, 7 tables
Abstract:Athlete monitoring data may be recorded minute by minute throughout a match or training session, while injury information may only indicate whether the entire session was injury-associated. This creates a modelling problem: assigning the same session-level label to every minute would imply that injury status is known at each exact time, even though within-session injury onset is unknown. Our novelty is a fixed-landmark, one-representation-per-athlete-session formulation that directly addresses this mismatch. Instead of labelling every minute, we construct one representation per athlete-session at each landmark using information observed up to that point. This keeps the target at the session level and avoids unsupported minute-level injury supervision. A landmark is a fixed time point within the same session, such as 10, 20, or 30 minutes. At each landmark, we assess whether the whole session is injury-associated or non-injury-associated and examine how discrimination changes as more within-session information becomes available. Using 2020 SoccerMon data, we analyse 3,743 athlete-sessions from 48 elite women’s football athletes, including 22 injury-associated sessions from five athletes. We evaluate pre-session, cumulative, dynamic, and combined representations with athlete-disjoint validation, athlete-cluster bootstrap uncertainty, common-cohort sensitivity analysis, alternative negative-athlete fold allocations, equal-athlete weighting, and Logistic Regression, Random Forest, and XGBoost benchmarks. Primary CUM+DYN Logistic Regression yields ROC-AUC 0.367-0.607 and PR-AUC 0.0080-0.0150 across landmarks, with wide uncertainty. PRE-containing representations show higher point estimates at several landmarks but remain uncertain. Comments: 12 pages, 3 figures, 7 tables Subjects: Machine Learning (cs.LG) Cite as: arXiv:2609.03790 [cs.LG] (or arXiv:2609.03790v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.03790 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Evangelos Chatzidimitriou [view email] [v1] Thu, 3 Sep 2026 13:00:23 UTC (843 KB) Full-text links: Access Paper: View a PDF of the paper titled Landmark-Based Discrimination of Injury-Associated Athlete-Sessions from Minute-Resolution Multimodal Football Monitoring Data, by Evangelos Chatzidimitriou and 1 other authorsView PDFHTML (experimental)TeX Source view license Current browse context: cs.LG 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?) 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
[LG-19] From Nowcasting to Forecasting: Adapting a Reanalysis-Trained
链接: https://arxiv.org/abs/2609.03763
作者: Mikko Partio,Leila Hieta,Ossi Laine
类目: Machine Learning (cs.LG); Atmospheric and Oceanic Physics (physics.ao-ph)
*备注:
Abstract:Accurate cloud-cover forecasts are important for temperature prediction, radiation forecasting, and solar-power operations. Short-range forecasting methods can preserve observed cloud placement during the first forecast hours, but their skill decreases when cloud fields evolve through formation, dissipation and deformation. Longer lead times require accounting for atmospheric evolution, but operational numerical weather prediction (NWP) forecasts may not accurately represent the satellite-observed cloud state at initialization. We develop CloudCast v2, a machine-learning model for 12-hour cloud-cover forecasting from observation-based initial conditions. The model is first trained on the Copernicus European Regional Reanalysis (Ridal2024) to learn cloud-evolution dynamics, and is then adapted to satellite-derived cloud fields using conditional flow matching (Lipman2023), a generative method that transforms noise into cloud-cover forecasts conditioned on the observed initial cloud fields and NWP inputs. CloudCast v2 reduces mean absolute error by 10% relative to its predecessor, CloudCast v1 (Partio2025), over the 1-12 h range. It also overtakes CloudCast v1 in fractions skill score, a neighborhood-based measure of spatial agreement, after approximately 3-6 h, depending on the cloudiness category. These results show that observation-initialized machine-learning forecasts can extend beyond the usual 1-3-hour nowcasting range while retaining spatial detail from satellite cloud fields.
[LG-20] Projected Riemannian Gradient Descent for the Bures-Wasserstein Barycenter: Dimension-Independent Linear Convergence at Unit Step Size
链接: https://arxiv.org/abs/2609.03762
作者: A. Afham
类目: Machine Learning (cs.LG); Optimization and Control (math.OC); Quantum Physics (quant-ph)
*备注: 33 Pages, 3 figures. Comments welcome!
Abstract:The computation of the Bures-Wasserstein (BW) barycenter of an ensemble of positive definite matrices arises throughout machine learning, optimal transport, and quantum information. Riemannian gradient descent (RGD) at unit step size – the fixed-point iteration used in practice – converges rapidly, yet existing analyses present a dichotomy: unit-step guarantees carry worst-case exponential dependence on the dimension, while dimension-independent guarantees require small step sizes that forfeit the empirical speed. We resolve this dichotomy, not by improving the guarantees for unit-step RGD, but by proposing a Projected RGD algorithm that achieves dimension-independent linear convergence at unit step size. The achieved rate, (1 - \kappa^-3/2) , where \kappa is the condition number of the ensemble, also polynomially improves on the best small-step guarantee ( \kappa^3/2 versus \kappa^5/2 iteration complexity). The crux is a novel Projection Lemma: clipping the eigenvalues of a positive matrix to an interval [\alpha, \beta] is the closed-form, non-expansive (1-Lipschitz) BW-metric projection onto the set \S : \alpha I \leq S \leq \beta I\ – a statement which, unlike its known one-sided counterpart, does not follow from convexity. The projection is moreover free: it reuses an eigendecomposition the next iteration must perform in any case, so the projected and unprojected iterations cost the same per step. The same analysis covers the invariant matrix projection problem of Brahmachari et al. (2025), whose fixed-point algorithm we identify as unit-step RGD on a totally geodesic submanifold, thereby extending the dimension-independent guarantee to that setting verbatim.
[LG-21] Genetic Algorithms for Tractable Bayesian Network Fusion via Pre-Fusion Edge Pruning GECCO2025
链接: https://arxiv.org/abs/2609.03724
作者: Pablo Torrijos,José A. Gámez,José M. Puerta,Juan A. Aledo
类目: Neural and Evolutionary Computing (cs.NE); Machine Learning (cs.LG)
*备注: 9 pages. Presented at the Genetic and Evolutionary Computation Conference (GECCO 2025)
Abstract:Bayesian Network (BN) fusion combines multiple input networks into a single structure, balancing dependency preservation with computational tractability. While unrestricted fusion retains all dependencies, it often results in overly complex networks with high treewidth, which affects inference scalability. Limited fusion mitigates this by pruning edges to control treewidth but risks overfitting to input-specific noise and omitting dependencies from the original BNs. This paper introduces a consensus framework that prioritizes shared structures among input networks while enforcing treewidth constraints, ensuring a good consensus. We propose genetic algorithms with advanced initialization, specialized operators, and a tailored fitness function. Additionally, we adapt existing methods to this problem and implement greedy baselines for benchmarking and further optimization. Experiments on synthetic and real-world BNs show the superiority of the proposed genetic algorithms over the adapted methods and greedy baselines.
[LG-22] Federated Causal Discovery via Regression-Directed Cumulants
链接: https://arxiv.org/abs/2609.03705
作者: Pablo Torrijos,Fabio Stella,José A. Gámez,José M. Puerta
类目: Machine Learning (cs.LG)
*备注: Accepted at the 12th International Conference on Probabilistic Graphical Models (PGM 2026)
Abstract:In this paper we study linear non-Gaussian acyclic models (LiNGAM) when used in federated environments. These causal models allow one to go beyond Markov equivalence. However, in many domains data are scarce, and increasing the sample size by centralising data from different clients is not advisable due to regulations such as the GDPR. The federated environment offers an attractive option to balance privacy and causal discovery accuracy. Unfortunately, the standard centralised estimator in the LiNGAM setting, i.e., DirectLiNGAM, cannot be straightforwardly federated. Higher-order cumulant tensors offer a way around this obstacle: they depend only on the joint distribution of the variables involved and add exactly across independent sample groups, so a single communication round suffices in horizontal, vertical, and hybrid partitions. However, FedISHC, i.e., the current federated method along these lines, breaks down under near-symmetric noise. To overcome the above limitation, we introduce the FedRCD family of causal discovery algorithms, and investigate three variants that trade off communication rounds against algebraic noise; two of them are exact federated counterparts of the centralised high-order cumulant (HC) and HC-LiNGAM algorithms, and the single-round variants further effectively support exact unlearning at any granularity, from a single observation to a whole client. Numerical experiments show that at sample sizes typical of real deployments, the entire cumulant-based federated family does not actually rank variables by the population asymmetry that the scores encode at zero. It ranks them by a variance ladder induced by the DAG along its directed paths, the cumulant counterpart of varsortability. Marginal standardisation collapses every cumulant method to near-random ordering, while scale-invariant DirectLiNGAM, not federable under this protocol, is unaffected. Comments: Accepted at the 12th International Conference on Probabilistic Graphical Models (PGM 2026) Subjects: Machine Learning (cs.LG) Cite as: arXiv:2609.03705 [cs.LG] (or arXiv:2609.03705v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.03705 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-23] Resolution-Aware Experimental Design under Partial Identifiability
链接: https://arxiv.org/abs/2609.03686
作者: Sofianos Panagiotis Fotias
类目: Machine Learning (cs.LG)
*备注:
Abstract:Experimental design is commonly framed as choosing the experiment expected to provide the most information. Under partial identifiability however, persistent nuisance uncertainty can make the same observation carry different structural meanings. We introduce Resolution-Aware Experimental Design (RAED), which selects an experiment by the smallest expected nonempty structural candidate set achievable subject to false-exclusion control. We prove an exact cross-nuisance aliasing separation: an experiment can be preferred by structural and full-latent information gain, average classification, and nuisance-marginalized informativeness while having arbitrarily poorer valid structural resolution. RAED nevertheless preserves the expected ordering under a genuine composite Blackwell comparison. To make this criterion operational, we develop a learned score-based implementation with finite-sample nuisance-average and positive-tail calibration, and characterize a rare-tail sample-complexity obstruction. Under constrained sensing, two subsurface-flow benchmarks exhibit genuine RAED–expected-information-gain (EIG) experiment-selection disagreements, with the clearest and largest held-out resolution differences in WCA. In a fluvial benchmark, tail protection changes the selected physical experiment and replaces hard-region false exclusions primarily with explicit ambiguity. In a mechanistic methane-oxidation benchmark, a prospectively specified 5% false-exclusion tolerance also yields a nontrivial finite-sample population guarantee for tail-sensitive nuisance risk, with 95% joint confidence across all three structural families.
[LG-24] Extracting Forgotten Prompts from Targeted Unlearned Models
链接: https://arxiv.org/abs/2609.03662
作者: Au Ashley Hoi-Ting,Meghdad Kurmanji,William F. Shen,Nicholas D. Lane,Ligang He
类目: Machine Learning (cs.LG)
*备注:
Abstract:Recent unlearning methods (e.g. NPO, DPO, LUNAR) make use of refusal alignment to suppress forgotten data. However, it has been shown that refusal responses might leave traces of unlearning, and recent attacks have been able to successfully recover some of the unlearned knowledge. In this paper, we uncover a new vulnerability. Existing attacks typically assume that the forgotten prompts are already known to the adversary and focus on recovering their answers. However, we show that the forgotten prompts themselves can be extracted by using the retained data and black-box access to the model. Our attack, Targeted Active Search (TAS), first identifies the forgotten entities by constructing canonical templates and entity pool, and selectively querying the model using the most informative template-entity pair under a limited query budget. Once the entities are identified, TAS instantiates prompt templates with those entities to probe the unlearned model and reconstruct the forgotten prompts. Experiments across three unlearning methods with three datasets and three LLMs shows that TAS recovers the forgotten entity with 100% accuracy and reconstructs up to 95% of forgotten prompts, all while using up to 99.7% fewer queries than naive probing.
[LG-25] Relative Prime Factorization and Finite-State Presentations under Fixed Finite-Monoid Observation
链接: https://arxiv.org/abs/2609.03643
作者: Takayuki Kuriyama
类目: Formal Languages and Automata Theory (cs.FL); Machine Learning (cs.LG)
*备注: 47 pages; reproducible verification code and a machine-readable certificate for the 36-element witness are available via the fixed GitHub snapshot cited in the paper
Abstract:Let L\subseteq\Sigma^* and fix a morphism h:\Sigma^*\to M into a finite monoid. We study exact factorization and canonical presentation in the relative syntactic congruence \theta_L,h:=\equiv_L\cap\ker h . We separate unique factorization from finite direct presentation. An exhaustively computer-checked 36 -element quotient has a unique exact prime factorization for every live non-unit class, yet its valid prime-return rules contain an infinite family, so unique factorization does not imply the finite relative presentation property (FRP), even for a finite quotient. We lift the same defect to a nonregular context-free language with an infinite relative quotient and finite prime spectrum. To isolate the obstruction, we introduce the finite-state relative presentation property (FSRP), in which canonical valid right-hand-side languages are represented by finite residual controllers, and prove \mathrmFRP\subsetneq\mathrmFSRP . We then introduce prime-target left-division determinism (PTLD), which implies unique exact factorization, tail exactness, tail determinism, and a quadratic bound on valid rules. A nonregular deterministic context-free example with a finite group observer satisfies PTLD while lying outside every fixed (k,\ell) -substitutable class. Finally, for fixed h we give a strong positive-data learner for the canonical PTLD presentation with polynomial-time hypothesis updates and a finite characteristic sample, together with a limit reconstruction of the canonical FSRP controller from weakly behaviorally correct CFG-valued learners. Comments: 47 pages; reproducible verification code and a machine-readable certificate for the 36-element witness are available via the fixed GitHub snapshot cited in the paper Subjects: Formal Languages and Automata Theory (cs.FL); Machine Learning (cs.LG) MSC classes: 68Q45, 68Q70, 68Q32, 20M35 ACMclasses: F.4.3; I.2.6 Cite as: arXiv:2609.03643 [cs.FL] (or arXiv:2609.03643v1 [cs.FL] for this version) https://doi.org/10.48550/arXiv.2609.03643 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-26] Residual neural networks overcome the curse of dimensionality for semilinear heat equations
链接: https://arxiv.org/abs/2609.03626
作者: Ilkhom Mukhammadiev,Diyora Salimova
类目: Numerical Analysis (math.NA); Machine Learning (cs.LG); Analysis of PDEs (math.AP); Probability (math.PR)
*备注: 29 pages, 1 figure
Abstract:Rigorous results show that feedforward neural networks can overcome the curse of dimensionality in the numerical approximation of high-dimensional partial differential equations (PDEs), but comparatively little is known about residual neural networks (ResNets) in the nonlinear PDE setting. We prove that ResNets overcome the curse of dimensionality in the numerical approximation of solutions of semilinear heat equations with globally Lipschitz continuous, gradient-independent nonlinearities: under polynomial growth and network approximability hypotheses on the PDE data, there exist \eta\in(0,\infty) and ResNets \Psi_d,\varepsilon , d\in\mathbbN , \varepsilon\in(0,1] , with at most \eta d^\eta\varepsilon^-\eta parameters whose realizations approximate the solution in dimension d with an L^2 -error of at most \varepsilon . The proof represents one deterministic realization of a multilevel Picard estimator by a ResNet whose shortcut connections transmit the spatial variable and a scalar accumulator, while the residual branches successively add the summands of the estimator. For ridge-sum initial conditions, admissible sigmoidal activations, and globally Lipschitz truncations of the nonlinearity, we obtain, for every \xi0 , the explicit bound C_\xi d^4+\xi\varepsilon^-(3+\xi) on the number of parameters.
[LG-27] Neural-Network Maxent: a general extension with learned nonlinearity applied to time-series for Desert Locust distribution modelling
链接: https://arxiv.org/abs/2609.03603
作者: Alessandro Grassi,Edoardo Kimani Bellotto,Wassim El Azami,Sabrina Outmani,Maximilien Houel
类目: Machine Learning (cs.LG); Image and Video Processing (eess.IV); Data Analysis, Statistics and Probability (physics.data-an); Populations and Evolution (q-bio.PE)
*备注: 22 pages, 8 figures, preprint
Abstract:Species Distribution Modelling (SDM) is essential for understanding how environmental conditions shape biodiversity, particularly for destructive pests such as the Desert Locust (Schistocerca gregaria), whose breeding dynamics are tightly coupled to rapidly evolving environmental conditions. Maxent has become the dominant method for presence-only data, but its reliance on a linear combination of hand chosen feature transforms limits its ability to capture the nonlinear, temporal relationships common in ecological monitoring, where covariates such as precipitation, soil moisture, and vegetation indices evolve meaningfully over time. Standard implementations flatten time-series covariates into independent features, discarding sequential structure that carries critical signal. We introduce RNN Maxent, an extension of the Maxent framework that replaces the fixed feature dictionary with a neural network, specifically a Gated Recurrent Unit (GRU), trained end to end via backpropagation. The approach preserves Maxent’s presence only statistical foundations, background normalization, and probability calibration, differing only in that the nonlinearity is learned from data rather than fixed in advance. We apply RNN Maxent to map suitable habitat for the Desert Locust using 50 day environmental time series derived from ERA5 Land, MODIS, and Sentinel 3, maintaining a 7 day gap between covariates and presence records to yield forecasting behavior. Compared against standard Maxent, RNN Maxent improves performance across metrics (ROC AUC 0.862 std 0.036 vs. 0.792; F1 0.671 std 0.056 vs. 0.590).
[LG-28] WeatherNext 3: Increasing resolution and performance of global weather models with raw observations
链接: https://arxiv.org/abs/2609.03582
作者: Stephan Rasp,Boris Babenko,Dominic Masters,Andrew El-Kadi,Samier Merchant,Guy Shalev,Ilan Price,Fred Zyda,Remi Lam,Sasha Shysheya,Matthew Willson,Stratis Markou,Shreya Agrawal,Suhani Vora,Mohammed Alewi Hassen,Sunny Mak,Tom R. Andersson,Megan Bela,Akib Uddin,Nofar Peled Levi,Ben Gaiarin,Ferran Alet,Aaron Bell,Peter Battaglia,Alvaro Sanchez-Gonzalez
类目: Machine Learning (cs.LG)
*备注:
Abstract:State-of-the-art AI weather models have shown impressive medium-range forecast skill and computational efficiency, but suffer two key shortcomings: their forecasts have lower spatial and temporal resolution than the best physics-based models and they are exclusively initialized with and trained on analysis data. As a result, they cannot directly make use of observations, and any biases in the analysis are inherited by the forecast. WeatherNext 3 addresses these shortcomings and establishes a new state-of-the-art for probabilistic medium-range forecasting skill. First, WeatherNext 3 generates new forecasts every hour (rather than every 6 hours like traditional global models) by ingesting low-latency geostationary satellite data. Second, WeatherNext 3’s temporal and spatial resolution are on par with physics-based global models, with hourly time steps and 0.1 degree resolution for single-level variables, including solar radiation and cloud cover. Third, WeatherNext 3 moves beyond traditional analysis variables by learning to predict satellite-derived precipitation estimates, as well as tropical cyclone and station observations. Modelling sparse station data allows WeatherNext 3 to make 2m temperature and dewpoint predictions at any location and time, conditioned on local geographical features, with substantially lower error than competing global models, even when evaluated against unseen stations. Together, WeatherNext 3’s capabilities move operational AI-based weather forecasting beyond emulating the traditionally distinct stages of data assimilation, forecasting and post-processing, which helps to further push the frontier of performance and granularity for global weather prediction.
[LG-29] Coupled Scaling: A Representational Accessibility Framework for Neural Scaling Laws
链接: https://arxiv.org/abs/2609.03533
作者: Jie Wang
类目: Machine Learning (cs.LG)
*备注: 35 pages, 2 figures. Code and reproducibility artifacts: this https URL
Abstract:Existing theories derive neural scaling from data geometry or a specified data-model spectrum, but systems trained on the same data can scale differently when architecture or optimization changes the representations they can efficiently reach. We introduce Coupled Scaling, a task-conditioned framework in which finite-budget scaling depends on the relation between task structure and the geometry accessible to an architecture-optimization system. In a solvable mode-truncation model, loss separates into target energy outside architectural support and an unresolved supported tail. For an arbitrary priority order, the residual lies between the best-N supported tail and the tail beyond the largest completed high-value prefix. If the cumulative-tail and coverage log-rates are \gamma_A,T and \rho_A,O,T , the residual exponent lies in [\rho_A,O,T\gamma_A,T,\gamma_A,T] . Under bounded off-prefix gain, the completed prefix is rate-determining and \alpha_A,O,T=\rho_A,O,T\gamma_A,T ; for a_A,T,j\asymp j^-b_A,T , this gives \alpha_A,O,T=\rho_A,O,T(b_A,T-1) . A fixed-kernel specialization derives the training-time exponent from the near-zero tail of a task-weighted spectral measure defined independently of the loss fit. The framework separates architectural support from finite-budget acquisition and motivates two tests: static task-relevant geometry should track loss at a common budget, while multiscale geometry should track coupling-specific exponent ordering, including reversal across contrasting tasks. An audit of released emergence trajectories identifies the controls needed for a direct factorial test that measures geometry separately from the scaling fit.
[LG-30] An Adversarial Zero-Shot Learning Approach for Anomaly Detection in Multivariate IoT Traffic Data
链接: https://arxiv.org/abs/2609.03505
作者: Mahshid Rezakhani,Tolunay Seyfi,Fatemeh Afghah
类目: Machine Learning (cs.LG); Networking and Internet Architecture (cs.NI)
*备注:
Abstract:Anomaly detection in Internet of Things (IoT) networks presents unique challenges due to the diversity of devices, lack of labeled data, and domain variability across environments. In this paper, we propose a novel framework for multivariate time-series anomaly detection that leverages adversarial learning and contrastive loss within a sequence-based Variational Autoencoder (VAE) architecture. Our method enables zero-shot domain adaptation by jointly optimizing domain-invariant latent representations and semantically structured embedding spaces, without requiring labeled data or raw feature transfer. To address the heterogeneity of IoT deployments, we introduce encoder and decoder adaptor layers that align feature distributions across domains while preserving contextual semantics. Additionally, we propose a destination-based segmentation strategy to better model real-world communication structures in IoT traffic. Our framework is comprehensively evaluated on six distinct datasets spanning industrial, enterprise, general-purpose, smart home, and military automation domains across 44 transfer scenarios. Experimental results demonstrate strong zero-shot generalization in several cross-domain settings and competitive performance against a contrastive domain-adaptation baseline under realistic, heterogeneous, and privacy-constrained IoT conditions.
[LG-31] Restricted Eigenvalues Beyond Gaussian Width: Threshold Occupancy under Heavy Tails
链接: https://arxiv.org/abs/2609.03504
作者: Shi Fu,Huibo Xu,Qixin Zhang,Dacheng Tao
类目: Machine Learning (cs.LG)
*备注:
Abstract:Restricted eigenvalue (RE) bounds govern stable recovery by norm-regularized estimators. For isotropic sub-Gaussian measurements, the benchmark sample size is 1+w(A)^2 , where w(A) is the Gaussian width of the normalized descent cone. The COLT 2015 open-problem note (Banerjee et al., 2015) asked whether the same law follows for heavy-tailed designs from a uniform small-ball condition alone. We give an explicit and systematic negative answer to the general question as formulated there: the proposed law fails in its full dimension-free, arbitrary-set form, and the missing obstruction is simultaneous threshold occupancy. A constant-width polyhedral descent cone with fixed small-ball constants has zero empirical RE on every sample path up to half the ambient dimension. More generally, every finite range space admits exact threshold encoding in an arbitrarily narrow spherical cap and a lift to a full polyhedral descent-cone section. For every fixed threshold VC dimension d , as \beta\downarrow0 , the sharp worst-case sample complexity is \Theta(\beta^-1[d\log(1/\beta)+\log(1/\delta)]) . The separation persists under exact isotropy and all finite moments: on the same constant-width cone, Gaussian measurements succeed with O(1+\log(1/\delta)) samples, whereas an isotropic heavy-tailed design fails pathwise for n\lesssim\sqrtp/\log p . Gaussian smoothing yields an everywhere-positive C^\infty density while retaining arbitrarily poor RE. Under isotropy, a distribution-free fallback governed by affine dimension times squared enclosing radius is sharp on this family.
[LG-32] Spectral characteristics of autoencoder parameters as a vector representation of data
链接: https://arxiv.org/abs/2609.03495
作者: Maria Nikitina,Anton Bishuk,Oleg Bakhteev
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注: 13 pages, 6 figures. This is a shortened (theorem proofs are skipped) and translated version of the paper published in a Russian-language peer-reviewed journal, the citation is in the paper footnote
Abstract:This paper examines the relationship between the parameters of autoencoder models and the statistical properties of the data on which they are trained. Autoencoders are defined as models with an encoder-decoder architecture, trained to reconstruct input data through a compressed latent representation. It is proposed that the model parameters can be viewed as a dense vector representation of the corresponding sample. To test this hypothesis, a theoretical and experimental study is conducted in which a vector representation is formed based on the spectral characteristics of the autoencoder parameter matrices. Theoretical analysis shows that the singular values of the model parameter matrices are related to the eigenvalues of the covariance matrix of the training data, ensuring the transfer of information between the data space and the parameter space. Experimental results on the CIFAR-10 and FashionMNIST datasets confirm that the resulting vector representations allow for a high degree of accuracy in distinguishing between models trained on different data subsets, without resorting to complex vector generation algorithms or using the original samples. These results suggest that the parameters of trained autoencoders can be viewed as sample representations.
[LG-33] Mind the Gap: Robustness Risks in PII Detection Systems
链接: https://arxiv.org/abs/2609.03464
作者: Adeel Zafar,Slawomir Nowaczyk
类目: Machine Learning (cs.LG)
*备注:
Abstract:Personally Identifiable Information (PII) detection is a foundational component of data protection infrastructure where missed entities constitute direct privacy and security risks. Although modern PII systems report strong performance on standard benchmarks, we show that these evaluations mask substantial robustness failures under realistic distribution shifts encountered in deployment. Rather than comparing state-of-the-art accuracy, we study how different PII detection paradigms fail under noisy, unstructured, and informal inputs. We construct a stress test benchmark spanning seven categories of natural distribution shift and evaluate representative systems from three widely deployed architectural families: encoder-based NER (SpaCy), rule-based hybrid detection (Presidio), and generative LLM extraction (Qwen2.5-3B). All three exhibit significant degradation on out-of-distribution inputs, but with distinct and complementary failure modes. Encoder models primarily fail on unseen surface forms and boundary detection, rule-based systems fail on non-standard formats, and LLMs exhibit entity-type confusion and generation instability. These results show that aggregate benchmark scores obscure deployment-critical weaknesses and that no single architecture is uniformly reliable across PII categories. Motivated by these findings, we propose a hybrid detection pipeline with a QA-driven feedback loop for iterative risk mitigation, and release our benchmark to support OOD-aware evaluation of PII systems. Subjects: Machine Learning (cs.LG) Cite as: arXiv:2609.03464 [cs.LG] (or arXiv:2609.03464v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.03464 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-34] A Two-Stage Forecasting System for CPU Workload Prediction in Private Clouds
链接: https://arxiv.org/abs/2609.03457
作者: Ashir Javeed,Anton Borg,Håkan Grahn,Lars Lundberg,Dhyey Patel,Sogand Shirinbab
类目: Machine Learning (cs.LG)
*备注: 28 pages, 4 figures
Abstract:Accurate cloud resource forecasting is essential for proactive resource provisioning, maintaining Quality of Service (QoS), and reducing operational costs in dynamic cloud environments. The existing forecasting approaches predominantly estimate future CPU workload directly from historical resource traces, which often overlook the relationship between customer service demand and subsequent resource consumption. This study proposes a two-stage integrated forecasting model that explicitly models this dependency by first forecasting customer service requests, expressed as Transactions Per Second (TPS), and subsequently estimating future CPU workload from the TPS forecast. Both the forecasting component and resource prediction component employed the XGBoost model within a cascaded learning architecture, complemented by adaptive online retraining using an expanding-window strategy to address concept drift in continuously evolving cloud workloads. The proposed work was evaluated using real-world traces collected from a private cloud environment comprising ten applications. Experimental results demonstrate robust forecasting performance by achieving Symmetric Mean Absolute Percentage Error (SMAPE) below 7% for most applications, with the best-performing application achieving an MAE of 0.7372 , RMSE of 1.1866 , SMAPE of 3.57% , and an R2 of 0.9185 . Horizon-wise drift analysis confirmed stable recursive forecasting behavior with controlled error accumulation across a 60-step prediction horizon. Compared with the conventional direct CPU forecasting method, the proposed two-stage integrated model gives improved forecasting robustness, computational efficiency, and interpretability, making it well-suited for proactive resource management and intelligent auto-scaling in cloud computing environments.
[LG-35] Beyond Straightness: Non-Crossing Flow Matching via Quantile AlignTree Coupling
链接: https://arxiv.org/abs/2609.03443
作者: Junyi Lin,Mengyu Li,Jingxuan Hu,Kejun He,Cheng Meng
类目: Machine Learning (cs.LG)
*备注:
Abstract:The performance of Flow Matching largely depends on the quality of the coupling between the source and target distributions. However, independent coupling often leads to path crossings and local velocity ambiguity, while OT-based couplings typically incur high construction costs. To address this challenge, we propose Quantile AlignTree Flow Matching (QAT-FM), an efficient structured coupling strategy that constructs a hierarchical coupling between a Gaussian prior and the target data distribution via a quantile-aligned tree structure. QAT-FM constructs the coupling in \mathcalO(Nd\log N) time and supports per-pair source sampling with \mathcalO(d) complexity, enabling scalable training for large-scale high-dimensional generative tasks. Theoretically, we prove that the QAT coupling satisfies marginal consistency, induces non-crossing linear interpolation paths, and consistently improves path separation at intermediate times compared with independent coupling, thereby alleviating local velocity ambiguity. QAT-FM further extends naturally to conditional generation, enabling structured conditional coupling while preserving global Gaussian alignment. Experiments across diverse benchmark datasets demonstrate that QAT-FM achieves competitive generative performance while substantially reducing coupling construction cost.
[LG-36] Guide Not Bind: Why Defeasible Priors Fail in Augmented Lagrangian Causal Discovery
链接: https://arxiv.org/abs/2609.03442
作者: Sairam Sundararaman,Sara Girdhar,Manit Narasimha Murthy,Samrudh N,Bhaskarjyoti Das
类目: Machine Learning (cs.LG)
*备注: 29 pages, 6 figures
Abstract:Differentiable causal discovery methods increasingly encode expert priors as forbidden-edge constraints enforced by an Augmented Lagrangian (ALM) penalty, on the assumption that a data-adaptive relaxation mechanism will discount and eventually override a rule the data consistently contradicts. We show this design, which we call \emphguide, not bind, fails for two independent, precisely characterized reasons, and that directly repairing both restores it only partially. First, sequential penalty-ramping ALM suppresses a wrongly-forbidden true edge before any counterfactual check can detect it: we give three necessary conditions any adaptive relaxation must satisfy to avoid this (Proposition~\refprop:conditions), prove that DADU—the natural relaxation rule this paper introduces as the object of study—violates all three (Corollary~\refcor:dadu_failure), and confirm the failure across 3,072 training runs spanning graphs from 4 to 32 nodes, where a single wrong prior suppresses a true edge in 87–97% of trials under DADU. Second, and independent of any fix to the mechanism, we prove in closed form that the standard correlation-matching objective ties a true edge and its reverse to an identical cost of exactly 2r^2 (Lemma~\reflem:tie), not because the underlying equal-variance model is unidentifiable, but because normalizing to correlation discards exactly the variance information that would make it identifiable; covariance matching instead separates the two directions by a provable margin of at least w_0^4 (Lemma~\reflem:separation).
[LG-37] IGPO: Temporal Instance-Graph Policy Optimization for Long-Horizon LLM Agents
链接: https://arxiv.org/abs/2609.03383
作者: Jinwei Gan
类目: Machine Learning (cs.LG)
*备注:
Abstract:Graph-based policy optimization improves credit assignment for long-horizon LLM agents by organizing rollout trajectories into state-transition graphs. However, existing methods construct graphs independently within each policy update, discarding transitions discovered by earlier policies and limiting advantage estimation to small, batch-local rollout groups. We propose \emphTemporal Instance-Graph Policy Optimization (TIGPO), which extends graph-based credit assignment across policy updates. TIGPO maintains a persistent transition graph for each task, allowing valid transitions discovered by different policy versions to jointly determine credit for current rollouts. To actively reconnect current exploration with historical experience, TIGPO allocates a fixed rollout budget between Exploration slots for ordinary task sampling and Revisit slots for delayed reattempts of previously explored tasks. For each revisit, TIGPO pairs the current rollout group with its corresponding earlier Exploration group to construct a cross-temporal reference. The enlarged reference is designed to stabilize relative advantage estimation under small rollout groups, while comparison on the same task directly captures policy improvement across training stages. Historical transitions and scores serve only as structural and detached statistical references and are never replayed in the policy loss. Experiments on ALFWorld and WebShop demonstrate that TIGPO consistently outperforms prior group-based and graph-based policy optimization methods.
[LG-38] RecurTrace: Adaptive Latent Reasoning with Loop-Time Memory
链接: https://arxiv.org/abs/2609.03379
作者: Yuxiang Wang,Kunyu Feng,Yingda Shen,Haoning Xu,Junyu Wang,Zhizheng Wu
类目: Machine Learning (cs.LG)
*备注:
Abstract:Repeating a small block of middle layers increases a language model’s effective inference depth without adding parameters or generating extra tokens, and recent work shows that this latent recurrence improves reasoning. However, two design choices limit these gains. Each iteration sees only the previous output and cannot directly access earlier computations. Moreover, a fixed loop count wastes depth on easy inputs while leaving hard ones with too little computation. We introduce RecurTrace, which addresses both limitations using the loop’s own trajectory. Specifically, Loop Memory Attention lets each looped layer attend to its own states from previous iterations along the loop-time axis, so the model can revisit earlier computations instead of relying on the latest state alone. A halting head then reads the loop state and predicts whether to continue, with supervision from an oracle that identifies when additional depth still reduces loss. In a controlled MathQA comparison on the same looped backbone, RecurTrace achieves 56.9% accuracy with an average of 2.0 loops, exceeding the best fixed loop depth by 2.2 points at matched compute. By comparison, ACT and PonderNet collapse to one loop, and CALM reaches only 54.1% with 5.6 loops, while the stronger LoopUS-Conf and TaH-Mismatch baselines reach 55.3% at 3.2 loops and 55.7% at 2.1 loops. Finally, RecurTrace improves generation accuracy over same-budget fine-tuned baselines at 0.6B, 1.7B, 4B, and 8B, with the gain growing with model size from 0.6 to 3.4 points.
[LG-39] SimpleDesign: A Joint Model for Protein Sequence and Structure Codesign
链接: https://arxiv.org/abs/2609.03377
作者: Jiarui Lu,Yuyang Wang,Yizhe Zhang,Jiatao Gu,Navdeep Jaitly,Joshua M. Susskind,Miguel Ángel Bautista
类目: Machine Learning (cs.LG); Biomolecules (q-bio.BM)
*备注: Published in Transactions on Machine Learning Research (TMLR), 2026. this https URL
Abstract:Proteins are fundamental to biological processes, with their function determined by the complex interplay between the amino acid sequence and the three-dimensional structure. Developing generative models capable of understanding this intrinsically multi-modal relationship is crucial for fields like drug discovery and protein engineering. Existing models often rely on a multi-stage training process where autoencoders that tokenize data into latent representations are trained in a first stage. Secondly, a generative model is trained on the latent representation of the autoencoder(s), i.e., generative modeling in a latent space. We hypothesize that this multi-stage training is not necessary to obtain performant co-design models and thus present SimpleDesign, an effective multi-modal protein design model trained directly in the data space. SimpleDesign leverages a single-stage end-to-end objective that combines discrete cross-entropy for sequences and a regression objective for structures. In order to effectively model the difference in sequence and structure modalities, we develop a Mixture-of-Transformer architecture that allows modality-specific processing while keeping global self-attention over both modalities. We train SimpleDesign on over 2M sequence-structure pairs achieving strong performance across co-design and unconditional sequence/structure generation benchmarks.
[LG-40] me Without Timesteps: Simulating Coupled Dynamical Systems via Self-Consistency
链接: https://arxiv.org/abs/2609.03358
作者: Liyu Zerihun,Mark Shinyoung Lee
类目: Machine Learning (cs.LG)
*备注:
Abstract:Numerical simulation of dynamical systems is usually organized as a causal march through time: each state is computed from the previous one. We explore a different formulation for coupled systems. For each subsystem type we train a neural surrogate mapping a full driving trajectory and initial condition directly to a full output trajectory; following classical waveform relaxation, coupled systems are assembled by enforcing self-consistency among these trajectories: simulation becomes a fixed-point problem over complete trajectories rather than a stepwise rollout. On coupled van der Pol oscillators and Hodgkin-Huxley neuron networks, sequential depth becomes the number of solver iterations: 4-10 Newton iterations where the reference integrator takes 1500 steps. The gradient likewise loses its time recursion: it becomes a linear system solved by GMRES at memory independent of solver depth. A single scalar measured from the learned operator, the spectral radius of its Jacobian, predicts in advance where the coupled solve will converge; past that boundary, unrolled backpropagation diverges and a Neumann adjoint fails, while the implicit gradient remains correct to 0.04%. We report where the approach succeeds and where surrogate error degrades it.
[LG-41] Efficient Constant Optimization for Symbolic Regression with GPU-Accelerated Tree-Based Genetic Programming
链接: https://arxiv.org/abs/2609.03352
作者: Hao Mao,Xu Tony Liu,Shuai Lu,Peng Zhao,Wenzheng Jiang,Yuntian Chen
类目: Neural and Evolutionary Computing (cs.NE); Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG); Mathematical Software (cs.MS)
*备注: Accepted at the 30th Annual IEEE High Performance Extreme Computing Conference (HPEC 2026), 14-18 September 2026. To appear in IEEE Xplore
Abstract:Constant optimization refines the numerical coefficients of candidate expressions in tree-based genetic programming for symbolic regression. But its per-generation cost has led modern GPU-accelerated frameworks to omit it or restrict it to lightweight forms. We present a GPU-resident, batched Levenberg–Marquardt solver that optimizes constants across a structurally heterogeneous population of expression trees using a fixed number of population-wide CUDA launches per iteration. Reverse-mode automatic differentiation assembles the per-tree Jacobian in one backward sweep, making the dominant per-iteration cost independent of the number of constants per tree, and a double-precision delivery guard guarantees that returned constants are never worse than their initial values. On early-generation populations, the solver sustains up to 5.1\times10^5 trees per second on an NVIDIA A100; at a GPU-saturated benchmark configuration it delivers roughly 9.9\times the throughput of Operon running on a 64-core EPYC 7763, while matching fp64-reference quality. Integrated in-process into EvoGP, the solver enables end-to-end search to recover governing equations on 10 of 18 constructed problems versus 0 for stock EvoGP. Our code is at this https URL.
[LG-42] Learning Informative Prior with Infinite-Dimensional Continuous Normalizing Flow for Bayesian Inverse Problem
链接: https://arxiv.org/abs/2609.03343
作者: Yang Zhao,Junxiong Jia,Tao Zhou
类目: Numerical Analysis (math.NA); Machine Learning (cs.LG)
*备注: 41 pages
Abstract:This paper addresses infinite-dimensional Bayesian inference for inverse problem of partial differential equations with model parameters in infinite-dimensional Hilbert space. To effectively incorporate prior information, we propose a novel continuous normalizing flows based infinite-dimensional model. Specifically, by introducing a well-defined neural ordinary differential equation in infinite-dimensional space, a simple reference measure can be transformed into a more complex measure which encodes the prior information. A corresponding theoretical framework is established to ensure the well-posedness of our proposed Bayesian prior in infinite-dimensional space. We also provide training methods of the prior for two distinct data settings, along with two sampling algorithms for the resulting Bayesian posterior. The proposed framework is applied to three representative inverse problems: the simple smooth inverse problem, inverse scattering problem, and the inverse heat conduction problem. Numerical experiments support the theoretical analysis and demonstrate the efficiency of the proposed algorithms.
[LG-43] Gradients Know What Outcomes Dont: Unlocking Reinforcement Learning for LLM Reasoning with Gradient-Aligned Rewards
链接: https://arxiv.org/abs/2609.03342
作者: Leqi Zheng,Jinbo Su,Fang Niu,Chaokun Wang,Weiping Wang,Jiajun Zhang,Shannan Yan,Jie Wu,Zhaolu Kang,Rong Fu,Hang Zhang
类目: Machine Learning (cs.LG)
*备注:
Abstract:Reinforcement learning from verifiable rewards (RLVR) drives chain-of-thought reasoning in large language models, yet its binary outcome reward cannot distinguish among correct trajectories. Existing dense reward alternatives, from surface heuristics to process reward models, either ignore the expert solutions already present in training corpora or require expensive offline annotation. We propose Gradient-Aligned Reward (GAR), which operates in the policy’s own gradient space: truncated backpropagation through the output projection layer extracts a compact gradient vector for each rollout, and cosine similarity with an expert-anchor gradient yields a dense, reasoning-aware reward with less than 9% wall-clock overhead. We prove that this cosine admits a multiplicative decomposition into prediction-error and activation-pattern factors, providing a concrete characterization of what the alignment signal measures. On Qwen3-4B and Qwen3-8B, GAR consistently improves over GRPO and other baselines on competition-level math benchmarks and transfers to GPQA Diamond and MMLU-Pro without domain-specific data. Code and data are available at this https URL.
[LG-44] A Large Open Multi-Energy Corpus of Soil Compaction Tests with Machine-Learning Baselines
链接: https://arxiv.org/abs/2609.03337
作者: Sompote Youwai,Chana Phutthananon,Warat Kongkitkul
类目: Machine Learning (cs.LG)
*备注:
Abstract:Every engineered fill is specified by a maximum dry density and an optimum moisture content. Each determination needs a full Proctor test. Published correlations rest on one to four hundred specimens, usually from one laboratory at one compactive energy, and are seldom released. This paper releases a corpus without those limits. It holds 2,854 laboratory compaction tests from six public sources, across 162 provenance groups and four Proctor energy levels, with fines from 1.5 to 100%. Every record is audited to the Proctor method its source names, and no energy is inferred. Screening on the zero-air-voids condition removed 11.8% of harmonised records, and 5.7% of those with a measured specific gravity. A material share of published compaction data is physically impossible. The optimum degree of saturation over the corpus is 0.815 at a coefficient of variation of 11%. That is a baseline, not a constant. Both parameters are then estimated from one classification suite and the compaction standard. A tabular foundation model reaches R2 0.824 for density and 0.784 for water content under random folds. It reaches 0.727 and 0.696 with folds drawn around provenance, and 0.520 and 0.614 with a whole source held out. Compactive energy is negligible marginally yet decisive conditionally. Density on the 66 modified-Proctor records is predicted at R2 0.740 with it and -0.651 without. Symbolic regression yields closed forms coupled through a phase relation. No predicted pair can then exceed the zero-air-voids line. The predictions are for screening, not acceptance.
[LG-45] DE-Venus: A Data-Efficient RLVR Framework for Large Language Models
链接: https://arxiv.org/abs/2609.03324
作者: Shenzhi Yang,Guangcheng Zhu,Kai Tang,Zhengqing Zang,Xing Zheng,Haobo Wang,Yingfan Ma,Bowen Song,Bo Han,Bo An,Lei Feng,Weiqiang Wang,Junbo Zhao,Gang Chen
类目: Machine Learning (cs.LG)
*备注:
Abstract:Reinforcement learning with verifiable rewards (RLVR) improves large language model reasoning, but its practical scaling is constrained by expensive on-policy rollouts and the cost of obtaining reliable targets at scale. Existing methods address sample selection, incomplete supervision, or noisy labels separately, often entangling supervision logic with distributed training and hindering controlled comparison and reuse. We present DE-Venus, a unified framework for data-efficient RLVR that treats supervision as evolving state across data preparation and policy optimization. It organizes this lifecycle into three modules: Active Data Selection allocates training and annotation budgets; Weak Supervision Construction derives learning signals from unlabeled examples; and Training-Time Supervision Refinement filters or corrects unreliable supervision. DE-Venus supports seven representative methods and a data-selection pipeline by expressing method-specific decisions as offline dataset transitions or online transformations of targets, rewards, batches, and advantages while preserving verl’s distributed execution contracts. Across public benchmarks and three business scenarios, separate configurations preserve or improve model quality with only 10% of labels or as little as 13% of relevant data; selected business configurations also reduce observed convergence steps by 63%–75%. DE-Venus thus reduces annotation and training costs without sacrificing scalable RL execution.
[LG-46] Beyond .WAV: Design and Software Verification of VocalCap a Traceable Browser-Based Audio Capture System for Vocal Biomarker Research
链接: https://arxiv.org/abs/2609.03320
作者: Augusto Camargo
类目: ound (cs.SD); Machine Learning (cs.LG)
*备注:
Abstract:Remote voice studies often retain a final audio file with limited evidence about how it was captured, transferred, processed, and accepted. This paper presents VocalCap, an institution-controlled, browser-based system for self-guided capture of voice and related acoustic signals by participants without technical training. A versioned protocol drives the workflow. Each accepted recording retains a browser-native object, a client-lossless Float32 WAV derived from the same MediaStream, and a server-canonical mono PCM16 WAV, linked to evidence of capture execution, technical quality, byte-level integrity, recovery, and transformation provenance. IndexedDB preserves accepted browser artifacts until server confirmation, while session completion requires successful verification of every task and artifact. Software tests challenged the acquisition contracts with malformed or altered objects, exact-zero interruptions, channel-topology variants, and interrupted or repeated operations. A post hoc technical audit of 39 consented pilot recordings found 25 sample-identical stereo files and 14 files with signal confined to the left channel. Topology-aware active-channel selection limited the canonical root-mean-square level difference to less than 0.001 dB in all 14 affected files; equal-weight stereo averaging would have introduced approximately 6.02 dB of attenuation. Production end-to-end verification completed two five-task profiles in Chromium and WebKit, yielding 10 accepted recordings and 30 retained artifacts that passed server-side integrity and format checks. The results verify VocalCap’s software behavior under the tested browser-engine conditions. Device-level acoustic agreement, target-population usability, clinical validity, and biomarker performance remain subjects for separate studies.
[LG-47] Geometry-Aware Graph Construction via Adaptive Spectral Bandwidth Control
链接: https://arxiv.org/abs/2609.03306
作者: Ecem Bozkurt,Antonio Ortega
类目: Machine Learning (cs.LG); Signal Processing (eess.SP)
*备注: Accepted at IEEE MLSP 2026. 6 pages, 4 figures
Abstract:Kernelized graph methods - spectral clustering, diffusion maps, and sparse kernel -regression graphs - that use Gaussian kernels depend on the choice of Gaussian bandwidth sigma, which governs the spectral character of the local kernel operator. When sigma is too small, the kernel overestimates local complexity and treats each sample as an independent direction; when sigma is too large, the kernel collapses multiple directions together, the condition number diverges, and all geometric discrimination is lost. We propose a choice of scale to make the spectral complexity of the kernel consistent with the intrinsic complexity of the underlying manifold. We propose a per-node bandwidth criterion that operationalizes this principle by jointly matching the kernel’s effective rank to the local intrinsic dimension estimated via minimum spanning tree, anchoring the search in the manifold-consistent log-log scaling regime. We evaluate SSL embeddings from six encoders on CIFAR-100, showing that adaptive bandwidth consistently improves leave-one-out (LOO) classification and label propagation (LP) accuracy over fixed-bandwidth methods and competing adaptive methods.
[LG-48] Latent Energy Action Planning with World Models
链接: https://arxiv.org/abs/2609.03294
作者: Phu Pham,Aniket Bera
类目: Machine Learning (cs.LG)
*备注: 9 pages, 5 figures, 8 tables
Abstract:Latent world models support efficient model predictive control from high-dimensional observations, yet optimizing a single learned latent objective can favor action sequences whose decoder-predicted terminal descriptor does not match the goal descriptor. We introduce Latent Energy Action Planning (LEAP), which treats the complete action horizon as a differentiable variable and optimizes it through a frozen LeWorldModel (LeWM). LEAP couples terminal latent goal matching with a terminal-window state energy. Low energy requires the predicted terminal latent to agree with the goal latent and the decoder-predicted terminal descriptor to agree with the goal descriptor. A frozen goal-conditioned proposal initializes the search, a quasi-Newton solver refines actions through the autoregressive rollout, and post-optimization projection enforces the admissible action range. Across four control domains using the officially released LeWM checkpoints, the complete LEAP planning system raises mean success from 77.5% for LeWM planned with the cross-entropy method (LeWM+CEM) to 94.8% under a matched protocol, a 17.3-percentage-point improvement, while retaining the frozen LeWM representation and predictor.
[LG-49] Selective Hypergraph Refinement for Frozen Graph Clustering
链接: https://arxiv.org/abs/2609.03265
作者: Zimo Si
类目: Machine Learning (cs.LG)
*备注: 36 pages, 9 figures; includes appendices
Abstract:Existing graph-clustering methods typically improve clustering performance by optimizing model parameters and node representations. Effective means of further improving the clustering results of an already trained and frozen model, however, remain limited. We study post-processing for frozen graph clustering. After checkpoint fixation, the procedure uses no labels and updates neither model parameters, node representations, nor the original graph structure. Instead, it exploits an attribute hypergraph to supplement higher-order relations that ordinary graphs cannot readily express, thereby refining existing cluster assignments. Because global hypergraph refinement can yield both performance gains and erroneous updates, we propose Selective Hypergraph Refinement (SHR). The method generates candidate residual directions from the hypergraph and evaluates their reliability using graph structure, node attributes, and matched-null evidence. It updates only nodes with sufficient support and otherwise retains their original assignments. Further analysis shows that whether a node changes cluster is jointly governed by its native assignment gap and the directional strength of the refinement. In a controlled common-suite evaluation, 13 of 15 backbone-dataset cells had a positive mean macro gain, one produced exact no-action, and one was negative. The cell-equal macro gain was 0.066 pp (95% bootstrap CI, [0.030, 0.107] pp), while only 0.209% of hard assignments changed on average. A broader 15-combination native-interface evaluation yielded a macro gain of 0.137 pp at a mean change ratio of 0.375%. These results indicate that frozen clustering outputs retain a limited but measurable refinement space after training. The effect is heterogeneous across backbone-dataset pairs, and broader coverage also increases exposure to negative transfer.
[LG-50] B2B Customer Conversion Prediction: A Document Representation Graph Theory and CatBoost Driven Methodology KDD2023
链接: https://arxiv.org/abs/2609.03239
作者: Tianqi Wang,Sheikh Shams Azam,Wan Eih Huang,Anton Wiranata,Christopher G. Brinton,Jan P. Allebach
类目: Machine Learning (cs.LG)
*备注: 10 pages, 6 figures. Presented at the 2nd Workshop on End-End Customer Journey Optimization at KDD 2023
Abstract:In the one-time selling B2B context, the buying cycle may last months or even years. During the long process, targeting customers that have a high potential to make purchases and recommending personalized campaigns accordingly are important for effective marketing. For this goal, we study the following problems, B2B customer data aggregation, customer feature generation, and prediction of whether a B2B customer would show interest in making a purchase (i.e., prediction of conversion into sales funnel). We propose an algorithm to aggregate individual contacts to the B2B customer level based on multiple keys. For non-standardized keys such as company names, we propose a novel architecture to cluster them in a domain encompassing irregularities such as spelling mistakes and spelling variants. We then define and generate a set of features and apply the CatBoost model for customer conversion prediction. Our framework achieves 91% prediction accuracy. Based on the prediction results and analysis of the model, we then discuss personalized campaign recommendations to foster conversion.
[LG-51] he 2026 PNPL Competition: Word Classification and Efficient Cross-Subject Generalisation in LibriBrain100
链接: https://arxiv.org/abs/2609.03231
作者: Francesco Mantegna,Gereon Elvers,Dulhan Jayalath,Gilad Landau,Tasha Kim,Miran Özdogan,Luisa Kurth,Teyun Kwon,SungJun Cho,Benjamin Ballyk,Alex Fung,Anna Greer,Pratik Somaiya,Christian Herff,Yorguin Mantilla Ramos,Hamza Abdelhedi,Karim Jerbi,Greg Farquhar,Brendan Shillingford,Mark Woolrich,Oiwi Parker Jones
类目: Machine Learning (cs.LG)
*备注: 16 pages, 2 figures
Abstract:The ambition of the 2025 PNPL competition (Landau et al., 2025) was to launch a multi-year curriculum for non-invasive speech decoding. Designed to progress from foundational tasks toward the linguistic complexity required for a practical brain-computer interface (BCI), it set the stage with speech detection and phoneme classification tasks. Winning submissions reached F1-macro scores of 95.6% and 73.6% on the respective tasks (Elvers et al., 2026), highly significant advances. This success was built on the LibriBrain dataset (Özdogan et al., 2025), the largest within-subject MEG dataset recorded at the time with \sim50 hours of data for one subject. However, while within-subject scale drives strong decoding performance, a practical BCI must generalise to new users from minutes of data, not hours. The 2026 PNPL competition responds to this challenge with LibriBrain100 (Mantegna et al., 2026), an extended LibriBrain dataset with 32 additional subjects ( \sim40 minutes each) plus even more within-subject data ( \sim80 hours). Advancing the curriculum of tasks to focus on word classification, two complementary tracks are presented in this competition: the Deep track targets within-subject word classification at scale, aiming at the best possible performance; the Broad track targets cross-subject generalisation, progressively reducing the amount of subject-specific fine-tuning data from \sim40 to \sim20 to \sim10 minutes, a duration that falls within a clinically feasible range and brings us a step closer to a non-invasive BCI capable of restoring communication to people living with profound paralysis. Comments: 16 pages, 2 figures Subjects: Machine Learning (cs.LG) Cite as: arXiv:2609.03231 [cs.LG] (or arXiv:2609.03231v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.03231 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-52] Language-encoded network topology enables large language models to reason about complex networks
链接: https://arxiv.org/abs/2609.03229
作者: Ucchwas Talukder Utsha,Sakib Mostafa,James Zou,Md Tauhidul Islam
类目: Machine Learning (cs.LG)
*备注:
Abstract:Networks describe systems in biology and beyond, from protein interactions and social relationships to power grids and citation records. Reasoning about such systems requires understanding their structure: which elements are central, which connections bridge separate communities, and how it changes when elements are removed. Although large language models (LLMs) excel at natural language, they struggle with such questions when networks are given as edge lists, sentences or measurement tables, because their structural meaning must be inferred. Here we introduce BioGlyph, which compiles network topology into an interpretable and transferable language of structural roles. BioGlyph combines graph partitioning and structural measurements to identify roles such as hubs, community cores and cross-community connectors, and fixed rules to translate them into a universal vocabulary. The representation describes each element through its structural role, supporting evidence and semantic consequences, leaving both the network and the LLM unchanged. Across twenty networks spanning five domains, BioGlyph substantially improves open LLMs’ ability to answer structural reasoning questions, outperforming edge-based, numerical and learned representations by up to 26 percentage points in system accuracy. Ablations show that the gain comes from explicitly encoding structural roles in semantically interpretable terms. The gain is more prominent in dense, community-structured networks and diminishes in sparse networks whose topology is more readily inferred from text. In a budding-yeast protein-interaction network, BioGlyph exposes biological organization: cross-community connectors are enriched for essential genes, whereas peripheral proteins are depleted. BioGlyph thus provides an interpretable representation for both language models and scientists to reason about network structure.
[LG-53] Coupled Tensor-Tensor Completion Method with Applications in Drug Repurposing
链接: https://arxiv.org/abs/2609.03190
作者: Maryam Bagherian,Albert Hung,Ivo Dinov,Joshua Welch
类目: Numerical Analysis (math.NA); Machine Learning (cs.LG)
*备注:
Abstract:Many biomedical challenges can be posed as tensor completion problems where the observed entries of a multidimensional array (a tensor) are used to impute the missing values. In such settings, incorporating side information about the modes of the tensor, such as gene-gene similarity, can significantly enhance the solutions of the completion problem. Most existing tensor completion methods can only incorporate side information in the form of matrices. In this study, we introduce a novel framework to incorporate side information in the form of tensors. Our new approach, called Coupled Tensor-Tensor Completion (CTTC), leverages the hidden connections among multimodal tensors to improve tensor completion performance. In addition to practical utility, CTTC has theoretical foundations in distance metric learning and group theory. We derive an alternating algorithm to solve the CTTC optimization problem and establish its convergence to a stationary point. Finally, we show that CTTC outperforms state-of-the-art tensor completion methods at predicting drug effects. Results: Compared with other tensor completion methods, including HaLRTC, CTRC, Cell, and NTDDR, CTTC demonstrates superior run-time and RSE tensor completion accuracy on two benchmark datasets, DTD and LINCS. Subjects: Numerical Analysis (math.NA); Machine Learning (cs.LG) Cite as: arXiv:2609.03190 [math.NA] (or arXiv:2609.03190v1 [math.NA] for this version) https://doi.org/10.48550/arXiv.2609.03190 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-54] Portable Causal Fairness Across Synthetic Data Generator Families
链接: https://arxiv.org/abs/2609.03180
作者: Steven Golob,Sikha Pentyala,Martine De Cock
类目: Machine Learning (cs.LG)
*备注:
Abstract:When a statistical agency or regulator releases synthetic data in place of sensitive records, it chooses the generator that produces the table, and can shape that generator so unfair pathways are absent. DECAF made this concrete on one non-private GAN: three fairness definitions become three sets of edge cuts on the generator’s causal graph. Whether the mechanism belongs to DECAF, or to causal factorisation itself, was untested. We port all three definitions to nine generators from three unrelated families (marginals-based, GAN, and diffusion, each with differentially private variants), across three levels of formal privacy guarantee, over 2,520 matched-pair runs on Adult and COMPAS datasets. The mechanism transfers everywhere, and our new causal diffusion backbone yields the fairest release of any family we tested, at fidelity close to the marginals tier. Applying the cut barely moves fidelity, only costs a downstream classifier about 0.07 to 0.15 AUC on average, and adding privacy guarantees don’t make the data less fair.
[LG-55] Frontier LLM s are effective batch optimizers: Assessing reasoning models in continuous and discrete settings
链接: https://arxiv.org/abs/2609.03177
作者: Frank Hu,Shriram Chennakesavalu,David Graff
类目: Machine Learning (cs.LG)
*备注:
Abstract:Frontier large language models (LLMs) have become attractive priors for optimization due to their large-scale pretraining that enables them to navigate a variety of optimization settings. However, the effectiveness of modern reasoning LLMs in batch optimization settings remains underexplored. Here we investigate the performance of the current generation of frontier LLMs as batch optimizers in both continuous and discrete settings. We find that while LLMs are competitive zero-shot batch optimizers for numerical test functions, their performance is brittle compared to classical non-LLM optimization approaches. However, LLM priors are significantly better in semantically rich settings, indicating that their batch optimization behavior is highly effective when navigating and reasoning over the discrete spaces most similar in structure to their pretraining data.
[LG-56] BASP: Communication-Efficient Batch-Aware Sequence Parallelism for LLM Training
链接: https://arxiv.org/abs/2609.03151
作者: Bigyan Ghimire,Jon C. Calhoun
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG)
*备注: Accepted by IEEE HPEC 2026
Abstract:Long-context reasoning for large language models (LLMs) is becoming increasingly important, but training over long sequences remains challenging due to massive memory and communication requirements. Sequence parallelism has emerged as an essential technique for addressing bottlenecks in long sequence LLM training. However, we observe that existing sequence parallelism methods are batch-agnostic and apply uniform sequence partitioning across all batch sizes, resulting in inefficient communication. In this paper, we introduce Batch- Aware Sequence Parallelism (BASP), a sequence parallelism approach that leverages batch structure to reduce communication overhead. BASP exploits batch structure by partitioning GPUs into disjoint sequence-parallel groups according to the micro- batch size. This design reduces the all-to-all communication group size, thereby localizing communication and improving training efficiency. Experimental results on an NVIDIA A100 cluster show that BASP improves end-to-end training time by up to 1.17 - 1.31x in Llama and Qwen models compared to standard sequence parallel baselines, while preserving identical model accuracy and memory usage.
[LG-57] RACE-AIMC: Selective Inference for Heterogeneous Analog In-Memory Accelerators at the Edge
链接: https://arxiv.org/abs/2609.03149
作者: Osama Yousuf,Martin Lueker-Boden
类目: Emerging Technologies (cs.ET); Hardware Architecture (cs.AR); Machine Learning (cs.LG)
*备注: 6 pages, 2 figures, 3 tables
Abstract:Analog in-memory computing (AIMC) speeds up neural-network inference by doing the arithmetic directly inside a memory array, instead of shuttling weights back and forth between memory and a processor. This saves energy, but the physical devices that store the weights are imperfect: programming errors, electrical noise, limited-resolution converters, and outright broken cells all distort the computation, and every physical chip is distorted in its own way. A designer with several such chips available faces an uncomfortable choice: run all of them and combine the answers (safe, but wasteful of energy), or trust a single chip blindly (cheap, but with no guarantee on how often it is wrong). This paper introduces RACE-AIMC (Risk-Aware Certified Ensemble for AIMC), a framework that resolves this choice with statistics rather than guesswork. Offline, RACE-AIMC studies a pool of physical accelerators, picks the single best one for a given energy budget, and computes a mathematically exact upper bound on how often that accelerator will be wrong when it chooses to answer. Online, only that one accelerator is switched on; a lightweight check decides whether to accept its answer or defer to a fallback. In our simulations using a noisy weight mapping and multiple independent test runs, every certified bound stayed under a 10% error target (mean bound 7.83% ± 0.89%, with 70.88% ± 0.98% of inputs answered directly). The resulting system matches the accuracy of a clean digital baseline while cutting modeled energy use by 69.02% relative to always running every accelerator in the pool.
[LG-58] Feasible but Not Safe: Constraint Violations and Report-Channel Attacks in Learned Cell-Free ISAC Association
链接: https://arxiv.org/abs/2609.03147
作者: Mehdi Zafari,Iman Mohammadi,A. Lee Swindlehurst
类目: Networking and Internet Architecture (cs.NI); Machine Learning (cs.LG); Signal Processing (eess.SP)
*备注: Accepted to the 17th ACM Workshop on Wireless of the Students, by the Students, for the Students (S3 '26) at MobiCom 2026. 4 pages, 3 figures
Abstract:Learning-based schedulers have been proposed to provide real-time user, target, and access point (AP) association in distributed cell-free integrated sensing and communication systems. In a typical approach, a graph neural network (GNN), trained on labels from a mixed-integer linear program, maps lightweight per-AP statistics to decisions on AP clustering, user and target scheduling, and mode selection in one forward pass. Such solutions assume that hard constraints, enforced only as soft training penalties, hold at inference, and that the self-reported statistics are truthful. Using our ASSENT algorithm as an example, we find that despite high F_1 scores, many solutions violate at least one hard constraint, demonstrating that high prediction accuracy does not ensure joint feasibility. Projecting the GNN output onto a feasible solution restores constraint satisfaction with low utility loss, even with a simple greedy repair procedure. We further show that feasibility alone does not guarantee robustness to false data injection attacks. A single malicious AP that reports false information cannot substantially increase its user associations, but can greatly increase the rate of infeasible solutions. The effect of such attacks depends on the type of information being falsified. Misreporting information that affects the objective can largely be mitigated through feasibility projection, whereas falsifying information that affects the constraints cannot. The latter can, however, be detected using a low-complexity cross-AP consistency check. These results show that learned ISAC schedulers should be evaluated using constraint-aware feasibility metrics in addition to conventional accuracy measures.
[LG-59] CRAW: Codec Robust Audio Watermarking
链接: https://arxiv.org/abs/2609.03107
作者: David Chernin,Ethan Fetaya
类目: ound (cs.SD); Cryptography and Security (cs.CR); Machine Learning (cs.LG)
*备注:
Abstract:Recent advances in generative speech models have made it increasingly difficult to distinguish authentic from synthetic audio, enabling new forms of fraud and misinformation. Audio watermarking offers a promising defense by embedding an imperceptible signal into generated speech that can later be detected to verify its provenance. However, recent studies have shown that existing post-hoc watermarking methods fail under neural codecs and denoisers, transformations routinely applied during real-world storage, transmission, and processing, severely limiting their practical utility. Here we introduce CRAW, a codec-robust audio watermarking framework that jointly improves robustness against neural re-synthesis while maintaining high perceptual quality. CRAW combines distortion-aware training with an attention-based pooling mechanism, inference-time perceptual mask- ing, and an error-correcting code to recover the fidelity lost during robust training. Experiments demonstrate that CRAW achieves state-of-the-art robustness against neural codecs, denoisers, and vocoders while maintaining perceptual quality comparable to existing post-hoc watermarking methods. The code is available at this https URL.
[LG-60] Scaling Laws Tabular Data and Actuarial Ratemaking Models
链接: https://arxiv.org/abs/2609.03106
作者: Ronald Richman
类目: Machine Learning (cs.LG); Risk Management (q-fin.RM)
*备注:
Abstract:Scaling laws in modern deep learning describe how held-out loss improves as model capacity, training data, and compute increase, often following power-law trends. We investigate whether analogous scaling regularities arise in actuarial ratemaking, where data are tabular, heterogeneous, and noisy, and where classical models such as GLMs remain strong baselines. Using a real-world motor insurance portfolio, we train models from different families across increasing fractions of the training data and multiple random seeds, evaluating out-of-sample Poisson deviance, a likelihood-based loss for Poisson count predictions in which lower values indicate better held-out fit. We find that all model families improve with additional data, but scaling exponents differ substantially: TabM exhibits markedly stronger data scaling than purely supervised tabular Transformers and standard MLP baselines. Transformer variants show weak parameter scaling unless augmented with additional inductive biases (TabM-style adaptation or self-supervision). These results provide quantitative guidance on model selection by data regime and suggest that effective scaling on actuarial tabular tasks depends on architecture and loss function objective design, with simple increases in Transformer size providing limited gains.
[LG-61] Distilling deep optical flow stereo methods to retrieve dense three-dimensional wind fields
链接: https://arxiv.org/abs/2609.03100
作者: Thomas J. Vandal,Dong L. Wu,James L. Carr,Derek J. Posselt,Elise Penn,Tristan Ballard,August Posch,Kate Duffy
类目: Machine Learning (cs.LG); Atmospheric and Oceanic Physics (physics.ao-ph)
*备注:
Abstract:Geostationary atmospheric motion vectors (AMVs) provide the dense horizontal wind vectors (u,v) and heights ingested into data assimilation systems. Traditional AMVs track features using window-based cross-correlation and estimate heights via infrared brightness temperatures paired with numerical weather prediction (NWP) background states, creating a circular dependency that yields inaccurate heights, high computational cost, and sparse retrievals. Stereo winds from GEO-GEO and GEO-LEO geometrically resolve heights from parallax shifts across different poses, eliminating NWP dependence and improving accuracy, but they remain computationally heavy with limited coverage. In this work, we replace window-based tracking in stereo matching with deep optical flow for efficient, improved retrieval. Fine-tuning balances a self-supervised geometric residual loss with supervised radiosonde reconstruction. To eliminate multi-satellite overlap requirements, we distill the stereo teacher into a single-satellite student model. Chi-square and height uncertainties from the teacher are emulated by the student for quality assurance. The student generates winds across full-disk GEO imagery globally. Validation compares stereo and student models against radiosondes, operational AMVs, ERA5 reanalysis, and EarthCARE cloud profiles. Results through triple collocation show that stereo winds improve performance beyond operational AMVs for water vapor bands (6.2, 6.9, and 7.3 \mum), wit degradation in the long-wave infrared (11.2 \mum) band.
[LG-62] he Gradient Does Not See Rank: Rank-Indifference in Matrix-CODI on ProsQA ICML2026
链接: https://arxiv.org/abs/2609.03090
作者: Samuel Larson(Pebble ML)
类目: Machine Learning (cs.LG)
*备注: Accepted at the ICML 2026 Mechanistic Interpretability Workshop, this https URL . 9 pages. Corrects a data-entry error in the workshop version: the seed-1337 accuracy in the three-seed replication was reported as 80.47% (a control run); the archived value is 78.91%, so the three-seed mean is 81.0 +/- 2.0pp (was 81.5 +/- 1.2pp). All other results are unchanged
Abstract:Continuous chain-of-thought models compress reasoning into latent tokens. Matrix-valued variants, which route each latent token through a d x d matrix bottleneck, introduce rank as a single-sample structural observable on the latent matrix Z. If matrix latents carry parallel reasoning paths via superposition, rank should track them, and truncating Z to low rank should hurt accuracy on tasks whose solutions plausibly require multiple components. Across four training regimes of a matrix-CODI model (three on ProsQA, one on GSM8K-Aug below the learning threshold), the rank-k projection ablation curve is flat to within 0.6 percentage points. A three-seed replication yields 81.0 +/- 2.0 percentage points accuracy while the final effective rank of Z spans 4, 12, 13; the loss does not reward any particular rank. To test whether rank-blindness arises from the flatten-then-project readout alone, we trained four readouts: a bilinear reparametrization, a bilinear-plus-GELU readout nonlinear in Z, an SVD-augmented readout feeding singular values through an MLP, and a quadratic readout in Z Z^T. All four rank-k curves remain flat (Spearman p-values 0.63, 0.14, 0.82, 0.46). The flat curves persist for readouts nonlinear in Z. A linear probe on Z underperforms a raw pretrained hidden state at target prediction (AUC 0.673 vs. 0.846). A negative control on vanilla GPT-2 SFT (no matrix bottleneck, no Z, three seeds, n=500) reproduces a flat rank-k curve under the same intervention paradigm with pooled-mean range 0.20pp, and a random-h sensitivity floor lands at the same accuracy: the rank-k ablation alone conflates rank-blindness with position-irrelevance.
[LG-63] LeanStream: A Speculate-and-Refine Streaming Framework for Efficient on-Device LLM Inference
链接: https://arxiv.org/abs/2609.03079
作者: Renyuan Liu,Yuyang Leng,Kaiyan Liu,Yuzhou Zhong,Shaohan Hu,Chun-Fu(Richard)Chen,Peijun Zhao,Heechul Yun,Shuochao Yao
类目: Machine Learning (cs.LG)
*备注: 6 pages, 13 figures. To appear in the Proceedings of the 32nd Annual International Conference on Mobile Computing and Networking (MobiCom '26)
Abstract:On-device LLM inference is attractive for privacy and responsiveness, but remains challenging on mobile and embedded devices because model weights far exceed available DRAM. Prior systems exploit activation sparsity and offload weights to SSD or flash storage, but face a fundamental systems trade-off: accurate sparse execution decisions require the latest context, whereas efficient computation-I/O overlap requires early prediction. As a result, existing designs either serialize execution or incur redundant weight fetches, extra computation, and large cache overheads. We present LeanStream, a streaming speculate-and-refine framework for efficient on-device LLM inference. LeanStream progressively refines computation, loading, and cache-retention priorities using partial GPU results, enabling fine-grained overlap between GPU execution and storage I/O. We implement LeanStream on both mobile and embedded platforms. Compared with prior on-device LLM inference systems, LeanStream reduces memory usage by 4.8 \times to 7.5 \times at the best throughput achieved by prior work, while further improving token generation throughput by 1.6 \times to 2.1 \times .
[LG-64] Learnable composition for neural operators
链接: https://arxiv.org/abs/2609.03069
作者: Zituo Chen,Baiming Zhang,Sili Deng
类目: Machine Learning (cs.LG)
*备注:
Abstract:Neural operators are fast, differentiable surrogates for physical simulation, but their accuracy often degrades when domain geometry, size, or operating conditions differ from training. Supervised adaptation can recover accuracy, but even a small target set requires costly high-fidelity simulations. We therefore ask how pretraining and transfer can be designed together to reduce this deployment cost. LatentDDM first pretrains a neural operator to predict fields on small subdomains. For a new setting, it freezes this operator and trains only a lightweight module that composes the local predictions. We evaluate our method on two complementary problems: steady Darcy flow, where long-range pressure coupling must extend across increasingly large porous domains, and unsteady incompressible flow around a pitching airfoil, where rollout errors compound as target pitching frequencies exceed the training range. Compared with the capacity-matched models that process the full domain at once, LatentDDM’s error is 36-56% lower on larger Darcy domains after adaptation with 16 target simulations. It also improves 20-step field rollouts in fast-pitching airfoil flow, both zero-shot and after few-shot calibration. These results identify the co-designed local pretraining and composition-level transfer as a promising design principle for physical foundation models.
[LG-65] Differentially private federated learning with Byzantine-robust aggregation: A cross-domain framework for secure model training in banking and healthcare systems
链接: https://arxiv.org/abs/2609.03064
作者: Srikumar Nayak
类目: Cryptography and Security (cs.CR); Machine Learning (cs.LG)
*备注: 9 pages, Journal research article
Abstract:Federated learning allows banks, hospitals, and other regulated organizations to train a shared model without moving raw records off their own servers, which is attractive wherever data protection law or competitive sensitivity rules out pooling data centrally. Two problems limit how far this promise can be trusted in practice. First, the parameter updates that clients exchange still leak information about local records through gradient inversion and membership inference attacks. Second, an honest averaging rule such as FedAvg has no defense against a subset of clients that submit corrupted or adversarial updates, so a small number of malicious or compromised participants can quietly steer the shared model off course. This paper presents a federated learning framework, DP-BR-FedAvg, that combines a Gaussian-mechanism differential privacy layer with a coordinate-wise trimmed-mean Byzantine-robust aggregation rule, evaluated on a simulated cross-institutional classification task resembling fraud and clinical-risk scoring. Across sixty communication rounds with twenty clients, a quarter of them Byzantine, plain FedAvg collapses on the minority class (F1-score 0.030) while the proposed framework recovers substantially more of the signal (F1-score 0.119) while bounding the privacy loss of any single client’s contribution. A Byzantine-robust aggregator with no privacy layer performs best in raw accuracy, quantifying the cost privacy imposes on robustness. The results show that privacy and robustness mechanisms interact rather than simply add, and that system design for regulated, adversarial, cross-institutional settings needs to budget for that interaction.
[LG-66] Population-Calibrated Graph Screening at 835-Million-Address Scale with Label-Free Transfer to New Chains
链接: https://arxiv.org/abs/2609.03036
作者: Yury Korolev
类目: Cryptography and Security (cs.CR); Machine Learning (cs.LG)
*备注: 25 pages, 9 tables. Code and evaluation harness: this https URL
Abstract:Compliance screening of blockchain addresses is, in practice, a lookup against sanctions registries plus clustering heuristics; it fails on unlabelled addresses and on chains with no label coverage at all. We describe a deployed system that scores an address by its position in a multi-chain transaction graph rather than by its presence in a list. The substrate is a single graph of 835,330,427 addresses and 15,826,261,934 edges across five EVM chains; a shared inductive encoder with per-chain normalisation feeds two scoring heads. Decision thresholds are exact quantiles of the score distribution over the full population, scanned per chain segment, so the alert volume is known in advance. We report: label-free transfer: heads trained on two chains recall 0.8598 / 0.8182 / 0.9967 of held-out positives on Base, Arbitrum and Gnosis at a 10^-3 population alert rate, with no target-chain labels in head training; a static lead-time replay over 68 external registry events: 40 of 68 (58.8%) flagged at the 0.1% budget, \times 152 over an event-level random-flagging baseline, with first on-chain appearance a median of 528.8 days (Ethereum) / 647.8 days (Tron) before public designation; a serving path whose score is bit-identical to the offline artefact at end-to-end p50 151 ms, gated by a 2,882-address drift panel; and an adversarial harness of eight recurrent reinforcement-learned archetypes that passes an 8-criterion degeneracy audit and, on a detector-independent snapshot, exposes a measured blind spot of the deployed heads against synthesised behaviour.
[LG-67] Causal Foundation Models
链接: https://arxiv.org/abs/2609.03003
作者: Christopher Stith,Hossein Rahmani,Jesse C. Cresswell
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注: Code is available at this https URL
Abstract:Causal inference is the practice of estimating the effect of a treatment or intervention from data. It traditionally requires a bespoke pipeline for every new problem: first proposing a causal mechanism, selecting a compatible estimator, and finally training it. Meanwhile, across diverse settings and modalities, much of machine learning has shifted to the paradigm of foundation models: networks pretrained once at scale and applied to new tasks without fine-tuning. Causal foundation models (CFMs) bring this paradigm to causal inference. CFMs are pretrained neural networks that estimate causal quantities, such as the average treatment effect, on entirely new datasets using in-context learning without requiring model updates. This work provides a practical introduction to this emerging area. We summarize the necessary background in causal inference and machine learning before discussing CFMs. Throughout, we include example code and Jupyter notebooks.
[LG-68] No-Regret Bayesian Optimization with Finite-Library Input-Warped Kernels
链接: https://arxiv.org/abs/2609.02993
作者: Edvin Ketabati Augustinsson,Robert A. Bridges
类目: Machine Learning (cs.LG); Optimization and Control (math.OC); Machine Learning (stat.ML)
*备注:
Abstract:Gaussian-process Bayesian optimization (GP-BO) excels at black-box optimization of costly functions, e.g., hyperparameter optimization (HPO) and multi-agent system (MAS) design. Convergence-rate guarantees exist for select methods, notably GP upper confidence bound (GP-UCB), but require a fixed kernel. Critically, the kernel encodes how input proximity affects objective value similarity. When raw coordinates poorly match this geometry - as with log-scaled hyperparameters or localized peaks - input warping can greatly improve sample efficiency, yet known GP-UCB proofs require a fixed kernel. We propose Finite-Library Input-Warped Bayesian Optimization (FLIWBO), which selects warps from a finite library of smooth input maps by any history-dependent rule. It adapts the input geometry to accelerate learning while retaining high-probability convergence guarantees under mild hypotheses, with an explicit \sqrt(N_\varepsilon) library-size cost. Controlled diagnostics show that finite-library warping repairs planted geometry mismatches and identify FLIWBO failure cases. Across four repeated benchmarks - warped synthetic objectives, a confidence-fence trap, and Fashion-MNIST HPO - FLIWBO-UCB beats raw-coordinate GP-UCB under misspecified geometry, escapes traps that defeat even oracle-warp expected improvement, and recovers much of the gain from manual log scaling, while leading the tested methods that admit a matching regret guarantee. A 20-dimensional MAS design study further shows feasibility under costly noisy evaluations. Code for experiments is available: this https URL.
[LG-69] RACE: Spatiotemporal Contact Memory Graph Network Simulator for Granular Dynamics
链接: https://arxiv.org/abs/2609.02991
作者: Changjian Zhou,Negin Yousefpour,Jie Qi,Junfeng Fang,Guillermo A. Narsilio,Hans Petter Jostad
类目: Machine Learning (cs.LG); Numerical Analysis (math.NA); Machine Learning (stat.ML)
*备注:
Abstract:Learned graph simulators provide an efficient alternative to high-fidelity solvers for granular dynamics. However, granular motion depends strongly on inter-granular contact history, which is difficult to preserve when particle contacts form, break, and rearrange. Existing simulators mainly store temporal information in node features or node-level memory. Here we introduce TRACE, a graph-network simulator that stores interaction history directly on contact edges. Each edge maintains a persistent memory updated by attention-based message passing and a gated recurrent unit, while an edge-identity dictionary preserves this memory as the contact graph changes. A physics-structured decoder predicts inter-granular normal and tangential contact forces, enforces the Coulomb friction limit, and applies equal-and-opposite internal forces. The model is trained with single-step pretraining followed by autoregressive rollout fine-tuning. We evaluate TRACE on 2D and 3D granular column-collapse benchmarks. In both cases, TRACE produces stable, physically consistent long-horizon rollouts, closely reproducing the final deposit geometry and the kinetic energy released during collapse. Compared with graph network simulator (GNS) and node-memory graph neural simulator (NMGNS), TRACE reduces long-rollout position error by 31-62% and final-deposit error by 58-89% across the two benchmarks, while using fewer parameters and maintaining near-zero particle interpenetration. TRACE also achieves 12.2 \times and 8.9 \times speedups over the material point method (MPM) reference solver in 2D and 3D, respectively. Our code is available at this https URL.
[LG-70] Mesh-Native Physics-Informed Graph Surrogates for TCAD-in-the-Loop Design Space Exploration
链接: https://arxiv.org/abs/2609.02988
作者: Leonid Popryho,Ayoub Sadeghi,Inna Partin-Vaisband
类目: Machine Learning (cs.LG); Hardware Architecture (cs.AR); Computational Engineering, Finance, and Science (cs.CE)
*备注: 9 pages, 4 figures, 2 tables. Accepted at IEEE/ACM International Conference on Computer-Aided Design (ICCAD 2026)
Abstract:High-fidelity TCAD simulation of drift-diffusion transport remains the workhorse of emerging FinFET device design, but it is computationally expensive, especially for 3D structures where runtime escalates steeply with mesh complexity. This sharply limits multi-objective design space exploration. Existing machine-learning surrogates map a fixed set of design parameters to a few scalar device metrics, discarding the underlying physics and losing transferability across device geometries and families. A physics-informed graph attention network (GAT) surrogate is proposed. It operates directly on the tetrahedral TCAD mesh and predicts, at every mesh node, the electrostatic potential together with the electron and hole quasi-Fermi levels, the fundamental unknowns of the drift-diffusion system. Training combines a data loss with finite-volume current-continuity residuals, embedding carrier-transport physics into the objective. Operating on the mesh as a graph, the surrogate inherits size generalization: a model trained on few-fin meshes applies unchanged to substantially larger arrays, bounded at inference only by GPU memory. Per-node uncertainty from a deep ensemble drives an active-learning loop that screens large candidate pools in seconds and forwards only the most informative designs for full simulation. Benchmarked against Sentaurus Device on multi-fin tri-gate FinFETs, the surrogate reproduces the three drift-diffusion fields with sub-volt per-field RMSE and reaches a per-design throughput orders of magnitude higher than the full simulator. The advantage grows with device size: on large multi-fin arrays that are prohibitively slow to simulate directly, inference still completes in under a second per device, enabling Pareto-front exploration across device scales infeasible for direct TCAD sweeps.
[LG-71] ail-Likelihood Reinforcement Learning
链接: https://arxiv.org/abs/2609.02987
作者: Shrinivas Ramasubramanian,Daman Arora,Fahim Tajwar,Guanning Zeng,Qingyang Wu,Zhongzhu Zhou,Chenfeng Xu,Haiwen Feng,Yuda Song,Aarti Singh,Ruslan Salakhutdinov,J. Andrew Bagnell,Jeff Schneider,Andrea Zanette
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:Reinforcement learning typically optimizes average reward. For generative policies, the average can hide an important distinction: two policies can achieve the same mean reward while having very different chances of producing a rare but high-reward rollout. This matters as sampling increases during training and inference, since its benefit depends on retaining probability mass on high-reward outcomes. We propose to optimize this coverage directly. Rather than considering only expected reward, we consider all of its upper tails: for each reward threshold, how likely is the policy to exceed it? This turns a continuous reward into a family of binary success events. We introduce Tail-Likelihood Reinforcement Learning (TailRL), which maximizes the log-probability of exceeding a randomly chosen reward threshold. Its gradient gives more weight to rare, high-reward rollouts and can be interpreted as a mixture of Best-of-(k) gradients. TailRL requires only a simple modification to the advantage function, making it compatible with existing reinforcement learning pipelines. Across object localization, maze navigation, GUI grounding, and code optimization, TailRL leverages rare high-reward training samples to avoid suboptimal solutions and yields models that benefit more from additional samples at inference time.
[LG-72] Modern Transformers Are Implicit Hybrids: From Functional Differentiation to Principled Hybrid Architecture Design
链接: https://arxiv.org/abs/2609.02986
作者: Runlin Shi,Bojian Yin,Guoqi Li
类目: Machine Learning (cs.LG)
*备注: 24 pages, 15 figures, 8 tables
Abstract:Hybrid architectures combining Full Attention (FA) and Linear Attention (LA) are increasingly prominent, yet their allocation remains heuristic. We seek an evidence-grounded basis in head-level functional organization learned by RoPE-based Transformers. Behavioral probes do not yield a complete taxonomy, so we propose two intervention metrics: RoPE Frequency Importance Score (RFIS), measuring how each frequency affects a head’s attention distribution, and RoPE Positional Dependence (RPD), isolating dependence on rotary positional modulation. On Qwen3-series models and Llama3.1, RFIS suggests and RPD verifies a complete taxonomy of retrieval and positional heads separated by a salient mid-low-frequency band. Controlled Transformers show that this boundary follows the training-length positional scale; we term it the Global Positional Band (GPBand). The analysis suggests a potential cause of zero-shot length-extrapolation failure and yields two principles: positional modeling should operate only locally, with global access through position-independent retrieval; and both functions should be assigned at head granularity with layer-specific allocation. We instantiate them in Head-wise Hybrid Architecture (HwH), using NoPE FA for global retrieval and LA for local positional modeling. With an FA-to-LA ratio below 1:3, HwH retains strong language modeling and commonsense reasoning while improving retrieval and substantially strengthening zero-shot long-context extrapolation over Transformer, LA, and a layer-wise hybrid baseline. Ablations validate both principles and component roles, highlighting principled hybrid architecture design as a promising route toward future foundation models.
[LG-73] Equation Recast for Canonical Operator Learning Across Parametric PDEs
链接: https://arxiv.org/abs/2609.02982
作者: Qiyun Cheng,Valentin Duruisseaux,Cesar F. Clauser,Md Hossain Sahadath,Huihua Yang,Shaowu Pan,Nathaniel Ferraro,Anima Anandkumar,Wei Ji,Cristina Rea
类目: Machine Learning (cs.LG); Computational Physics (physics.comp-ph); Plasma Physics (physics.plasm-ph)
*备注:
Abstract:Learning solution operators across broad parameter ranges can require substantial coverage of both input functions and physical parameters, particularly for purely data-driven parametric models. In addition, the resulting models may fail silently outside the training distribution. We introduce equation recast, which reformulates parametric operator learning as the learning of a single canonical operator. Parameter-induced operator variations are derived analytically from the governing equation and absorbed into effective sources, enabling zero-shot prediction across new parameter regimes. Across multi-parameter, nonlinear, and singular PDE settings, equation recast supports extrapolation, integrates sparse heterogeneous datasets in a shared canonical representation, and uses loss of convergence as an internal warning signal for failure of the recast iteration. In high-fidelity tokamak simulations for nuclear fusion, the framework unifies electron-temperature data across four device geometries through canonical-domain mapping within one jointly trained operator. Equation recast provides a route toward reusable neural PDE solvers combining equation-guided transfer, data efficiency, and monitorable inference.
[LG-74] Privacy Leakage in Federated Learning: Gradient-Based Client Identity Inference and Defenses for Inertial Sensing in Vehicular Edge Networks
链接: https://arxiv.org/abs/2609.02971
作者: Ali Akarma(1 and 2),Toqeer Ali Syed(1),Muhammad Khan(3),Qurat-ul-ain Mastoi(3),Adeel Ahmad(1) ((1) Islamic University of Madinah, Madinah, Saudi Arabia, (2) King Fahd University of Petroleum and Minerals, Dhahran, Saudi Arabia, (3) University of the West of England, Bristol, U.K.)
类目: Cryptography and Security (cs.CR); Machine Learning (cs.LG); Networking and Internet Architecture (cs.NI)
*备注: Accepted for presentation at IEEE VTC2026-Fall, Boston, MA, USA, September 6-9, 2026
Abstract:As vehicular networks move toward 5G/6G edge intelligence, federated learning (FL) is widely promoted as a privacy-preserving way for vehicles and infrastructure to train shared models without exposing raw sensor data. Yet the updates clients transmit still leak enough information to identify who sent them, which threatens the anonymity that safety-critical V2X applications assume and adds to existing concerns over adversarial ML, model poisoning, and backdoor attacks. We study server-side client identity inference from transmitted weight deltas using inertial (IMU) measurements, evaluated on the UCI Human Activity Recognition (HAR) benchmark as an accessible proxy for the IMU streams produced onboard connected vehicles. Across five attack classifiers and five non-IID partitions, an honest-but-curious server recovers client identity with near-perfect accuracy (approximately 1.000) from undefended updates, confirming a concrete identifiability risk. We then quantify the privacy-utility trade-off of a lightweight clip-then-noise defense by sweeping Gaussian noise (sigma in 0.00, 0.05, 0.10, 0.20, 0.50, 1.00) at fixed clipping (C=1.0), and report formal (epsilon, delta)-DP budgets through Renyi accounting. A practical region (sigma in [0.1, 0.2]) drives attack accuracy to near-random while costing under 5% relative FL accuracy. Ensemble FL supplies complementary structural privacy with a 1/K anonymity-set bound and no noise penalty. Results are supported by cryptographic (SHA-256) train/evaluation gradient disjointness, three seeds, and a count-normalized attacker-advantage metric. We position HAR explicitly as a proxy and discuss what validation on true vehicular telemetry would require.
[LG-75] Statistical Feature Augmentation for Anomaly Detection in Dynamic Graphs
链接: https://arxiv.org/abs/2609.02965
作者: Philipp Schlinge,Jean-Luc Schnipper,Martin Atzmueller
类目: ocial and Information Networks (cs.SI); Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:Dynamic networks are being applied in many domains, from social media to logistics systems, each with their own set of special characteristics. A model employed on this type of data must capture the duality between temporal/structural and feature-based information. Yet state-of-the-art deep learning models often struggle to learn especially short-term behavioral interaction signals, such as sender intensity or interaction inertia, directly from raw event streams. To address this gap, we propose a statistical feature augmentation method that explicitly encodes behavioral interaction statistics into the input feature space. We evaluate our proposed method on an anomaly detection task across three real-world datasets (Reddit, Wikipedia, MOOC) and seven models spanning both continuous-time and discrete-time architectures. As a baseline, we apply the same models trained on the original embeddings. Our results show, that augmentation consistently improves detection performance. Beyond performance, the enriched input enables fine-grained post-hoc analysis of behavioral importance, since each statistic occupies a dedicated input dimension. In particular, this work showcases a promising approach for merging classical network analysis with deep learning.
[LG-76] Evaluating GNNs for Success Prediction in Artist Collaboration Networks
链接: https://arxiv.org/abs/2609.02920
作者: Wiktor Dowgiałło
类目: ocial and Information Networks (cs.SI); Computers and Society (cs.CY); Machine Learning (cs.LG)
*备注: 23 pages, 7 figures, 5 tables
Abstract:As the music industry becomes an increasingly collaborative effort, understanding the underlying structures of the artist network has become a focal point in cultural data analytics. This study expands on the previous analyses of the Italian and Danish networks by introducing a novel dataset of the Polish music scene. By utilizing methodologies used in the prior studies, this work enables a direct comparison between three distinct European music landscapes and allows to merged the created networks into one. Furthermore, this research introduces a framework to test the efficacy of Graph Neural Networks (GNNs) for artist popularity predictions based on the metadata and the position in the network. The statistical analysis revealed that the Polish and tri-national network exhibit similar properties and clustering behaviours, consistent with prior models. An evaluation of the predictive architectures reveals that while GNN models achieve a comparable F1-macro scores to the Multilayer Perceptron (MLP) in specific cases however the MLP remains a superior model regarding the success metric. The results suggest that internal node features - such as genre and label affiliation might carry more predictive capabilities than the topology of the network. The higher performance of the GNN models in the tri-national network might also suggests that the relational features become more informative when the network spans multiple linguistic and geographic boundaries, with the GNNs potentially capturing complex ‘bridge’ structures between the merged networks.
[LG-77] Parameterised graph theory for tensor networks: entanglement rerouting structural simplification and agnostic tomography
链接: https://arxiv.org/abs/2609.04165
作者: Matthias C. Caro,Natalie McHugh,Sergii Strelchuk
类目: Quantum Physics (quant-ph); Data Structures and Algorithms (cs.DS); Machine Learning (cs.LG)
*备注: 74 + 12 pages; 6 figures
Abstract:Parameterised graph theory studies how the complexity of graph-theoretic problems depends on structural parameters of the input graph. This perspective has proved useful in analysing tensor-network simulation (Markov and Shi, 2008). Its implications for tensor-network representations and tomography are less well understood. In particular, which graph parameters determine whether a tensor-network state (TNS) admits a tractable matrix product state (MPS) or tree tensor network (TTN) representation, and which control the complexity of learning the state? We address these questions using parameterised graph theory. First, we show that cutwidth and tree-cutwidth bound the bond dimension overhead required to represent a TNS as an MPS or TTN. In the TTN case, tree-cutwidth also bounds the local dimension of the grouped subsystems. The proofs are based on entanglement rerouting, a tensor-network analogue of rerouting information in a classical network. Second, we derive graph-dependent upper bounds on the sample and computational complexity of realisable TNS tomography, with exponents that depend on cutwidth, tree-cutwidth, and a new graph parameter, learning complexity, which we bound in terms of degree and treewidth. We obtain these results by extending the disentangling MPS learner of (Cramer et al., 2010), as analysed further in (Bakshi et al., 2025; Lin et al., 2025), to TTNs and to tensor networks on arbitrary known graphs. Finally, we extend the framework beyond the realisable setting. For an arbitrary input state, our agnostic learner outputs a pure state whose fidelity is within additive error \epsilon of the optimum over tensor-network states on the given graph with a given bond dimension, with explicit graph-dependent bounds on sample and computational complexity. Comments: 74 + 12 pages; 6 figures Subjects: Quantum Physics (quant-ph); Data Structures and Algorithms (cs.DS); Machine Learning (cs.LG) Cite as: arXiv:2609.04165 [quant-ph] (or arXiv:2609.04165v1 [quant-ph] for this version) https://doi.org/10.48550/arXiv.2609.04165 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-78] Cooperative Multi-Task Semantic Communication for Joint Classification and Regression Tasks
链接: https://arxiv.org/abs/2609.03977
作者: Ahmad Halimi Razlighi,Mohammad Siddiqur Rahman,Maximilian H. V. Tillmann,Edgar Beck,Armin Dekorsy
类目: ignal Processing (eess.SP); Machine Learning (cs.LG); Image and Video Processing (eess.IV); Machine Learning (stat.ML)
*备注: This work has been submitted to the IEEE for possible publication
Abstract:Multi-Task semantic communication (SemCom) prioritizes simultaneous execution of multiple tasks over bit-accurate reconstruction in future intelligent networks. In our prior work [1], we introduced the cooperative multi-task SemCom (CMT-SemCom) framework, in which the semantic encoder is divided into a common unit (CU) and multiple specific units (SUs) to facilitate cooperative multi-task processing. However, CMT-SemCom has been evaluated on homogeneous classification tasks on simplistic datasets, limiting its applicability to real-world perception systems. In this paper, we extend our CMT-SemCom to jointly handle heterogeneous classification and regression tasks on the complex Cityscapes dataset. We adopt the information maximization (InfoMax) principle so that it accommodates mixed discrete and continuous semantic variables. In particular, we benchmark the proposed framework against independent single-task training, a conventional task-agnostic digital transmission, and single-encoder multi-decoder SemCom. Additionally, we investigate the impact of CU capacity on joint task performance, providing design insights. Extensive evaluations demonstrate that CMT-SemCom significantly outperforms the benchmarks.
[LG-79] Correlated initialization of deep residual networks
链接: https://arxiv.org/abs/2609.03589
作者: Felix Benning,Ivan Nourdin,Giovanni Peccati
类目: Probability (math.PR); Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:We study the large-depth behavior of residual networks whose weights are correlated across layers at initialization. Our results confirm and extend a conjecture of Marion et al. [2025], according to which correlated initializations should interpolate continuously between the Brownian stochastic differential equation arising from independent initialization and the ordinary differential equation arising from perfectly correlated initialization. When the initialization is obtained from the application of a feature function to a stationary Gaussian sequence with regularly varying correlation, we prove that there exists a unique critical scaling such that the infinite-depth limit is the solution of a Young differential equation driven by a Hermite process. Hermite processes reduce to the fractional Brownian motion if the feature function generating the initialization has Hermite rank one, which is the case for the identity function, for example. We show that the critical scaling and asymptotic limit are uniquely determined by the decay of correlations together with the Hermite rank of the feature function. Consequently, the correlation structure and Hermite rank of the initialization represent meaningful hyperparameters in the asymptotic regime. By contrast, under finite-variance iid initialization, the asymptotic driver is universally Brownian up to normalization regardless of the choice of distribution. Our proofs rely on a collection of novel results establishing a robust stability theory for Young differential equations in Banach spaces. Subjects: Probability (math.PR); Machine Learning (cs.LG); Machine Learning (stat.ML) MSC classes: 60G18, 60G22, 60H10, 60L20, 60L90, 60F17, 68T07 Cite as: arXiv:2609.03589 [math.PR] (or arXiv:2609.03589v1 [math.PR] for this version) https://doi.org/10.48550/arXiv.2609.03589 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-80] owards a Statistical Understanding of Mixture-of-Experts
链接: https://arxiv.org/abs/2609.03501
作者: Siyuan He,Bokai Yang,Jie Hu,Ziwen Gao,Yuhong Yang
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Statistics Theory (math.ST)
*备注: 166 pages, 2 figures
Abstract:Mixture-of-experts (MoE) architectures increase model capacity by combining a collection of expert predictors through input-dependent routing, while often activating only a small subset of experts for each input. Despite their growing importance in modern large-scale models, the statistical roles of their design choices, especially routing, sparse activation, and shared experts, remain only partially understood, as existing theory has largely focused on parametric or correctly specified MoE models. In this paper, we view MoE as a form of localized aggregation and show how this localization reshapes the approximation-estimation-computation tradeoff. We derive oracle risk bounds for learning dense and sparse routing with evolving experts, separating approximation, expert-learning, and router-estimation errors, and characterize how sparse Top-K routing can retain the benefits of localized aggregation while controlling per-input computation. We also interpret gating through the geometry of input space, relating routing performance to regions of local expert advantage, and show how shared experts, as adopted in architectures such as DeepSeekMoE, can extract common predictive structure so that routed experts focus on residual local variation. Together, these results provide a unified statistical framework for understanding MoE through input-dependent expert aggregation, in which expert specialization and computational tradeoffs are governed by local predictive structure.
[LG-81] Computing stable configurations of confined smectic liquid crystals with a deep variational framework
链接: https://arxiv.org/abs/2609.03389
作者: Yuchen Xie,Baoming Shi,Yucen Han,Lei Zhang
类目: oft Condensed Matter (cond-mat.soft); Machine Learning (cs.LG)
*备注: 13 pages, 6 figures
Abstract:Smectic liquid crystals are layered liquid-crystalline phases characterized by orientational order and periodic density modulation. Although their structures can be modeled using continuum theories, computing stable configurations remains challenging in complex geometries, particularly when the high-frequency density modulations associated with smectic layering should be resolved. We propose a deep variational framework (DVF) for computing these configurations within the modified Landau–de Gennes model, in which the coupled orientational and positional order parameters are represented on a regular reference domain while physical confinement is incorporated through coordinate mappings. A warmup penalty mitigates the spectral bias of neural networks toward smooth, nonlayered fields, enabling robust recovery of oscillatory smectic states. Comparisons with a neural-network baseline and finite-difference relaxation demonstrate the essential role of this penalty and the numerical stability of the resulting layered states. The DVF reproduces experimentally established smectic-A defect structures and layer morphologies across diverse confinement geometries and further predicts a chevron-like smectic-C state in a tangent-anchored sphere. Together, these results demonstrate the applicability of the DVF to computing stable smectic configurations across experimentally relevant confinement geometries and anchoring conditions.
[LG-82] SurgeGen: A Hybrid Generative Diffusion Framework for Storm Surge Scenario Synthesis
链接: https://arxiv.org/abs/2609.03382
作者: Shunan Zheng,John J. Hasenbein
类目: Dynamical Systems (math.DS); Machine Learning (cs.LG)
*备注:
Abstract:Predicting storm surge induced by landfalling tropical cyclones is crucial for flood mitigation and coastal risk management. Traditionally, physics-based numerical models simulate storm surge by solving the Navier–Stokes equations using numerical methods, but these simulations are computationally expensive. Generative models are promising for storm surge emulation because they can generate diverse realizations rather than producing a single deterministic prediction. However, their use for storm surge emulation remains largely unexplored. In this paper, we leverage diffusion models for storm surge surrogate modeling, combining a baseline prediction stage with conditional generation to provide a more interpretable modeling framework. We develop SurgeGen, a two-stage generative framework for generating storm surge scenarios conditioned on hypothetical storms with parameters defined in a continuous space. First, a baseline model produces a coarse estimate of the storm surge height. This estimate then conditions a diffusion model, which generates refined storm surge scenarios that better capture spatial patterns and variability. We demonstrate that our approach can generate realistic and diverse storm surge scenarios under conditions both within and outside the training distribution.
[LG-83] Grassmann–Plücker Parametrization of Convolutional Filter Subspaces: Regularity and Closed Embeddings
链接: https://arxiv.org/abs/2609.03361
作者: Hongyu Yuan,Huaiqing Zuo
类目: Algebraic Geometry (math.AG); Machine Learning (cs.LG)
*备注: 24 pages; includes a symbolic computational example using Singular
Abstract:We propose a geometric parametrization of the filters in a single convolutional layer: the parameter is no longer an ordered family of filter vectors, but a fixed-dimensional subspace of the filter space. For one-dimensional finite-stride convolution, the filter-to-convolution-operator correspondence gives an injective linear map \mathcalC:\mathcalK\to H . This map sends filter subspaces in \mathrmGr(q,\mathcalK) to operator subspaces in \mathrmGr(q,H) ; composing it with the Plücker embedding yields a projective parametrization \Phi:\mathrmGr(q,\mathcalK)\to\mathbbP(\bigwedge^q H) . Using T_U\mathrmGr(q,\mathcalK)\cong\mathrmHom(U,\mathcalK/U) , we compute the differential of the induced Grassmannian map and show that the differential of \Phi is injective at every point. We then use the vanishing equations for Plücker coordinates and standard affine coordinates on a Grassmannian to prove that \mathrmGr(q,\mathcalC(\mathcalK))\hookrightarrow\mathrmGr(q,H) is a closed embedding, and hence that \Phi is a closed embedding. Consequently, the parameter space is isomorphic to its projective image, the parametrization is finite and birational onto its image, every fiber is a singleton, and the resulting projective neural variety is smooth. For k=4 and q=2 , we also use Singular to recover the image ideal and check its dimension, degree, chart rank, and smoothness. This computation illustrates, rather than replaces, the general proof. Finally, we discuss possible connections with filter redundancy and low-rank convolution, while distinguishing the proved geometric results from application proposals requiring numerical validation.
[LG-84] Introducing SINFONIA: Symplectic slimplectic and Magnusian (Neural) Flows for Orbital Numerical Integration and Acceleration
链接: https://arxiv.org/abs/2609.03329
作者: Lidia J. Gomes Da Silva
类目: General Relativity and Quantum Cosmology (gr-qc); High Energy Astrophysical Phenomena (astro-ph.HE); Instrumentation and Methods for Astrophysics (astro-ph.IM); Machine Learning (cs.LG); Computational Physics (physics.comp-ph)
*备注: 21 pages, 1 table, 5 figures. Comments welcomed
Abstract:Long-duration gravitational-wave modelling must resolve fast orbital motion together with slow dissipative evolution while preventing small numerical errors from accumulating into secular phase drift. Here we ask whether the finite-time evolution map itself can be learned as an explicit, differentiable, structure-preserving object and then repeatedly composed through a complete inspiral. We construct three neural-flow architectures: a symplectic and slimplectic flow on Galley’s doubled phase space, [SINFONIA-J0]; a Taylor-anchored flow, [SINFONIA-J1]; and a Magnusian flow that learns the finite-time dissipative correction in the interaction picture, [SINFONIA-J2]. Applied to a 2.5PN neutron-star inspiral, all three expose the same controlling mechanism: long-time accuracy is governed not by pointwise map error alone, but by its signed projection onto a single secular channel fixed by energy–angular-momentum balance. Encoding this structure allows the learned maps to remain accurate through 10^2 – 10^5 window compositions to coalescence at timesteps of a full orbital period and beyond, reaching chained phase errors orders of magnitude below a benchmark slimplectic integrator at lower cost. The same secular structure can also be exploited for physics inference: when the channel is left unconstrained, the accumulated phase retains enough information to recover an un-modelled dynamical-friction-like force, both parametrically and as a learned function of separation. Network-off controls isolate the contribution of learning from the analytic structure already built into each map. These results establish a proof of concept for structure-preserving learned evolution maps as tools for fast long-duration integration and physics inference in gravitational-wave source modelling.
[LG-85] What is Smoothness?
链接: https://arxiv.org/abs/2609.03246
作者: Zachary P Bradshaw
类目: Representation Theory (math.RT); Machine Learning (cs.LG)
*备注: 26 pages, 5 figures
Abstract:Smoothness of a function on the real line is reflected in the decay of its Fourier transform, which suggests that smoothness of a function in L^2(G) for a group G should mean concentration of the Fourier coefficients at low frequency. Such a reading presupposes an ordering of the irreducible representations of G , but for non-abelian G , no ordering is canonical. Given a symmetric generating set S , the Laplacian of the associated Cayley graph is block diagonal over the dual, and we order the irreps by the mean of the eigenvalues in each block. This produces an ordering function \omega:\widehatG\to\mathbbR that depends only on the pair (G,S) . This function is bounded between zero and two, vanishing only at the trivial representation and achieving the upper bound exactly when the Cayley graph is bipartite. We then ask how much freedom the construction has. Within the class of operators satisfying natural axioms, the induced orderings are exactly the real functions on the dual vanishing at the trivial representation and agreeing on conjugate pairs, and the orderings coming from inversion orbits of conjugacy classes form a basis for them. We cut the freedom down further by requiring two additional inputs: nonnegativity of the class weights and a declaration of which group elements count as uniform incremental changes, which pins the operator to the Cayley-Laplacian up to positive scale. We observe that the construction persists for compact groups even though the Cayley graph does not, and we extend the theory to finite sets carrying a transitive group action, where the acting group selects which frequencies exist and the generating set orders them. The answer to the title question is therefore that smoothness is a property of a function together with a choice of group and generating set, not of the function alone.
[LG-86] Improving precipitation forecasts in an AI weather model using observational data
链接: https://arxiv.org/abs/2609.03210
作者: Julian F. Schmitt,Bertrand Delorme,Robert C. King,Yashica Patodia,Tapio Schneider,Aditi Sheshadri,Ravi Jain
类目: Atmospheric and Oceanic Physics (physics.ao-ph); Machine Learning (cs.LG)
*备注: 16 pages, 4 figures. Submitted to Science
Abstract:Artificial intelligence weather prediction (AIWP) systems now surpass state-of-the-art physical models for medium-range weather forecasting. Current global AIWP models are trained almost exclusively using one reanalysis dataset, ERA5, but it has known biases, particularly for precipitation. Here we fine-tune a graph-transformer architecture with IMERG precipitation data at 0.25° resolution. The resulting model improves medium-range continuous ranked probability scores by up to 19%, while also demonstrating superior skill for tropical storms and drizzle events. Our model exceeds the Brier skill score of state-of-the-art operational models on extreme rainfall prediction by 57% globally; however, a physics-based operational model remains more reliable for the heaviest precipitation events. Our results demonstrate that incorporating observations-based precipitation data directly into training can substantially improve precipitation forecasts.
[LG-87] Generative Nested Sampling of Atomistic Thermodynamic Landscapes
链接: https://arxiv.org/abs/2609.03193
作者: Alessandro Coretti,Nico Unglert,Sebastian Falkner,Georg K. H. Madsen,Christoph Dellago
类目: atistical Mechanics (cond-mat.stat-mech); Machine Learning (cs.LG); Computational Physics (physics.comp-ph)
*备注:
Abstract:Nested sampling (NS) resolves the thermodynamics of an atomistic system from a single simulation, but its practical reach is limited by the Markov-chain updates needed to decorrelate walkers within each likelihood-constrained ensemble. Flow-based NS has removed this bottleneck for gravitational-wave (GW) inference, yet its transfer to atomistic systems is not merely a change of application. Comparing a GW150914-like binary-black-hole likelihood with an eight-particle two-dimensional Lennard-Jones (LJ) system of comparable dimensionality, we show that the two landscapes differ fundamentally: atomistic multimodality is discrete and combinatorial, generated by particle permutations separated by hard collision walls, and its coordinate coupling is dense and collective, whereas the GW posterior exhibits smooth degeneracies and localized parameter coupling. Guided by this diagnosis, we introduce NS-Flows: a single conditional normalizing flow, conditioned on the NS energy bound and trained on a sliding window of recent live sets, that replaces MCMC by direct parallel draws corrected by importance-weighted rejection resampling. Live sets supply data self-consistently, allowing flow training without structured priors or a pre-existing dataset. For LJ disks in PBC, the algorithm reduces energy evaluations by over two orders of magnitude and wall-clock time by roughly one third, an advantage that becomes increasingly favorable as the cost of the potential grows. The flow’s generation efficiency further acts as a physical diagnostic: it varies non-monotonically along the annealing trajectory, is lowest in the dense disordered regime, and is quantitatively captured by the constrained ensemble’s internal mode complexity together with target drift across the training window, identifying liquid-like ensembles, rather than prior-target separation, as the hard case for current flow architectures.
[LG-88] A Closed-Form Formula for Consistent Lipschitz Regression on Metric Spaces with Sparse Neural Network Realizations
链接: https://arxiv.org/abs/2609.03129
作者: Ruiyang Hong,Hrad Ghoukasian,Anastasis Kratsios
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注: 75 pages, 17 figures
Abstract:Several classical machine-learning methods, such as KRRs and SVRs, are both computationally and analytically tractable since their estimators either admit closed-form expressions or are obtained by minimizing convex training objectives; neither feature is generally available for deep neural networks. We address this by introducing a simple closed-form ``two-stage’’ compositional formula \hatf for reconstructing an unknown Lipschitz function f:\mathcalX\to \mathbbR on a metric space (\mathcal X,\rho) from N i.i.d. noisy observations. Our main result is a high-probability uniform ( L^\infty ) recovery guarantee that jointly controls approximation and statistical errors while enjoying an optimization error of zero; in particular, we do not assume oracle access to an approximate ERM. Our secondary main results establish the optimality of our formula in three complementary senses. 1) Function space: On Ahlfors-regular metric spaces, the hypothesis class parameterized by our formula attains the optimal fat-shattering dimension. 2) Parameter space: Its dependence on the parameters is maximally numerically stable, in the sense that a smaller approximation error cannot be achieved with a smaller Lipschitz dependence on the model parameters. 3) Forward pass: Its dependence on the input is maximally regular, matching the Lipschitz constant of the target function f . When \mathcal X=[0,1]^d is equipped with the \ell^\infty norm, \hatf admits algorithmic ReLU-MLP and exact ReLU-multi-head transformer realizations of depth \mathcalO(\log(N)) with \mathcalO(N) nonzero parameters. Comments: 75 pages, 17 figures Subjects: Machine Learning (stat.ML); Machine Learning (cs.LG) Cite as: arXiv:2609.03129 [stat.ML] (or arXiv:2609.03129v1 [stat.ML] for this version) https://doi.org/10.48550/arXiv.2609.03129 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-89] Occupancy-based Quantile Risk Control
链接: https://arxiv.org/abs/2609.03104
作者: Zihao Shi,Huajun Xi,Bingyi Jing,Hongxin Wei
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:Conformal risk control is an emerging framework for the safe deployment of machine learning models with finite-sample guarantees. To accommodate a broader class of risk notions, quantile risk control extends this framework to quantile-based risk measures. However, existing methods either suffer from excessive conservatism or lack rigorous finite-sample guarantees. To address these limitations, we introduce Occupancy-based Quantile Risk Control (OQRC), a novel method that provides tight risk control bounds with finite-sample validity. Our key idea is to formulate risk control as a finite-occupancy problem by partitioning the loss space with the ordered calibration losses. Specifically, we estimate the distribution of test losses across the resulting bins and upper-bound the risk by the maximum loss attained within each bin. We then select the parameter \lambda such that this upper bound does not exceed a predefined threshold \alpha with high probability 1-\delta . Theoretically, we establish a finite-sample guarantee showing that OQRC yields tight risk control bounds that converge to the optimal bounds at a provable rate of \mathcalO_ p(n^-1/2) . Extensive experiments demonstrate the effectiveness of our method, reducing the risk gap by up to 78.64% on common benchmarks.
[LG-90] Advances in Machine Learning for Directed Evolution: A Five-Year Retrospective
链接: https://arxiv.org/abs/2609.03046
作者: Bruce J. Wittmann
类目: Biomolecules (q-bio.BM); Machine Learning (cs.LG)
*备注:
Abstract:The last five-plus years have seen many protein engineering disciplines transformed by advances in machine learning (ML), but the same cannot be said for directed evolution. Reflecting on a previously co-authored perspective, I discuss why I believe this to be the case, arguing that a disconnect between the goals of machine-learning-assisted directed evolution (MLDE) researchers–“identify an optimal protein”–and the goals of directed evolution more broadly–“identify a sufficient protein given time and resource constraints”–is a principal culprit. As an example, I highlight how nearly all current MLDE methods neglect to account for the cost of DNA synthesis, resulting in strategies that have limited practical applicability regardless of the underlying models’ capabilities. I close by discussing recent works that are exceptions to this overarching trend, and emphasize that the last five years of efforts in ML-assisted protein engineering and the prescribed reframe of MLDE objectives need not be mutually exclusive.
[LG-91] Physics-Informed Neural Network Surrogate for Oxygen Vacancy Dynamics in epitaxial mathrmSrTiO_3 on Si memristors via Dynamic Spectral Optimization
链接: https://arxiv.org/abs/2609.02966
作者: Rodion Podorozhny,Nikoleta Theodoropoulou,Jelena Tešić
类目: Materials Science (cond-mat.mtrl-sci); Machine Learning (cs.LG)
*备注:
Abstract:Physics-informed neural networks (PINNs) offer a promising framework for modeling semiconductor devices, yet standard architectures struggle with severe numerical stiffness and multiscale spatial discrepancies inherent to oxide heterostructures. Here, we demonstrate a cascaded PINN architecture coupled with a custom second-order Chebyshev second kind polynomial spectral optimizer (DSO V2 Hybrid) to model ion-electronic drift-diffusion transport in Pt/SrTiO _3 /Si memristive heterostructures across a 20 nm STO film on a 380 \mu m Si substrate. By isolating potential, carrier density, and vacancy transport into four sequentially trained sub-neural-networks, our model circumvents condition numbers exceeding 10^16 without operator splitting. The trained surrogate reproduces experimental conductive-AFM current-voltage hysteresis ( R^2 0.96 ) while ensuring strict Poisson consistency across continuous space. Compared to conventional finite-element solvers (e.g., COMSOL), the PINN surrogate enables differentiable inverse parameter estimation and linear time inference.
[LG-92] SurfSpec: Enhancing Off-Target-Agnostic Specificity by Bounding Pocket-Ligand Geometric Mismatch
链接: https://arxiv.org/abs/2609.02963
作者: Minyeong Hwang,Yoorim Gang,Ziseok Lee,Wooyeol Lee,Young Bin Park,Jae-Mun Choi,Kyungsu Kim,Eunho Yang
类目: Quantitative Methods (q-bio.QM); Machine Learning (cs.LG)
*备注:
Abstract:Lead optimization in structure-based drug design aims to improve target binding while avoiding unintended interactions with off-target pockets. However, existing affinity-driven methods do not explicitly control specificity, whereas current specificity-aware approaches commonly require prior knowledge of off-target structures. We address off-target-agnostic specificity-aware lead optimization by analyzing the geometric mismatch between a ligand and the target pocket. We provide a conservative specificity lower bound for geometrically separated off-targets without requiring access to off-target structures. By metricizing pocket–ligand mismatch, the triangle inequality shows that reducing target–ligand mismatch improves a conservative lower bound on mismatch to a separated off-target class, which can be translated into a specificity lower bound through an empirical geometry–affinity calibration. Motivated by this analysis, we introduce SurfSpec, an off-target-agnostic lead optimization framework that iteratively grows ligands toward under-occupied regions of the target pocket surface. SurfSpec alternates between linker generation toward selected target-surface patches, which provides geometric pseudo-labels, and refinement under a pocket-conditioned ligand prior, which restores these pseudo-labels into valid ligands. On the CrossDocked2020 test set, SurfSpec reduces geometric mismatch and outperforms evaluated off-target-agnostic lead optimization baselines in empirical specificity, while maintaining competitive target-affinity improvement.
[LG-93] FrOGS: Discrete Neural Sampler for Independent Alloy Configurations Across Chemical Conditions NEURIPS2026
链接: https://arxiv.org/abs/2609.02948
作者: Kyucheol Min,Elyssa Hofgard,Tess Smidt
类目: Materials Science (cond-mat.mtrl-sci); Statistical Mechanics (cond-mat.stat-mech); Machine Learning (cs.LG)
*备注: Submitted to the AI4Mat workshop at NeurIPS 2026
Abstract:Predicting the thermodynamic properties of an alloy requires sampling its configurations across many chemical conditions and recovering free energies on a common absolute scale. Markov chain Monte Carlo (MCMC) is the standard tool, but it requires separate simulations at different conditions, and auxiliary free-energy methods such as thermodynamic integration are used to place results on a common absolute scale. Modern discrete neural samplers typically use reverse KL divergence as the objective and can be mode-seeking or biased. We present Free energy Offering Generative Sampler (FrOGS), a hybrid discrete neural sampler that couples an autoregressive model to a continuous-time Markov chain (CTMC) to be trained jointly under a single shared loss. FrOGS draws i.i.d. configurations, returns an unbiased estimate of the partition function, and gives consistent estimates of thermodynamic observables. We train a single model across a wide range of chemical conditions to produce estimates on a common absolute free-energy scale. FrOGS matches exact finite-size results on the 2D Ising model and reference phase diagrams for AgPd and CuAu, without mode collapse. We additionally compare to SEGAL, a published autoregressive baseline, and find that only FrOGS recovers the stability range of the CuAu _3 phase.
[LG-94] Hadronic Mono-Z Dark Matter Sensitivity with Flow Matching on CMS Open Data
链接: https://arxiv.org/abs/2609.02923
作者: Hitesh Rasineni(1),Bhavishya Chebrolu(2) ((1) VIT-AP University, Amaravati, India, (2) Mohan Babu University, Tirupati, India)
类目: High Energy Physics - Experiment (hep-ex); Machine Learning (cs.LG); High Energy Physics - Phenomenology (hep-ph); Instrumentation and Detectors (physics.ins-det)
*备注: Companion study to arXiv:2607.13771v1 (leptonic channel)
Abstract:We present a projected sensitivity study for hadronic mono- Z dark-matter production using CMS Run~2015D HTMHT open data corresponding to 2.256382381~\invfb, from which 1,439,523 events satisfy the hadronic mono- Z selection. Backgrounds are modelled with a conditional flow-matching continuous normalizing flow trained on the selected HTMHT events and evaluated on a held-out validation split reweighted to the full selected population. To mitigate artifacts from missing-object features and avoid in-sample scoring bias we apply sentinel imputation for undefined angular features, persist the train/validation split indices, and enforce a minimum reported background yield of 20 events when selecting the working point. A signal-side offline trigger proxy is applied to the simulated signal before scoring. Under this procedure the baseline analysis yields expected significances of 2.89 \sigma , 7.62 \sigma , and 7.41 \sigma for three simplified-model benchmarks. An ablation study that removes the detailed extra-jet kinematics reduces the expected significance by 53–71%, indicating that extra-jet topology carries substantial discriminating power in the hadronic mono- Z channel. These results are projected sensitivities (no unblinding performed); the limitations and reproducibility of the study are discussed in Sections limitations and reproducibility.
附件下载
