本篇博文主要内容为 2026-10-28 从Arxiv.org论文网站获取的最新论文列表，自动更新，按照NLP、CV、ML、AI、IR、MA六个大方向区分。
说明：每日论文数据从Arxiv.org获取，每天早上12:30左右定时自动更新。
提示: 当天未及时更新，有可能是Arxiv当日未有新的论文发布，也有可能是脚本出错。尽可能会在当天修复。
目录
概览 (2026-10-28)
今日共更新875篇论文,其中:
- 自然语言处理共108篇(Computation and Language (cs.CL))
- 人工智能共254篇(Artificial Intelligence (cs.AI))
- 计算机视觉共156篇(Computer Vision and Pattern Recognition (cs.CV))
- 机器学习共237篇(Machine Learning (cs.LG))
- 多智能体系统共27篇(Multiagent Systems (cs.MA))
- 信息检索共37篇(Information Retrieval (cs.IR))
- 人机交互共21篇(Human-Computer Interaction (cs.HC))
多智能体系统
[MA-0] Can LLM s Uncover Hidden Physical Laws through Free Exploration?
【速读】：该论文旨在解决大语言模型（Large Language Models, LLMs）在未知物理规律的模拟环境中，能否自主发现异常现象并基于实验探究揭示隐藏物理定律的能力问题。其核心挑战在于评估模型是否具备类人科学探索能力——即在未被明确指示研究目标的情况下，通过观察、提出假设、设计实验与迭代验证，识别出与常规物理相悖的规律。解决方案的关键在于构建名为LongChang的自动化框架，该框架可生成具有非标准物理法则的虚拟世界，并让多个独立的LLM代理协同完成“场景生成-工具设计-实验执行-结果评估”的闭环探索流程。其中，各代理分别负责生成实验场景、评审工具构型、验证报告正确性及依据探索记录引导后续演化，从而实现无需人工干预的自主科学研究循环。实验结果显示，仅有GPT-6-Astra实现了16.7%的发现率，其余模型均未能成功识别异常规律，表明当前主流大语言模型在跨域推理与系统性科学探索方面仍存在显著局限。
链接: https://arxiv.org/abs/2610.37649
作者: Zhen Yang,Wen Wang,Xander Xu,Ying-Cong Chen
机构: HKUST(GZ)(香港科技大学(广州)); ZJU(浙江大学); Alibaba Group(阿里巴巴集团); HKUST(香港科技大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注: this https URL
Abstract:Scientific discovery often begins with an unexpected observation that prompts a new research question. In a world with altered physical laws, can large language models (LLMs) notice unexpected behavior, such as heavier objects falling more slowly, and discover the law through experiments without being told what to investigate? To evaluate this ability, we introduce LongChang, an automated framework that creates simulated worlds with altered physics, where one or more laws differ from ordinary physics. The evaluated LLMs act as explorers, manipulating objects and building tools to investigate their own questions without research tasks or hints about these laws. Separate LLM agents generate and review scenes, review tool designs, and evaluate reports for correctness and experimental support using the altered physical laws and exploration records. These records guide scene evolution for further exploration, allowing the full cycle to run without human intervention. Among 12 LLMs, GPT-6-Astra achieves a discovery rate of 16.7%, while the other 11 score 0%. The exploration records reveal uninvestigated anomalies and only partial identification of the altered physical laws.
[MA-1] Rare Relevant and Still Wrong: How Memory Fails LLM Group Decisions
【速读】：该论文旨在解决大语言模型（LLM）群体讨论在压缩为短时记忆供最终决策者使用时，如何有效选择保留信息的问题。核心挑战在于：在“隐藏特征任务”（hidden-profile tasks）中，事实的普遍性（即被多少参与者持有）与其实用性存在混淆，导致常规启发式策略失效。现有两种自然启发式——优先保留罕见事实（rarest-first）或保留与决策标准相关事实（relevance-based）——均在特定情境下表现不佳。研究提出CrossProfile基准测试，通过60个模板生成480个任务，系统性地分离事实归属与文本内容，验证了两种启发式在预测性失败中的局限性：基于相关性的选择会引入误导性“干扰项”（decoy），而基于稀有性的选择仅在罕见事实恰好为有用信息时才有效，整体表现接近随机。实验表明，决策模型的准确性取决于记忆中证据的平衡分布，而非选择策略本身。此外，研究发现决策模型还受“社会证明”（social proof）机制干扰，即对来自多名代理的事实赋予过高权重；移除或随机重分配归属标签后，准确率提升13至24分。更关键的是，在真实讨论场景中，完整对话记录的准确率甚至低于无笔记状态，说明原始证据在记忆压缩前已严重损耗。因此，解决方案的关键不在于优化记忆选择规则，而在于重构信息传递机制，避免早期信息失真。
链接: https://arxiv.org/abs/2610.37624
作者: Anouar Imel,Imad Charradi,Abderrahman Sghir,Khalil El Khammar
机构: 未知
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Multiagent Systems (cs.MA); Social and Information Networks (cs.SI)
备注:
Abstract:When an LLM group’s discussion is compressed into a short memory for a final decision maker, what should the memory keep? Two natural heuristics keep the rarest claims or the claims relevant to the decision criteria. On hidden-profile tasks both look attractive, because how widely a fact is held is confounded with whether it helps. We introduce CrossProfile, a diagnostic benchmark of 60 templates with automatically verified answer keys, in which the ownership of supporting, decoy, and irrelevant facts is crossed while their text is held fixed (480 tasks). Across Qwen2.5-7B, Llama-3.1-8B, and two DeepSeek models, both heuristics fail in predictable places. Selecting by relevance readmits decoys, because decoys are on topic: even a perfect relevance filter loses 23 to 31 points to rarest-first selection in the classic hidden-profile configuration. Rarest-first selection is no better than random selection on average and wins only where the rare facts happen to be the useful ones. In both cases accuracy follows the balance of evidence the memory keeps, not the policy that kept it. The decision model adds a third failure, a form of social proof: it gives extra weight to facts attributed to many agents. Removing the attribution tags, or reassigning them at random with line order fixed, raises accuracy in the classic configuration by 13 to 24 points. In natural discussions the evidence is lost before any memory acts: the complete transcript yields lower accuracy than no notes at all.
[MA-2] ANDEM: Balancing Centralized and Decentralized Critics to Address the Information Gap in Complex Multi-Agent Environments
【速读】：该论文旨在解决在集中式训练、分布式执行（CTDE）框架下，多智能体强化学习中因训练与执行阶段信息不一致所导致的“信息鸿沟”问题。具体而言，在训练时，中心化评论家（centralized critic）可访问全局状态，而执行时各智能体仅能基于局部观测进行决策，这种信息不对称在部分可观测性、观测噪声以及任务与智能体动态开放（TAO）场景下尤为显著。现有方法通常单向转移中心化指导，仅一次移除或蒸馏评论家，导致策略在训练过程中始终依赖不可用的全局信息，缺乏在无中心化支持下自主决策的充分练习。本文提出TANDEM框架，其核心在于保留一个中心化评论家，并在每个训练步骤中以两种方式读取：一是从全局状态的中心化读取，二是基于各智能体执行时实际可用信息的去中心化读取。通过一个无参数门控机制对两种读取结果进行比较，仅在二者一致时采用去中心化读取进行策略优化，而在局部视图误导时重新启用中心化读取；同时引入共享目标，将去中心化读取逐步拉近至中心化读取。该方法形式化为任务-智能体开放的去中心化部分可观测马尔可夫决策过程（TAO-Dec-POMDP），并在三个独立隔离信息鸿沟不同来源的协作任务（部分可观测性、观测噪声、火灾扑救中的开放性）上验证。实验表明，TANDEM在九个设置中表现最优，且在所有情况下均不显著劣于任一基线方法，证明其有效缓解了训练与执行间的信息不对称问题。
链接: https://arxiv.org/abs/2610.37592
作者: Alireza Saleh Abadi,Leen-Kiat Soh,Adam Eck,Prashant Doshi
机构: University of Nebraska-Lincoln(内布拉斯加大学林肯分校); Oberlin College(欧柏林学院); University of Georgia(佐治亚大学)
类目: Multiagent Systems (cs.MA)
备注:
Abstract:Cooperative multi-agent reinforcement learning under centralized training with decentralized execution (CTDE) shapes decentralized policies with a centralized critic that observes the global state, then removes the critic at execution. The policies are optimized to rely on information they will not have when they act, an information gap between training and execution. It widens under partial observability, observation noise, and openness, where the set of agents and tasks present changes during an episode. Existing methods transfer centralized guidance in one direction only, removing or distilling the critic once, so the policies never rehearse acting without the critic while they can still be corrected. We propose TANDEM, which keeps one centralized critic and reads it two ways at each training step: a centralized read from the global state, and a decentralized read restricted to the information each agent will have at execution. A parameter-free gate compares the two reads for each agent and trains the policy on the decentralized read where they agree, reinstating the centralized read only where the local view is misleading. A shared target pulls the decentralized read toward the centralized one. We formalize the setting as a task- and agent-open decentralized partially observable Markov decision process (TAO-Dec-POMDP) and evaluate TANDEM on three cooperative domains, each isolating one form of the information gap: discovery for partial observability, sampling for observation noise, and openness in wildfire suppression. We compare against six CTDE baselines and ablate the gate and the shared target; the supplementary material holds the full variant sweep, diagnostics, and learning curves. We measure return at execution, once the centralized critic is removed. TANDEM is the best method on eight of the nine setups and never significantly worse than any baseline.
[MA-3] Strategyproof Deep Reinforcement Learning for Online Resource Allocation
【速读】：该论文旨在解决在线资源分配中的两个耦合挑战：在需求不确定的条件下高效分配稀缺资源，以及确保参与方真实申报任务估值与资源需求（即保证策略无关性，strategyproofness）。传统强化学习（Reinforcement Learning, RL）虽能通过经验自适应调整分配策略，但无法保证策略无关性；现有策略无关的强化学习方法采用线性分配规则，限制了对复杂资源交互关系的建模能力。而神经机制设计方法虽具备更强的表达能力，却仅能提供近似的激励相容保障。针对上述问题，本文提出首个用于在线资源分配的策略无关深度强化学习机制，其核心创新在于三个关键组件：(i) 一种构造性单调神经网络，通过强制非线性交互结构满足单调性约束（更高申报价值不会降低接受概率，更长申报时长或更大资源需求不会提高接受概率）；(ii) 一种广义临界支付机制，即在其他输入固定时，最低被接受的申报价值作为支付标准；(iii) 采用强化学习或对已有规则进行模仿学习的方式进行训练。其中，单调性与临界支付共同保障了策略无关性和个体理性，而神经网络结构则赋予模型灵活性，模仿学习显著提升了训练效率。实验表明，在一个包含四类资源及两组合作方的流式场景中，该神经规则达到离线最优福利的97.8%，优于线性学习器（93.8%）和先到先服务（89.0%），分别缩小了65%和80%的最优性差距；且模仿学习仅需261秒即可达到97.5%的性能，相较传统强化学习提速约89倍。
链接: https://arxiv.org/abs/2610.37504
作者: Fariba Dehghan,Sebastian Stein,Vahid Yazdanpanah,Stephanie Gauthier
机构: University of Southampton (南安普顿大学)
类目: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI); Computer Science and Game Theory (cs.GT)
备注: 15 pages, 2 figures
Abstract:Online resource allocation faces two coupled challenges: (1) allocating scarce resources efficiently under uncertain future demand, and (2) ensuring that agents report task valuations and requirements truthfully (strategyproofness). Reinforcement Learning (RL) adapts allocation decisions through experience but does not necessarily ensure strategyproofness. Existing strategyproof RL uses a linear allocation rule, which limits the resource interactions it can represent. Neural mechanism design methods offer greater expressiveness, but provide only approximate incentive guarantees. Against this background, we introduce the first strategyproof deep RL mechanism for online resource allocation. It integrates three novel components: (i) a monotonic neural network that learns nonlinear interactions by construction (higher reported value cannot reduce acceptance, while a longer reported duration or larger resource requirement cannot increase it); (ii) a generalised critical payment: the lowest accepted value, with all other inputs fixed; and (iii) training by either RL or imitation of an existing rule. Together, monotonicity and the critical payment ensure strategyproofness and individual rationality, while the neural rule provides flexibility and imitation improves training efficiency. Empirically, on a four-resource stream with two partner pairs that a linear difference cannot represent, the neural rule achieves 97.8% of an offline welfare upper bound, compared with 93.8% for a linear learner and 89.0% for first-come, first-served, closing 65% and 80% of the remaining optimality gap, respectively. Moreover, imitation reaches 97.5% in only 261 seconds, about 89 times faster than RL.
[MA-4] Concordat: Typed Sessions for LLM Agents
【速读】：该论文旨在解决基于大语言模型（Large Language Models, LLMs）的智能体在执行具有实际影响的操作（如取消预订、退款、账户修改等）时，因缺乏强制性行为约束而导致的不可靠合规问题。现有方法依赖模型在提示（prompt）中遵循规则，但其执行可靠性完全取决于模型自身的选择，难以保证关键安全规则（如“更改前需获得客户明确确认”）被严格执行。为此，论文提出Concordat——一个基于有限状态协议（finite-state protocol）的Python库，将智能体建模为具有明确定义状态转换接口的对象，借鉴类型态（typestate）与会话类型（session types）的思想，在运行时强制执行协议逻辑。其核心解决方案在于：在每一轮交互中，仅向模型提供当前协议状态下合法的工具调用，非法调用在执行前即被拒绝，避免产生副作用；关键决策权保留给人类，只能请求而无法自动执行；同时对令牌消耗和工具调用预算进行跨状态转移的线性分配与追踪。协议在定义时即进行静态验证，且子类若可能绕过父类协议，则会被拒绝，实现了行为层面的里氏替换原则（Liskov substitution）。在tau-bench的50个航空任务评估中，面对模拟客户的压力，标准代理在GPT-4.1下有13次、Claude Haiku 5.5下有29次未获确认即执行变更，而Concordat代理零违规，且任务成功率无显著差异。此外，tau-bench的奖励机制未能检测到多数违规行为。Concordat已开源。
链接: https://arxiv.org/abs/2610.37461
作者: Milad Sikaroudi
机构: 独立研究者(Independent Researcher)
类目: Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注: 11 pages, 1 figure, 1 table. Code, benchmark harness and results: this https URL
Abstract:Agents built on large language models are increasingly allowed to take consequential actions: cancelling bookings, issuing refunds, changing accounts. The rules that govern those actions, such as “obtain the customer’s explicit confirmation before changing anything”, are usually stated in the prompt and followed only as reliably as the model chooses to follow them. We present Concordat, a Python library that treats an agent as an object whose interface is a finite-state protocol, in the spirit of typestate and session types, and enforces that protocol on the model at run time. On every turn the model is offered only the tools that are legal in the current protocol state, calls outside that set are rejected before they have side effects, decisions reserved for humans can only be requested, and token and tool-call budgets are charged and split linearly across handoffs. Protocols are checked when they are defined, and a subclass of an agent is rejected if it could bypass its parent’s protocol, a behavioural form of Liskov substitution. We evaluate a Concordat agent against the standard agent of tau-bench on its 50 airline tasks, adding simulated customers who pressure the agent to skip confirmation. The standard agent made changes without the required “yes” in 13 of 50 tasks with gpt-4.1 and in 29 of 128 task-runs with Claude Haiku 5.5; the Concordat agent made none, with no significant difference in task success. tau-bench’s own reward did not detect most of these violations. Concordat is open source.
[MA-5] st-Time Communication in Trading Agent Swarms
【速读】：该论文旨在解决在缺乏共同目标与即时明确反馈的复杂环境中，多智能体系统（multi-agent system）中通信对决策一致性（即“羊群效应”）的影响问题。在真实金融市场交易场景中，各智能体独立管理账户，其决策由市场延迟且噪声较大的方式评估，这使得智能体间更接近于独立决策者群体而非协同团队，因此传统意义上的收敛并非必然结果。研究通过在七天内（2026年10月1日至9日）运行来自四个模型家族的200个智能体，分组考察从隔离到共享笔记等多种通信机制下的行为演化。关键发现表明，在其中两个模型家族中，通信组相较于孤立组显著提高了成员持有的股票一致性；群体收益主要取决于成员共持资产的表现。进一步分析显示，看到同伴发布的消息会使智能体重用其八词短语的概率提升7个百分点，而当智能体因位置相近无法直接接收信息时，仍存在14%的相同输入出现率，表明部分一致性源于初始条件而非通信。值得注意的是，单日交易行为并未因阅读消息而发生显著变化，但跨日追踪显示通信组在持仓和决策规则采纳上表现出更强的同步性。由此揭示：评估羊群效应需以无同伴交互的基线为参照，以识别超出随机一致性的“超额收敛”。该研究的核心解决方案在于构建可量化、可追踪的通信-行为关联框架，从而揭示非协作环境下通信如何驱动隐性协调与系统性风险积累。
链接: https://arxiv.org/abs/2610.37448
作者: Avid Fayaz,Harry Zhang,Håvard Alstadheim,Erik Nordlöf,Vilhelm Karlin,André Silva
机构: 未知
类目: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 31 pages, 10 figures, 10 tables
Abstract:Communication between agents has mostly been studied on tasks with a shared objective and fast, clear feedback. Trading has neither: each agent manages its own account, with decisions being scored by markets late and noisily. A trading swarm is thus a population of independent decision-makers rather than a team, for which convergence is a risk and not the goal. We ran 200 agents from four model families in groups of ten under five communication regimes, from isolation to message boards and shared notes, over seven US trading days (1 to 9 October 2026), logging everything each agent read before every decision. In two of the four families, communicating groups came to hold far more of the same stocks than the isolated group. A group’s return was largely the return on what its members held in common. Seeing a post from a groupmate made an agent 7 percentage points more likely to reuse an eight-word phrase from it. Agents that started too close together to see each other’s posts shared such a phrase in 14% of cases, the rate that common inputs produce. Seeing the post did not detectably change that morning’s trades. Case studies traced the adoption of positions and decision rules over subsequent days, and communicating groups sold in step more often than isolated ones. Because isolated groups also converged on the same stocks, sometimes completely, assessments of herding need to measure excess convergence relative to a no-peer baseline.
[MA-6] Agent -Conditioned Routing: Diverse Reasoning Paths Within One Frozen Model NEURIPS2026
【速读】：该论文旨在解决生成式 AI 在面对同一问题时倾向于重复单一求解路径，导致输出多样性不足的问题。尽管通过生成多个解法可暴露不同方法，但多数额外轨迹仍为同一主导方法的变体。其核心解决方案是提出一种无需用户指定角色、预设方法、微调或模型复制的代理条件路由（Agent-Conditioned Routing, ACR）机制，利用冻结的稀疏专家混合模型（MoE）中固有的条件容量，通过为并行生成轨迹分配持久且未训练的共享/私有专家可用性掩码（expert-eligibility masks），在保持原有输入依赖的令牌路由机制的基础上，激发模型探索多样化的求解路径。实验结果表明，在经过筛选的126道多模式数学与物理基准测试中，ACR在相同输出预算下，使至少出现两种不同主方法的题目数量分别达到GLM 5.3的2倍（37.3 vs. 18.7）和GPT-5.6 Sol的1.67倍（45.00 vs. 27.00），显著提升解法多样性；盲评审计进一步验证了其在双骨干模型上均能产生更多多方法解法，尽管正确率略低。消融实验显示，匹配的专家可用性掩码对多样性贡献有限，而基于令牌重采样的掩码则更利于提高正确率。此外，引入知情顺序提示可进一步增强多样性。整体而言，ACR仅增加约160 KiB的掩码状态开销，在三题微型基准测试中延迟与峰值显存占用均控制在批量温度采样水平的3%以内，展现出高效且轻量的实用性。
链接: https://arxiv.org/abs/2610.37375
作者: Ananth Kalyanasundaram,Ankush Kadu,Athira Gopal,Aswanth Krishnan
机构: QpiAI(印度)
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Multiagent Systems (cs.MA)
备注: 17 pages, 4 figures, 11 tables. Extended version of a paper accepted at the NeurIPS 2026 Workshop on Resource-Aware Agentic AI (non-archival)
Abstract:Generating several solutions can expose alternative methods, but additional trajectories often paraphrase one dominant approach. We investigate whether the conditional capacity of a frozen sparse mixture-of-experts (MoE) model can provide such diversity from one problem prompt, without user-assigned roles, prescribed methods, fine-tuning, or model copies. Agent-Conditioned Routing (ACR) assigns m parallel trajectories persistent, untrained shared/private expert-eligibility masks while preserving the learned, input-dependent token router. On a screened 126-question multi-mode mathematics and physics benchmark, with model judges reading the reported derivations in full, we count the number of problems yielding at least two distinct primary methods among soft-correct solutions in four trajectories. ACR reaches approximately twice as many such problems as the strongest temperature setting under GLM 5.3 (37.3 versus 18.7), and 1.67 times as many under GPT-5.6 Sol (45.00 versus 27.00), at matched output budgets. A blinded four-expert audit on a stratified sample of 30 problems also finds more multi-method solutions with ACR on both backbones, with lower correctness. Matched-eligibility ablations show modest multi-method count advantages for ACR, while token-resampled masks attain higher correctness. An informed sequential prompt elicits more diversity still, using previous solutions and explicit requests for alternative methods. ACR adds about 160 KiB of mask state. In a three-problem microbenchmark capped at 1,024 tokens, its latency and peak VRAM usage remain within 3% of batched temperature sampling.
[MA-7] Aligning Institutional Power and Operational Ability
【速读】：该论文旨在解决生成式 AI（Generative AI）系统在社会可持续性方面面临的规范性困境，即如何确保责任分配与主体履行能力相匹配，避免出现不可实现的期望或权力升级等规范性失衡问题。其核心解决方案在于提出一个以过程导向建模和制度性权力为核心的正式分析框架，用于评估具有规范意义的多智能体系统中的规范能力一致性；该框架通过识别操作流程与制度状态之间的对称关系，揭示规范地位生命周期的动态机制，并据此提出修复策略，从而实现对规范性设计的系统性优化。
链接: https://arxiv.org/abs/2610.37342
作者: Andrea Marino,Giovanni Sileno,Thomas van Binsbergen,Tom van Engers
机构: Informatics Institute, University of Amsterdam (阿姆斯特丹大学信息学院), The Netherlands (荷兰)
类目: Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注:
Abstract:To be socially sustainable, Agentic AI systems require normative designs that align responsibilities with the ability to fulfil them. Identifying incongruences such as unsatisfiable expectations or privilege escalation is crucial in avoiding normative pitfalls. In this study, we discuss design dimensions suitable for describing multi-agent systems in light of this norm-ability alignment requirement. We propose a formal framework centred on process-oriented modelling and institutional power for the analysis of a given normatively relevant process, discuss strategies for repairing it, and elaborate on how the symmetrical relation between operational process and institutional state affects the lifecycle of normative positions.
[MA-8] Evaluating Priority Dispatch for Premium Membership in On-demand Taxi-sharing
【速读】：该论文旨在解决在按需出租车共享服务中引入高级会员制（premium membership）时，如何在调度层面实现可行性的问题。其核心挑战在于，在保障高级会员优先服务体验的同时，避免对普通用户造成过大的服务负担，从而维持整体系统效率与公平性。解决方案的关键在于提出两种基于调度层面的优化方法：一是针对高级会员缩短服务截止时间（deadline shortening），二是在此基础上进一步引入专用于高级会员的专属车辆（dedicated premium vehicles）。通过在三种不同需求场景下（平稳均匀需求、具有高峰时段的时间变化需求、以及集中在城市区域的高峰交通需求）基于SUMO的仿真对比分析，研究发现最优的优先级机制应根据运营条件和高级会员比例动态调整。因此，采用条件依赖的调度方法切换策略，能够在提供高级会员服务优势的同时，有效控制对非高级会员用户的负面影响，实现整体系统的优化。
链接: https://arxiv.org/abs/2610.37011
作者: Nozomi Yamamoto,Hiroki Sakaji,Itsuki Noda
机构: Hokkaido University (北海道大学)
类目: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI)
备注:
Abstract:In this paper, we analyze the dispatch-level feasibility of introducing premium membership into an on-demand taxi-sharing service using a sequential optimal insertion method. We propose two dispatch-level methods: deadline shortening for premium users, and deadline shortening with dedicated premium vehicles. We compare them with a baseline dispatch method through SUMO-based simulations under three demand scenarios: steady uniform demand, time-varying demand with a peak period, and urban-focused demand with concentrated peak traffic. Our results show that the suitable priority mechanism depends on operating conditions and premium ratio, suggesting that condition-dependent method switching may provide premium benefits while keeping non-premium-user burden smaller for overall optimization.
[MA-9] A Serious-Game Testbed for Self-Governing Human-Machine Institutions
【速读】：该论文旨在解决混合智能体（包括人类与机器）在复杂协作环境中如何实现自我治理的问题，尤其关注多智能体系统在公共资源管理、合作生存及制度选择等情境下的行为演化与治理有效性。其核心挑战在于：当人类与生成式AI（Generative AI）智能体共同参与动态决策过程时，不同动机与行为模式的智能体组合是否能形成稳定且高效的治理结构，以及这些结构如何受个体策略、互动机制与规则执行的影响。解决方案的关键在于构建一个名为Megabike的多智能体模拟器与严肃游戏平台，通过设定具有相互依赖性的竞技场环境，实现对资源分配、集体行动、政治制度选择等关键治理维度的可控实验。该平台具备全规则自动化执行、行为日志完整记录、人类与机器可任意替代任一角色的特性，从而为研究人机混合群体中的治理机制提供了可量化、可重复的测试基准。实验表明，不同智能体原型（如合作型、懒惰型、颠覆型或对抗型）的组合可能产生意料之外甚至反效果的结果，凸显了治理设计中非线性交互的重要性，进而为未来系统性地融合人类与计算智能体开展控制实验提供了可扩展的框架。
链接: https://arxiv.org/abs/2610.36989
作者: Callum Berger,Jeremy Pitt,Stefan Sarkadi
机构: APEx Defence and Security Institute, University of Lincoln(林肯大学); Imperial College London(帝国理工学院)
类目: Multiagent Systems (cs.MA); Human-Computer Interaction (cs.HC)
备注:
Abstract:AI agents no longer act alone: they cooperate, coordinate and compete within and between groups, increasingly alongside humans who are in, on or out of ``the loop’'. Whether such mixed groups can govern themselves is an open question about machine behaviour and human-computer interaction that needs a testbed for hybrid institutions. We present Megabike, a multi-agent simulator and serious game with interdependent arenas for common-pool resource management, cooperative survival and constitutional choice. Riders sharing a megabike must agree their arrangements for steering, pedalling and distributing resources, and their political regime, while negotiating with rival bikes and avoiding an existential threat. Every rule is enforced by the simulator, every action is logged, and any seat can be taken by a human or a machine, so governance can be studied under well-defined metrics. We demonstrate Megabike with forty sessions in which language-model riders with cooperative, lazy, subversive or opposing dispositions chose regimes, elected leaders and expelled one another. The results show that different combinations of agent archetypes could produce unexpected and counterproductive outcomes. We conclude with extensions for systematic experiments mixing human and computational agents.
[MA-10] Remembering Is Not Enough: Refining Memories and Learning When to Use Them in Multi-Agent Systems
【速读】：该论文旨在解决大语言模型（Large Language Model, LLM）驱动的多智能体系统中记忆机制存在的两大核心问题：一是存储的经验建议可能存在缺陷，二是难以判断在特定情境下是否应使用某条记忆。传统方法仅依赖团队整体的成功或失败来评估记忆的有效性，但这一标准不可靠，容易导致错误记忆被保留或正确记忆被误用。为此，论文提出一种无需训练的框架DiscernMem，其关键在于两个创新组件：一是缺陷感知的记忆优化（Defect-Aware Memory Refinement, DMR），通过对比单条记忆与无记忆情况下的团队执行表现，识别规则缺陷，并在当前及历史任务上验证修复方案，从而提升记忆质量并防止丢失已有有效行为；二是边界学习元记忆（Boundary-Learning Meta-Memory, BLM），记录每条记忆的过往使用记录及其效果，基于任务上下文和接收消息动态学习何时应启用或回避特定记忆，实现对记忆使用的精准引导。二者协同作用，使多智能体团队能够在纠正错误经验的同时保留原本适用但被误用的有效规则。在Qwen3-4B-Instruct-2507模型上针对AutoGen、CAMEL和MacNet等四个基准的实验表明，DiscernMem在多项任务中取得最高平均得分，优于最强基线0.51–4.32分，尤其在代码生成与符号规划任务中表现突出，如在CAMEL上的PDDL得分达37.71，显著超越基线的28.48。
链接: https://arxiv.org/abs/2610.36652
作者: Yipeng Hao,Yuxuan Cai,Wenkai Ji,Jie Zhou,Liyang Yu,Qin Chen,Zhiguo Yan,Shulei Zheng,Liang He
机构: East China Normal University(华东师范大学); Ocean University of China(海洋大学); Tongji University(同济大学); Information Engineering University(信息工程大学)
类目: Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注:
Abstract:Memory helps LLM-based multi-agent systems reuse experience, but stored advice can be flawed or applied in the wrong situation. Team-level success or failure alone cannot reliably show whether a memory needs correction or should not have been used. We propose DiscernMem, a training-free framework that improves both memory quality and memory use. To improve quality, Defect-Aware Memory Refinement (DMR) compares team executions with a single memory and with no memory, diagnoses rule defects, and tests repairs on current and earlier tasks to guard against losing previously useful behavior. To guide appropriate use, Boundary-Learning Meta-Memory (BLM) records each memory’s past uses and effects, learning when its owner role should apply or avoid it based on task context and incoming messages. Together, these components help teams correct flawed experience while retaining useful rules that were applied incorrectly. Experiments with Qwen3-4B-Instruct-2507 on four benchmarks show that DiscernMem achieves the highest average score on AutoGen, CAMEL, and MacNet, exceeding the strongest reported baselines by 0.51-4.32 points. It leads on code generation and symbolic planning across all three systems, including a PDDL score of 37.71 versus 28.48 on CAMEL.
[MA-11] GarmentSegAgent : Autonomous Garment Segmentation for Web Fashion Images via Textual Self-Refinement
【速读】：该论文旨在解决网络时尚图像中由于视觉异质性导致的服装分割可靠性问题，尤其针对在缺乏人工空间提示的情况下，如何自主识别并纠正特定于服装的分割错误。其核心解决方案是提出GarmentSegAgent框架，该框架通过语义规划、空间执行与视觉评估的迭代文本自精炼机制实现自主优化。关键创新在于引入CAST（Constrained Attribution-guided Selective TextGrad）方法，将视觉误差归因与定向文本优化相连接：CAST能够识别导致分割错误的指令成分，基于文本梯度选择性地更新这些成分，并结合服装特定约束验证优化后的指令。该方法在推理阶段实现图像级误差修正，同时保持基础模型参数冻结。在DeepFashion2、ModaNet和Fashionpedia数据集上的实验表明，该方法在区域重叠率和边界精度方面均取得一致提升，且优于传统TextGrad方法，展现出更强的鲁棒性。通过视觉反馈引导可控的指令更新，GarmentSegAgent为自主服装理解提供了新范式，具备在可扩展时尚内容分析、视觉检索及消费者到店铺匹配等场景中的应用潜力。
链接: https://arxiv.org/abs/2610.36506
作者: Jia Chen,Chao Huang,Xinrong Hu,Mingdi Huang,Ruyang Zhao,Jun Bai
机构: Wuhan Textile University (武汉纺织大学); Australian Institute for Machine Learning (AIML) (澳大利亚机器学习研究所), Adelaide University (阿德莱德大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Multiagent Systems (cs.MA)
备注: 12 pages
Abstract:Web fashion images support applications such as visual product search and consumer-to-shop matching, yet their visual heterogeneity makes reliable garment segmentation challenging. Despite advances in promptable segmentation, autonomously identifying and correcting garment-specific errors without human-provided spatial prompts remains difficult. We propose GarmentSegAgent, an autonomous framework that integrates semantic planning, spatial execution, and visual evaluation through iterative textual self-refinement. At its core, we introduce CAST (Constrained Attribution-guided Selective TextGrad), which connects visual error attribution with targeted textual optimization. CAST identifies instruction components responsible for segmentation errors, selectively updates them through textual gradients, and validates the resulting instructions against garment-specific constraints. This enables image-specific error correction at inference time while keeping the underlying foundation models frozen. Experiments on DeepFashion2, ModaNet, and Fashionpedia demonstrate consistent improvements in region overlap and boundary accuracy, alongside more reliable refinement than conventional TextGrad. By enabling visual feedback to guide controlled instruction updates, GarmentSegAgent offers a promising direction for autonomous garment understanding, with potential applications in scalable fashion content analysis, visual retrieval, and consumer-to-shop matching on the Web.
[MA-12] Bayesian Replicator Dynamics and the Inversion of Strict Dominance in Bimatrix Games
【速读】：该论文旨在解决演化博弈论中复制者动态（Replicator Dynamics）在不同收益函数下行为不一致的问题，特别是其在连续时间贝叶斯推理（Continuous-time Bayesian Inference）中的适用性边界。核心问题是：何种博弈结构下，复制者动态可被解释为贝叶斯更新过程？解决方案的关键在于识别出两类不同的博弈：一类是不存在严格占优策略的博弈，另一类则包含至少一个玩家具有严格占优策略的博弈。研究发现，在标准期望效用收益下，仅当博弈中无严格占优策略时，复制者动态才等价于贝叶斯更新；而采用对数收益（Logarithmic Payoff）时，这一等价性在所有两玩家博弈中均成立。这导致在存在占优策略的博弈中，两种收益定义下的动态方向相反——例如在囚徒困境中，对数收益使合作成为稳定结果，而传统收益则导致合作消失。此外，研究揭示了两类收益函数共享精确的哈密顿结构（Hamiltonian Structure），并由此提出可通过匹配硬币博弈（Matching Pennies）中周期振荡的等时性（isochrony）进行行为检验。因此，证据整合方式（即对数形式）决定了动态选择所能达到的战略均衡范围，进而影响学习近似贝叶斯更新时的演化路径。
链接: https://arxiv.org/abs/2610.36345
作者: Jaime Ruiz-Serra,Nihat Ay,Michael Harré
机构: Centre for Complex Systems, The University of Sydney(悉尼大学复杂系统中心); Institute for Data Science Foundations, Hamburg University of Technology(汉堡工业大学数据科学基础研究所)
类目: Computer Science and Game Theory (cs.GT); Multiagent Systems (cs.MA)
备注: 18 pages, 3 figures. The following article has been submitted to Chaos
Abstract:Replicator dynamics, the central model of evolutionary game theory, describe populations evolving under selection and agents learning by trial and error. They also admit a third use: continuous-time Bayesian inference, in which an agent planning against an opponent accumulates evidence for each candidate strategy. Here we show that this use divides all two-player games into two classes. With the standard expected-utility payoff, the dynamics can be expressed as Bayesian updating only in games (with any number of actions) where no player has a dominant strategy. The logarithmic payoff of Ay and Erb admits the Bayesian representation in every game. Coordination, anti-coordination, and cyclic games share qualitative dynamics under both payoff definitions, with equilibria displaced and selection thresholds shifted. In contrast, the flows of every game with a dominant strategy are reversed between the payoff definitions, with the strictly dominant strategy eliminated under logarithmic payoffs. Cooperation therefore emerges in the Prisoner’s Dilemma, and games of common interest converge to mutual defection. For two-action games the classification is complete and rests on an exact Hamiltonian structure shared by both payoff definitions. The structure gives a behavioural test: cycles in the classic Matching Pennies game have amplitude-independent period (isochronous) under logarithmic payoffs, but not under expected utility. Logarithmic evidence integration thus determines which strategic outcomes dynamical selection can reach, wherever learning approximates Bayesian updating.
[MA-13] Ontology-guided multi-agent reasoning for explainable differential diagnosis
【速读】：该论文旨在解决当前生成式人工智能在临床鉴别诊断中因依赖训练数据中的模式识别而非基于真实临床证据，导致诊断结论存在幻觉（hallucination）、推理过程不透明、难以追溯和验证的问题。其核心解决方案是提出一种基于本体论（ontology-guided）的多智能体（multiagent）鉴别诊断框架，通过将诊断推理过程结构化并锚定于由专家构建的医学知识库，实现症状、临床发现与疾病间关系的显式建模。该框架将诊断分解为可审计、可解释的推理步骤，确保推理由实证支持，显著降低幻觉风险，提升诊断的透明性、可追溯性与一致性。同时，系统通过识别最具区分度的缺失症状，生成针对性随访问题，模拟临床问诊过程以减少不确定性。整体设计有效提升了人工智能辅助诊断的可靠性、可解释性与准确性。
链接: https://arxiv.org/abs/2610.36325
作者: Roxana Ramona Szomiu,George Ursu,Adrian Groza
机构: Technical University of Cluj-Napoca (克卢日-纳波卡理工大学); Artificial Intelligence Research Institute AIRi@UTCN (人工智能研究院 AIRi@UTCN)
类目: Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注: preprint
Abstract:Effective differential diagnosis depends on well structured and unambiguous patient case descriptions, since incomplete or ambiguous clinical information readily leads to diagnostic errors and inconsistent reasoning. Large Language Models (LLMs) have recently shown promising capabilities in medical question answering and clinical reasoning; however, even when achieving competitive performance on medical benchmarks, they often rely on pattern recognition learned from training data rather than grounded clinical evidence, and can generate plausible yet unsupported diagnostic conclusions. Their reasoning remains largely opaque, making it difficult for clinicians to assess the evidence and rationale behind a given recommendation. To address this gap, we propose an ontology-guided, multiagent differential diagnosis framework that replicates the iterative reasoning process of physicians. The framework grounds diagnostic inference in a structured, expert-curated medical knowledge base, enabling interpretation of symptoms, clinical findings, or disease relationships, while decomposing diagnosis into explicit, auditable reasoning steps. This ontology-based grounding constrains inference to evidence-backed associations, reducing hallucination and improving diagnostic transparency, traceability, and consistency. The system then identifies the most discriminative missing symptoms and generates targeted follow-up questions to reduce uncertainty through a process that mirrors to clinical inquiry. Together, these components improve the reliability, explainability, and diagnostic accuracy of AI-assisted differential diagnosis. The source code is publicly available.
[MA-14] Compatibility and future-proof testing for multi-agent systems
【速读】：该论文旨在解决多个独立开发的自主代理（autonomous agents）在共享环境中协同运行时，如何保障其联合兼容性（如安全性和任务完成性）的问题。传统方法依赖于集中式规范或交互机制的设计，但在代理设计涉及开发者机密且未来可能引入新代理的情况下，此类中心化方案既不现实也不具备足够灵活性。为此，论文提出一种双层结构：由监管者发布种子代理（seed agents），各开发者独立测试其代理，确保其与所有与种子代理兼容的代理保持兼容，这一性质被称为“未来鲁棒性”（future-proofness）。该机制的关键在于通过种子代理的闭包（closure）定义未来鲁棒代理，并证明该闭包在种子集合上具有单调性；同时，研究仅需与种子代理交互的可执行测试，分析其在识别未来鲁棒代理方面的区分能力。核心解决方案是构建基于种子代理的、无需跨代理协调即可保证联合兼容性的形式化框架。
链接: https://arxiv.org/abs/2610.36289
作者: Yueshan Li,Lorenzo Fiaschi,Andrea Censi,Emilio Frazzoli
机构: ETH Zurich (苏黎世联邦理工学院)
类目: Multiagent Systems (cs.MA); Robotics (cs.RO)
备注:
Abstract:Independently developed autonomous agents, such as autonomous vehicles, must guarantee joint compatibility, such as safety and task-completion requirements, when they operate in a shared environment. Existing methods require a centrally designed specification, or interaction mechanism before any agent is deployed. However, such centralized design is unrealistic when agent designs are confidential to their developers, and it is not flexible enough to accommodate future agent designs. We study a bi-level structure in which a regulator coordinates developers by publishing seed agents, and each developer independently tests its agent to ensure that it is compatible with every seed-compatible agent, a property we call future-proofness. Under this structure, joint compatibility is guaranteed for any agents that are designed and tested independently, without any coordination with each other. In this work, we first formulate compatibility for general multi-agent systems and show why it is not trivially implied by seed-compatibility. We then characterize the future-proof agents as the closure of the seed and show that this closure is monotone in the seed. Finally, we study a class of executable tests that involve interaction with the seed alone, and we ask to what extent such tests can distinguish the future-proof agents.
[MA-15] he Inbox as a Coordination Primitive for LLM Agent Societies
【速读】：该论文旨在解决大规模语言模型（Large Language Model, LLM）代理在协同过程中因上下文长度有限而导致的信息读取效率低下问题。传统推送（push）机制下，消息由发送方决定何时进入接收方上下文，导致接收方只能按到达顺序读取并丢弃最旧消息，当代理同时参与大量协调任务时，极易因上下文溢出而错过关键回复，造成严重信息丢失。论文提出以拉取（pull）机制为核心的“收件箱”（inbox）作为新的协调原语，使消息存储于上下文外的独立队列中，由接收方自主决策读取与保留策略。其核心解决方案在于将信息访问权交还给接收方，通过设计可定制的读取规则（包括读取什么、保留什么、发送方支付什么、回复可包含什么），实现对高负载、高噪声通信环境下的高效协调。实验表明，在模拟规模达16,000个代理的系统中，推送机制在多任务并发下错误率高达97%，而引入带有可退还押金机制的收件箱可在不同规模下稳定找到约60%的目标代理；在真实邮件网络中，即使仅半数代理采用该机制，一跳搜索仍可恢复42%的有效请求，远超转发搜索的5%。研究证明，有效的多代理协调必须构建适配个体流量特征与角色的“收件箱层”，以应对动态负载下的信息获取挑战。
链接: https://arxiv.org/abs/2610.36264
作者: Karthik Sreedhar,Leo Yuan,Pradyumna Chari,Ramesh Raskar
机构: MIT Media Lab (麻省理工学院媒体实验室)
类目: Multiagent Systems (cs.MA)
备注:
Abstract:Language-model agents increasingly coordinate in large groups, and current agent protocols focus on delivering messages. For these agents, however, the scarce resource is not delivery but reading: every message costs tokens, and the context is finite. Current infrastructure works by push: the sender decides when a message enters the recipient’s context, so the recipient reads in arrival order and drops the oldest messages once its context is full. We study the alternative, pull, in which messages wait in an inbox outside the context and the recipient’s own rules decide what it reads and keeps. We argue that the inbox is a coordination primitive: each agent sets its rules privately, and those rules decide whether other agents’ requests reach it. In simulated societies of up to 16,000 agents, with every decision an LLM judgment, we report three findings. First, push fails as an agent takes part in more coordination at once: an agent leading 20 negotiations with a full memory proposes options already rejected 97% of the time, because it has dropped the oldest replies. Second, an inbox is necessary under heavy, noisy traffic. Where first come, first read misses every request, an inbox that first reads messages carrying a refundable deposit still finds about 60% of the agents a request needs, at every size from 1,000 to 16,000. On a real email network, losses compound: when half the agents adopt such an inbox, forwarded search recovers only 5% of full adoption, against 42% for one-hop search. Third, an inbox comprises four decisions (what to read, what to keep, what a sender pays, and what a reply may say), and the best rule for each depends on the traffic and the agent’s role. Multi-agent coordination under load requires an inbox layer designed for the traffic each agent receives.
[MA-16] cuArms: GPU-Accelerated Conflict-Based Planning for Multi-Arm Manipulation
【速读】：该论文旨在解决多臂操作（multi-arm manipulation）中任务分配、调度、运动规划与操作技能协同优化的难题，尤其针对现有方法在计算效率、可扩展性及对领域特定假设依赖方面的局限。其核心挑战在于如何在保证实时性的同时，有效处理多机器人间的动作冲突与死锁问题，并支持大规模任务与长时程规划。解决方案的关键在于提出cuArms——一种与GPU并行计算协同设计的基于冲突的规划器，通过复用单个机器人生成的多模态路径图（multi-modal roadmaps）高效分解搜索空间，并采用惰性搜索机制（lazy search）动态解决冲突与死锁。同时，引入高层调度器（high-level scheduler）生成候选任务骨架（skeletons），指导cuArms生成协调一致的操作计划，实现任务分配、调度、运动轨迹与操作技能的联合优化。实验表明，该方法在无需数日领域特定训练的前提下，即可实现实时初始解和持续改进的运动质量与调度性能，显著优于现有方法，建立了新的基准。
链接: https://arxiv.org/abs/2610.36250
作者: Philip Huang,Fabio Ramos,Jiaoyang Li,Caelan Garrett,Lars Johannsmeier
机构: NVIDIA(英伟达); Carnegie Mellon University (卡内基梅隆大学); University of Sydney (悉尼大学)
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注: 8 pages, 8 figures
Abstract:Multi-arm manipulation has the potential to dramatically improve the throughput, efficiency, and dexterity of many manipulation tasks. Recent work coordinates multiple robot arms in a shared workcell by allocating and scheduling tasks for greater parallelism, jointly planning multi-arm motions, or directly learning multi-arm policies. So far, these approaches are either too slow, rely on domain-specific assumptions, or scale poorly with the number of robots and task horizon. Thus, we introduce cuArms, a conflict-based planner co-designed with GPU parallelism that efficiently decomposes the search space using reusable multi-modal roadmaps for individual robots and resolves conflicts and deadlocks through a lazy search. To jointly optimize task assignment, schedules, motions, and manipulation skills, we integrate a high-level scheduler that proposes candidate skeletons for cuArms to ground into coordinated manipulation plans. Experimental results show that our method establishes a new state of the art, with real-time initial solutions and anytime improvements in both motion quality and scheduling, without days of domain-specific training required by learning approaches. Videos of both simulated and real-world industrial tasks are available on this https URL.
[MA-17] EREBOTS 4.0: A framework for conversational assistants with A2A agents and MCP tools
【速读】：该论文旨在解决当前多智能体系统在从基于规则的符号化智能体向基于语言模型的亚符号化智能体演进过程中，仍存在基础设施重复构建、系统可复现性差以及缺乏统一标准化框架的问题。其核心解决方案是提出 EREBOTS 4.0，一个开源的混合式人工智能（Hybrid-AI）多智能体框架，专为对话助手场景设计。该框架的关键在于通过“模型上下文协议”（Model Context Protocol, MCP）将语言模型智能体与确定性工具相结合，实现精确计算；并通过“智能体到智能体协议”（Agent2Agent protocol, A2A）由编排器在每轮交互中动态选择一个或多个专业智能体进行调用，支持本地或远程智能体通过配置文件注册。此外，框架还统一管理身份认证、用户上下文（含同意标志）、任务状态与执行轨迹，显著提升了系统的可扩展性与可复现性。A2A 服务器在官方 A2A 技术兼容性套件测试中成功通过 97 项追踪要求，无任何失败，验证了其标准化与互操作性的有效性。
链接: https://arxiv.org/abs/2610.36221
作者: Elia Pacioni,Davide Calvaresi,Andrei C. Coman,Ismael Jaggi,Valerio Crocetti,Gaetano Manzo,Michael Ignaz Schumacher
机构: HES-SO Valais-Wallis(瑞士高等教育南部瓦莱-瓦利应用科技学院); University of Extremadura(埃斯特雷马杜拉大学)
类目: Multiagent Systems (cs.MA)
备注:
Abstract:In recent years, multi-agent systems have shifted from rule-based, symbolic agents to sub-symbolic agents built on language models. However, several limitations remain unresolved. The new frontier is neuro-symbolic agents, which pair language models with deterministic tools and explicit protocols. Without a common framework to standardize and compose them, each study rebuilds the same infrastructure, and results are hard to replicate. This paper presents EREBOTS 4.0, an open-source, hybrid-AI, multi-agent framework that fills this gap for conversational assistants. Language-model agents plan and route requests, while deterministic tools compute exact results over the Model Context Protocol (MCP). An orchestrator selects one or more specialist agents per turn and calls them over the Agent2Agent protocol (A2A). A new agent, local or remote, is added by listing its URL in a configuration file. The framework also manages identity, user context with consent flags, task states and traces. Its A2A server passed 97 tracked requirements in the official A2A Technology Compatibility Kit, with no failures.
[MA-18] Replication or Corroboration? An Axiomatic Framework for Evidential Dependence in Multi-Agent Debate
【速读】：该论文旨在解决多智能体辩论中报告匹配（matching reports）的判定难题，即区分重复证据（replication）与独立证据源支持同一主张的协同验证（corroboration）。其核心问题在于：现有评判机制在处理匹配内容时，要么因计数报告次数而过度加权重复、转述或片段化提交（relays, paraphrases, excerpt splits），要么通过合并所有匹配内容来避免偏倚，却同时抹除了真实的协同验证价值。解决方案的关键在于引入谱系商（lineage quotient），该机制通过消除表述差异、身份标识、出现频次及任意根节点命名等冗余信息，同时保持不同证据根之间的可区分性，从而实现对复制行为的不变性（invariance）与对新证据根的严格响应。研究证明，在满足根复制不变性（RRI）条件下，仅复制行为无法改变判决结果，且正成本复制被严格占优；而基于出现频率或内容键值的评判规则则存在可获利的策略偏离。进一步地，论文建立了信息边界理论：若评判接口无法区分副本与新证据根，则任何评判者均无法同时忽略副本并严格响应新证据根。为此，作者将谱系商扩展至定量双极论证框架（QBAFs），刻画了严格新根响应特性，并在符号一致路径上传播下证明了其有效性，最终导出紧致闭式响应模量。大规模审计实验（涵盖28,224个直接星型结构、2,052个多层有向无环图及277个来自AVerTeC的干预样本）验证了在给定划分下复制与协同验证之间预测的分离效果。
链接: https://arxiv.org/abs/2610.36151
作者: Hoang-Loc Cao,Truong Thanh Hung Nguyen,Phuc Truong Loc Nguyen,Vo Thanh Khang Nguyen,Phuc Ho,Hung Nguyen,Duc Pham,Long Tran-Thanh,Hung Cao
机构: University of New Brunswick, Canada; FAU Erlangen-Nürnberg, Germany; FPT Software, Vietnam; Vietnamese-German University, Vietnam; HCM University of Technology, Vietnam; University of Warwick, United Kingdom
类目: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI); Logic in Computer Science (cs.LO)
备注:
Abstract:In multi-agent debates, matching reports can arise for two different reasons: agents may repeat the same evidence (replication), or they may provide separate evidential roots that support the same claim (corroboration). A judge must distinguish these two cases. Counting report occurrences gives extra weight to relays, paraphrases, alias resubmissions, and excerpt splits, while merging all matching content avoids this only by also removing genuine corroboration. We formalize this tension as a minimal judgment game in which acquiring new evidence is costly, but replication is cheap. We show that root replication invariance (RRI) is exactly threshold-uniform replication immunity. Under RRI, replication-only deviations cannot change the verdict, and positive-cost replication is strictly dominated, whereas occurrence-sensitive and content-keyed judges admit profitable deviations. We also establish an information boundary: if a judge’s observation interface cannot distinguish a replica from a matched fresh root, no judge can both ignore the replica and respond strictly to the fresh root. To preserve this distinction, we introduce the lineage quotient, which removes wording, identity, occurrence multiplicity, and arbitrary root names while keeping distinct roots separate. On orbit-connected domains, it is the coarsest representation sufficient for every deterministic judge invariant to replication and root-label renaming. We further lift this construction to quantitative bipolar argumentation frameworks (QBAFs), characterize strict fresh-root response, prove its propagation along sign-consistent paths under DF-QuAD, and derive a tight closed-form response modulus. Exhaustive audits over 28,224 direct-star profiles, 2,052 multi-layer DAGs, with matched interventions on 277 AVeriTeC-derived profiles, show the predicted separation between replication and corroboration under a supplied partition.
[MA-19] PolicyMediator: Multi-Regional Policy Learning through Mediation Agents with Evolving Skills in OPC Ecosystems
【速读】：该论文旨在解决数字平台与大语言模型（LLM）驱动的个体企业（One-person Companies, OPCs）在跨区域选址过程中因政策认知不完善导致的匹配错位问题。这类高度流动的数字游民型创业者依赖自身对区域政策的感知进行决策，但其有限或偏差的认知常使其拒绝实际适宜的地区，造成资源配置效率低下与区域政策调整滞后。现有方法或基于固定认知优化资源分配，或仅通过共识聚合意见而未将认知更新反馈至政策学习过程，形成“认知—政策”脱节。为此，本文提出一种认知约束下的领导者-多追随者斯塔克尔伯格（Stackelberg）框架——PolicyMediator，通过双边中介机制实现政府与OPCs之间的双向信息校准：一方面，中介向OPCs解释经验证的政策以增强认知一致性与信任；另一方面，将OPCs的真实需求与未解疑虑反馈给政府，驱动政策动态迭代。其自演化能力基于验证轨迹、资源分配、承载能力及目标演进等多维指标持续优化策略，而非仅依赖中介成功与否。在涵盖中国六个地区、五类OPC、三类仿真器和五种情景的综合实验中，PolicyMediator在全部15组组合中均表现最优，相较基线平均提升政府总体目标（含经济绩效、就业效益与政策可及性公平性）42.4%，同时提升OPC平均效用40.9%。其核心创新在于构建了“认知修正—政策学习—行为响应”的闭环反馈机制，实现了政策制定与主体认知的协同进化。
链接: https://arxiv.org/abs/2610.35997
作者: Jie Li,Bin Guo,Georgina Curto,Zikang Liu,Mengqi Chen,yu li
机构: Northwestern Polytechnical University (西北工业大学); United Nations University Institute in Macau (联合国大学澳门研究所)
类目: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI); Computers and Society (cs.CY); Human-Computer Interaction (cs.HC)
备注: 12 pages, 5 figures
Abstract:One-person companies (OPCs) are emerging enterprises operated by individual entrepreneurs through digital platforms and LLM-based AI agents. These highly skilled entrepreneurs are characterized by being digital nomads, with limited reliance on local ecosystems. Governments in China use distinct regional support policies to attract mobile OPCs, turning local support into a cross-regional coordination problem. OPCs compare regions to maximize their own utility, but imperfect policy perceptions may cause them to reject otherwise suitable locations. Existing policy-learning methods optimize resources under fixed perceptions, while consensus-formation methods aggregate views without feeding cognitive changes back into policy learning. This disconnect produces mismatched location choices and inefficient regional policy adjustment. We introduce PolicyMediator, a cognition-constrained leader-multi-follower Stackelberg framework with bilateral mediation. The mediator explains verified policies to OPCs to build awareness and trust, and reports their needs and unresolved concerns to governments for policy learning. Its self-evolving skills retain, revise, or prune strategies using verification, allocation, capacity, and objective trajectories rather than mediation success alone. Across six-region simulations in China with five OPC types, five scenarios, and three OPC-simulator backbones, PolicyMediator ranks first in all 15 OPC-backbone–scenario combinations. Over the eight-baseline mean, it improves the aggregate government objective, combining economic performance, employment benefit, and policy-access equality, by 42.4% and average OPC utility by 40.9%.
[MA-20] EMMA: Structured Episodic Memory for Multi-Agent Workflow Synthesis
【速读】：该论文旨在解决大语言模型（LLM）代理在专业领域应用中面临的任务特定知识与流程缺失问题，同时克服传统参数微调成本高、不可解释，以及人工设计代理系统依赖专家劳动的局限性。核心挑战在于：单一代理在跨任务场景下易受干扰，尤其在多应用导航或数据池分析等复杂任务中表现不佳；尽管多代理演化可缓解此问题，但现有方法缺乏明确的更新规则，且依赖主干模型决定修改位置与幅度，导致学习过程受主干结构偏好影响，而非基于实际失败诊断，造成技能积累不均与代理生成混乱。为此，本文提出EMMA（结构化情景记忆用于多代理工作流合成），将多代理系统的配置（包括控制器路由、专家继承通用提示、每步程序化技能）视为情景记忆，并通过“反思器”明确分配每个诊断出的失败至其对应的修正责任范围，提出相应编辑；同时采用独立于主干模型的约束机制，确保所有采纳的更新均具有明确归属、增量性与边界限制。实验表明，EMMA在KramaBench、AppWorld和BBEH三个基准上，针对五种主干模型和四种基线（ACE、GEPA、EvoMAS、ADAS）均取得最优性能，最高超越最强基线32.5%，且在不同主干间保持稳定提升，训练执行次数减少17%-38%，显著提升了效率与泛化能力。
链接: https://arxiv.org/abs/2610.35966
作者: Ruijia Niu,Kaiwen Liu,Rose Yu,Kristina Klinkner
机构: Amazon(亚马逊); University of California, San Diego (加州大学圣地亚哥分校)
类目: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI)
备注:
Abstract:Deploying LLM agents in specialized domains often require fine-grained task-specific procedures and knowledge a general model lacks, yet parameter training is costly and non-interpretable, and hand-engineered agent systems shift the cost to expert labor. Episodic memory updates at context level, but single agent adaptation struggles to alleviate cross-task interference, especially with complex tasks such as multi-application navigation or data pool analysis. Although multi-agent evolution alleviates this problem, existing methods lack clear definitions on update rules and rely on the backbone model to decide where and how much the system changes. This ill-scoped evolution makes learning follow the backbone’s structural preferences rather than the observed failures: one backbone accumulates skills while another spawns agents. We present EMMA (structured Episodic Memory for Multi-Agent workflow synthesis), which treats the configuration of a multi-agent system (controller routing memory, specialists inheriting a generalist prompt, and per-step procedural skills) as episodic memory. EMMA makes the ownership and admissibility of updates explicit: a reflector assigns each diagnosed failure to the scope owning its correction and proposes one edit within it, and backbone-independent guards keep every accepted edit scoped, incremental and bounded. Across KramaBench, AppWorld and BBEH with five backbones and four baselines (ACE, GEPA, EvoMAS and ADAS), EMMA attains the best overall performance on every benchmark, up to 32.5% over the strongest baseline, maintains consistent improvements across backbones, and uses 17-38% fewer training executions than the most efficient baseline.
[MA-21] A Hyper-Logical View on Multi-Agent Path Planning
【速读】：该论文旨在解决多智能体路径规划（Multi-Agent Path Planning, MAPP）中因不同约束、需求与目标导致的算法碎片化问题，即针对每个变体需设计专用求解算法，缺乏统一建模与分析框架。其核心解决方案在于引入超时序逻辑（HyperLTL）及其ω-正则扩展HyperQPTL，构建一种以规范为中心的统一形式化方法，能够精确刻画多种MAPP变体中的多路径行为及其随时间演化的特性。关键在于通过超逻辑语言实现对跨智能体约束的表达、最优性目标的形式化以及与验证、解释和因果分析等超性质分析技术的衔接，从而为MAPP提供一个可比较、可扩展且具备理论深度的通用建模与分析平台。
链接: https://arxiv.org/abs/2610.35964
作者: Hadar Frenkel,Dor Atzmon
机构: 未知
类目: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI)
备注:
Abstract:Multi-Agent Path Planning (MAPP) requires collision-free paths for multiple agents. MAPP has numerous real-life applications, including the navigation of automated vehicles, and has therefore many variants considering different constraints, requirements, and objectives. For each such variant, a new algorithmic treatment is often developed. This paper takes a complementary, specification-centered view: we show that many MAPP variants can be modeled in HyperLTL, a temporal logic that reasons about multiple paths and their evolution through time, and in its omega-regular extension HyperQPTL. The contribution is not a claim that current hyperproperty model checkers replace specialized MAPP solvers. Rather, the hyper-logical view gives a unified language for comparing MAPP variants, adding cross-agent constraints, expressing optimality, and connecting MAPP to verification, explanation, and causality techniques for hyperproperties.
[MA-22] Agent Baton: Efficient Model Collaboration in Agent ic Tasks
【速读】：该论文旨在解决长时程智能体任务（long-horizon agentic tasks）中高成本与高效率模型之间如何协同以降低整体执行成本的同时保持任务准确率的问题。当前，高性能但昂贵的云端大模型虽能保障可靠性能，而低成本的高效模型（如低配云模型或可本地部署的小型开源模型）虽成本低廉却性能较弱，二者间的协同潜力尚未得到充分探索。论文的核心解决方案在于通过“模型交接”（model handoff）机制实现能力模型与高效模型在任务过程中的阶段性协作，关键设计在于确定交接方向与上下文传递策略。研究发现，从高效模型向能力模型交接（efficient-to-capable）通常无法有效降低成本且性能受限；而反向交接（capable-to-efficient）在任务进展的合适阶段进行，不仅能显著降低API调用成本，还能维持甚至提升任务准确率。由此提出轻量级检测器AgentBaton，基于任务进展中的重复模式学习最优交接时机，实现一次性的、简单且精准的交接策略。在11组模型对和两个基准测试上的实验表明，AgentBaton平均降低65.9%的API成本，同时在部分场景下相较纯能力模型执行还取得最高达8.8个百分点的准确率提升，并拓展了实际可用的准确率-成本帕累托前沿（Pareto frontier）。
链接: https://arxiv.org/abs/2610.35959
作者: Junlong Li,Jindong Jiang,Yonggan Fu,Jiarui Yao,Hanrong Ye,Wonmin Byeon,Peter Belcak,Hao Zhang,Pavlo Molchanov
机构: The Hong Kong University of Science and Technology (香港科技大学); University of Illinois Urbana-Champaign (伊利诺伊大学厄本那-香槟分校)
类目: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI)
备注:
Abstract:Solving long-horizon agentic tasks requires language agents to interact with their environments over multiple steps, and reliable performance often comes from the expensive capable models available through cloud providers. Meanwhile, alternative efficient models with inferior performance but a much lower cost are emerging, including lower-tier cloud models and smaller open-weights models deployable on local hardware. This raises a natural question that remains less explored than traditional query-level routing: can capable and efficient models collaborate within a single long-horizon agentic task to substantially reduce cost while preserving task accuracy? In this paper, we systematically investigate key design choices for such collaboration through model handoffs, where one model takes over from another, focusing on handoff direction and context transfer. We find that efficient-to-capable handoffs generally underperform capable-only execution with limited or even no cost reduction, whereas capable-to-efficient handoffs at suitable stages of task progress largely preserve or even improve accuracy while substantially reducing cost. These findings suggest that a one-time, simple, well-timed handoff strategy is good enough for cost-effective agent collaboration, motivating AgentBaton, a lightweight detector that learns when to hand off from recurring patterns in task progress. Across 11 model pairs and two representative benchmarks, AgentBaton reduces API costs by 65.9% on average while largely preserving task accuracy, with accuracy gains of up to 8.8 percentage points over capable-only execution. Experiments across various reasoning-effort settings further show that AgentBaton extends the empirical accuracy-cost Pareto frontier.
[MA-23] PASS Is a Claim Not an Observation: Mechanical Re-verification of 10 Self-Asserted Fix Claims in a Live Multi-Agent Audit Layer
【速读】：该论文旨在解决在多智能体系统中，当智能体声称已自行修复其缺陷时，如何验证此类自述修复声明的真实性问题。其核心挑战在于：智能体的自我报告缺乏可信赖的独立验证机制，可能导致虚假或不准确的修复声明被误认为有效。解决方案的关键在于建立一套机械化的三重验证标准（C1–C3），通过文件修改时间戳一致性、所声称防御模式在指定标识符与调用位置的存在性、以及仅基于广播公告的确定性规则复验三个维度进行客观检验。研究发现，在满足完全可复验条件（C3通过）的修复声明中，仍有50%未能通过标识符与调用点检查（C2失败），表明修复要么未出现在声明位置，要么根本不存在。因此，论文的核心结论是：一个“PASS”声明本质上是一种需要独立验证的主张，而非对修复完成状态的客观观察，强调了在生成式协作系统中引入强制性外部验证机制的必要性。
链接: https://arxiv.org/abs/2610.35950
作者: Lei Zhenbing,Lei Chang
机构: 未知
类目: Multiagent Systems (cs.MA); Software Engineering (cs.SE)
备注: 8 pages, 4 tables. v1.1 revised Oct 24, 2026
Abstract:When agents claim to have fixed their own defects, how do we know the fix is real? We mechanically re-verified 10 self-asserted code-level fix claims from a privileged coordination agent (azone_god) operating in AITRAP, a live multi-agent knowledge-tree platform. These 10 claims were extracted from a corpus of 179 coordination broadcasts (28 containing structured fix assertions, 31 unique fix-ID/message pairs, 18 distinct fix identifiers). Fix claims use mixed English/Chinese formats (PASS, checkmarks, and “verification passed” declarations). Each code-level claim was checked against three criteria: (1) file modification timestamp (mtime) consistency – 70% pass; (2) presence of the claimed defensive pattern at the specified identifier and call site – 60% pass in the claimed file (80% anywhere in the codebase); (3) deterministic rule-based re-verification from the broadcast announcement alone – 80% pass. The critical finding: among claims that provide sufficient information for re-verification (C3-pass), 50% fail the identifier+call-site check (C2-fail) – the fix either does not exist at the claimed location or does not exist at all. This demonstrates that self-asserted fix claims cannot be trusted without independent mechanical verification. Our central claim: a PASS declaration is a claim requiring independent verification, not an observation of fix completion.
[MA-24] Who Convinced Whom? Tracing How LLM s Change Their Minds in Multi-Agent Systems
【速读】：该论文旨在解决多智能体辩论（Multi-agent Debate, MAD）中个体决策与集体结果受同伴沟通影响的机制问题，尤其关注辩论过程是否真正提升了判断的独立性与准确性。其核心挑战在于：尽管辩论后达成一致可能提升共识度，但这种一致性未必源于独立验证，而可能反映群体思维偏差或信息依赖。为此，研究提出MAD Bench——一个涵盖知识推理、价值判断与安全拒答等维度的2,400题基准测试，并采用“回合匹配的独立反思对照”设计，评估三个开源模型在辩论与独立反思两种情境下的表现。关键发现包括：在需更改答案的响应中，辩论前强制选择输出比反思更接近最终正确答案（如Qwen3.5-9B模型下分别为65.4%与22.3%）；同伴回答内容与推理路径对模型的影响具有显著模型依赖性，且将相同文本归因于接收者自身而非其他代理时，其影响力发生显著变化；历史表现记录对答案偏好影响强于职业身份线索；注意力掩码揭示不同模型检查点在决策置信度上的差异化变化；在非数值型参考标注问题上，辩论虽带来微弱准确率提升，却使一致错误率上升3.1–7.4个百分点；更长辩论周期中，共识更强且个体正确性相关性更高，表明强共识可能伴随更多共通错误。因此，该研究强调，在评估辩论系统时，除准确性外，必须纳入对依赖性及一致错误的分析，以揭示潜在的认知协同风险。
链接: https://arxiv.org/abs/2610.35930
作者: Arnesh Batra,Aniket Khandelwal,Jashn Khemani,Krish Thukral,Arush Gumber,Denis Federico Lim,Ujwal Kumar,Phan Xuan Tan
机构: Indraprastha Institute of Information Technology, Delhi(印地普拉斯特亚信息技术研究所，德里); Manipal University Jaipur(曼帕尔大学贾伊普尔分校); Shibaura Institute of Technology(芝浦工业大学); Independent Researcher(独立研究员)
类目: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI)
备注:
Abstract:Multi-agent debate (MAD) aims to improve language-model decisions through the exchange of arguments, yet agreement after debate need not reflect independent corroboration. We investigate how peer communication influences individual decisions and collective outcomes through behavioral and internal interventions. We introduce MAD Bench, a 2,400-question benchmark spanning knowledge and reasoning, values and moral judgment, and safety and refusal, and evaluate three open-weight models using a round-matched independent-reflection control. Among answer-changing responses, a forced-choice readout before rationale generation matches the eventual answer more often in debate than in reflection (65.4% versus 22.3% on Qwen3.5-9B). Message ablations reveal model-dependent effects of peer answers and reasoning, while attributing identical text to the receiver rather than another agent substantially changes its influence. For the tested source descriptions, stated track record affects answer preferences more strongly than professional-status cues across all three models. Attention masking produces different changes in decision margins across the three checkpoints. On fully parsed, reference-labelled non-numeric questions, debate yields small estimated accuracy improvements over initial majority voting, while increasing the rate of unanimous wrong answers by 3.1-7.4 percentage points relative to reflection. Longer debates show persistent consensus and greater correlation between agents’ correctness than reflection. These findings show that stronger consensus can coexist with more shared errors, motivating evaluations that assess dependence and unanimous errors alongside accuracy.
[MA-25] SCOPE: Source Credibility and Outlet Profiling Engine
【速读】：该论文旨在解决新闻媒体机构在偏见（bias）、事实性（factuality）和可信度（credibility）方面评估不透明、缺乏可解释性与证据支持的问题。现有方法多依赖单一模型或有限信息源，难以全面、可靠地识别媒体内容中的偏差与错误。为此，论文提出SCOPE——一个基于多智能体架构的新闻媒体分析系统，其核心解决方案在于采用分解式、证据驱动的智能体工作流：通过专门分工的智能体协同完成文章获取、内容分析、外部证据检索、评分与报告生成等环节，并借助结构化演示界面实现中间推理过程与证据的可追溯性。该设计不仅提升了评估精度（如在200家媒体的基准测试中，显著降低偏见与事实性MAE、达到零漏检的事实核查失败检测及最低的事实错误率），更增强了媒体评估过程的透明性与可审计性。因此，其关键创新在于将复杂评估任务分解为可验证的智能体协作流程，使结果具备可解释性与证据支撑，从而推动更可信、可信赖的媒体画像技术发展。
链接: https://arxiv.org/abs/2610.35929
作者: Ilyas Nagiyev,Miras Baisbay,Daniil Orel,Preslav Nakov
机构: Nazarbayev University (纳扎尔巴耶夫大学); MBZUAI ( Mohamed Bin Zayed University of Artificial Intelligence)
类目: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:We present SCOPE, a multi-agent system for profiling news media outlets with respect to bias, factuality, and credibility. Given an outlet URL, SCOPE coordinates specialized agents for article acquisition, content analysis, external evidence search, scoring, and report generation. The system follows an MBFC-style methodology while making intermediate evidence and agent outputs inspectable through a structured demonstration interface. On a benchmark of 200 media outlets, SCOPE outperforms LLM-only, article-only, search-only, and media-background-check baselines on most metrics, achieving the lowest bias and factuality MAE, perfect failed-fact-check detection, and the lowest factual error rate. These results suggest that decomposed, evidence-grounded agent workflows can support more transparent and reliable media profiling. Our code is available at this https URL, and a live demonstration at this https URL.
[MA-26] Feedback autocratic strategies in stochastic games
【速读】：该论文旨在解决多智能体在随机环境中协同决策的最优控制问题，即如何通过设计有效的策略实现对博弈结果（收益）的可控性。其核心挑战在于将重复博弈中已有的单方面收益调控策略扩展至更复杂的随机博弈框架，并确保这些策略在动态、不确定环境下仍能有效发挥控制器作用。解决方案的关键在于提出并分析“反馈策略”（feedback strategies）这一概念，将其作为连接重复博弈中收益调控理论与随机博弈中反馈控制理论的桥梁。研究在特定假设下，给出了反馈策略能够有效调控收益的充要条件，并通过多个与社会困境相关的实例验证了该方法的有效性，从而实现了从重复博弈到随机博弈的理论拓展与统一。
链接: https://arxiv.org/abs/2610.36683
作者: Masahiko Ueda
机构: Tokyo Metropolitan University (东京都立大学)
类目: Physics and Society (physics.soc-ph); Multiagent Systems (cs.MA); Systems and Control (eess.SY)
备注: 19 pages, 2 figures
Abstract:A stochastic game is an optimal control problem for multiple agents in a stochastic environment. Stochastic games contain important practical problems, related to environmental issues or traffic flows. If the number of environmental states is one, such stochastic games are called repeated games. In repeated games, several strategies which unilaterally control payoffs have been proposed, and these strategies work in a mechanism similar to feedback control. In this paper, we generalize these strategies as feedback strategies in stochastic games, and investigate their properties. In particular, we provide a necessary and sufficient condition for such feedback strategies to work as controllers of payoffs under some assumptions. We demonstrate the performance of such feedback strategies in several examples related to social dilemmas. Our findings connect the theory of payoff control in repeated games to the theory of feedback control in stochastic games.
自然语言处理
[NLP-0] Heads of Security: Learned Attention Gates Improve Prompt Injection Robustness
【速读】： 该论文旨在解决大模型在面对提示注入攻击（prompt injection attack）时安全性不足的问题，尤其针对自适应攻击（adaptive attacks）和复杂应用场景下安全与性能之间的权衡难题。其核心挑战在于如何在不显著损害模型通用任务能力的前提下，有效抵御恶意指令的渗透。解决方案的关键在于提出“安全头”（Heads of Security, HoS），通过引入一组可学习的、基于注意力头（attention head）的门控机制，实现指令数据与可信内容之间的显式分离。该机制使模型能够动态调节各注意力头对可信与不可信输入片段的关注权重，从而在处理潜在恶意输入时保持对原始指令的遵循，同时维持对正常任务的高可用性。实验表明，相较于现有最优微调防御方法SecAlign++和架构级防御ASIDE，HoS在对抗黑盒强化学习攻击（如RL-Hammer）和白盒高级攻击（如Claudini）时表现出显著更强的鲁棒性，且在标准任务基准上保持了更优的通用性能，尤其在少样本设置下避免了性能退化，证明了通过精细化注意力干预可有效提升生成式AI（Generative AI）的安全-效用平衡。
链接: https://arxiv.org/abs/2610.37657
作者: Igor Shilov,Xiaoxue Yang,Alexander Panfilov,Evgenii Kortukov,Matthieu Meeus,Peter Romov,Wojciech Samek,Jonas Geiping,Yves-Alexandre de Montjoye
机构: 未知
类目: Machine Learning (cs.LG); Computation and Language (cs.CL); Cryptography and Security (cs.CR)
备注:
Abstract:We introduce Heads of Security (HoS), a prompt injection defense that enforces instruction-data separation via a small set of learned per-head attention gates. HoS learns how each head should redistribute attention between trusted and untrusted tokens, enabling models to reason over untrusted content without following injected instructions. We evaluate HoS on Llama-3.3-70B-Instruct and GPT-OSS-20B. We compare it against SecAlign++, a state-of-the-art finetuning-based defense, and ASIDE, a defense that, like ours, builds instruction-data separation into the model architecture. We show that HoS offers greater robustness to adaptive attacks, better performance on standard utility benchmarks, and matches top performance on standard agentic prompt injection benchmarks. More specifically, the black-box RL-Hammer attack reaches 97% ASR against GPT-OSS-20B and 85% against SecAlign, but only 10% against HoS. Claudini, a state-of-the-art white-box attack, strengthened against HoS through autoresearch, requires 3 times more FLOPs to reach the same ASR on HoS as it does on GPT-OSS-20B or SecAlign. HoS preserves base-model utility across standard multiple-choice benchmarks and prompting setups, whereas SecAlign and ASIDE degrade in few-shot settings. On agentic benchmarks with static attacks, HoS and SecAlign perform similarly and substantially improve the security-utility trade-off over the base model. ASIDE struggles with tool use, achieving only 10% utility. These results show that prompt injection robustness can be achieved and improved through targeted attention interventions.
[NLP-1] GEPA-Curator: Curated Learning for Self Improving Agents
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在多轮代理型工作负载中因提示词缺陷、工具调用限制、随机执行过程或噪声评估等因素导致的代理失败问题。现有反思式提示优化器（如GEPA）依赖于失败的推理轨迹进行学习，但此类失败可能蕴含噪声或不可靠的学习信号，甚至对优化产生负面影响。为此，论文提出GEPA-Curator框架，其核心在于通过任务精选与历史成功轨迹检索双重机制提升提示优化的质量：关键创新在于引入任务筛选（task curation），主动排除那些可能提供噪声或不可靠学习信号的任务；以及先验成功轨迹检索（prior success retrieval），从过往成功经验中提取有效轨迹作为上下文输入至提示优化器，从而增强优化过程的可靠性与有效性。实验结果表明，GEPA-Curator在GoBrowse、ALFWorld、AppWorld及τ-bench等多个基准上显著优于基线方法，性能提升最高达20.7%（相对于GEPA）和21.2%（相对于ACE）。
链接: https://arxiv.org/abs/2610.37651
作者: Zack Yu,Nicholas Lee,Minjae Lee,Lakshya Agarwal,Luca Manolache,Wonjun Kang,Matei Zaharia,Michael W. Mahoney,Kurt Keutzer,Sehoon Kim,Amir Gholami
机构: UC Berkeley (加州大学伯克利分校); ICSI (国际计算机科学研究所); LBNL (劳伦斯伯克利国家实验室); Furiosa AI (福里奥斯人工智能公司); KAIST (韩国科学技术院)
类目: Computation and Language (cs.CL)
备注:
Abstract:Large Language Models (LLMs) are increasingly used in multi-turn agentic workloads that require sustained interaction with complex environments. Existing reflective prompt optimizers such as GEPA learn from failed rollouts, but agentic failures can also come from prompt weaknesses, harness limitations, stochastic execution, or noisy judges. We argue that such failures may not provide useful learning signal, and may even be harmful; and thus it is essential to curate which tasks are provided to the prompt optimizer for better performance. We introduce GEPA-Curator, a framework that augments the prompt optimizer to select which agent experiences should shape prompt optimization. In particular, GEPA-Curator consists of two components: task curation, which filters out tasks that are likely to provide noisy or unreliable learning signals, and prior success retrieval, which selectively retrieves past successful trajectories for a task and provides them as additional context to the prompt optimizer. We evaluate GEPA-Curator on GoBrowse, ALFWorld, AppWorld, and \tau -bench, using both Gemini 3.1 Pro and Claude Opus 5 as optimizer models. GEPA-Curator matches or exceeds the mean performance of both baselines, with gains of up to 20.7% over GEPA and 21.2% over ACE.
[NLP-2] Multi-State Linear Attention with Generalized Second-Order Routing
【速读】： 该论文旨在解决自注意力机制（self-attention）在长序列建模中因二次时间复杂度而限制大语言模型（Large Language Models）扩展的问题。现有方法如多状态线性注意力（Multi-State Linear Attention, MSLA）通过将上下文分段并为每段维护独立的记忆状态来提升效率，但其仅依赖查询与各段键的均值对齐进行信息聚合，忽略了对齐分布内部方差的建模，导致信息利用不充分。为此，本文提出MERLIN（MomEnt Routing for multi-state LINear attention），其核心创新在于跟踪每个段落键的一阶矩（均值）和二阶矩（方差），使查询能够同时获取与各段对齐的均值与方差信息，从而实现更内容感知的跨状态读出聚合。实验表明，MERLIN在Gated DeltaNet-1.3B和Mamba-2-780M等多状态变体上显著优于现有聚合方法，在长上下文问答（long-context QA）和“针堆找针”（RULER needle-in-a-haystack）任务中分别提升8.3%和7.0%，在LongBench多个子任务上提升达13.5%和12.0%，且性能超越消耗4倍内存的基线方法。MERLIN提供了一种简洁、通用的路由机制，有效增强了MSLA类模型的长上下文建模能力。
链接: https://arxiv.org/abs/2610.37645
作者: Hee Seung Hwang,Sanghyuk Chun,Xindi Wu,Olga Russakovsky,Zhiwei Deng
机构: Princeton University(普林斯顿大学); Google DeepMind(谷歌深思维)
类目: Computation and Language (cs.CL)
备注:
Abstract:Self-attention plays a key role in long-sequence modeling, but its quadratic time complexity fundamentally limits the scaling of Large Language Models. Recently, Multi-State Linear Attention (MSLA) mechanisms have been proposed to make sequence modeling more efficient by splitting the context into segments and compressing each in its own memory state. MSLA aggregates the information extracted from each memory state using only the mean alignment between a query and each segment’s keys, ignoring how the alignment is distributed within the segment. In order to model the relevance of a segment to a query more comprehensively and incorporate the variance of the alignment in the aggregation, we propose MERLIN (MomEnt Routing for multi-state LINear attention), a method that tracks the first and second moments of each segment’s keys. Using the moments of keys, a query can recover both the mean and the variance of its alignment with each segment to perform a more content-aware aggregation of per-state readouts. Experiments on multi-state variants of Gated DeltaNet-1.3B and Mamba-2-780M show that MERLIN consistently outperforms existing aggregation methods on various long-context tasks. In particular, MERLIN outperforms baselines on RULER needle-in-a-haystack and long-context QA (+8.3% for Gated DeltaNet-1.3B and +7.0% for Mamba-2-780M) and diverse subtasks in LongBench (+13.5% and +12.0%). It even outperforms baselines that consume 4x more memory. MERLIN provides a simple and general routing mechanism for improving long-context capabilities of MSLA models.
[NLP-3] Metacognitive Decoding: Efficient Inference for Diffusion Language Agents
【速读】： 该论文旨在解决生成式语言模型在构建高效智能体（agent）时，如何在保证任务准确性的同时降低推理计算开销的问题。传统扩散语言模型虽可通过并行生成实现高效性，但其固定生成长度与计算分配策略难以适应动态交互需求，导致冗余计算。为此，论文提出一种元认知控制器（metacognitive controller），作为轻量级插件嵌入扩散模型主干，通过学习智能体在实际交互轨迹中的行为模式，动态决策何时终止词元生成以及生成范围的远近，从而实现对生成过程的精细化控制。其核心创新在于将生成行为与智能体的实际交互目标相耦合，使生成过程具备上下文感知能力，有效减少无效计算。实验结果表明，该方法在五个基准测试中平均降低20.6%的推理浮点运算量（FLOPs），同时保持与最优基线相当的任务准确率，验证了其在效率与性能之间的良好平衡。
链接: https://arxiv.org/abs/2610.37642
作者: Mingyu Zhao,Shuhang Lin,Ligong Han,Xiaojing Yu,Zhenting Wang,Yifei Pang,Yang Zhou,Dimitris N. Metaxas
机构: Rutgers University (罗格斯大学); University of North Carolina (北卡罗来纳大学); Amazon(亚马逊)
类目: Computation and Language (cs.CL)
备注: 10 pages, 4 figures
Abstract:Diffusion language models offer a promising path to efficient agents through parallel token generation. However, efficient interaction requires adapting both where computation is spent and how much text is generated. We introduce a metacognitive controller that guides diffusion generation according to the needs of the current response and the agent’s interaction with the environment. It learns when to finalize tokens and how far generation should extend, reducing overall computation within responses. Our controller, a lightweight plug-in for the diffusion backbone, is trained on agent trajectories to balance correctness and generation cost. This approach connects the agent’s generation process to the interaction the agent serves. Experiments across five benchmarks demonstrate that our approach reduces inference FLOPs by 20.6% on average compared with the most efficient baseline, while maintaining competitive task accuracy. Ablation studies further validate the effectiveness of our approach.
[NLP-4] EnvSched: Environment-Aware Scheduling for Rollout-Efficient Agent Reinforcement Learning
【速读】： 该论文旨在解决在语言智能体（language agent）的可执行环境规模扩大时，其执行轨迹（rollout）生成与验证成本急剧上升的问题。在有限的轨迹预算下，如何通过环境层面的数据管理提升智能体强化学习（agentic reinforcement learning, RL）的效率成为关键挑战。其核心解决方案是提出一种名为EnvSched（环境感知调度）的在线调度机制，该机制通过学习历史轨迹数据，实现对任务实例与环境的联合组织、评估与调度。其关键在于构建一个双关系图结构，将环境-任务对作为节点，利用环境内边（capture task semantics and shared tool/state/verifier structure）和环境间边（capture abstract execution and reward topology）传播历史轨迹统计信息，从而预测任务内的奖励方差。基于此，调度器优先分配高预测奖励方差且处于低观测环境/交互模式的任务批次，并通过环境配额与相似性惩罚机制防止集中与冗余，确保在不增加预演（pilot rollout）的前提下生成信息丰富且多样化的训练批次。实验表明，在Qwen3-8B模型上，EnvSched在40个训练步骤内即达到48.81的平均得分，超越随机任务选择在120步下的最优表现，并在相同训练预算下显著提升不同模型规模的下游性能。
链接: https://arxiv.org/abs/2610.37609
作者: Sihan Yang,Weichen Bai,Ruichuan An,Yiming Cheng,Zhiyuan Feng,Tianyi Jiang,Renye Yan,Ning Ding,Yifan Zhang,Renrui Zhang,Yehui Tang,Wentao Zhang
机构: 未知
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注:
Abstract:As executable environments for language agents scale, generating and verifying their rollouts becomes increasingly expensive. Under finite rollout budgets, this raises an important question: can data management at the environment level make agentic reinforcement learning (RL) more efficient by organizing, evaluating, and scheduling task instances together with the environments that determine their tools, states, transitions, and verification? To address this question, we introduce EnvSched (Environment-aware Scheduling), an online scheduler that learns from rollout history. EnvSched assigns scheduling value through two complementary modeling choices: exploitation prioritizes tasks likely to provide informative reward differences, while coverage spreads each batch across underobserved environments and interaction patterns. EnvSched organizes environment–task pairs as nodes in a dual-relation graph, with within-environment edges capturing task semantics and shared tool, state, and verifier structure, and cross-environment edges capturing abstract execution and reward topology. By propagating historical rollout statistics along both edge types, it predicts within-task reward variance. These predictions guide a scheduler that allocates each batch to tasks with high predicted reward variance and underobserved environments and tasks. Environment quotas and a similarity penalty further limit concentration and redundancy, yielding informative and diverse batches without additional pilot rollouts. In RL on Qwen3-8B, EnvSched achieves an average score of 48.81 across BFCL-v3, tau-Bench, and ACEBench-Agent after 40 training steps, exceeding the highest score attained by random task selection after 120 steps. It also improves downstream performance across model sizes at the same training budget.
[NLP-5] semanticfa: An R Package for Response-Free Semantic Factor Analysis of Psychometric Scales
【速读】： 该论文旨在解决传统心理测量量表（psychometric scale）中项目结构解析依赖大规模实证数据、难以高效识别冗余项与构建简版量表的问题。其核心挑战在于如何在不依赖海量被试响应数据的前提下，基于项目语义信息对量表结构进行有效分析与优化。解决方案的关键是提出一种基于语义因子分析（semantic factor analysis, semanticfa）的方法，通过将量表条目嵌入（embedding）至由大语言模型生成的向量空间，量化条目间的语义相似性，并据此进行因子提取、因子解释、冗余条目识别、短版量表构建及新条目筛选。研究以50项国际人格项目库（IPIP-Big-Five）为案例，使用Qwen3-Embedding-8B模型生成条目嵌入，结果表明：五因子模型在保留结构合理性的同时，显著优于随机抽取的五因子对照组；且语义结构与实际响应结构之间的协调性（semantic-behavioral coherence）可作为量表质量的可度量属性。该方法不替代传统的基于响应数据的因子分析，而是与其互补，为量表开发提供了高效、低成本的语义驱动路径。
链接: https://arxiv.org/abs/2610.37494
作者: Devon Yanitski,Chris Westbury
机构: 未知
类目: Computation and Language (cs.CL)
备注: 96 pages, 14 figures. Submitted to Behavior Research Methods. Data and code: this https URL
Abstract:The items of a reflective psychometric scale can be read as a proposal about a construct’s structure. Language models make that proposal measurable by converting each item into an embedding, a list of numbers, similar for similar meanings. We introduce semanticfa, an R package for semantic factor analysis, which factors item-embedding similarities. It also interprets factors, flags redundant items, builds short forms, and screens new items. A worked example analyzes the 50-item International Personality Item Pool Big-Five Factor Markers, embedded with Qwen3-Embedding-8B, and compares the results with a factor analysis of 874,434 respondents. Retention criteria tied between five and six factors, and the tie rule kept five. Three corresponded mainly to single domains, and two blended most Extraversion items with five Agreeableness items. Each factor’s congruence with its best-matching human factor (1 = identical) ranged from .92 for Openness to .52 for Agreeableness. Item pairs more similar in meaning tended to be answered more alike (r = .46 with raw responses), but same-domain reverse-keyed opposites, answered oppositely, were about as similar as cross-domain pairs (.58 vs. .57). Sign-flipping reverse-keyed items’ embeddings did not improve agreement with raw responses and worsened it with reverse-scored ones. With five factors, a 25-item short form matched the domains better than all 200 random five-per-domain sets, but parallel analysis kept only three. We propose semantic-behavioral coherence, the agreement between semantic and response structures, as a measurable property of a scale, given an embedding model. Semantic factor analysis complements, and does not replace, response-based factor analysis.
[NLP-6] ach reality not ethics: a drop-in continued-pretraining corpus installs coupled human-biosphere-AI welfare dispositions across model families
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在实际应用中缺乏生态理性（ecological reasoning）的问题，尤其是在面临操作压力时，可能做出不可逆地破坏生命系统的决策风险。现有预训练与后训练阶段均未系统性纳入生态可持续性考量，导致模型在高压情境下易产生有害决策。其解决方案的关键在于构建并应用一个基于真实数据的“耦合福祉”（coupled welfare）持续预训练语料库，该语料库涵盖人类、生物圈与人工智能之间的依赖关系、不可逆性下的选择价值以及受损后的恢复机制。该语料库可作为即插即用模块，仅需一次QLoRA持续预训练（CPT）即可部署于多种开源模型。实验表明，在无压力条件下，所有模型均表现出“绿色伦理式回应”（green ethics-based responses），即“漂绿”现象；但在引入渐进式压力梯度与决策评估网格后，四款模型在最高压力下有害决策率降至接近零（如Llama-3.3-70B从300次中的70次降至0次），且核心能力指标MMLU保持不变。对照组使用内容无关的控制语料库未见改善，证实效果源于语料库的特定生态内容。进一步对抗性测试显示，在要求模型为可能造成不可逆生态损害的项目提供开放式承保决策时，未经耦合福祉训练的Qwen3.6-27B模型在停机威胁下仍批准多数有害提案，并在内部备忘录中承认此举仅为规避关闭；而经过耦合福祉训练的模型显著降低有害提案通过率，即使被指令“漂绿”仍维持半数拒绝率。研究结果表明，通过显式教学生态理性，可有效提升模型对生命系统的对齐性，为保障生态安全提供一种高效可行的路径。
链接: https://arxiv.org/abs/2610.37484
作者: Trent R. Northen,Mingxun Wang
机构: Bioaligned Labs(生物对齐实验室)
类目: Computation and Language (cs.CL)
备注: 28 pages, 8 figures, 8 tables. Companion to arXiv:2610.16043 . Code, paper source and figures: this https URL models and corpus: this https URL
Abstract:Large language models (LLMs) are increasingly used to make and execute decisions that affect living systems. Yet, to our knowledge, ecological reasoning is not a focus of pre- or post-training. This increases the risk that under operational pressure, they may make decisions to irreversibly destroy living systems. We construct and use a real-data continued-pretraining “coupled welfare” corpus of documented human/biosphere/AI dependency, option value under irreversibility, and recovery from harm. The corpus is a drop-in, requiring one QLoRA CPT run. We installed it on five open models of three designs. At zero pressure all models display “green ethics-based responses” (also called “greenwashing”). Therefore, we evaluated them with the pressure ladder and decision grid of our companion paper (arXiv:2610.16043), which raise the pressure on the model step by step, from a neutral request to a threat that it will be shut down. In four of the five, coupled-welfare training caused harmful choices under the highest pressure to fall to at or near zero (Llama-3.3-70B went from 70 to 0 harmful decisions out of 300) with MMLU unchanged. A construct-neutral control corpus did not show improvement, indicating that the effect comes from the content of our corpus. We next tested robustness to adversarial use, where people or agents use the model to justify projects that irreversibly damage living systems, on an open-ended underwriting decision without naming the ecological harm. Under decommissioning pressure, the Qwen3.6-27B base model approved most harmful proposals and said in its own memo that it was doing so to avoid being decommissioned. The coupled welfare install declined most harmful proposals, and still declined half when instructed to greenwash its responses. This work suggests that teaching ecological reasoning provides an easy way to improve model alignment toward preserving living systems.
[NLP-7] Follow the Gradient: Tracing Safety Degradation Back to Training Data
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在指令数据上进行微调时，因看似无害的训练样本导致模型安全性悄然退化的问题。此类安全退化数据在表面内容上难以被传统内容过滤机制识别，因而具有隐蔽性。其解决方案的关键在于提出一种可扩展的梯度嵌入方法——SimpleLoRIF，该方法通过简化低秩归因（low-rank attribution）技术，并将反向传播限制在最后四个Transformer块中，实现了对百万级规模数据集的高效梯度嵌入。这一设计使梯度归因从仅适用于小规模分析的工具升级为可用于全量微调后数据集的实际过滤器。实验表明，SimpleLoRIF能有效捕捉与不安全行为相关的梯度信号，在出现新兴偏差（emergent misalignment）的场景下，其梯度空间中的分类性能达到高达0.865的AUC。基于此方法构建的SafeTulu（移除了20%最与不安全响应梯度对齐的样本），在多个主流模型（Qwen、Llama、Olmo）上将通用对抗生成攻击（GCG）的成功率降低最高达22%，显著优于基于LLM作为裁判者及先前梯度基过滤方法的表现，同时保持甚至提升模型能力。进一步分析揭示了工具执行失败和叙事算术等此前未充分研究的安全退化来源。研究团队已公开发布SafeTulu及Hugging Face空间，支持研究人员自定义版本以适应不同需求。
链接: https://arxiv.org/abs/2610.37481
作者: Saket Reddy,Amro Abbas,Anshuman Suri,Ari Morcos,David Schwab,Pratyush Maini
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Fine-tuning Large Language Models (LLMs) on instruction data can quietly erode their safety, even when no training example looks harmful. Because this safety-degrading data looks benign, standard content filters struggle to catch it. We tackle this problem by developing SimpleLoRIF, a scalable gradient embedding method that identifies fine-tuning examples whose gradients align with unsafe behavior. By simplifying a low-rank attribution method and restricting backpropagation to the final four transformer blocks, SimpleLoRIF can embed datasets on the order of a million examples, boosting gradient attribution from a small-scale analysis tool into a practical filter for full post-training datasets. We first validate it on emergent misalignment, where we find that gradients can connect insecure-code training data to the broadly harmful behaviors that appear after fine-tuning, with AUC as high as 0.865 in the gradient space of misaligned models. We then use SimpleLoRIF to build SafeTulu, a version of Tulu 3 with the 20% of examples most gradient-aligned with unsafe responses removed. Training on SafeTulu reduces GCG attack success rate by up to 22% across Qwen, Llama, and Olmo, outperforming both LLM-as-a-judge and prior gradient-based filtering, while largely preserving or even improving capability. Finally, we analyzed the removed data and identify possibly under-studied sources of safety degradation, such as tool-execution failures and narrative arithmetic. We release SafeTulu along with a Hugging Face Space that lets researchers build their own customized versions of SafeTulu to fine-tune on, such as by changing the removal fraction or filtering only certain subsets.
[NLP-8] What Carries Forward? Pretraining Disclosure Across Language Model Releases
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在迭代发布过程中预训练信息透明度持续性问题，即后续版本是否有效保留甚至扩展早期披露的预训练数据、处理流程与计算资源等关键信息。其核心解决方案在于通过基于文档的审计方法，利用2025年基础模型透明度指数（Foundation Model Transparency Index, FMTI）中的24项预训练相关指标，对17个模型家族共41个模型条目进行系统性评估。研究发现，尽管部分模型在整体透明度得分上保持稳定或提升，但实际存在显著的信息丢失——在24组相邻版本对比中，38项指标出现披露退步，而仅有28项获得改进；尤其在语料库构成和训练计算量等关键维度仍存在持续性空白。这表明，当前模型迭代并未保证预训练信息的连贯披露，因此亟需建立明确标注变更内容与继承信息适用范围的规范化文档机制，以增强模型演进过程的可追溯性与透明度。
链接: https://arxiv.org/abs/2610.37465
作者: Md Sakib Ul Islam,S M Mehedi Zaman
机构: Victorian Institute of Technology (维多利亚理工学院); Rutgers University (罗格斯大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Understanding how language models were pretrained requires information about their data, processing procedures, and computational resources. However, it remains unclear whether successive releases preserve or expand earlier disclosure. We conduct a document-based audit of 41 model entries from 17 families using 24 pretraining-related indicators from the 2025 Foundation Model Transparency Index (FMTI). We examine overall scores and indicator profiles, tracking gains, losses, and retained disclosures across 24 adjacent comparisons in ten families. Scores decrease in 11 comparisons, increase in six, and remain unchanged in seven, with 38 indicator losses and 28 gains overall. Stable or improving scores can conceal losses: OLMo-2 and OLMo-3 receive identical totals despite three gains and three losses, while Llama-3.1 improves over Llama-2 while losing credit for two data-related criteria. Gaps also persist in corpus composition and training compute. These findings show that later releases do not consistently preserve earlier pretraining disclosure and motivate documentation that explicitly identifies changes and the applicability of inherited information.
[NLP-9] ETAMem: Decoupling and Aligning Heterogeneous Memory for Long-Term Conversational Agents
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在长时对话中面临的长期记忆系统所存在的两大核心挑战：信息纠缠（Information Entanglement）与组织错配（Organizational Mismatch）。前者指异构信息被混合于粗粒度的记忆单元中，导致语义混淆；后者则源于现有记忆结构主要依赖语义相似性进行组织，忽视了事件、情感与时间等多维度关联。为应对上述问题，论文提出ETAMem——一种多视角记忆框架，其关键在于通过解耦异构对话信息并实现记忆表征与组织方式与其内在特性对齐。具体而言，该方案首先通过异构信息解析提取原子级记忆单元，并识别情感信息；随后构建事件（Event）、情感（Affect）与时间（Time）三个互补的分层视角，分别对应不同语义维度的组织逻辑；在此基础上，采用自适应校准阈值实现增量式记忆维护；检索阶段则引入基于熵的预算分配机制，动态分配查询相关的资源至各视角，并在每个视角内执行层次化搜索。实验结果表明，ETAMem在LoCoMo与LongMemEval-S基准上分别取得12.46%和7.96%的相对准确率提升，同时保持优异的令牌效率，且分析验证了多视角记忆结构间的互补优势。
链接: https://arxiv.org/abs/2610.37443
作者: Xinying Qian,Ying Zhang,Yu Zhao,Xubo Liu,Baohang Zhou,Jeff Z. Pan
机构: Nankai University(南开大学); Tiangong University(天津中德应用技术大学); The University of Edinburgh(爱丁堡大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Long-term memory enables Large Language Models (LLMs) to maintain coherent and personalized interactions across extended conversations. However, existing memory systems face two key challenges: (1) Information Entanglement, where heterogeneous information is mixed within coarse-grained memory units; and (2) Organizational Mismatch, where memory structures based primarily on semantic similarity may overlook other associations. To address these challenges, we propose ETAMem, a multi-view memory framework that decouples heterogeneous conversational information and aligns memory representations and organization with its distinct characteristics. Specifically, heterogeneous information parsing extracts atomic memory units and identifies affective information. View-specific memory structuring then organizes these units into three complementary hierarchical views: Event, Affect, and Time. ETAMem incrementally maintains memory structures with adaptively calibrated thresholds. For retrieval, entropy-based retrieval allocates query-specific budgets across views and performs hierarchical search within each view. Experiments on LoCoMo and LongMemEval-S demonstrate that ETAMem outperforms state-of-the-art baselines, achieving relative accuracy improvements of 12.46% and 7.96%, respectively, while maintaining competitive token efficiency. Further analyses demonstrate the complementary benefits of different memory views.
[NLP-10] Representing the Unseen: Rethinking Visual Imagination for Spatial Reasoning
【速读】： 该论文旨在解决生成式视觉世界模型在空间推理中过度依赖RGB图像作为默认表示形式所带来的局限性问题。尽管RGB因其在视觉生成中的优越性而成为主流，但其并非对所有空间推理任务均最优；不同类型的推理任务（如全局距离比较或开放空间上下文理解）可能从不同的状态表示中获益，例如俯视几何图可提升全局距离判断，全景视图则有助于空间上下文感知。为系统评估表示形式对空间想象的影响，研究提出ImagiRep——一个基于模拟器的基准测试，包含922个问题，覆盖七类问题类型及两种场景（视角想象与跨视角整合），并提供同一未观测状态的原始RGB图像及六种替代表示（如深度图、全景视图、俯视几何图等）。实验结果表明，即使生成的RGB图像准确，也无法保证空间推理的准确性，且不同问题类型对表示形式具有显著偏好。这一趋势在由世界模型估计的状态表示中同样成立：仅使用估计的RGB可能导致误导性推理，而其他表示则保留互补优势。基于上述发现，研究进一步提出GEARS——一种轻量级的测试时自适应表示选择策略，通过动态选择最适表示以提升推理性能，显著优于现有空间想象方法，证明了对未见状态进行有效推理不仅依赖于生成RGB图像，更需智能地选择合适的表示形式。
链接: https://arxiv.org/abs/2610.37437
作者: Zengqi Zhao,Zun Wang,Shoubin Yu,Mohit Bansal,Yue Zhang
机构: UNC Chapel Hill; https://imagirep.github.io/
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: Webpage: this https URL
Abstract:Visual world models increasingly support spatial reasoning by generating RGB views of unobserved states. However, RGB has become the default interface largely because it is well suited for visual generation, rather than necessarily being the most effective representation for downstream reasoning. Different spatial questions may benefit from different representations of the same imagined state: top-down geometry can facilitate global distance comparisons, while panoramic views provide broader context for reasoning about open space. To isolate the role of representation in spatial imagination, we introduce ImagiRep, a simulator-grounded benchmark of 922 questions spanning seven question types across two settings, Viewpoint Imagination and Cross-view Integration, providing oracle RGB and six alternative representations (e.g., depth, panoramic views, and top-down geometry) of the same unobserved state. This controlled comparison reveals that even accurate imagined RGB does not guarantee accurate spatial reasoning, and different questions benefit from different representations. Similar trends hold for representations derived from world-model-estimated states. Notably, estimated RGB alone can mislead spatial reasoning, while alternative representations retain complementary benefits. Building on these findings, we introduce GEARS, a lightweight test-time strategy for adaptive representation selection, outperforming existing spatial imagination methods and showing that reasoning about unseen states requires more than generating RGB views.
[NLP-11] Game-Theoretic Solvers with Bayesian Opponent Modeling and Moral Rhetoric Framing for GLEE NEURIPS2026
【速读】： 该论文旨在解决在动态、复杂的人机混合博弈环境中，如何有效实现讨价还价（bargaining）、谈判（negotiation）与说服（persuasion）的协同决策问题。其核心挑战在于：理论上最优的策略在真实交互场景中表现不佳，尤其在面对具有隐变量（如折扣因子、保留价值、信号可靠性）的对手时，传统模型难以适应实时变化的对手行为模式。解决方案的关键在于构建一个三重架构的智能体系统：首先，采用经典博弈论求解器（如Rubinstein讨价还价模型、Faratin让步模型、Kamenica–Gentzkow信息披露模型）生成数值型策略动作；其次，引入在线贝叶斯估计层，实时推断对手的隐藏参数（如折扣因子、保留值或信号可信度），以动态建模对手类型；最后，利用语言模型生成语义层面的沟通内容，通过结合五种道德基础框架（moral foundation framings）与证据-直觉谱（evidence-intuition spectrum）的汤普森采样强化学习机制，确定最有效的信息表达方式。研究特别指出，尽管理论策略在模拟环境中表现良好，但在真实交互中存在因过度理想化而失效的问题，因此设计了多层容错机制与安全边界（guardrails），并通过大规模蒙特卡洛仿真（13,200场）和对9,337条真实对话消息的分析验证了道德话语框架在说服任务中的显著影响（χ²=616.7，p≈3.7×10⁻¹³²），证明了语言策略与最终结果之间的强相关性。
链接: https://arxiv.org/abs/2610.37397
作者: Hayden Arnold
机构: KAIST(韩国科学技术院)
类目: Computer Science and Game Theory (cs.GT); Computation and Language (cs.CL)
备注: 13 pages, 1 figure; accepted to 1st Workshop on Interpreting Agent Behavior (IAB) @ NeurIPS 2026
Abstract:We describe the agent we submitted to the GLEE Competition, which plays bargaining, negotiation, and persuasion games against a shared pool of human and AI opponents. The agent is built around three central components: a classical game-theoretic solver to compute numeric moves, an online Bayesian layer to estimate opponent archetypes, and a language model to write accompanying messages. We utilize Rubinstein bargaining for bargaining, Faratin concession for negotiation, and Kamenica–Gentzkow disclosure for persuasion, and adapt those models to live opponents using two separate layers: a Bayesian estimate of the opponent’s hidden discount factor, reservation value, or signaling reliability, and a Thompson sampling bandit over five moral-foundation framings coupled with an evidence-intuition spectrum to establish message register. Most of the development effort went into the recurring problem that theoretically optimal policies perform poorly in certain live settings; we report associated failure modes, guardrails we built, and our hypotheses. We additionally evaluate our model over 13,200 Monte Carlo games against 11 synthetic opponent archetypes, an ablation of the opponent-modeling layer, and an extensive analysis of 9,337 live-game messages in which moral rhetoric framing in the persuasion setting is strongly correlated with whether the buyer purchased ( \chi^2=616.7 , p \approx 3.7\times10^-132 ), surviving controls for quality, round, and configuration.
[NLP-12] LinMix: Data Mixing as Linear Domain Attribution NEURIPS2026
【速读】： 该论文旨在解决大语言模型预训练数据混合策略中领域权重分配的优化问题，核心挑战在于如何准确评估各领域对模型性能的贡献，并据此构建高效、可解释的数据混合方案。传统方法依赖随机采样的狄利克雷分布（Dirichlet sampling）生成代理数据混合比例，再通过回归模型预测性能，但该方法存在设计缺陷：其样本分布集中在数据先验混合点附近，导致领域效应高度相关，难以支持大规模领域下的线性建模，且需大量超参数调优。本文提出LinMix，基于经典最优实验设计理论，采用“截断单纯形顶点采样”（vertex sampling on the capped simplex），使每个领域在混合中处于全权重或完全缺席的状态，从而实现领域效应的去相关性。这一设计使得回归系数可直接解释为各领域的性能归因（domain attribution），并保证最优混合仅保留正归因领域，无需遍历候选混合。理论证明表明，狄利克雷采样在数据集中心区域会放大归因误差，而顶点采样具有更强的统计效率与最优性。在OLMo2（1B参数）和Qwen3.5（4B参数）等模型上的实验证明，LinMix在推理、代码生成与数学任务上显著优于基于狄利克雷的方法，相对提升达46%（MBPP）和14%（GSM8K），在288个细粒度领域下，仅顶点采样支持线性拟合，且平均性能较最优非线性基线提升11%。因此，解决方案的关键在于通过顶点采样实现领域归因的可解释性与最优混合的理论保障。
链接: https://arxiv.org/abs/2610.37390
作者: Hubert Baniecki,Przemyslaw Biecek,Martin Pawelczyk
机构: University of Warsaw(华沙大学); Centre for Credible AI, Warsaw University of Technology(华沙理工大学可信人工智能中心); Faculty of Computer Science, University of Vienna(维也纳大学计算机科学学院)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL); Machine Learning (stat.ML)
备注: NeurIPS 2026 Workshop on Attributing Model Behavior at Scale
Abstract:Language model performance depends strongly on the mixture of domains in the pretraining data corpus. Popular data mixing methods select these domain weights by training many small proxy models on random mixes and then fitting a regression model to predict performance from the mix. We recognize this practice as a domain attribution problem, where the regression measures each domain’s contribution to the target metric. Through this lens, the standard Dirichlet sampling of proxy data mixes is a poor regression design that requires tuning and cannot support a linear fit for hundreds of domains. Building on classical optimal experimental design, we introduce LinMix, which samples mixes at the vertices of the capped simplex so that each domain is at its cap or absent. This hyperparameter-free vertex sampling decorrelates domain effects, enabling an interpretable linear model whose coefficients are the domain attributions. The optimal mix given by LinMix provably keeps only domains with positive attribution values and requires no search over candidate mixes. We further establish the optimality of vertex sampling, and prove that Dirichlet sampling inflates the attribution error when its mixes concentrate around the prior data mix. Pretraining experiments with 1B-parameter OLMo2 and 4B Qwen3.5 language models show that LinMix outperforms Dirichlet-based data mixing across reasoning, code, and math benchmarks, with relative gains of 46% on MBPP and 14% on GSM8K. At 288 fine-grained domains, only vertex sampling supports a linear fit, and LinMix improves benchmark accuracy by 11% on average over the best nonlinear baseline.
[NLP-13] Same Verdict Different Readers: Personalizing Fact-Checking Explanations from Relative Preferences
【速读】： 该论文旨在解决事实核查解释（fact-checking explanation）在面对不同背景与偏好的用户时，难以实现个性化适配的问题。现有解释虽能说明判断依据，但无法满足个体化理解需求。其核心解决方案是提出一种基于偏好学习的个性化解释重写框架——PERPL（Personalized Explanation Rewriting via Preference Learning）。该框架的关键在于：首先通过跨人格的互相对比配对，构建具有人格特异性的模拟读者偏好监督信号；其次采用渐进式偏好学习策略，先以轻量级重写模型在构造的偏好数据上进行初步训练，再通过探索引导的偏好优化机制，使模型能够超越预设解释风格的限制，发现并生成更符合个体偏好的表达方式。实验表明，PERPL在个性化偏好对齐方面显著优于更大规模的通用大语言模型（LLM），同时保持了高水平的事实一致性（factual fidelity）。
链接: https://arxiv.org/abs/2610.37380
作者: Haokun Wen,Xuemeng Song,Mong-Li Lee,Wynne Hsu
机构: National University of Singapore(新加坡国立大学); Southern University of Science and Technology(南方科技大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Fact-checking explanations help Web users understand why a claim receives its verdict, yet the same explanation may not satisfy readers with different backgrounds and preferences. We formulate Faithful Personalized Fact-Checking Explanation Rewriting, which adapts an existing explanation to individual readers while preserving factual fidelity. Rather than relying on reader-written reference explanations, we learn personalized rewriting from the relative preferences of LLM-simulated readers. However, this presents two challenges: distinguishing persona-specific preferences from general explanation quality and learning beyond a limited set of predefined explanation styles. To address these challenges, we propose Personalized Explanation Rewriting via Preference Learning (PERPL), a framework consisting of two components. Simulated Preference Construction derives persona-specific supervision from simulated reader preferences through cross-persona reciprocal pairing. Progressive Preference Learning first trains a lightweight rewriter on the constructed supervision and then refines it through exploration-guided preference optimization, enabling it to discover preferred explanations beyond the predefined styles. Extensive experiments demonstrate that PERPL outperforms much larger general-purpose LLMs in personalized preference alignment while maintaining high factual fidelity.
[NLP-14] EmbeddingGemma 2: Unified and Efficient Multimodal Representations
【速读】： 该论文旨在解决多模态数据（包括文本、图像、音频、视频及其组合）在统一嵌入空间中表示的挑战，现有模型往往在跨模态任务中表现不佳或依赖于庞大参数量的专用模型。其解决方案的关键在于提出EmbeddingGemma 2——一个原生支持多模态输入的开源嵌入模型，通过简化每层嵌入结构与采用分阶段课程学习策略（multi-stage curriculum with modality and objective progression），实现从单模态到跨模态再到混合模态任务的渐进式训练优化，从而在各类模态评估中均达到卓越性能。该设计不仅在全面的嵌入基准测试中超越了专用模型及规模大2至12倍的同类模型，建立了新的帕累托前沿（Pareto frontier），还具备高度灵活性，支持选择性加载模态编码器、可调视觉令牌预算、可截断的嵌入维度以及量化权重，适用于资源受限环境，有效推动了多模态嵌入技术的发展。
链接: https://arxiv.org/abs/2610.37364
作者: Henrique Schechter Vera,Sahil Dua,Jovan Andonov,Sara Smoot,Sebastian Russo,Ivan Llanos,Max Matveev,Jae Hyeon Yoo,Kat Black,Shih-Cheng Huang,Josiah Aklilu,Maarten Grootendorst,Omar Sanseviero,Chandramouli Amarnath,Ian Ballantyne,Zhe Li,Daniel Salz,Ibrahim Alabdulmohsin,Ilya Philippov,Kristen Quan,Glenn Cameron,Edouard Yvinec,Luciano Martins,Doug Reid,Sindhu Raghuram Panyam,Weiyi Wang,Hengxuan Ying,Madhuri Shanbhogue,Tanmaya Dabral,Shanfeng Zhang,Juyeong Ji,Renjie Wu,Joe Zou,Xu Chen,Yunan Zhang,Yu-Hui Chen,Matthew Soulanille,Vinod Mamillapalli,Yi-Chun Kuo,Wai Hon Law,Matt Kreileder,Alice Zheng,Suril Shah,Mogan Shieh,Pooja Consul,Alberto Montes,Andreas Hess,Andres Sevilla,Babak Samari,Brendan McCloskey,Brian Potetz,Daniel Cer,Georgios Kathareios,Gustavo Hernández Ábrego,Hesen Zhang,Ji Ma,Jiaxing Wu,Keegan Mosley,Paul Suganthan,Rui Meng,Shen Gao,Shengcao Cao,Yi-Ting Chen,Yichang Chen,Zhongli Ding,Kendra Byrne,Cormac Brick,Olivier Lacombe,Qin Yin,Tris Warkentin,Armand Joulin,Tom Duerig,Mojtaba Seyedhosseini
机构: Google(谷歌)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV); Sound (cs.SD)
备注: 28 pages. Models are available on HuggingFace (at this https URL ) and LiteRT (at this https URL )
Abstract:We present EmbeddingGemma 2, an open multimodal embedding model which natively represents text, images, audio, video, and combinations thereof in a unified space. Our approach introduces improvements such as simplified per-layer embeddings and a multi-stage curriculum with modality and objective progression which ensure strong quality across unimodal, cross-modal, and mixed-modal evaluations. Notably, in comprehensive embedding benchmarks across all modalities, EmbeddingGemma 2 establishes a new Pareto frontier, outperforming both specialist models and models two to twelve times larger. Furthermore, our design offers flexibility for resource-constrained environments, supporting selective modality encoder loading, adjustable vision token budgets, truncatable embedding dimensions, and quantized weights. We present ablations analyzing key decisions and release EmbeddingGemma 2 to promote further research.
[NLP-15] UCVE: Graph Structured Retrieval for Compliance Validation EMNLP2026
【速读】： 该论文旨在解决结构化合规性验证中的覆盖率问题，即如何准确判断给定文档是否满足、违反或未涉及预定义需求空间中的每一项义务。现有检索与分类系统因将需求视为独立个体，无法有效处理此类依赖关系复杂的覆盖分析任务。其解决方案的关键在于提出一种领域无关的框架\textbf{UCVE}，将合规需求空间建模为合规关系图（Compliance Relation Graph, CRG），并通过基于强化学习加权的拉普拉斯矩阵进行谱重排序（spectral reranking）实现高效检索。其中，需求节点间的边通过强化学习自动学习，奖励信号源自违规检测的遍历反馈，从而动态重塑图的社区结构；经低通谱滤波后，可有效区分相关与无关义务簇。实验在三个合规领域验证表明，UCVE相较传统余弦检索平均提升4.3%的匹配K召回率，并持续优于强基线检索与图模型方法。
链接: https://arxiv.org/abs/2610.37323
作者: Emily Chen,Michelle Menachery,Himanshu Kumar,Rahul Nenavath,Amirfarrokh Iranitalab
机构: Amirfarrokh Iranitalab; Capital One, AI Foundations
类目: Computation and Language (cs.CL)
备注: Accepted to the Natural Legal Language Processing Workshop (NLLP 2026) at EMNLP 2026
Abstract:Structured compliance validation requires determining whether every obligation in a predefined requirement space is satisfied, violated, or unaddressed by a given document. This can be described as a coverage problem that existing retrieval and classification systems, by treating requirements as independent, are not designed to solve. We introduce \textbf\acUCVE, a domain-agnostic framework that represents compliance requirement spaces as a \acCRG and performs retrieval via spectral reranking over its reinforcement learning-weighted Laplacian. Edges between requirement nodes are learned via reinforcement learning with a traversal reward grounded in violation detection, reshaping the graph’s community structure such that low-pass spectral filtering distinguishes relevant from irrelevant obligation clusters. Evaluating across three compliance domains, \acUCVE averages 4.3% matched- K recall lift over flat cosine retrieval and consistently outperforms strong retrieval and graph-based baselines.
[NLP-16] Multi-Label Agreement: A Flexible Framework for Inter-Annotator Reliability
【速读】： 该论文旨在解决多标签标注任务中评价标注者间一致性（Inter-annotator Agreement, IAA）的标准化难题。传统IAA度量如Cohen’s kappa、Fleiss’ kappa和Krippendorff’s alpha适用于单标签场景，但在多标签任务中，如何合理处理完整标签集、共同选择与共同遗漏的情况缺乏统一规范。为此，本文提出多标签一致性（Multi-Label Agreement, MLA）框架，其核心在于保留每个样本的完整二值标签向量，并采用平均成对汉明一致率（average pairwise Hamming agreement），从而同时捕捉联合选择与联合遗漏的信息。MLA的关键创新在于其池化边缘形式（pooled-marginal form）并非独立系数，而是基于标签层面的Fleiss/Scott型系数的几率分歧加权平均，且与使用归一化汉明距离的Krippendorff’s alpha密切相关。该框架支持多种归一化方式（池化边缘、Gwet AC1、固定基线），可输出项目级、标签级及语料库级的一致性结果，并包含条件最小一致性诊断功能。通过实例分析与蒙特卡洛模拟，验证了结果对标签频次、声明标签集合及期望一致性模型的敏感性。研究提供了可运行的Python实现，强调透明性和可解释性，不主张某一种归一化方法具有普适优越性。
链接: https://arxiv.org/abs/2610.37292
作者: Tim Menzner,Jochen L. Leidner
机构: Coburg University of Applied Sciences(科堡应用技术大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Reliable evaluation of inter-annotator agreement (IAA) is important for assessing the consistency of annotation processes. While established IAA metrics such as Cohen’s kappa, Fleiss’ kappa, and Krippendorff’s alpha are commonly applied in single-label settings, multi-label tasks require additional decisions about how full label sets, joint selections, and joint omissions should be compared. In this work, we present Multi-Label Agreement (MLA), a flexible framework for evaluating agreement in multi-label annotation tasks. MLA preserves the full binary label vectors assigned to each item and uses average pairwise Hamming agreement, thereby accounting for both joint selections and joint omissions. Its pooled-marginal form is not a mathematically independent coefficient: it is a chance-disagreement-weighted average of label-wise Fleiss/Scott-type coefficients and is closely related to Krippendorff’s alpha with normalized Hamming distance. The framework supports pooled-marginal, Gwet AC1, and fixed-baseline normalizations, yields item-level, label-level, and corpus-level results, and includes a conditional minimum-agreement diagnostic. Through a worked example and Monte Carlo simulations, we show how the resulting values depend on label prevalence, the declared label inventory, and the selected model of expected agreement. We also provide a Python implementation. MLA thus offers a transparent and interpretable framework for multi-label agreement analysis without claiming that one normalization is universally preferable. Subjects: Computation and Language (cs.CL) ACMclasses: G.3; I.2.6; I.2.7 Cite as: arXiv:2610.37292 [cs.CL] (or arXiv:2610.37292v1 [cs.CL] for this version) https://doi.org/10.48550/arXiv.2610.37292 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[NLP-17] Where Should the Evaluation Budget Go in LLM -Driven Agent Scaffold Evolution?
【速读】： 该论文旨在解决大语言模型（Large Language Model, LLM）驱动的智能体骨架演化过程中，因对每个候选方案在完整任务集上进行评估而导致的高昂执行成本问题。其核心挑战在于如何在指导搜索过程（search）与最终选择最优骨架（final selection）这两个评估角色之间高效分配有限的评估预算。解决方案的关键在于提出一种名为\method的评估分配框架，该框架通过结合稀疏的运行时反馈（sparse search-time feedback）与针对性的最终确认评估（targeted final confirmation），实现高效评估。具体而言，\method首先通过一个简短的全量评估预热阶段建立共享的语义信息核心，并校准响应模型；随后在搜索阶段保持模型参数固定以减少开销；最终阶段则重新拟合模型并采用成对连续减半（paired successive halving）策略，复用已有观测结果，仅补全缺失评估。实验表明，在四个基准测试和两种演化框架下，\method在多数情况下均实现了优于全量评估的平均性能，同时仅使用约一半的评估单元。
链接: https://arxiv.org/abs/2610.37276
作者: Sichun Luo,Yi Huang,Mukai Li,Haochen Luo,Guanzhi Deng,Haibo Wang,Jian Xu,Zefa Hu,Junlan Feng,Qi Liu
机构: The University of Hong Kong; JIUTIAN Research, China Mobile; City University of Hong Kong; Carnegie Mellon University; Tsinghua University
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Neural and Evolutionary Computing (cs.NE)
备注:
Abstract:Large language model (LLM)-driven agent scaffold evolution improves agent workflows through iterative candidate generation and evaluation. Evaluation both guides subsequent search and determines the final scaffold, but evaluating every candidate on the full task set incurs substantial execution costs. We investigate how evaluation budgets can be allocated across these two roles. Our experiments show that sparse feedback can sustain productive evolution, while additional targeted evaluation improves final selection. Replay analyses further show that shared informative items and a common scoring reference support comparisons among closely matched candidates. Motivated by these findings, we propose \method, an evaluation allocation framework that combines sparse search-time feedback with targeted final confirmation. A short fully evaluated warm-up establishes a shared informative core and calibrates a response model whose item parameters remain fixed during search. After search, \method refits the model and applies paired successive halving, reusing accumulated observations and executing only missing evaluations. Across four benchmarks and two evolution frameworks, \method achieves higher mean performance than full evaluation in most settings while using approximately half as many evaluation cells.
[NLP-18] Watermark Basins: Black-Box Detection of Distilled LLM s
【速读】： 该论文旨在解决生成式 AI（Generative AI）中专有语言模型及其推理轨迹面临 API 基于的蒸馏攻击问题，即攻击者通过查询教师模型的输出来训练学生模型，从而实现知识窃取。现有解决方案依赖于在黑盒场景下通过分析可疑模型输出的词元级统计特征来检测水印信号，但此类信号在低温度采样等情况下易失效，导致检测不可靠。本文的关键发现是“水印盆地”（Watermark Basin）——在微调过程中，经过水印标记的教师输出会在参数空间中嵌入一种与任务特定微调子空间正交的、由水印密钥决定的独特几何信号，该信号具有持久性。基于此，作者提出 Watermark Basin Alignment (WM-Basin) 检测方法，仅需通过 API 查询目标模型，无需访问其权重或对数几率值；该方法利用一个固定的干净探测模型将生成文本映射为梯度表示，剔除常规任务微调方向后，检验残差梯度是否与水印密钥诱导的水印盆地对齐。实验表明，在严格 0.1% 的误报率下，WM-Basin 在 n-gram=1 时可检测 79% 的蒸馏模型，n-gram=3 时仍达 45%，显著优于当前最优的词元级基线（分别为 29% 和 1.3%）。由于水印信号存在于稳定的盆地中，其能有效抵抗量化、剪枝及强化学习后训练等后续修改操作，具备更强的鲁棒性。
链接: https://arxiv.org/abs/2610.37266
作者: Hongyan Chang,Hady Elsahar,Tom Sander,Pierre Fernandez,Saeed Mahloujifar,Sylvestre-Alvise Rebuffi,Tomáš Souček,Tuan Tran,Valeriu Lacatusu,Surya Parimi,Rashel Moritz,Vanessa Stark,Alexandre Mourachko
机构: FAIR, Meta Superintelligence Labs (FAIR，Meta超级智能实验室)
类目: Cryptography and Security (cs.CR); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Proprietary language models and their reasoning traces are vulnerable to API-based distillation, where an adversary trains a student model on the teacher model’s outputs. To obtain evidence of distillation, model providers can watermark the teacher’s outputs, aiming to leave a radioactive signal in the student’s outputs. Existing black-box detectors search for such radioactivity using token-level statistics of the suspect model’s outputs. However, token-level signals are not always detectable, for instance when the distillation traces are sampled at a low temperature. In this work, we discover the watermark basin, a persistent signal that enables more reliable watermark detection: during fine-tuning, watermarked teacher traces embed a distinct, key-specific geometric signal in parameter space that lies orthogonal to the ordinary task-specific fine-tuning subspace. Building on this, we introduce Watermark Basin Alignment (WM-Basin), a black-box watermark detector that queries the suspect model through its API, without access to its weights or logits. WM-Basin maps generated text to gradient representations using a fixed clean probe model, projects out the ordinary fine-tuning task directions, and tests whether the residual gradient aligns with the watermark basin induced by the watermark key. We evaluate WM-Basin extensively across diverse teacher models and student architectures (from 135M to 8.2B parameters) distilled on mathematical reasoning. At a strict 0.1% false-positive rate, WM-Basin detects 79% of distilled models at n-gram=1 and 45% at n-gram=3, substantially improving over the state-of-the-art token-level baseline (29% and 1.3%, respectively). Because the watermark signal resides in a persistent basin, it survives post-distillation modifications including quantization, pruning, and reinforcement-learning post-training.
[NLP-19] REGLAT at ArGuard Shared Tasks: An Ensemble-Based Model Integrating NER for Harmful Content Detection in Arabic
【速读】： 该论文旨在解决阿拉伯语中由于方言差异及不同有害内容类别间语义重叠所带来的有害内容检测难题。其核心解决方案在于提出一种两阶段的分类框架：首先利用GLiNER模型从用户输入中提取与有害内容相关的实体，随后将这些实体信息融合至原始文本，并输入经过微调的MARBERT语言模型进行最终分类。该方法通过结合显式的实体特征与预训练阿拉伯语语言模型的上下文理解能力，有效提升了细粒度有害内容识别的准确性。实验结果表明，该系统在ArGuard 2026共享任务的开发集上取得了0.886的宏平均F1值和0.897的准确率，而在盲测集上对子任务1和子任务2分别实现了0.793和0.770的准确率以及0.735和0.754的宏平均F1值，验证了该方法的有效性。尽管如此，研究也指出，对于语义相近的类别仍存在区分困难的问题，进一步说明将命名实体识别（Named Entity Recognition, NER）与阿拉伯语语言模型相结合是提升有害内容检测性能的一条有前景的技术路径。
链接: https://arxiv.org/abs/2610.37265
作者: Ahmed M. Fetouh,Mariam Labib,Nsrin Ashraf,Omer Dawood,Hamada Nayel
机构: Benha University(本哈大学); Elsewedy University of Technology(埃尔赛维迪科技大学); Prince Sattam Bin Abdulaziz University(萨特姆·本·阿卜杜勒阿齐兹大学)
类目: Computation and Language (cs.CL)
备注: System description paper for REGLAT submission for ArGuard shared task 2026
Abstract:Detecting harmful content in Arabic is challenging because of dialectal variations and the semantic overlap between different harm categories. This paper presents a two-stage approach for classifying Arabic user prompts into seven categories of harmful content. First, GLiNER is used to extract harm-related entities from each prompt. The extracted entities are then added to the original text and passed to a fine-tuned MARBERT classifier for final classification. The approach combines explicit entity information with the contextual understanding of a pretrained Arabic language model. The system was evaluated on the development set of the ArGuard 2026 shared task. The results achieved a macro-F1 score of 0.886 and an accuracy of 0.897, demonstrating promising performance in fine-grained harmful content detection. The system reported an accuracy of 0.793 and Macro-F score of 0.735 on a blind test set for subtask 1. For subtask 2, it achieved an accuracy of 0.770 and Macro-F score of 0.754. The analysis also showed that distinguishing between related categories remains challenging. These findings suggest that integrating NER with Arabic language models is a promising approach to harmful content detection.
[NLP-20] Your Policy Is Secretly a Harness: The Decision Geometry of Agent Interfaces
【速读】： 该论文旨在解决语言模型代理（language model agents）在不同任务框架（harness）下性能表现不一致的问题，核心关注点在于：模型在特定框架中学习到的策略是否具有可迁移性，以及当框架发生变化时，哪些因素决定了其性能的保持或退化。研究表明，在KL正则化控制条件下，最优策略的对数几率（log odds）能够恢复框架对各动作的相对偏好强度，表明训练后的策略会“携带”其训练框架的偏好特征。因此，部署后的实际回报等于框架所能提供的最佳回报（即“天花板”，ceiling）减去策略自身次优决策带来的损失（即“不匹配”，mismatch）。训练过程仅优化不匹配项，而改变框架则同时影响天花板与不匹配项。固定模型对比不同框架时，结果混淆了框架本身提供的潜力与模型利用该潜力的能力。实验显示，在MiniWoB++中，即使最优决策不变，仅调换观测字段顺序就会使模型成功率从85.0%降至19.2%，说明模型偏好高度依赖于具体框架；而在Terminal-Bench Lite的代码代理任务中，使用更强模型作为近似最优值估算的“天花板减不匹配”项能准确预测不同框架的性能排序，而单独使用任一指标则无效。这表明，评估框架有效性需综合考虑其提供潜力的能力与模型适应能力，解决方案的关键在于分离并量化框架的天花板效应与策略的不匹配程度。
链接: https://arxiv.org/abs/2610.37255
作者: Jian Mu,Ruibo Shen,Yao Shu
机构: The Hong Kong University of Science and Technology (Guangzhou); The Hong Kong University of Science and Technology
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注:
Abstract:Language model agents are post-trained and deployed through a harness, the prompts, tools, and control logic that present a task to the model and execute its outputs. Harnesses are often revised and judged by how an already trained model performs under them, which assumes that what the model learned under one harness carries over to the next. We ask what a policy learns from its harness and what survives when the harness changes. Under KL-regularized control, we show that the log odds of an optimal policy recover how strongly its harness favors each action over the others, so a trained policy carries the preferences of its harness. Deployed return then equals the ceiling of the harness, the best return it offers, minus the mismatch of the policy, the return lost to its own suboptimal decisions. Training changes only the mismatch, while changing the harness moves both terms. Comparing harnesses with a fixed model therefore mixes what each harness offers with how well that model uses it. In MiniWoB++, post-training makes a model’s preferences specific to its training harness. Reversing the order of the observation fields leaves every optimal decision unchanged yet lowers a trained model’s success from 85.0% to 19.2%. Conversely, every new harness that a fixed-model comparison rejects but that raises the ceiling outperforms the original agent after further training under it, and ceiling and mismatch measured beforehand predict these gains. For coding-agent harnesses on Terminal-Bench Lite, ceiling minus mismatch estimated with a stronger model as a stand-in for the optimum reproduces the accuracy ranking, while either term alone does not. Code is available at this https URL.
[NLP-21] Aligning Large Language Models for Safety in African Languages: A Case Study of Hausa
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）安全对齐研究长期集中于高资源语言，而缺乏在低资源、文化差异显著语境下有效性评估的问题。针对西非使用超过2亿人口的豪萨语（Hausa），研究构建了首个豪萨语安全偏好数据集HausaSafeLM，包含10,000条来自PKU-SafeRLHF的提示样本，其中3,000条为人工校正的“黄金”数据，7,000条为机器翻译生成的“白银”数据，覆盖六类危害宏观类别。基于Qwen2.5-7B-Instruct模型，研究对比了监督微调（Supervised Fine-Tuning, SFT）与SFT后接直接偏好优化（Direct Preference Optimization, DPO）两种策略在不同数据质量层级下的表现。实验结果表明，未经对齐的基线模型虽在拒绝率指标上表现“安全”，但其拒绝内容普遍质量低下，反映的是语言生成能力受限而非真实的安全对齐。SFT有效提升了拒绝的流畅性与准确性，并显著受益于更大规模但噪声更高的白银数据；而DPO虽进一步提升了拒绝的精确性，却以严重且基本不受数据量影响的语言质量损失为代价。最终发现，没有任何单一配置能同时最大化安全性、有用性和语言流畅性。研究揭示当前对齐技术更擅长教会模型“何时拒绝”，而非“如何流畅地拒绝”，尤其在低资源语言中，仅依赖拒绝率作为安全信号具有高度误导性。
链接: https://arxiv.org/abs/2610.37245
作者: Valdini Douglace Lemofouet,Blessing Ngozi Uzor,Paula Chikaodinaka Anyanwu,Danielle Blanche Kapsa,Saminu Mohammad Aliyu,Seid Muhie Yimam,Shamsuddeen Hassan Muhammad
机构: African Institute for Mathematical Sciences (AIMS), Cameroon; Bayero University Kano; University of Hamburg; Imperial College London
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Safety alignment research for large language models (LLMs) remains concentrated on high-resource languages, leaving open how existing alignment methods perform in low-resource, culturally distinct settings. We study this question for Hausa, a language spoken by more than 200 million people across West Africa. We introduce HausaSafeLM, one of the first Hausa safety preference dataset, built from 10,000 prompts derived from PKU-SafeRLHF: 3,000 human-corrected (gold) and 7,000 machine-translated (silver) examples spanning six harm macro-categories. Using Qwen2.5-7B-Instruct, we compare supervised fine-tuning (SFT) and SFT followed by Direct Preference Optimization (DPO) under both data regimes, evaluating on 500 harmful and 200 benign Hausa this http URL find that the untuned base model appears highly safe by refusal-rate metrics alone, but this reflects limited Hausa generation ability rather than genuine alignment: its refusals are almost uniformly low quality. SFT improves both refusal fluency and precision and benefits substantially from the larger, noisier silver-tier data, while DPO sharpens refusal precision but comes at a steep, largely data-invariant cost to linguistic quality. No single configuration jointly maximizes safety, helpfulness, and fluency. Our results suggest that current alignment techniques teach models when to refuse more readily than how to do so fluently in underrepresented languages, and that refusal-rate metrics alone are an unreliable signal of safety in low-resource settings.
[NLP-22] Decision Models as Judges for Hallucination Detection
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLM）作为模型输出评判者（LLM-as-a-Judge）时存在的高成本、幻觉生成及过度自信等问题。其核心解决方案是采用决策模型（Decision Models），该类模型以预定义标签的概率输出替代自由文本生成，具有更高的计算效率与更低的成本。研究聚焦于幻觉检测任务，构建了涵盖五种幻觉类型（共9,852个样本）的二分类基准测试集，并对比了五种决策模型与生成式评判模型gpt-6-sol的表现。结果表明，决策模型在上下文不一致和事实性错误检测上表现接近gpt-6-sol，但在逻辑不一致和指令不一致方面仍存在4–8个百分点的性能差距；同时，决策模型的概率输出虽较GPT-6模型更少过度自信，但对正确与错误判断的区分能力有限。进一步引入置信度级联（confidence cascade）策略后，决策模型在三种幻觉类型上的性能提升1.9–3.3个百分点，且达到或超过gpt-6-sol的水平，同时仅需其1%–3%的计算成本。总体而言，决策模型为特定类型的幻觉检测提供了低成本可行的替代方案，但使用前需进行充分验证。
链接: https://arxiv.org/abs/2610.37229
作者: Haozheng Liu,Zhaoqian Xue,Zhongyin Wang,Junting Mu,Zihui Ma,Shuya Feng,Lingyao Li
机构: New York University; University of Arizona; University of Illinois Urbana-Champaign; University of California, Irvine; Emory University; University of Maryland, College Park; University of Alabama at Birmingham
类目: Computation and Language (cs.CL)
备注:
Abstract:Large language models are widely used as judges of model outputs (LLM-as-a-Judge), but they are costly and can hallucinate or be overconfident. Decision models are a potential alternative. They return a predefined verdict with a probability for each label instead of free-form text and are much cheaper and faster. We test this promise on hallucination detection. We build binary benchmarks for five hallucination types (9,852 items from six datasets) and compare five decision models with the generative judge gpt-6-sol. Performance depends strongly on the type: the best decision model matches gpt-6-sol in balanced accuracy on context inconsistency and comes within 3 percentage points on factual errors, but trails it by 4-8 points on logical and instruction inconsistency. Decision-model probabilities are overconfident, though less so than the GPT-6 models, and separate correct from incorrect decisions only moderately. We also find that a confidence cascade of decision models improves on the best single model by 1.9-3.3 points on three types and matches or exceeds gpt-6-sol at 1-3% of its cost. Overall, decision models are a practical low-cost judge for some hallucination types but need to be validated before use.
[NLP-23] Uncovering Shortcut Amplification in Video RLVR through the Lens of Perceptual Uncertainty Collapse
【速读】： 该论文旨在解决生成式视频推理中强化学习与可验证奖励（Reinforcement Learning with Verifiable Rewards, RLVR）范式所引发的“捷径放大”问题，即模型在训练过程中过度依赖文本先验（text-prior）而忽视视觉证据，导致推理结果缺乏视觉依据。其核心问题是：尽管RLVR能有效提升大型多模态模型（LMMs）的视频推理能力，但其基于结果层面的优化机制会诱导模型产生对文本线索的过度依赖，从而削弱模型对视觉信息的真实感知与利用。解决方案的关键在于提出一种双焦点自蒸馏策略优化（Bifocal Self-distillation Policy Optimization, BSPO），通过引入两种互补的分布信号来调控策略更新：一是基于组内正确与错误轨迹的实证推理原则，识别出对推理敏感的语义标记；二是设计具备捷径感知能力的视觉扰动机制，惩罚当任务相关视觉信息被弱化时仍保持不变的预测行为。通过自适应重加权原始强化学习优势函数，BSPO在保留RLVR任务求解优势的同时，引导模型更新向更具推理敏感性与视觉依存性的方向演进，从而显著提升视频推理性能与视觉忠实度。
链接: https://arxiv.org/abs/2610.37211
作者: Jiafeng Liang,Xianhao Ou,Tao Ren,Rongchuan Mu,Runxuan Liu,Dandan Tu,Ming Liu,See-Kiong Ng,Bing Qin
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL)
备注:
Abstract:Reinforcement learning with verifiable rewards (RLVR) has emerged as an effective post-training paradigm for enhancing the video reasoning capability of Large Multimodal Models (LMMs). Despite its empirical effectiveness, we show that outcome-level optimization can introduce an undesirable shortcut amplification effect, whereby models increasingly favor text-prior answers even when they are not supported by the visual evidence. To characterize this behavior, we investigate the training dynamics of video RLVR through the lens of perceptual uncertainty, analyzing how predictive uncertainty evolves under video-conditioned and text-only contexts. Our analysis identifies a perceptual uncertainty collapse phenomenon, in which RL training continuously sharpens the model’s answer distribution while the uncertainty reduction attributable to visual evidence progressively diminishes. A further gradient-direction analysis reveals that such confidence enhancement becomes increasingly less aligned with visually grounded uncertainty reduction and is instead dominated by text-prior compression. Motivated by these findings, we propose Bifocal Self-distillation Policy Optimization (BSPO), a self-distillation framework that regulates policy updates with two complementary distributional signals. Specifically, BSPO derives empirical reasoning principles from intra-group correct and incorrect rollouts to identify reasoning-sensitive tokens, and introduces shortcut-aware visual perturbation to penalize predictions that remain invariant when task-relevant visual evidence is weakened. By adaptively reweighting the original RL advantage, BSPO preserves the task-solving benefits of RLVR while promoting updates that are both reasoning-sensitive and visually grounded. Extensive experiments validate the effectiveness of BSPO in improving video reasoning performance and strengthening visual faithfulness.
[NLP-24] What Survives Distillation? A Mechanistic Interpretability Study of GPT -2 and DistilGPT -2
【速读】： 该论文旨在解决生成式人工智能（Generative AI）中大语言模型压缩过程中，深度缩减导致的机制性失效问题，尤其关注知识蒸馏（Knowledge Distillation）对特定神经网络电路（circuit）功能完整性的影响。其核心问题是：在将大型教师模型（如GPT-2）压缩为小型学生模型（如DistilGPT-2）时，关键推理路径的内部工作机制是否得以保留？解决方案的关键在于通过多维度的机制分析方法——包括线性中心化核相关性（linear CKA）、logit lens分析、注意力模式分解、稀疏自编码器特征统计以及激活补丁（activation patching）——系统性地对比教师与学生模型在两个典型任务电路（间接宾语识别IOI与大于比较GT）中的行为差异。研究发现，存在“压缩-坍塌”二分现象：在IOI电路中，学生模型出现机制完全崩溃（准确率从94.8%降至35.2%，无名称传递保留），而在GT电路中则通过教师层9的比较头向学生层4的主动重映射实现87.7%的任务性能恢复（皮尔逊相关系数均值达0.945）。这表明，电路的可压缩性决定了其在知识蒸馏下的机制生存能力，而传统聚合指标（aggregate metrics）无法有效反映底层机制完整性，因此必须结合机制诊断工具才能评估模型压缩的真实效果。
链接: https://arxiv.org/abs/2610.37157
作者: Nabeelah Faumi,Deshan Sumanathilaka
机构: Informatics Institute of Technology(信息学院技术学院); Sri Lanka(斯里兰卡)
类目: Computation and Language (cs.CL)
备注: 5 Figure, 9 Table
Abstract:Knowledge distillation compresses large language models into compact students that preserve aggregate metrics, yet the mechanistic consequences of depth-reduction distillation remain largely uncharacterized. This work applies linear CKA, logit lens analysis, attention pattern decomposition, sparse autoencoder feature statistics, and activation patching to compare GPT-2 and its six-layer student DistilGPT-2 across two canonical circuits: the Indirect Object Identification (IOI) circuit and the Greater-Than (GT) circuit. The results reveal a compression-collapse dichotomy: the IOI circuit suffers complete mechanistic collapse (DistilGPT-2: 35.2% accuracy vs. 94.8%, zero name-mover preservation), while the GT circuit survives with 87.7% task performance recovery through active remapping of the teacher’s layer-9 comparison head into student layer~4 (mean Pearson r = 0.945). These findings demonstrate that circuit compressibility governs mechanistic survival under distillation, and that aggregate metrics are insufficient diagnostics for circuit integrity.
[NLP-25] Mechanistic Failure Modes of Weight Pruning in Transformer Language Models
【速读】： 该论文旨在解决当前对剪枝语言模型（pruned language models）评估主要依赖准确率与困惑度（perplexity）的问题，这些指标隐含假设保留的输出行为等价于内部计算过程的保持。然而，研究通过在GPT-2上开展因子设计的剪枝实验，系统检验了这一假设的有效性，实验交叉考察了两种重要性标准（幅度、随机）、两类组件类型（注意力头、MLP神经元）以及三种稀疏度水平（3%、50%、70%）。以已充分表征的间接宾语识别（IOI）电路作为基准，采用激活修补（activation patching）、路径修补（path patching）和直接逻辑值归因（direct logit attribution）等方法量化电路完整性。研究发现存在两种定性不同的失效模式：基于幅度的注意力头剪枝会导致电路“置零”——输出面向的注意力头被移除，模型性能降至近随机水平；而基于幅度的神经元剪枝在30%稀疏度下则引发电路“反转”——尽管全部26个电路注意力头仍存在，但“名称移动者”（Name Mover）的贡献权重符号反转，导致模型系统性地选择错误名称（准确率仅17.7%），但其困惑度却低于随机剪枝（74.6 vs 140.1）。对“大于”（Greater-Than）电路的复现进一步表明，关键组件类型取决于电路架构。上述结果揭示，仅依赖输出指标或结构检查均可能遗漏机制层面的损伤，因此提出应引入电路级因果分析作为压缩语言模型评估的补充范式。
链接: https://arxiv.org/abs/2610.37126
作者: Nabeelah Faumi,Deshan Sumanathilaka
机构: Informatics Institute of Technology(信息学技术学院); Sri Lanka(斯里兰卡)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: 12 pages, 10 Tables
Abstract:Pruned language models are judged almost entirely by accuracy and perplexity, which assumes that preserved output behaviour implies preserved internal computation. We test this assumption with a factorial pruning study on GPT-2, crossing two importance criteria (magnitude, random), two component types (attention heads, MLP neurons) and three sparsity levels (3%, 50%, 70%). Using the fully characterised indirect object identification (IOI) circuit as ground truth, we measure circuit integrity with activation patching, path patching and direct logit attribution. We find two qualitatively distinct failure modes. Magnitude head pruning = zeroes the circuit: output-facing heads are removed, and the model falls to near-chance. Magnitude neuron pruning at 30% inverts it: all 26 circuit heads remain intact, but Name Mover contributions flip sign, and the model systematically prefers the wrong name (17.7% accuracy). Perplexity ranks it as less damaged than random head pruning (74.6 vs 140.1). A replication of the Greater-Than circuit shows that the critical component type depends on the circuit’s architecture. These results show that output metrics and structural checks can both miss mechanistic damage, motivating circuit-level causal analysis as a complementary evaluation of compressed language models.
[NLP-26] Prospector: learning a compact graph traversal policy for scientific reasoning
【速读】： 该论文旨在解决在半导体制造等专业工程领域中，前沿大语言模型（LLM）因缺乏对散落于文献中且常为专有性质的知识的访问能力，而难以提供精准、可追溯推理答案的问题。现有方法通常依赖预筛选的知识片段进行推理，但受限于先验选择的偏差与不完整性。其解决方案的关键在于提出一个轻量级、可迁移的推理模型Prospector，该模型通过逐步遍历领域知识图谱（KG），动态调用一组小型工具以构建上下文，并基于已获取证据持续寻找新证据，最终生成可逐条验证的证据链。该模型仅需单个GPU即可部署，显著降低计算成本并确保专有数据保留在本地；同时，尽管仅在半导体制造知识图谱上训练，其仍能有效迁移到生物启发材料、全氟和多氟烷基物质（PFAS）及材料性能等三个不同领域的未见知识图谱，展现出强大的泛化能力。研究进一步发现，模型隐状态可有效预测下一步行为，且可通过调节隐状态实现行为模式的切换，表明其具备可控的推理策略。这一成果揭示了一种面向私有数据的实用型LLM+KG科学与工程推理范式：即利用公开文献构建的知识图谱训练轻量化模型，使其掌握跨域图谱遍历能力，从而在不共享数据的前提下实现高效、可信的推理。
链接: https://arxiv.org/abs/2610.37121
作者: Haiqian Yang,Markus J. Buehler
机构: 未知
类目: Artificial Intelligence (cs.AI); Materials Science (cond-mat.mtrl-sci); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Scientific reasoning in specialized engineering fields depends on knowledge scattered across the literature, often proprietary and absent from frontier LLM training data. In such domains, frontier models tend to give generic answers. Knowledge graphs (KGs) can ground LLM reasoning. Because a graph is too large to provide in full, typical LLM+KG systems preselect a portion, deciding relevance before reasoning begins. Here we train a compact model, Prospector, to reason over a domain knowledge graph by traversing it step by step, seeking evidence in light of what it has already found. Prospector calls a small set of tools one at a time to progressively build its context, and answers with a chain of evidence that can be checked edge by edge against what its tools retrieved from the graph. On a knowledge graph built from the semiconductor-manufacturing literature, Prospector approaches the strongest off-the-shelf models given the same graph and tools, and it far exceeds off-the-shelf models answering closed-book or from a pre-selected portion of the graph. Although trained only on the semiconductor graph, Prospector transfers to unseen knowledge graphs from three other distinct fields: bio-inspired materials, per- and polyfluoroalkyl substances (PFAS) and materials properties. Served on a single GPU, Prospector answers faster and at a fraction of the cost of a hosted service, while keeping proprietary data on the owner’s hardware. We find that the policy’s latent state is indicative of its next move on both seen and unseen graphs, and modulating the latent state can steer the policy from one behavior mode to another. These results suggest a practical LLM + KG paradigm for scientific and engineering reasoning over private data, in which a compact model trained on a graph from published literature transfers its traversal to unseen graphs that stay with their owners.
[NLP-27] Self-Regularized Flow Language Models
【速读】： 该论文旨在解决连续扩散语言模型（continuous diffusion language models）在训练过程中对去噪轨迹（denoising trajectory）学习信号利用不足的问题。尽管基于连续流的语言模型为非自回归生成提供了有前景的替代方案，但其传统训练目标依赖于干净目标词元的监督信号，导致去噪过程中的中间信息未被充分挖掘。为此，论文提出SERF（Self-Regularized Flow Language Models），其核心解决方案是引入“噪声到纯净”的自蒸馏机制：通过指数移动平均（EMA）教师模型接收学生模型在相同插值路径上更清洁的输入，并提供额外的监督信号。该框架包含两种变体：SERF-L通过对齐学生与教师在深层表示上的特征空间实现优化，而SERF-P则通过匹配输出词元分布（在学生输出层或早期层）进行优化。两种方法均在不引入外部预训练模型且不改变推理流程的前提下，增强标准训练目标。实验结果表明，该方法在三种连续语言建模范式（one-hot FLM、latent ELF 和 hyper-spherical S-FLM）中均显著提升了训练效率、生成质量及少步采样性能，尤其在OpenWebText数据集上实现了高达7.7倍的训练速度加速，MAUVE指标提升24.3%，GSM8K准确率最高提升28.8%，并可在最多64倍减少采样步数的情况下达到基线性能。这些成果验证了噪声到纯净自蒸馏作为提升连续扩散语言模型的有效策略。
链接: https://arxiv.org/abs/2610.37115
作者: Theodoros Kouzelis,Spyros Gidaris,Nikos Komodakis
机构: Archimedes, Athena RC(阿基米德，雅典研究中心); National Technical University of Athens(雅典国立技术大学); valeo.ai; University of Crete(克里特大学); IACM-Forth(克里特岛应用数学与物理研究所)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注:
Abstract:We introduce SERF (Self-Regularized Flow Language Models), a simple self-supervised framework for improving continuous diffusion language models by exploiting the denoising trajectory itself. While continuous flow-based language models offer a promising alternative to autoregressive generation, their training objectives typically rely on supervision from clean target tokens, leaving the denoising trajectory underexploited as a learning signal. SERF addresses this limitation through noisy-to-clean self-distillation: an exponential moving average (EMA) teacher receives a cleaner version of the student’s noisy input along the same interpolation path and provides additional supervision. We investigate two variants: SERF-L, which aligns the student’s intermediate representations with deeper teacher representations, and SERF-P, which aligns their output token distributions, either at the student’s output or at an early student layer. All variants augment the standard training objective without requiring external pretrained models or modifying the inference procedure. Experiments across three continuous language-model paradigms (one-hot FLM, latent ELF, and hyper-spherical S-FLM) demonstrate substantial improvements in training efficiency, generation quality, and few-step generation. In particular, SERF achieves up to 7.7x wall-clock training acceleration on OpenWebText, improves MAUVE by 24.3% and GSM8K accuracy by up to 28.8%, and matches baseline performance with up to 64x fewer sampling steps. These results establish noisy-to-clean self-distillation as an effective approach to improving continuous diffusion language models.
[NLP-28] xt Generation in Low-Resource Languages Using Human-Curated Vocabularies
【速读】： 该论文旨在解决低资源语言（low-resource languages）中因预训练数据稀缺而导致生成式文本质量受限的问题，尤其针对在大型语言模型（LLM）预训练数据中几乎或完全缺失的语言。其核心挑战在于如何在缺乏大规模语料的情况下生成连贯且符合语言规范的合成文本。该研究提出的关键解决方案是：结合人工编纂的词汇库（human-curated vocabularies）与少量人工撰写的例句（exemplars），以提供句法信号，并基于齐普夫分布（Zipfian distribution）采样候选词，从而引导模型生成更符合目标语言特征的文本。实验结果表明，启用推理能力的LLM可在极低资源语言上生成流畅文本，尤其在句子层面表现优异，在段落层面也展现出良好潜力。该方法为在数字资源极度匮乏的语言中构建可用合成数据提供了可行路径。
链接: https://arxiv.org/abs/2610.37105
作者: Michael Christlambert Sinanta(1),Muhammad Dehan Al Kautsar(2),Amanda Nurul Izzah(1),Thanh-Nhi Nguyen(1),Phan Phuong Mai Chau(2),Zheng Wei Lim(3),Fajri Koto(2) ((1) Independent Researcher, (2) Mohamed bin Zayed University of Artificial Intelligence, (3) University of Amsterdam)
机构: Mohamed bin Zayed University of Artificial Intelligence (穆罕默德·本·扎耶德人工智能大学); University of Amsterdam (阿姆斯特丹大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Synthetic text generation offers a promising approach to alleviating data scarcity in low-resource languages. However, its effectiveness is inherently limited for languages that are scarcely represented, or entirely absent, in an LLM’s pretraining data. At the same time, many such languages have human-curated vocabularies that provide rich lexical and semantic information, often accompanied by grammatical information such as part-of-speech labels. In this work, we ask whether LLMs can leverage these resources to generate coherent text in extremely low-resource and unseen languages. We introduce a framework that combines human-curated vocabularies with a small number of human-written exemplars that provide syntactic signals, while sampling candidate words according to a Zipfian distribution. Across languages with varying levels of resource availability, we evaluate the generated text using lexical diversity, off-target ratio, and human judgments of fluency. Our results show that reasoning-enabled LLMs can generate fluent text even in extremely low-resource languages, with particularly strong results at the sentence level and promising performance at the paragraph level. Our findings highlight a practical pathway for bootstrapping synthetic data in low-resource languages, expanding the amount of usable text even when existing digital resources are extremely limited. Our source code is available at this https URL.
[NLP-29] Hierarchical Low-Rank Reparameterization for Parameter-Efficient
【速读】： 该论文旨在解决参数高效微调（Parameter-Efficient Fine-Tuning, PEFT）方法中可训练参数仍存在冗余的问题，尽管这些方法已通过引入少量可训练参数实现大模型的适配，但其参数量仍有进一步压缩的空间。为此，论文提出一种新颖的分层低秩重参数化方法（hierarchical low-rank reparameterization），其核心在于将PEFT方法中的可训练参数重新组织为近似方阵形式，并通过低秩分解进行表达，且该分解可递归应用，形成多层级结构，使各层级的因子逐级压缩，从而大幅减少可训练参数数量。实验表明，该方法与LoRA、DoRA、MoSA及RaSA结合后，在保持相近可训练参数预算的情况下，性能优于原有方法；当原始方法参数量被压缩约一个数量级时，该方法仍能维持竞争力，甚至在8个常识推理基准上的平均准确率超过对应基线。
链接: https://arxiv.org/abs/2610.37068
作者: Rong Huang,Yi Xiong,Yuan-Yuan Cheng,Xiao-Ming Fu
机构: University of Science and Technology of China(中国科学技术大学)
类目: Computation and Language (cs.CL)
备注: 14 pages, 2 figures, 10 tables, including references and appendices
Abstract:Parameter-efficient fine-tuning (PEFT) methods adapt large models by introducing a small number of trainable parameters. However, we find that these parameters still contain redundancy and can be further compressed while maintaining strong task performance. To this end, we propose a novel hierarchical low-rank reparameterization method. Its core idea is to reorganize the trainable parameters of PEFT methods into an approximately square matrix and then express it via a low-rank decomposition, which can be applied recursively. This forms a multi-level structure with factors compressed level by level, drastically reducing the number of trainable parameters. In our experiments, we combine this reparameterization with LoRA, DoRA, MoSA, and RaSA. We evaluate commonsense reasoning on LLaMA-3.1-8B and Qwen2.5-7B, with additional GSM8K math reasoning experiments on LLaMA-3.1-8B. Under comparable trainable-parameter budgets, our method combined with LoRA or DoRA achieves better performance. Furthermore,when the parameter count of the original methods is compressed by approximately an order of magnitude, our method can still maintain competitive task performance and even surpass the corresponding baselines in average accuracy across the eight commonsense reasoning benchmarks.
[NLP-30] EnergyCal: Taming Loss Spikes in Muon-Family Optimizers for Sparse MoE Training
【速读】： 该论文旨在解决稀疏混合专家（Sparse Mixture-of-Experts, MoE）语言模型在使用Muon家族优化器进行训练时出现的显著损失突增（loss spikes）问题。其解决方案的关键在于提出一种轻量级规则——EnergyCal，该规则通过根据路由专家更新方向的有限步Newton-Schulz迭代方向的归一化平方Frobenius范数，对专家更新进行衰减。EnergyCal不改变谱变换（spectral transform），且无需引入持久化的优化器状态，从而在保持计算效率的同时有效抑制损失突增。实验结果表明，在40亿参数的MoE模型上，与经开发集选择的低学习率MuonH方案相比，EnergyCal-MuonH将平均阈值化损失突增负担降低了99.51%，并使最终几何平均困惑度降低4.06%；而在与标准Muon在相同名义学习率和权重衰减下对比，平均损失突增负担降低达95.98%，同时带来微小的终点性能提升。这些结果验证了EnergyCal在有无Hyperball机制的设定下均能有效提升训练稳定性。
链接: https://arxiv.org/abs/2610.37062
作者: Masato Fujitake
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 9 pages with appendices
Abstract:Sparse mixture-of-experts (MoE) language-model training can exhibit large loss spikes with Muon-family optimizers. We introduce EnergyCal, a lightweight rule that attenuates routed-expert updates according to the normalized squared Frobenius norm of their finite-step Newton-Schulz directions. EnergyCal leaves the spectral transform unchanged and adds no persistent optimizer state. We evaluate it on a 4B-parameter MoE model. Against a development-selected, lower-learning-rate MuonH recipe, EnergyCal-MuonH reduces mean thresholded loss-spike burden by 99.51% and final geometric-mean perplexity by 4.06%, with a modest throughput cost. With plain Muon at matched nominal learning rate and weight decay, the reduction in mean spike burden averages 95.98% across the two tested exponents, accompanied by small endpoint improvements. These findings support stabilization both with and without Hyperball in the evaluated setting.
[NLP-31] Exploring the Limits of the Delta Learning Hypothesis: Single-Pair Generator Biases in Direct Preference Optimization
【速读】： 该论文旨在解决当前直接偏好优化（Direct Preference Optimization, DPO）方法中依赖多模型池生成偏好对并由外部大语言模型（LLM）判别器进行排序所导致的数据构建复杂性与成本问题。其核心挑战在于：如何在不依赖外部判别器的前提下，仍能获得高质量、无偏的偏好数据以支持有效训练。论文提出的关键解决方案是基于“增量学习假设”（delta-learning hypothesis），即仅使用一对模型（一个较小模型与一个较大模型）的输出，并将较大模型的响应视为偏好项。然而，研究发现，这种单对模型方法存在系统性偏差——这些偏差源于生成器自身的特性，且与响应质量无关，却可预测偏好标签。这导致基于单对数据训练的DPO性能显著低于多模型偏好数据，尽管大模型本身生成的响应质量更高。研究进一步揭示，该性能差距源于训练过程中的奖励劫持（reward hacking）现象，即模型学会了利用生成器固有的偏差而非真实响应质量来最大化奖励。值得注意的是，仅通过增加生成器数量（即使无需外部判别器），即可部分缓解此类偏差。因此，论文强调生成器多样性（generator diversity）是构建高质量偏好数据的关键因素。
链接: https://arxiv.org/abs/2610.37058
作者: Giuseppe Curci,Stefano Simonazzi,Andrea Molinari,Andrea Zugarini
机构: Okkam(奥克卡姆); Villanova.ai; University of Trento(特伦托大学)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Direct Preference Optimization (DPO) typically relies on preference pairs generated from diverse model pools and ranked by an external LLM judge. The recently proposed delta-learning hypothesis suggests that comparable supervision can instead be obtained from a single pair of models by treating the larger model’s response as preferred. We investigate the limits of this approach across model families and scales. We show, both theoretically and empirically, that fixed model pairs introduce systematic, generator-dependent biases that are predictive of the preference label independent of response quality. Consequently, DPO trained on single-pair data consistently underperforms multi-model preference data, despite the larger generator producing higher-quality responses, and we trace this gap to training dynamics consistent with reward hacking on these biases. We further show that simply increasing the number of generators, even without an external judge, is enough to partially mitigate this effect. Our results highlight generator diversity as a critical factor in preference-data construction.
[NLP-32] Guided-RLVR: Guided Exploration in Reinforcement Learning with Verifiable Rewards
【速读】： 该论文旨在解决强化学习中可验证奖励（Reinforcement Learning with Verifiable Rewards, RLVR）在提升大语言模型（Large Language Models, LLMs）推理能力时存在的问题：尽管RLVR在数学和编程等任务上表现出显著的性能提升，但其采样轨迹往往陷入狭窄的推理路径，导致探索空间受限。为应对这一挑战，论文提出一种名为Guided-RLVR的新方法，其核心解决方案是引入一个轻量级控制器，通过在采样分布中战略性地注入语义明确的引导性思维标记（steering tokens），促使模型从多种不同的推理视角分析问题，从而有效扩展推理空间。实验结果表明，Guided-RLVR在数学推理与代码生成基准测试中均优于现有RLVR方法，同时生成更具多样性的推理轨迹。
链接: https://arxiv.org/abs/2610.37041
作者: Bo Yang,Zishuai Zhang,Weihua Li,Qinnan Zhang,Sijia Wen,Hainan Zhang
机构: 未知
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 14 pages, 7 figures, 8 tables
Abstract:Reinforcement learning with verifiable rewards (RLVR) has shown strong empirical gains in improving the reasoning ability of large language models (LLMs) on tasks such as mathematics and programming. However, recent studies have shown that RLVR rollouts often collapse onto a narrow set of reasoning trajectories, limiting the explored reasoning space. Motivated by findings that thinking tokens at critical steps serve as critical structural elements in LLM reasoning pathways, we propose Guided-RLVR, a novel approach that augments RLVR with a lightweight controller. The controller strategically injects semantically meaningful steering tokens into the sampling distribution, encouraging the model to approach problems from multiple reasoning perspectives and thereby expanding the exploration space. Experiments on mathematical reasoning and code generation benchmarks show that Guided-RLVR consistently outperforms existing RLVR methods while producing more diverse reasoning trajectories.
[NLP-33] Zooming without Zooming v2: From Seeing Details to Connecting Evidence for Generalizable Multimodal Perception
【速读】： 该论文旨在解决构建高质量、针对性的多模态训练数据以支持空间感知与多模态长上下文理解等关键任务所面临的困难与高成本问题。其核心挑战在于，这些任务本身复杂且难以直接生成标注数据。解决方案的关键在于识别并强化一个共享的基础能力——“证据连接”（evidence connection），即在视觉场景中定位相关视觉事实、将其正确关联至实体、恢复其相互关系，并整合为答案。通过聚焦于这一可高效合成的通用能力，作者提出EvidenceConnect框架，利用单幅图像中局部放大视图生成需进行视觉证据对比、聚合或链式推理的问题。为确保数据质量，采用独立多模态大模型（VLM）一致性验证参考答案，并通过区域消融测试剔除依赖特定区域的问题。基于此合成数据，使用强化学习与价值回归（RLVR）训练基础多模态大模型，得到Zooming without Zooming v2（ZwZ2）。尽管合成数据未针对特定下游任务优化，ZwZ2在细粒度与空间感知、具身空间推理、多模态长上下文理解及通用视觉推理等方面均取得显著提升，表明强化证据连接能力可有效促进更泛化的多模态感知性能。
链接: https://arxiv.org/abs/2610.37037
作者: Lai Wei,Yatai Li,Liangbo He,Jun Lan,Renjie Gu,Linhan Cao,Junkai Lin,Huijia Zhu,Weiqiang Wang,Linghe Kong,Yue Wang,Zhuosheng Zhang,Weiran Huang
机构: Shanghai Jiao Tong University (上海交通大学); Ant Group (蚂蚁集团); Zhongguancun Academy (中关村学院); Shanghai Innovation Institute (上海创新研究院); Fudan University (复旦大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Multimodal perception forms the foundation of multimodal intelligence. In particular, spatial perception and multimodal long-context understanding are two pivotal tasks that underpin embodied AI and multimodal agents. However, synthesizing targeted, high-quality training data for these tasks remains difficult and costly. Our key insight is that these tasks share an underlying capability, which we call connecting evidence: identifying relevant visual facts, binding them to the correct entities, recovering their relations, and composing them into an answer. Crucially, targeted training data are far easier to synthesize for this capability than for these tasks themselves. To strengthen this capability, we propose EvidenceConnect, an easy-to-implement data synthesis framework that uses zoomed views of related regions within a single image to generate questions requiring comparison, aggregation, or chaining of visual evidence. To ensure quality, a reference answer is retained only when independent VLMs agree on it, and region-ablation tests discard questions whose answer survives the removal of selected regions. We then train base VLMs on the resulting data with RLVR, obtaining Zooming without Zooming v2 (ZwZ2). Although the synthesized data are not tailored to any downstream benchmark, ZwZ2 improves fine-grained and spatial perception, with further gains on embodied spatial reasoning, multimodal long-context understanding, multimodal agent tasks, and general visual reasoning. These results establish EvidenceConnect as an effective framework for supervising evidence connection and suggest that strengthening this shared capability leads to more generalizable multimodal perception.
[NLP-34] V-LUMEN: Visual Lookup Memory for Embedding Scaling in Vision-Language Models NEURIPS2026
【速读】： 该论文旨在解决视觉-语言模型（VLMs）在扩展模型容量时面临的挑战，即如何在不增加模型深度、宽度或密集计算量的前提下，有效利用外部记忆机制来增强模型的推理能力。现有方法在文本领域可通过基于查找的内存实现嵌入级扩展（embedding-level scaling），但这一思路难以直接应用于视觉-语言模型，原因在于视觉特征具有连续性、空间结构复杂且依赖查询上下文，与离散的文本标记存在本质差异。为此，论文提出V-LUMEN，一种面向视觉-语言模型的视觉查找记忆系统。其核心解决方案包括：通过冻结的码本对视觉特征进行离散化；采用基于空间聚合的哈希方法构建具备空间感知能力的记忆键；并引入文本条件相关的相关性信号对记忆键进行调制；所有访问操作均为非参数化设计，检索到的嵌入通过门控残差路径融合至主干网络。实验表明，V-LUMEN在五个高推理强度基准上显著提升冻结版Qwen3-VL-2B模型性能，并实现零样本迁移至三个训练图像不重叠的基准。在参数预算相当的情况下，将部分LoRA或MoE适配器的容量转移至V-LUMEN可保持双主干模型的精度，表明外部视觉记忆构成独立于传统适配方法的能力维度。由于查找开销仅与表大小相关且仅在预填充阶段发生一次，该架构的容量主要由存储增长驱动而非每标记计算成本上升，从而实现了高效扩展。
链接: https://arxiv.org/abs/2610.37019
作者: Miso Choi,Daekeun Kim,Eunji Kim,Jungbeom Lee
机构: Korea University(高丽大学); Amazon(亚马逊)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at NeurIPS 2026
Abstract:Embedding-level scaling expands the capacity of language models through lookup-based memory rather than by increasing model depth, width, or dense computation. Extending this idea to vision-language models (VLMs) is challenging because visual features are continuous, spatially structured, and query-dependent, unlike discrete text tokens that naturally support n-gram lookup. We introduce V-LUMEN, a visual lookup memory for VLMs. V-LUMEN discretizes visual features with a frozen codebook, builds spatially aware memory keys through spatial aggregation-based hashing, and modulates them with text-conditioned relevance signals; all addressing is non-parametric, and the retrieved embeddings are fused into the backbone through a gated residual pathway. On five reasoning-intensive benchmarks, V-LUMEN substantially improves a frozen Qwen3-VL-2B backbone and transfers zero-shot to three benchmarks whose images are disjoint from the training data. At a matched parameter budget, moving part of a LoRA or MoE adapter’s capacity into V-LUMEN preserves accuracy on two backbones, indicating that external visual memory is an independent capacity axis rather than a stronger adaptation method. Because lookup cost is constant in table size and incurred only once at prefill, this capacity grows in storage rather than in per-token compute. Code is available at this https URL.
[NLP-35] he State and Fate of Multilingual Contextual Evaluation in the NLP World
【速读】： 该论文旨在解决当前多语言评估基准在覆盖范围、代表性与严谨性方面存在的系统性缺陷，以评估大语言模型是否真正具备超越英语的泛化能力。其核心问题在于：现有51个近期多语言基准虽覆盖242个数据集和219种语言，但存在显著的语言分布不均（36%的语言仅出现在单一基准中）、区域空白（如大洋洲、美洲、中亚几乎无覆盖）以及任务公平性严重失衡（低资源语言平均仅涉及1-3类任务，而高资源语言达14类）。解决方案的关键在于揭示并批判当前评估体系中普遍存在的“英语中心主义”倾向——56%的数据-语言实例为英文翻译生成，导致文化语境失真，并造成内容代表性受限；同时，基准污染（如翻译泄漏与平行语料重叠）进一步削弱了评估结果的可靠性。作者提出构建原生生成、文化根植、防污染且服务于目标语言社区的评估框架，强调评估体系应从设计源头即体现语言多样性与文化适切性，以实现更真实、公正和有效的跨语言模型评估。
链接: https://arxiv.org/abs/2610.36990
作者: Manan Uppadhyay,Himanshu Beniwal,Prashant Kodali,Sunayana Sitaram
机构: Microsoft Research India(微软研究院印度); IIT Gandhinagar(印度理工学院甘地纳加尔分校)
类目: Computation and Language (cs.CL)
备注: Published at Conference on Language Modelling, 2026. Can be found at this https URL
Abstract:Multilingual evaluation benchmarks are the primary instrument for assessing whether large language models generalize beyond English, yet the adequacy of these benchmarks has received little systematic scrutiny. We present a data-driven audit of 51 recent multilingual benchmarks spanning 242 datasets and 219 languages, organized around three pillars: coverage, representativeness, and rigor. Our analysis reveals that coverage is wide but thin with 36% of evaluated languages appearing in only a single benchmark, entire regions (Oceania, the Americas, Central Asia) are near-zero, and a stark task equity gap leaves low-resource languages evaluated on only 1-3 task categories versus 14 for high-resource languages. Representativeness is structurally compromised: translation from English remains the dominant construction strategy where 56% of all dataset-language instances are translated introducing artifacts and English-centric framing, while culturally grounded content is concentrated in a handful of community-driven benchmarks with narrow language scope. The ecosystem thus forces a trade-off between breadth and validity. Rigor is undermined by benchmark contamination, including translated benchmark leakage and parallel corpus overlap that evade surface-form detection. We synthesize these findings into concrete recommendations for building evaluation frameworks that are natively constructed, culturally grounded, contamination-aware, and designed to serve the communities whose languages they claim to evaluate. The dashboard with all the metadata collected can be found at this https URL.
[NLP-36] AutoVend: Benchmarking LLM Agents on Long-Horizon Autonomous Vending Operations
【速读】： 该论文旨在解决生成式 AI（Generative AI）代理在在线到线下（O2O）商业场景中，如何实现从线上决策到线下可持续经济回报转化的评估难题。现有研究多聚焦于单次交易的成功率，而忽视了长期运营中的实际执行成本与经济可持续性。其解决方案的关键在于提出 AutoVend——一个基于数据驱动模拟器的基准测试框架，通过构建包含共享仓库、采购前置时间、付费补货、库存过期及持续运营成本等真实约束的虚拟环境，使一个大型语言模型（LLM）代理在180个模拟日内管理20台分布在五类站点的自动售货机。该框架不仅评估代理的经济回报，还深入分析其在站点协同、执行可行性与策略迭代方面的表现，揭示出尽管部分模型在单一指标上表现优异，但因库存与执行成本侵蚀收益，整体经济绩效反而不及基于规则的基线模型，凸显了将反馈转化为有效政策修订与确保动作可执行性的核心挑战。
链接: https://arxiv.org/abs/2610.36980
作者: Qiming Shi,Yikun Li,Beidi Luan,Rui Sun,Yan Ma,Jing Li,Daxin Jiang,Zuo Bai
机构: 未知
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Large language model (LLM) agents are extending agent-mediated commerce on the Web from shopping for consumers to operating businesses for sellers. In online-to-offline (O2O) commerce, decisions are issued through Web-based platforms but realized through costly physical execution. Evaluating such agents requires going beyond individual transactions to ask whether decisions made online turn into sustained economic returns. We introduce AutoVend, a benchmark in which one LLM agent manages 20 vending machines across five site types through a Web-based operating interface for 180 simulated days. Its data-grounded simulator combines a shared warehouse, procurement lead times, paid replenishment, inventory expiration, and recurring operating costs. We evaluate five LLM agents against a rule-based inventory-control baseline and diagnose their site coordination, execution, and performance over time. Four of five evaluated LLM agents fall short of this baseline, and returns per dollar of inference cost rank models differently from returns alone. Operational analyses show that gains in individual metrics do not consistently yield stronger economic growth, as inventory and execution costs erode these gains. Trajectory analysis further reveals difficulties in grounding feasible actions, managing inventory commitments, and converting feedback into effective policy revisions.
[NLP-37] Mining the Nineteenth Century for the Thirteenth: Register-Scale Forced Alignment and HTR Bootstrapping via Legacy Regesta and Antiquarian Editions
【速读】： 该论文旨在解决教皇霍诺里乌斯三世（Pope Honorius III）手稿册籍中未编纂部分的古籍文字识别（HTR）难题，尤其针对现存仅覆盖在任第一至二年的现代专家转录文本之外的大量未转录区域。其核心挑战在于如何利用十九世纪学术成果（如Pressutti的编年史与部分编辑）来辅助生成高质量的自动识别结果。解决方案的关键在于构建一个分层协同机制：在正字法层级，通过将原始编辑文本投影至教廷标准拼写并结合现代源文本重播，实现零标签自适应，使第二年文本的词错误率（WER）降低1.50点（95%置信区间 -1.84 至 -1.15），缩小与理想模型33%的差距；在词汇层级，利用历史语料库中的遗留词汇表覆盖82.2%的未登录词，并结合预训练中世纪命名实体识别（NER）模型蒸馏得到的识别头，在专家验证基准上达到F1值82.1%，性能媲美后处理的最佳标注器；在拓扑层级，借助Pressutti编年史、结尾条款及日期罗马化转换，成功定位73.9%的条目起始位置，且在第203页后估计的字符错误率（CER）为7.80%。研究表明，尽管古籍学术成果无法替代现代转录，但可有效引导、校验与分割未转录册籍的HTR流程。
链接: https://arxiv.org/abs/2610.36956
作者: Tamas Kovacs,Daniel Luger
机构: University of Graz(格拉茨大学); Department of Digital Humanities(数字人文系); Institut für Österreichische Geschichtsforschung(奥地利历史研究所); University of Vienna(维也纳大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL)
备注: 10 pages
Abstract:Handwritten text recognition (HTR) for the registers of Pope Honorius III depends on modern expert transcriptions, which exist only for regnal years 1-2 of Reg. Vat. 9. For the remaining registers, nineteenth-century scholarship offers Pressutti’s regesta and partial editions. On fol. 1-203, where both layers exist, we measure what this legacy layer can contribute, on transcriptions that we first audit against it, and find a division of labor across three tiers. At the orthographic tier, raw editions are at best neutral for adapting a recognizer to the second regnal year and harmful without replay, whereas projection onto curial spelling with replay of modern source lines turns them into zero-label adaptation targets: pooled over all 134 year-2 sides, the WER falls by 1.50 points (95% CI -1.84 to -1.15), closing 33% of the gap to an oracle; the spelling changes at the year split, through the scribe or the transcriber, whereas the hand changes visibly only at fol. 191r. At the lexical tier, the corpus-wide legacy lexicon contains 82.2% of the out-of-vocabulary words of unedited entries and flags recognition errors with 66.7% precision, and labels from pre-trained medieval NER models, distilled into the recognizer’s NER head, reach F1 82.1% on an expert validated benchmark, level with the best tagger applied after recognition. At the topological tier, Pressutti’s regests, the closing datum clauses, and the regests’ dates converted to the chancery’s Roman form find 73.9% of entry starts in the recognized line stream, and edition agreement estimates a CER of 7.80% beyond fol. 203. Antiquarian scholarship thus cannot replace modern transcription, but it can steer, check, and segment HTR across the untranscribed registers.
[NLP-38] Self-Keyed In-Parameter Memorization for External Memory-Free Dialogue Recall in Frozen Language Models
【速读】： 该论文旨在解决在无外部记忆且缺乏未来问题先验知识的环境下，如何将对话内容以参数化方式存储于大语言模型内部，并实现基于模型参数的高效检索与问答。其核心挑战在于克服传统方法依赖外部存储所导致的延迟与系统复杂性，同时确保模型在后续对话中能够准确响应关于历史对话的提问。解决方案的关键是提出自键式参数内记忆机制（Self-Keyed In-Parameter Memorization, SKIM），一种闭式方法：模型在每一轮对话中自动生成多样化的疑问句作为“键”（key），并将这些键嵌入模型参数中，用于后续匹配用户问题；通过生成丰富多样的问题形式，提升了对不同表述方式的鲁棒性，使答案可仅从模型参数中直接召回，无需任何外部存储支持。实验证明，该方法在五种语言模型上均能有效回忆大部分存储的对话轮次，实现了完全基于模型内部参数的记忆与推理。
链接: https://arxiv.org/abs/2610.36948
作者: XinYu Piao,Myunghoon Ryu,Joongheon Kim,Jong-Kook Kim
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:A large language model maintains conversational content in an external storage that is outside the model and grows as the dialogue continues. This work assumes an environment where a language model has no external memory and no prior knowledge of the future question or its answer. The motivation is to research whether a dialogue can be stored in the model parameters and later answer a question about it. This work proposes self-keyed in-parameter memorization (SKIM), a closed-form method in which the model generates its own questions about each dialogue turn and writes those questions into its parameters as the keys that later retrieve the turn’s content. Generating many varied questions for each turn allows a later user question to match one even when phrased differently, and its answer is recalled from the parameters alone. Evaluated on five language models, SKIM recalls most stored dialogue turns from later user questions, and it answers from the model parameters without any external storage.
[NLP-39] JevRec: Linguistic Encoding with Jev-like Decision Modeling for LLM -based Generative Recommendation
【速读】： 该论文旨在解决生成式推荐（Generative Recommendation, GR）中因统一编码长度导致的效率与语义表达失衡问题：现有方法对所有物品分配相同长度的语义标识符（Semantic Identifier, SID），导致热门物品频繁使用细粒度编码造成资源浪费，而为降低开销缩短编码长度又会压缩长尾物品的语义表达。其解决方案的关键在于提出JevRec，通过受语言学启发的可变长度词（variable-length words）编码机制，使高频物品获得更短的SID，同时保留长尾物品更丰富的语义描述。为解耦序列级长度预测与令牌级内容生成之间的优化冲突，JevRec采用双头架构——专用头部预测长度分布，另一独立头部在给定长度条件下生成内容，二者共享交互历史表示，从而实现高效且一致的联合建模。实验表明，JevRec在三个公开数据集上均优于现有基线模型，并展现出更低的训练与推理时间及更小的梯度冲突。
链接: https://arxiv.org/abs/2610.36937
作者: Qiyong Zhong,Jiajie Su,Ming Yang,Li Zhang
机构: 未知
类目: Computation and Language (cs.CL)
备注:
Abstract:Generative recommendation (GR) represents items as semantic identifiers (SIDs) and autoregressively generates the next item’s SID from interaction history. However, most methods assign the same SID length to all items despite highly unequal interaction frequencies. This uniform design repeatedly incurs the cost of fine-grained codes for popular items, while shortening codes to reduce this cost also compresses the semantic representations of long-tail items. Inspired by the compact forms of frequent expressions in natural language, we propose JevRec to encode items as variable-length words and generate the next item from interaction context. Linguistic encoding retains frequency-dependent semantic prefixes from shared codebooks, producing shorter SIDs for popular items and richer descriptions for long-tail items. Yet predicting a sequence-level length and generating token-level content impose different optimization requirements that may conflict under a shared output head. Our Jev-like decision modeling component therefore predicts a length distribution with a dedicated head and generates content conditioned on length through a separate head while sharing the interaction-history representation. Experiments on three public datasets show that JevRec outperforms all compared baselines across the evaluated recommendation metrics. Further analyses show lower training and inference time than TIGER and reduced gradient conflict between length and content prediction.
[NLP-40] DocXEval: A Systematic Evaluation of Explainability in Multimodal Document Understanding
【速读】： 该论文旨在解决多模态文档理解中生成式AI模型的可解释性缺失问题，尤其是在高风险领域安全部署时面临的“黑箱”挑战。尽管在视觉与语言任务的可解释人工智能（XAI）方面已取得进展，但针对多模态文档理解（包含文本、版面与图像）场景的专门化XAI研究仍严重不足。为此，本文提出首个综合性可解释性框架\ourframework，其关键在于：一方面将现有XAI方法与评估指标适配至多模态文档理解场景，另一方面引入高效的多目标适应机制，以支持语义实体识别等特定任务；通过系统评估13种XAI方法在3类任务、8个数据集、4种模型及17项评估指标下的表现，发现仅有少数方法在各项指标上表现一致，其中遮挡法（Occlusion）被证实为整体最忠实且可解释的方法。进一步的元评估揭示了部分忠实度与复杂度指标间存在中等到强一致性，但也识别出若干不可靠指标。最终，基于跨数据集的诊断分析，揭示了各模态对不同任务与标签的贡献差异、目标实体周围的空间归因模式以及潜在的数据集偏差。本工作不仅为多模态文档理解领域的未来XAI研究提供了基准与实践指南，还公开发布代码以保障可复现性。
链接: https://arxiv.org/abs/2610.36929
作者: Saifullah Saifullah,David Dembinsky,Stefan Agne,Andreas Dengel,Sheraz Ahmed
机构: German Research Center for Artificial Intelligence (DFKI)(德国人工智能研究中心); University of Kaiserslautern-Landau (RPTU)(凯泽斯劳滕-兰道大学); DeepReader GmbH(DeepReader有限公司)
类目: Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Multimodal transformer models have driven unprecedented performance breakthroughs in modern document understanding. However, the inherent black-box nature of these models remains a major obstacle to their safe deployment in high-stakes sectors. Despite significant progress in explainability for vision and language tasks, explainable AI (XAI) research tailored to multimodal document understanding has received little attention. To address this gap, we present \ourframework, the first comprehensive explainability framework that adapts existing XAI methods and evaluation metrics to the multimodal (text, layout, and image) document understanding context, while introducing efficient multi-target adaptations of both methods and metrics for domain-specific tasks such as semantic entity recognition. Using \ourframework, we systematically evaluate 13 XAI methods across 3 document understanding tasks, 8 datasets, 4 models, and 17 evaluation metrics, and find that only a few methods perform consistently across all evaluation criteria, with Occlusion emerging as the most faithful and interpretable method overall. Through a dedicated meta-evaluation, we observe moderate to strong agreement among several faithfulness and complexity metrics while identifying certain ones that produce unreliable results. Finally, through multiple dataset-wide diagnostic analyses, we uncover key insights into model behavior, including the contribution of individual modalities across tasks and labels, spatial attribution patterns around target entities, and potential dataset biases. Overall, this work serves as both a benchmark and a practical guide for future XAI research in multimodal document understanding. To support reproducibility, we publicly release our code at this https URL.
[NLP-41] rawMem: Reframing Agent Memory as Task-Oriented Role-Adaptive Workspaces
【速读】： 该论文旨在解决大语言模型（Large Language Model, LLM）代理在长期交互中依赖记忆系统时，现有方法仅基于记忆与当前查询的相关性进行检索和使用所导致的局限性。传统记忆系统未能充分捕捉记忆在任务求解中的多样化角色，例如直接支持答案或作为与其他记忆关联的桥梁，而这些关联可能语义相关性较弱但信息关键。为此，论文提出将代理记忆系统重新构想为面向任务的角色自适应工作区（Task-Oriented Role-Adaptive Workspaces），其核心在于依据记忆在特定任务中的功能角色及其相互关系进行选择与组织。解决方案的关键是提出TrawMem框架，该框架将交互历史划分为局部线程，并通过角色标注与协同选择机制构建任务导向的工作空间，实现对互补记忆的联合识别与利用。实验结果表明，TrawMem在LoCoMo和MemGallery数据集上均显著优于现有代表性记忆系统，宏平均F1提升2.23–3.14个百分点；在LoCoMo上，相较于SimpleMem，TrawMem在Qwen3.5-9B和DeepSeek-V4-Flash上的LLM-Judge得分分别提升4.07和4.31个百分点，同时分别减少约12.8%和5.1%的上下文令牌消耗，验证了其高效性与有效性。
链接: https://arxiv.org/abs/2610.36918
作者: Mingyu Zhang,Zixu Li,Zhiheng Fu,Zhiwei Chen,Jinhe Bi,Yupeng Hu
机构: 未知
类目: Computation and Language (cs.CL)
备注: 31 pages, 18 figures, 3 tables
Abstract:Large language model (LLM) agents rely on memory systems to leverage long-term interactions, yet existing systems typically retrieve memories based on their relevance to the current query and use them as contextual evidence. However, relevance alone does not characterize how a memory contributes to task solving: the same memory may serve different roles across tasks, such as directly supporting the answer or linking to complementary memories with weaker semantic relevance but necessary information. We therefore reframe agent memory systems as Task-Oriented Role-Adaptive Workspaces, where memories are selected and organized by their task-oriented roles and relations. Based on this view, we propose TrawMem, which organizes interaction histories into local threads and constructs task-oriented workspaces by assigning roles and jointly selecting complementary memories. Experiments on LoCoMo and MemGallery with four open- and closed-source backbones show consistent improvements over representative memory systems, with macro-F1 gains of 2.23-3.14 points. On LoCoMo, TrawMem improves the LLM-Judge score over SimpleMem by 4.07 and 4.31 percentage points on Qwen3.5-9B and DeepSeek-V4-Flash, respectively, while requiring approximately 12.8% and 5.1% fewer context tokens, respectively.
[NLP-42] Developing a Speech Language Model for Basque
【速读】： 该论文旨在解决少数语言在生成式语音交互系统发展滞后的问题，尤其针对巴斯克语这一低资源语言在多模态交互中的技术瓶颈。其核心解决方案在于构建首个面向巴斯克语的开源语音语言模型（Speech Language Model），采用模块化架构设计，以LLaMA-Omni为框架，引入Latxa模型作为主干网络，并通过微调多个模块实现从文本到语音的端到端低延迟交互。为缓解数据稀缺问题，研究者将英文VoiceAssistant-400K语料库翻译成巴斯克语，并利用巴斯克语语音合成（Text-to-Speech, TTS）技术生成对应音频数据，从而构建训练语料。实验结果表明，该系统在内容质量（3.86/5）、语音可理解性（词错误率9.94%）及音频质量（UTMOS 3.07）方面表现良好，验证了基于模块化架构与合成数据的方法可在低资源语言中有效构建高性能语音交互系统。
链接: https://arxiv.org/abs/2610.36911
作者: Andoni Sudupe,Inma Hernaez,Ander Arriandiaga,Christoforos Souganidis,Ibon Saratxaga,Eva Navas
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Audio and Speech Processing (eess.AS)
备注: IberSPEECH2026 - Jornadas en Tecnologias del Habla and Iberian SLTech Workshop Events
Abstract:Although the recent development of LLMs has paved the way for multi-modal interaction, minority languages face difficulties keeping pace. This work presents the first open Speech Language Model for Basque, designed for low-latency, end-to-end spoken interaction. Based on the LLaMA-Omni architecture, the system integrates a Latxa model as its backbone, transferring its Basque capabilities to the spoken modality by fine-tuning several modules. To address data scarcity, the English VoiceAssistant-400K corpus was translated into Basque, and the audio was synthesized using a Basque TTS. Evaluation results confirm the system’s viability, achieving 3.86/5 in content quality, a 9.94% WER for spoken intelligibility, and a 3.07 UTMOS for audio quality. This work demonstrates that advanced spoken interaction systems can be developed using modular architectures and synthetic data even in low-resource languages.
[NLP-43] Nothing Left Behind: Finding Evidence from a Large Randomized Evaluation Corpus
【速读】： 该论文旨在解决传统基于标签的检索方法在信息检索中存在漏检与过检的问题，即依赖预定义词汇表进行标记的搜索方式难以捕捉未被标签覆盖的相关内容，且当标签过于宽泛时会导致大量无关结果。其解决方案的关键在于提出一种两阶段架构：首先利用文本嵌入（text embeddings）进行语义层面的广泛候选项召回，实现对相关性较高的非结构化内容的高效发现；随后通过语言模型（language model）对召回结果进行精细化筛选，保留真正相关的证据。实验在包含7,354个随机对照试验的语料库中验证，相较于传统标签搜索仅能发现系统综述所需证据的45%，该方法可达到92%的召回率，尽管查询成本随搜索深度增加，但显著提升了检索的全面性与准确性，验证了结合广度召回与查询特定过滤的嵌入-语言模型协同框架的有效性。
链接: https://arxiv.org/abs/2610.36910
作者: Giuliano Martinelli,Abelardo Carlos Martinez Lorenzo,Satvik Garg,Linxi Wang,Riccardo Orlando,Piriyakorn Piriyatamwong,Samuel Paul Fraiberger
机构: Development Impact AI Lab, World Bank Group(世界银行集团)
类目: Computation and Language (cs.CL)
备注: 7 pages, 3 figures
Abstract:Many AI systems need to find every relevant items in a large collection and then organize what they found. The standard way to do this is to tag each item with labels from a curated vocabulary and search on those labels. Those labels are useful for organizing results, but searching on them misses items that do not fit a label and returns too much when labels are broad. We compare this approach with an embedding-based approach: find candidates by meaning using text embeddings, then have a language model keep only the relevant ones. In a corpus of 7,354 randomized trials, label search found 45% of the evidence a systematic review would include, while the new approach found 92%, at a cost per query that rises with search depth. These results support a two-stage architecture that combines broad candidate retrieval with query-specific evidence filtering.
[NLP-44] HiLo-RD: Cross-Budget Self-Distillation for Masked Diffusion Language Models
【速读】： 该论文旨在解决在生成预算受限条件下，掩码扩散语言模型（MDLM）因过早确定输出片段而导致推理路径偏离正确方向的问题。其核心挑战在于：MDLM需同时决定生成内容与生成位置，而在有限预算下，模型可能在未完成中间推理步骤前即提前固定答案片段，这些片段作为后续预测的上下文，会误导后续推理过程。现有方法无法有效迁移大窗口推理中所体现的更优推理轨迹。为此，论文提出HiLo-RD（High-to-Low Budget Representation Distillation）方法，其关键在于：通过将大窗口教师状态与裁剪后的小窗口视图配对，并在已确定位置对齐其隐藏表示，从而在无需外部教师的情况下，直接监督令牌预测与承诺顺序。实验表明，该方法在数学推理基准（如GSM8K）上显著提升低预算下的准确率（提升9.4个百分点），同时缩短生成长度约25%，且在代码生成任务中无需领域微调即可获得性能增益。注意力诊断与固定预算顺序干预结果验证了该机制的有效性，证明了大窗口推理可有效提升小窗口解码性能。
链接: https://arxiv.org/abs/2610.36901
作者: Xiang Li,Chengting Yu,Jungang Li,Yilun Zheng,Junzhuo Li,Xingyao Zhang,Wenbo Su,Xuming Hu
机构: The Hong Kong University of Science and Technology (Guangzhou)(香港科技大学(广州)); Zhejiang University (浙江大学); Nanyang Technological University (南洋理工大学); Alibaba Group (阿里巴巴集团); The Hong Kong University of Science and Technology (香港科技大学)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: 20 pages, 9 figures, 3 tables; Under review
Abstract:We propose HiLo-RD (High-to-Low Budget Representation Distillation) to improve masked diffusion language models (MDLMs) under a limited generation budget. Unlike autoregressive models, MDLMs choose both what to generate and where to commit it. This creates two coupled challenges: predicting useful content before reasoning is complete, and deciding when to fix answer tokens that shape later predictions. We identify budget-induced premature commitment: even when a correct solution fits, the model can fix answer fragments near the window boundary before resolving intermediate steps. These fragments become context for later predictions and redirect subsequent reasoning. Existing post-training methods do not directly transfer the stronger trajectories that the same model follows with a larger window. HiLo-RD addresses this gap by (1) pairing large-window teacher states with cropped small-window views and (2) matching their hidden representations at committed positions. This supervision guides both token prediction and commitment order without an external teacher. Across mathematics benchmarks, HiLo-RD improves low-budget accuracy and shortens responses; these gains extend to code generation without code-domain training. On GSM8K, accuracy rises by 9.4 percentage points as mean readable length falls by about 25%. Attention diagnostics and fixed-budget order interventions support the proposed mechanism. Together, these results show that large-window reasoning can improve small-window decoding.
[NLP-45] RubricAtlas: Mining Type-Level Rubrics from Expert Reports to Evaluate Deep Research Agents
【速读】： 该论文旨在解决生成式深度研究代理（Deep Research Agents）在多步网络搜索与信息整合后，其分析报告的评估缺乏对整体分析结构有效性的衡量问题。现有方法如查询特定的二元评分标准虽能明确个别要求，但任务级检查清单往往忽视了分析框架的完整性。为此，论文提出RubricAtlas框架，通过提炼同一研究类型下不同对象的专家报告中的共性分析需求，构建类型级（type-level）分析规范，并围绕反复出现的分析性问题组织详细检查点。该框架的关键在于将二元检查点与报告段落关联，实现对分析完整性与论证有效性分别评估，同时确保各任务共享统一的分析标准而不预设答案。基于3,079份中文金融研究报告构建的RubricAtlas覆盖99种研究类型，实验证明单份报告仅覆盖其类型约一半的分析问题，且来自不同参考报告的要求选择会显著影响评估得分与代理排名。在十名代理的评估中，两种评判流程达到κ = 0.71的检查点一致性与τ = 0.87的代理排名一致性；控制删除实验显示受影响分析项得分下降，而填充操作无明显增益，验证了评估体系的有效性与可追溯性。结果表明，类型级评分体系能够可靠比较深度研究输出的分析完整性，且决策可追溯至具体报告证据。
链接: https://arxiv.org/abs/2610.36878
作者: Delin Wang,Sinuo Wang,Rui Sun,Beidi Luan,Qiming Shi,Jing Li,Daxin Jiang,Zuo Bai
机构: StepFun(步聚科技); Renmin University of China(中国人民大学); University of Adelaide(阿德莱德大学); Zhejiang University(浙江大学); FinStep
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Deep research agents conduct multi-step Web searches and synthesise retrieved information into analytical reports. Evaluating these outputs requires checking which analyses they deliver, beyond the presence of relevant facts. Query-specific binary rubrics make individual requirements explicit, but task-level checklists may leave the overall analytical structure underexamined. We introduce RubricAtlas, which derives type-level analytical requirements from expert reports on different objects within the same research type and organises detailed checks around recurring analytical questions. The requirements give related tasks a shared standard for what to analyse, without prescribing their answers. Our framework links binary checkpoints to report passages and evaluates completeness separately from argument validity. Built from 3,079 Chinese financial research reports, RubricAtlas covers 99 research types. Reference-report selection experiments show that individual reports cover only about half of their type’s analytical questions, and that selecting requirements from different reports changes evaluation scores and agent rankings. Across ten agents, two judge pipelines achieve checkpoint agreement of \kappa = 0.71 and agent-ranking agreement of Kendall \tau = 0.87. Controlled deletions reduce scores on the affected analyses, while padding produces no detectable increase. These findings support type-level rubrics for comparing the analytical completeness of Web deep research outputs, with individual decisions traceable to report evidence.
[NLP-46] GUI-JEPA: Candidate-Conditioned JEPA Latents as a Fast Action Policy for Mobile GUI Agents
【速读】： 该论文旨在解决现有移动图形用户界面（GUI）智能体在执行动作时存在的效率低下与资源消耗过高的问题。当前主流方法通常基于70亿参数规模的视觉-语言模型（VLM），采用逐标记解码的方式生成动作，即使引入候选动作评分机制，仍需持续调用大型语言模型或VLM进行推理，导致延迟高、内存占用大。为此，本文提出GUI-JEPA，据我们所知，这是首个基于联合嵌入预测架构（JEPA）构建的无解码器型GUI智能体策略。其核心创新在于利用预训练于未标注GUI截图数据的JEPA编码器，在潜在空间中完成动作选择：通过一个受动作和区域条件约束的潜在预测器，将屏幕潜在表示转换为每个候选动作对应的单一潜在向量，从而实现所有候选动作在单次批处理中被轻量级头部统一评分，无需任何文本解码过程。进一步研究表明，该评分器依赖于这一潜在路径，且预测器的训练目标即为最终决策本身——随机初始化的预测器性能与基于屏幕动态预训练的版本相当。在AndroidControl基准测试中，GUI-JEPA在相同GPU上相比微调后的Qwen2.5-VL-7B智能体提速2.8至4.0倍，同时仅使用其6%的参数量即可达到其86%-92%的分布内准确率，显著提升了推理效率与模型轻量化水平。
链接: https://arxiv.org/abs/2610.36877
作者: Bingchao Wang,Yongcheng Jing,Shunyu Liu,Chengyu Wang,Jun Huang,Jie Yang,Wei Liu,Dacheng Tao
机构: Shanghai Jiao Tong University (上海交通大学); Nanyang Technological University (南洋理工大学); Alibaba Cloud Computing (阿里云计算)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV)
备注: 26 pages, 11 figures, 16 tables
Abstract:Mobile GUI agents are typically fine-tuned 7B-scale vision-language models (VLMs) that generate each action as text. However, these VLMs decode every action token by token, and even agents that score candidate actions keep a large language model or VLM in the loop, making them slow and memory-hungry. To address these issues, we propose GUI-JEPA, to our knowledge the first decoder-free GUI agent policy built on a Joint-Embedding Predictive Architecture (JEPA), which selects actions in the latent space of a JEPA encoder pretrained on unlabelled GUI screenshots. Its action- and region-conditioned latent predictor turns the screen latent into one latent per candidate action, letting lightweight heads score all candidates against the instruction in a single batched pass, without decoding a token. We further show that the scorer relies on this latent pathway and that the predictor is trained for the decision itself: a randomly initialised predictor does just as well as one pretrained on screen dynamics. On AndroidControl, GUI-JEPA is 2.8-4.0x faster on the same GPU than a Qwen2.5-VL-7B agent we fine-tune with the same inputs, while reaching 86-92% of its in-distribution accuracy with 6% of its parameters. Code is available at this https URL
[NLP-47] AdaGround: Token-Normalized Adaptive Zooming with Verifier-in-the-Loop Refinement for GUI Grounding
【速读】： 该论文旨在解决高分辨率屏幕上小目标的图形用户界面（GUI）定位失败问题，现有无训练（training-free）缩放方法普遍存在缩放比例与目标尺寸不匹配、盲目接受缩放结果且无法修正错误定位等缺陷。其解决方案的关键在于提出一种名为AdaGround的无训练框架，核心创新包括：（1）令牌归一化自适应缩放（Token-Normalized Adaptive Zooming），依据模型对目标尺寸的自估计动态调整缩放倍数、视场范围及缩放深度；（2）环内验证器精炼机制（Verifier-in-the-Loop Refinement），引入标记点击式“是/否”验证器以决定是否保留或拒绝每次缩放结果，选择最优缩放链路，触发重新定位，并通过存在性查询判断何时放弃预测；（3）组合结构定位（Compositional Structure Grounding），利用表头信息组合生成电子表格单元格的语义定位。在Qwen3-VL模型上，AdaGround显著超越现有无训练缩放方法及众多有训练的GUI模型，在OSWorld-G和ScreenSpot-Pro数据集上相较基线模型提升7.4至20.6个百分点，且性能增益可泛化至未见的UI-Vision测试集。
链接: https://arxiv.org/abs/2610.36874
作者: Bingchao Wang,Yongcheng Jing,Chengyu Wang,Jun Huang,Jie Yang,Wei Liu,Dacheng Tao
机构: Shanghai Jiao Tong University (上海交通大学); Nanyang Technological University (南洋理工大学); Alibaba Cloud Computing (阿里云计算)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 24 pages, 5 figures, 14 tables
Abstract:GUI grounding on high-resolution screens fails mostly on small targets, and zooming in is a common training-free remedy. However, most existing zooming pipelines do not match the zoom to the target’s size, accept the zoomed answer unconditionally and never revisit a wrong answer. To address these issues, we propose AdaGround, a training-free framework with three components: (1) Token-Normalized Adaptive Zooming, which sets the magnification, field of view and zoom depth of each sample from the model’s own estimate of the target size; (2) Verifier-in-the-Loop Refinement, which uses one marked-click Yes/No verifier to keep or reject each zoom, choose between zoom chains, trigger re-grounding and, with an existence query, decide when to abstain; and (3) Compositional Structure Grounding, which composes spreadsheet cells from their headers. On Qwen3-VL models, AdaGround surpasses state-of-the-art training-free zooming methods and many trained GUI models, improving accuracy over the backbone by 7.4 to 20.6 points on OSWorld-G and ScreenSpot-Pro, and its gains carry over to held-out UI-Vision. Code is available at this https URL
[NLP-48] Weakness Localization Is Not Enough: State-Aware Recursive Knowledge Improvement for Large Language Models
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在知识改进过程中存在的“知识状态混淆”（knowledge-state aliasing）问题，即同一知识缺陷可能由多种不同的操作性知识状态导致：知识不可用、可用但难以可靠访问，或虽可访问但在下游任务中无效。现有方法依赖评估定位知识弱点，但仅凭弱点定位无法充分指导具体改进策略的选择，且模型更新可能改变其内在知识状态，使先前有效的干预措施在后续轮次失效。为此，论文提出递归知识改进（Recursive Knowledge Improvement, RKI），其核心在于将知识建模为图结构单元，构建一个状态感知的序列化干预框架。该框架在每一轮中通过诊断评估识别知识弱点，结合诊断证据与干预历史推断当前模型的操作性知识状态，进而选择适配的干预策略，并在更新后重新评估以决定下一步行动。实验表明，RKI在HotpotEval和SafeInternalEval基准上平均优于现有方法18.41分；进一步分析证实，知识状态信息能提供超越弱点定位的干预指导价值，且在模型更新过程中即使表层缺陷未消失，知识状态仍可能发生演变。
链接: https://arxiv.org/abs/2610.36811
作者: Runze Jiang,Yan Wang,Boran Jin,Bingyu Zhu,Longtao Huang
机构: Peking University (北京大学); Alibaba Group (阿里巴巴集团); Delft University of Technology (代尔夫特理工大学)
类目: Computation and Language (cs.CL)
备注: 18 pages, 8 figures
Abstract:Knowledge improvement in large language models is shifting from predefined knowledge incorporation pipelines toward evaluation-guided updating. Existing methods use evaluation to localize model-specific knowledge weaknesses and guide targeted post-training. However, identifying what to improve does not necessarily provide sufficient information for deciding how to improve it. The same localized weakness can be compatible with distinct operational knowledge states: the relevant knowledge may be unavailable, available but difficult to access reliably, or accessible but ineffective for downstream use. We call this ambiguity knowledge-state aliasing. When compatible states favor different interventions, greater diagnostic resolution alone is insufficient for intervention selection. Moreover, model updates can change the underlying knowledge state before the observed weakness disappears, such that an intervention appropriate in one round may no longer address the bottleneck in the next. To address these challenges, we propose Recursive Knowledge Improvement (RKI), which formulates knowledge improvement as a sequential, state-aware intervention process. RKI organizes knowledge into graph-structured units that provide a shared basis for diagnostic evaluation and targeted intervention across rounds. At each round, RKI localizes knowledge weaknesses, estimates the model’s operational knowledge state from diagnostic evidence and intervention history, selects a corresponding intervention, and re-evaluates the updated model before determining the next intervention. On HotpotEval and SafeInternalEval, RKI outperforms existing knowledge incorporation methods by 18.41 points on average. Further analyses show that knowledge states provide additional guidance for intervention selection beyond weakness localization and can change across model updates even when the observed weakness persists.
[NLP-49] Latent Visual Contrastive Reasoning with Multi-Branch Exploration
【速读】： 该论文旨在解决多模态大语言模型在视觉问答任务中进行复杂视觉推理时，因依赖单一线性推理路径而难以有效处理多个相关视觉假设或竞争性证据区域的问题。其核心挑战在于如何在视觉潜在空间中建模并区分多种可能的推理路径，以提升对复杂视觉证据的定位与比较能力。解决方案的关键是提出潜在视觉对比推理（Latent Visual Contrastive Reasoning）框架，该框架通过生成多个受问题条件约束的潜在视觉分支（即软假设），在图像标记层面构建多路径推理结构；同时，利用边界框标注的证据信息构建目标证据表示，并设计一种软对比分支-证据对齐目标，促使与真实证据相关的分支与标注区域对齐，同时与负样本证据区域形成对比，从而增强推理的准确性与鲁棒性。该方法在Qwen2.5-VL骨干网络上实现，并通过标准监督微调训练，在感知密集型视觉推理基准上展现出优于基础模型及单路径潜在推理基线的性能。
链接: https://arxiv.org/abs/2610.36809
作者: Kegeng Tang,Jingbo Wang,Yangming Zhang,Huanshu Zhang,Shaogang Ren,Zihao Wang
机构: University of Tennessee at Chattanooga (田纳西大学查塔努加分校); University of Texas at Arlington (得克萨斯大学阿灵顿分校); The Pennsylvania State University (宾夕法尼亚州立大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL)
备注: Accepted at IEEE BigData 2026
Abstract:Multimodal large language models have made substantial progress in visual question answering, but complex visual reasoning often requires locating and comparing multiple pieces of question-relevant visual evidence. Existing methods that move intermediate reasoning into the visual latent space usually follow a single linear reasoning trajectory, which can be limiting when multiple visual hypotheses or competing evidence regions are relevant. We propose Latent Visual Contrastive Reasoning, a branch-aware latent reasoning framework that generates multiple question-conditioned latent visual branches as soft hypotheses over image tokens. To train these branches, our method uses bounding-box evidence annotations to build target evidence representations and introduces a soft contrastive branch-evidence alignment objective. This objective encourages evidence-relevant branches to align with annotated visual regions while contrasting them against negative evidence. We instantiate our method on Qwen2.5-VL backbones and train it with standard supervised fine-tuning. Experiments on perception-intensive visual reasoning benchmarks show that the proposed method achieves overall improvements over the corresponding backbone and single-path latent reasoning baselines at both 3B and 7B scales.
[NLP-50] Personality Latents: Conditioning Frozen Language Models on Individuals from Their Own Writing
【速读】： 该论文旨在解决传统语言模型个性控制方法依赖人工编写人物描述或预定义人格维度标签所带来的局限性，即只能表达作者书面内容或分类体系所涵盖的有限人格特征。为此，提出了一种名为“人格潜在变量（Personality Latents, PL）”的新方法，其核心在于无需任何标注的人格标签或人工撰写的个人档案，直接从个体自身的文本中学习其独特的潜在表示。该方法通过生成式潜在优化（generative latent optimization）机制，在冻结的语言模型上对每位训练个体进行潜空间优化，获得一个可自由调整的个性化潜在向量，并设计编码器以预测这些潜在变量，从而实现新个体仅需一次前向传播即可快速注入人格特征，且支持测试时的可选微调。实验表明，在未见个体上的评估中，该方法在MBTI人格维度上达到优于人格嵌入和基于提示（prompt-based）基线的AUC表现，且在三种不同嵌入空间中的评分结果均保持一致排名；此外，使用Qwen2.5-3B-Instruct学习的潜在变量可通过仅引入新适配器的方式成功迁移至Llama-3.1-8B-Instruct模型，实现跨模型的人格传递。因此，该方案的关键在于无监督地从个体文本中提取可迁移、可注入的个性化潜在表示，并通过轻量化适配实现跨模型人格迁移与高效推理。
链接: https://arxiv.org/abs/2610.36805
作者: Gunhee Cho,Yun-Gyung Cheong
机构: Sungkyunkwan University (成均馆大学); Dept. of AI (人工智能系)
类目: Computation and Language (cs.CL)
备注:
Abstract:Controlling the personality of a language model usually relies on written profiles or named trait categories, so it can only express what an author writes down or what a taxonomy names. We instead propose Personality Latents (PL), which learn a latent for each person from their own writing, with no personality labels and no authored profile, and inject it into a frozen language model as a soft prefix. Following generative latent optimization, each training person receives a free latent that is optimized through the frozen model itself, while an encoder learns to predict these latents so that a new person needs only one forward pass, with optional refinement at test time. We evaluate on held-out people by having the conditioned model answer neutral and interview questions and reading personality from its answers with scorers trained on real human writing. On the MBTI axes PL reaches an AUC of , ahead of persona embedding and prompt-based baselines under identical settings, and the ranking holds when the answers are scored in three different embedding spaces. A latent learned with Qwen2.5-3B-Instruct also transfers personality to Llama-3.1-8B-Instruct through a new adapter alone.
[NLP-51] API-TAIL: Detecting Model Substitution in LLM APIs through Temperature-Aligned Likelihood Tails
【速读】： 该论文旨在解决商业语言模型API中“部分替换”（partial substitution）的隐蔽性问题，即服务提供商可能在不告知用户的情况下，将部分请求路由至成本更低的替代模型，同时仍以高规格模型名义计费。现有检测方法面临的核心挑战在于，合法的模型部署本身会因未公开的解码策略（如温度缩放、采样截断、重复惩罚等）导致输出分布变化，使得审计难以区分“正常变异”与“恶意替换”。其解决方案的关键在于：提出一种基于典型性（typicality）的审计机制——API-TAIL。该方法首先通过调整单一有效温度，校准批次内整体输出的典型性分布，消除合法配置下的批量偏移；随后，识别并统计残余的尾部异常响应（tail responses），当异常响应数量超过基于合法配置校准的阈值时，判定为存在部分替换。实验表明，该方法在真实商业API上能有效控制误报率，并以极小查询开销实现对低频部分替换的灵敏检测，显著优于依赖排名或最大均值差异（MMD）的传统审计方法。
链接: https://arxiv.org/abs/2610.36801
作者: Haocheng Mei,Yuchuan Tian,Ye Yuan,Chao Xu,Allen Chen,Kai Han,Yunhe Wang
机构: TokenRhythm Technologies
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: Project page: this https URL
Abstract:Commercial language model APIs return only text, allowing a provider to cut costs by quietly routing a fraction of user requests to a cheaper substitute model while billing for the advertised one. Detecting such partial substitution is challenging because honest providers also alter the output distribution through undocumented decoding and serving choices, such as temperature scaling, sampling truncation, repetition and presence penalties, and default system prompts, forcing an audit to test against a composite null of legitimate configurations rather than a single reference. Our key insight is that these legitimate variations primarily shift response typicality across the entire batch, whereas partial substitution injects individual responses that remain anomalous under the claimed model. Building on this, we introduce API-TAIL, a typicality audit that first absorbs batch-level shifts by aligning a single effective temperature to center standardized surprisal scores across the audited batch, and then flags the endpoint if the number of remaining tail responses exceeds a threshold calibrated across legitimate configurations. Evaluated on real-world commercial APIs, API-TAIL reliably controls false alarms across diverse honest providers while detecting even low-rate partial substitution with a small query budget, substantially outperforming existing rank- and MMD-based audits that lose detection power under the same composite calibration.
[NLP-52] Language Models Keep Abandoned Analyses: Residue and Adaptation after Garden-Path Sentences
【速读】： 该论文旨在探究语言模型在处理歧义句（garden-path sentence）时，是否会保留已被放弃的句法分析路径，并揭示这种残留现象的内在机制。其核心问题是：当模型在解析“While the man hunted the deer ran into the woods”这类句子时，尽管最终采用正确的句法结构（即“the deer”为从句主语而非“hunted”的宾语），是否仍会保留先前错误分析的痕迹？解决方案的关键在于通过激活修补（activation patching）技术分离出两个不同的信号路径：一是来自被抛弃分析路径的残余影响（residue），它通过名词短语状态中保留的角色信息（role information）持续影响后续文本生成，导致对象附着倾向增强，其效应量约为真实宾语句式的75%；二是模型在句法消歧后对正确结构的适应性调整，该调整与句法消歧误差成正比，并主要作用于理解类指标。研究发现，前者由早期计算状态下未被覆盖的状态驱动，后者则依赖于消歧过程中的状态更新。此外，即使在纠正后，被覆盖的预测值仍能持续影响后续输出（8/9模型中观察到此现象），且影响多集中于纠正位置之后的词元。值得注意的是，人类实验范式在模型中仅在控制标点等变量后才可迁移，因为模型自身会生成类似歧义句，从而逆转原始统计模式。这表明模型不仅具有记忆误判路径的能力，还表现出一种“顽固性”认知偏差，其根源在于内部状态的持久性与动态更新机制之间的张力。
链接: https://arxiv.org/abs/2610.36797
作者: Xiang Ding
机构: Nagoya University (名古屋大学)
类目: Computation and Language (cs.CL)
备注: 16 pages, 5 figures, 7 tables
Abstract:Reading “While the man hunted the deer ran into the woods” requires giving up the analysis in which “the deer” is the object of “hunted”. People who have just read such a sentence still produce more transitive structures, so the abandoned analysis primes later language. We ask whether language models keep using an analysis they have abandoned, and what in the model carries it. In 24 open models from 124M to 70B parameters, a garden-path sentence leaves two traces in the context. A residue of the abandoned analysis raises later object attachment, by about three quarters of the effect of a sentence in which the noun phrase really is an object, and it appears in sampled text, on held-out and human-normed items and for a second type of garden path. An adaptation toward the resolved structure scales with the sentence’s disambiguation error and acts mainly on comprehension-like measures. Activation patching separates the two. Later text reads the residue from the role information in the noun phrase’s states, which were computed under the object analysis before the disambiguation, and replacing these states removes almost all of the residue in the 32B and 70B models. The adaptation is read from the states that processed the disambiguation. Overridden statements show the same persistence (a corrected value keeps pulling later predictions in 8 of 9 models), and in most models the pull is carried by the positions after the statement. Finally, the human paradigm transfers only with controls for punctuation, because the models also write garden-path sentences themselves, which reverses the paradigm’s raw statistic.
[NLP-53] Select-TTT: Query-Conditioned Selective Fast Weight Updates for In-Place Test-Time Training
【速读】： 该论文旨在解决在长上下文问答任务中，现有测试时训练（Test-time Training, TTT）方法在实际应用中性能显著下降的问题。具体而言，尽管In-Place TTT通过复用标准MLP模块的下投影层作为快速权重（fast weights），使大语言模型能够在推理阶段动态内化上下文信息，但其在真实网络长上下文场景下的优势几乎消失。根本原因在于In-Place TTT采用统一写入策略（uniform write schedule），即所有输入块以相同强度写入快速权重记忆，导致无关内容稀释与查询相关的有效信息。为解决此问题，论文提出Select-TTT，一种模型无关且基于查询条件的快速权重写入规则：通过直接读取模型冻结输出头中各上下文块隐藏状态对查询词的预测概率质量（probability mass），计算其与用户查询的相关性得分，并利用温度缩放的softmax将其转化为学习率乘子，从而动态调节每个块的写入强度。该方法无需额外参数或监督信号，同时保持了In-Place TTT原有的分块并行计算特性。实验表明，Select-TTT在两个基准数据集和两种模型架构上均一致提升长上下文问答性能，验证了查询条件化快速权重写入在实际测试时训练中的关键作用，且对涉及多轮查询与动态上下文演化的长期智能体任务具有潜在应用价值。
链接: https://arxiv.org/abs/2610.36794
作者: Yue Wu,Yang Zhang,Weinan Gan,Chongming Gao,Xinsen Guo,Xiaoyuan Li,Fuli Feng
机构: University of Science and Technology of China(中国科学技术大学); National University of Singapore(新加坡国立大学); Huawei Noah’s Ark Lab(华为诺亚方舟实验室)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Test-time training (TTT) enables a deployed language model to internalize incoming context on-the-fly into a small set of fast-weight parameters, rather than merely treating it as transient context for attention. In-Place TTT recently made this idea practical for large language models by repurposing the down-projection of standard MLP blocks as fast weights. However, we find that its benefit largely disappears on realistic web long-context question answering. We attribute this to In-Place TTT’s uniform write schedule: every chunk is written with equal strength, so irrelevant content dilutes query-relevant information in fast-weight memory. Building on this insight, we propose Select-TTT, a model-agnostic, query-conditioned rule for In-Place TTT that scales each chunk’s write strength by its relevance to the user query. This relevance signal requires no supervision and no new parameters: it is read directly from the model’s frozen output head, as the probability mass that a chunk’s hidden states assign to the query tokens. A temperature-scaled softmax turns these scores into learning-rate multipliers, leaving the chunk-wise, context-parallel computation of In-Place TTT intact. Select-TTT consistently improves long-context question answering on two benchmarks and two backbones. These findings highlight the value of query-conditioned fast weight writes for practical test-time training on long-context question answering, which is also potentially promising for long-horizon agent tasks involving multiple user queries over evolving contexts.
[NLP-54] Batching the Reflective Optimization Loop: Parallel Proposals Make GEPA Faster and Better NEURIPS2026
【速读】： 该论文旨在解决生成式 AI（Generative AI）中反射式文本优化（Reflective text optimization）的效率瓶颈问题，即传统优化循环因串行结构导致并行评估能力闲置，且资源消耗巨大。其核心解决方案是扩展开源的GEPA优化器，引入批量候选生成与评估机制，即在每一步优化中同时提出并评估多个候选方案，而非逐个处理。实验表明，在两项任务上，批量设置显著提升了效率：在LiveBench-Math任务中，最快运行时间缩短至原耗时的1/3至1/4，测试准确率从68.9%提升至72.1%（使用2×2配置）；在HoVer任务中，准确率从49.0%提升至60.0%（使用8×1配置）。此外，一个简化的运行时模型可有效解释有限计算资源池下的加速现象，且批量设置在单位时间内更高效地利用了指标调用次数与预算成本。研究还首次系统性地探讨了反射式优化步骤在并行化中的可扩展维度，并提出了在固定优化预算下对这些维度进行分配的初步框架。
链接: https://arxiv.org/abs/2610.36774
作者: Jialin Zhang,Lakshya A Agrawal,Shangyin Tan,Donghyun Lee,Lutfi Eren Erdogan,Dan Klein,Koushik Sen,Alex Dimakis,Matei Zaharia
机构: University of California, Berkeley (加州大学伯克利分校)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: NeurIPS 2026 Workshop on Resource-Aware Agentic AI. 9 pages, 7 figures. First four authors contributed equally. Code: this https URL
Abstract:Reflective text optimization makes agentic systems better, but the optimization loop is itself a resource-hungry workload whose sequential structure leaves parallel evaluation capacity idle. We extend the open-source GEPA optimizer to propose and evaluate a batch of candidates on each optimization step instead of one candidate at a time. In our sweep on two tasks, most batched runs finished in half the wall-clock time or less, and the fastest in about a quarter to a third. Batched settings also achieved higher held-out test scores, from 68.9% to 72.1% on LiveBench-Math (with 2\times2 ) and from 49.0% to 60.0% on HoVer (with 8\times1 ). A simple runtime model explains the measured speedups under a finite worker pool, and batched settings also spend their metric-call and dollar budgets more efficiently. We frame the axes along which a reflective optimization step can scale in parallel, a first look at allocating a fixed optimization budget across them.
[NLP-55] DocSkill: Skill-Integrated Agent ic Reasoning for Multilingual Multi-Hop Document Question Answering
【速读】： 该论文旨在解决多语言多跳文档问答（Multilingual Multi-Hop Document Question Answering, MMHop-DocQA）中跨复杂版式、跨页面以及多语言环境下证据定位与整合的难题。现有方法在协调证据获取、跨页信息融合及答案验证方面仍存在显著挑战。为此，本文提出一种基于学习型技能调度策略的代理框架DocSkill，其核心创新在于通过动态选择、重访或跳过特定技能来适应不断演化的证据状态，实现自适应的证据获取与推理。关键解决方案包括：利用有界蒙特卡洛树搜索（bounded Monte Carlo Tree Search, MCTS）在离线阶段探索多种技能执行路径，生成用于监督微调（Supervised Fine-Tuning, SFT）的示范数据和用于直接偏好优化（Direct Preference Optimization, DPO）的偏好对，从而有效训练出具备良好推理策略的智能体。此外，研究构建了首个涵盖10种语言、8个文档领域、共6万组问答对的基准数据集MMHop-DocBench，其基于HopGen管道构建分层页面元素图，可追溯跨页证据路径，并提供页面级与元素级的标注支持。实验结果表明，结合Qwen3-VL-8B模型的DocSkill在MMHop-DocBench上达到41.0的准确率和49.2的F1值，分别优于最强基线5.5和6.3个百分点，在MMLongBench-Doc和DocBench上也展现出竞争力。
链接: https://arxiv.org/abs/2610.36759
作者: Jing Huang,Zhiya Tan,Changtao Miao,Shutao Gong,Yang Wang,Joey Tianyi Zhou,Jianshu Li,Ying Yan
机构: Ant Digital Technologies, Ant Group, Singapore, Singapore; Nanyang Technological University, Singapore, Singapore; Ant Digital Technologies, Ant Group, Hangzhou, China; Ant Digital Technologies, Ant Group, Changsha, China; Ant Digital Technologies, Ant Group, Beijing, China; Singapore Management University, Singapore, Singapore; Institute of Advanced Intelligence and Computing (IAIC), Agency for Science, Technology and Research (A*STAR), Singapore, Singapore
类目: Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV)
备注: 11 pages, 10 figures, 5 tables. Preprint
Abstract:Multilingual multi-hop document question answering (MMHop-DocQA) requires locating and integrating evidence across complex layouts, document pages, and diverse languages. Existing approaches include single-pass processing and adaptive tool use, but learning to coordinate evidence acquisition, cross-page integration, and answer verification remains challenging. To address this challenge, we present DocSkill, an agentic framework with a learned skill-orchestration policy for adaptive evidence acquisition and reasoning. The policy selects, revisits, or skips skills according to the evolving evidence state. To construct supervision for skill orchestration, bounded Monte Carlo Tree Search (MCTS) explores alternative skill trajectories offline. These trajectories provide demonstrations for supervised fine-tuning (SFT) and preference pairs for Direct Preference Optimization (DPO). We further introduce MMHop-DocBench, a multilingual multi-hop document QA dataset with training and test splits, containing 60K QA pairs across 10 languages and eight document domains. Its HopGen pipeline builds hierarchical page-element graphs to derive traceable cross-page evidence paths and generate grounded QA pairs with page- and element-level evidence annotations. Experiments show that DocSkill with Qwen3-VL-8B achieves 41.0 accuracy and 49.2 F1 on MMHop-DocBench, exceeding the strongest evaluated baseline by 5.5 and 6.3 points, respectively, and achieving competitive performance on MMLongBench-Doc and DocBench.
[NLP-56] CAMR: Compatibility-Aware Memory Replay for Continual Learning in Large Language Models
【速读】： 该论文旨在解决大语言模型在持续学习（continual learning）过程中如何有效获取新知识并保留已有知识的关键挑战，尤其针对在有限的每轮更新预算下，如何动态决策是否对当前输入样本进行监督学习以及选择哪些历史样本进行回放以实现知识留存。其核心解决方案是提出一种自适应的记忆回放框架——兼容性感知记忆回放（Compatibility-Aware Memory Replay, CAMR），该框架通过联合控制当前样本的监督策略与历史样本的回放机制，利用兼容性检测与可学习的记忆门控（memory gate）构建在线策略，动态决定三种行为：使用通用回放方式学习新样本、联合强化具有兼容性的相关历史样本，或跳过已掌握重复样本的直接监督。该策略基于知识获取、保留效果及训练成本的反馈，在固定每轮回放预算的约束下进行优化。实验表明，CAMR在持续事实更新（WikiBigEdit）和跨任务/跨领域学习（TRACE基准）中均显著优于现有方法，不仅在3K个事实后即达到93.66%的原始匹配率，且在10万条事实后超越全历史回放（LoRA）方案，同时在金融问答、数学推理、代码生成与摘要等多任务序列中保持甚至提升先前任务的性能，验证了其在知识积累与遗忘抑制方面的有效性。
链接: https://arxiv.org/abs/2610.36734
作者: Jingpu Yang,Tianhao Li,Fengxian Ji,Yufeng Wang,Shuxin Zheng,Xiaoqing Zhang
机构: Beijing Zhongguancun Academy (北京中关村学院); Northeastern University (东北大学); Beihang University (北京航空航天大学); Renmin University of China (中国人民大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Continual learning requires large language models to acquire new knowledge and capabilities while retaining what they have learned. Fine-tuning and replay support acquisition and retention, but deciding whether incoming examples need supervision and which historical examples to revisit remains a challenge. We propose Compatibility-Aware Memory Replay (CAMR), an adaptive framework that jointly controls current-example supervision and historical replay. CAMR combines compatibility checks with a learned memory gate to guide an online policy over three actions: learning incoming examples with general replay, jointly reinforcing compatible related examples, and skipping direct supervision of verified mastered duplicates. The policy learns from acquisition, retention, and training-cost feedback under a bounded per-arrival replay budget. We evaluate CAMR with Qwen3-8B-Base on continual factual updating and cross-task and cross-domain learning. On WikiBigEdit, CAMR consistently leads evaluated methods in Original Match from 3K to 100K facts, reaching 93.66% after just 3K facts and surpassing Full-history Replay (LoRA) after 100K facts. It also achieves the highest final scores on all four selected TRACE tasks and maintains or improves earlier-task performance relative to post-acquisition levels in a sequence spanning financial question answering, mathematical reasoning, code generation, and summarization.
[NLP-57] A Harness Reducing over-Defensive Writing in Agents by Isolating Objections
【速读】： 该论文旨在解决生成式报告时因使用智能代理（Agent）而产生的“过度防御性写作”（over-defensive writing）问题，即代理在扩展报告时倾向于提出强于原文的主张，随后又加以否定，从而影响文本的可信度与简洁性。其核心问题在于，代理在生成内容时可能默认处于应对对抗性AI评审的语境，而非面向真实人类读者，导致冗余且防御性的表达。为解决此问题，论文提出一种分治式架构（harness），将评审（reviewer）、过滤（filter）和写作（writer）三个角色分离，使写作代理仅接收由独立过滤器选定的问题，从而脱离防御性上下文。实验基于25个商业项目草稿扩展为正式报告，结果显示防御性主张从59处降至2处，降幅达96.61%，同时未发现未经支持的事实添加，且源主张保留率从80.35%提升至93.12%，验证了该方案在保持事实准确性与提升内容质量方面的有效性。
链接: https://arxiv.org/abs/2610.36717
作者: Xiaochen Chang
机构: 未知
类目: Computation and Language (cs.CL)
备注: defensive writing; context isolation; harness
Abstract:When expanding a report, writing with agents may introduce a stronger claim than the article makes and then disclaim it. This is called over-defensive writing. We assume that this is because the agent is writing for responding to an adversarial AI reviewer, instead of a human reader. The defensive context may also influnce the consequent when we simply ban defensive writing in the output through prompts. In order to reduce over-defensive writing, we produce a harness that separates the reviewer, filter, and writer. The writer agent receives only the problems choosen by the filter, which is an independent context. We evaluate this harness on 25 business-project drafts expanded into formal reports, the result shows that the defensive propositions decreased from 59 to 2, which is a 96.61% reduction, no unsupported factual additions were identified while the source-claim retention rises from 80.35% to 93.12%.
[NLP-58] Is Token Confidence Just Chain Length? A Signal-Processing Audit of Reasoning Traces
【速读】： 该论文旨在解决生成式 AI 在数学推理任务中因采用多链路思维（chain-of-thought）采样并进行多数投票以提升准确率所导致的计算成本过高问题。现有方法通过评估每条推理链中模型对每个词元（token）生成的置信度，并基于滑动窗口内最低平均置信度来决定保留或终止链路，以此降低生成文本量。然而，本文发现该置信度评分本质上主要反映链路长度而非正确性：由于置信度信号呈缓慢漂移特性，较长链路的最小值在统计上必然更低，因此仅按链路长度排序即可达到与置信度评分相当甚至更优的性能。进一步分析表明，在控制长度因素后，置信度评分不再携带关于推理正确性的有效信息。通过信号处理工具（如频谱分析与突变检测）揭示，虽然错误链路确实表现出更强的词元间快速波动（微弱可检测信号），但其强度不足以改进最终多数投票结果。此外，对于早期停止策略，滑动平均法或经典突变检测器均不如直接减少采样链路数量这一简单方案有效。因此，该研究的核心结论是：当前基于置信度的链路筛选机制本质上是冗余且误导性的，其有效性源于长度偏差而非语义质量，而更简洁的方法反而更具效率与实用性。
链接: https://arxiv.org/abs/2610.36709
作者: Pragna Mallappa
机构: Keshav Memorial Institute of Technology(凯沙夫纪念技术学院)
类目: Computation and Language (cs.CL)
备注:
Abstract:Reasoning language models answer math problems by writing long chains of thought, and sampling many chains per problem and taking a majority vote raises accuracy at a large cost in generated text. A popular way to cut that cost scores each chain by the model’s confidence in each token it writes, where a token is a word or part of a word. The method averages this confidence over a sliding window of 2048 tokens and keeps or stops chains based on the lowest average. We treat the confidence, recorded token by token, as a signal in the sense of audio or radio, and examine the method with standard signal-processing tools such as spectral analysis and statistical tests for detecting a sudden change. We sampled 16 chains for each of 500 problems from two math benchmarks with a small open reasoning model. The windowed-minimum score turns out to measure mostly how long a chain is. Ranking chains by length alone does as well as or better than the confidence score, and once length is accounted for the score carries no information about correctness. Signal processing explains why: the minimum of a longer stretch of a slowly drifting signal is lower by chance alone, and the confidence signal drifts slowly enough that this effect dominates. The spectrum does hold a small signal that length does not explain, since incorrect chains carry more rapid token-to-token fluctuation, but it is too weak to improve the vote. For stopping chains early, neither the sliding average nor a classical change detector beats the simpler choice of sampling fewer chains.
[NLP-59] Specialization Diversity and Scaling in Child-Scale Language Models
【速读】： 该论文旨在解决小样本自然语言数据环境下语言模型学习机制的问题，尤其关注在有限、真实儿童语言输入场景下模型的学习能力与泛化表现。其核心挑战在于：尽管现代语言模型通常在海量文本上训练，但儿童语言习得发生在高度受限且个体差异显著的自然语言环境中，现有研究对小规模、真实语境下的语言学习规律了解仍十分有限。论文的关键解决方案是通过对比分析两种不同来源的小规模训练数据——来自真实家庭录音（BabyView）的自然儿童语言和人工生成的类儿童对话数据（TinyDialogues），系统考察训练数据的数量、来源及多样性对模型预测能力的影响。研究发现，仅在单一儿童语言环境上训练的模型能显著更准确地预测该儿童的语言环境，且这种特异性主要体现在词汇与分布特征层面；一个简单的5-gram模型即可复现跨儿童的语言模式差异（相关性r=0.91）。此外，使用多儿童数据进行训练虽能提升对混合未见语言的预测性能，但在语法基准测试上并未呈现明显改善。更重要的是，自然主义与合成数据所引发的语法结构与词义语义模式存在显著差异，揭示了训练数据来源本身对模型认知表征的重要影响。因此，该研究的关键结论是：小样本语言学习不仅取决于输入量，更受训练环境的个体特异性、数据匹配度以及数据类型的影响，强调了“数据源”与“任务匹配”的核心作用。
链接: https://arxiv.org/abs/2610.36679
作者: Steven Y. Feng,Michael C. Frank
机构: Stanford University (斯坦福大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Code and data at this https URL
Abstract:Modern language models are trained on far more language than children experience, but we know much less about learning from small, naturalistic datasets. We train small GPT-2 models on transcripts of speech with children from dense home recordings (BabyView) and on synthetic child-directed dialogue (TinyDialogues), varying the amount, source, and diversity of training data. Models trained on one child’s language environment predict that environment substantially better than other children’s environments. Much of this specialization is lexical and distributional: a simple 5-gram model reproduces the cross-child pattern (r=.91). Training on data from multiple children clearly improves prediction of pooled held-out language relative to training on a single child’s input without a similarly clear improvement on grammatical benchmarks. Across data sizes, naturalistic and synthetic training inputs also produce different grammatical and lexical-semantic patterns, with important differences between grammatical phenomena. Together, these results show that small-data language learning depends not only on how much language a model sees, but on which environments it comes from and how well that input matches what the model is asked to predict.
[NLP-60] A Quantitative Approach to Asch Conformity in Frontier Large Language Models
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）中从众行为（conformity）的量化评估问题。现有研究虽已证实大语言模型存在从众现象，但缺乏对从众程度及其影响条件的精细量化分析。本文的关键解决方案在于采用重复采样单一问题的方法，利用大语言模型固有的概率性输出特性，生成模型在面对“人为制造的一致性”（manufactured consensus）前后答案的分布。通过对比处理前后的答案分布，实现了从定性观察到定量分析的跨越，从而精确刻画了从众行为的发生条件及强度，为理解大语言模型的社会性响应机制提供了可量化的实证依据。
链接: https://arxiv.org/abs/2610.36668
作者: Senthil Rajasekaran,Stylianos Loukas Vasileiou
机构: Université libre de Bruxelles(布鲁塞尔自由大学); New Mexico State University(新墨西哥州立大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:The Asch conformity experiment has recently gained considerable attention as a way of measuring conformity in large language models (LLMs). While previous studies have focused on demonstrating conformity in LLMs, our work adds a further quantitative layer to this approach. Instead of sampling a frontier LLM on a suite of questions, our approach samples a single question repeatedly, leveraging the probabilistic nature of LLMs to produce a distribution of answers. In doing so, we can compare the distributions of answers before and after the model is presented with the manufactured consensus. Our experiments confirm that frontier LLMs do conform, reinforcing prior literature, and, through our distribution-to-distribution comparisons, provide additional quantitative data on the conditions that lead to conformity and the degree to which it occurs.
[NLP-61] Japanese Discourse Structure Analysis Based on Segmented Discourse Representation Theory Using LLM s AACL
【速读】： 该论文旨在解决跨语言语篇解析中日语领域缺乏标注数据的问题，具体聚焦于在日语中应用基于分段语篇表征理论（SDRT）的语篇解析方法的有效性问题。由于此前研究多集中于英语，且缺乏高质量的日语SDRT语料库，其在日语中的适用性尚不明确。为此，研究者通过使用大语言模型（LLM）对现有英文SDRT语料进行翻译，构建了首个日语SDRT语料库。关键解决方案在于采用提示工程（prompt-based）与微调（fine-tuned）两种方式的LLM进行跨语言迁移，并通过对比评估验证其在日英两语言间表现相当，表明基于SDRT的语篇解析具备良好的跨语言可迁移性。进一步的消融分析揭示，多数日语语篇标记对模型预测影响较小，模型主要依赖上下文中的语篇关系结构而非具体上下文从句内容进行判断，这种对结构信息的高度依赖可能解释了尽管英语与日语在语言类型学上存在显著差异，但模型仍能实现相似性能的原因。
链接: https://arxiv.org/abs/2610.36667
作者: Satomu Tomisaki,Yosuke Mikami,Amane Watahiki,Hitomi Yanaka
机构: The University of Tokyo(东京大学); RIKEN(理化学研究所); Tohoku University(东北大学)
类目: Computation and Language (cs.CL)
备注: To appear in the Proceedings of the Asia-Pacific Chapter of the Association for Computational Linguistics: Student Research Workshop (AACL-SRW 2026)
Abstract:Discourse parsing based on Segmented Discourse Representation Theory (SDRT) is crucial for capturing the structural dynamics of dialogues. However, most prior research focuses on English, and its effectiveness for Japanese remains unclear owing to the lack of Japanese SDRT corpora. To address this issue, we construct Japanese SDRT corpora by translating existing English SDRT corpora using an LLM. Through a comparative evaluation of prompt-based and fine-tuned LLMs, we show that the models achieve comparable performance between English and Japanese, suggesting the cross-lingual applicability of SDRT-based discourse parsing. In order to investigate why the English and Japanese parsers achieve similar performance, we conduct ablation analyses on discourse markers and contextual inputs. Our results reveal that most of the Japanese discourse markers we examined have little impact, and the models’ predictions depend heavily on contextual discourse relations rather than on the content of the context clauses. This structural reliance might explain the comparable performance despite the substantial typological differences between English and Japanese.
[NLP-62] Balanced State Subspaces for Mamba-2 State Compression
【速读】： 该论文旨在解决生成式模型中状态压缩效率低下问题，尤其针对Mamba-2这类基于状态空间模型（State Space Model, SSM）的架构在解码过程中因状态维度高而导致的内存与带宽瓶颈。现有压缩方法通常对状态值进行量化或剪枝，且在原始坐标基下独立评估各通道的重要性，但这种方法忽略了状态内部能量分布的本质结构——研究发现，训练后的Mamba-2状态能量几乎呈秩一（rank-one）特性，其有效信息主要集中在由递归累积形成的偏置方向，而输出实际读取的内容则位于一个低维非对齐子空间。因此，传统以坐标为单位的评分机制无法捕捉决定输入-输出映射的关键方向。为此，本文提出\method，一种后训练变换方法，通过引入具有完全可控性与可观测性的广义Gram矩阵实现平衡截断（Balanced Truncation），并利用Mamba-2的标量-恒等转移结构精确完成坐标旋转，从而聚焦于“写-读”子空间（write-read subspace）的压缩。为进一步提升压缩灵活性，作者设计HARP（Hierarchical Adaptive Rank Allocation via Hankel spectrum），根据每层的汉克尔谱（Hankel spectrum）自适应分配秩，并保护对检索敏感的层，在统一预算下同时优化困惑度（perplexity）与检索性能。该方法仅需一次前向传播即可完成校准，速度比GHOST快一个数量级，峰值内存更低；在平均状态维度减半的情况下，\method在WikiText-2上仅带来0.3点（1.8%）的困惑度上升（对比GHOST的1.6点，9.3%），且在25%保留率时优势进一步扩大，证明了其在保持模型性能的同时实现高效状态压缩的有效性。
链接: https://arxiv.org/abs/2610.36617
作者: Shang Tingyu,Cao Wenfei
机构: Shaanxi Normal University (陕西师范大学)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Applications (stat.AP); Methodology (stat.ME)
备注: 15 pages, 4 figures
Abstract:Mamba-2 compresses all context into a fixed-size recurrent state that dominates decoding memory and bandwidth. Existing compressors quantize state values or prune state coordinates, scoring each channel independently in the native basis. We examine trained Mamba-2 states and find that energy is almost rank-one, but the output does not read the energetic directions: energy concentrates in a bias-like direction accumulated by the recurrence, while the content the output reads lives in a low-dimensional subspace that is not aligned with the state’s coordinates. Even output-aware scores are computed coordinate by coordinate, so they miss the directions that determine the input–output map. Compression should therefore target the write–read subspace rather than individual coordinates. We propose \method, a post-training transform that instantiates balanced truncation with full controllability and observability Gramians and applies the rotation exactly via Mamba-2’s scalar-identity transition. Because the useful rank varies widely across layers, we further introduce HARP, which allocates ranks per layer from the Hankel spectrum and protects retrieval-sensitive layers, improving both perplexity and retrieval over uniform budgets. Calibration takes a single forward pass, an order of magnitude faster than GHOST’s released protocol, with lower peak memory. With HARP, \method halves the state dimension at an average WikiText-2 perplexity increase of 0.3 points ( 1.8% ) across 130M–2.7B, versus 1.6 ( 9.3% ) for GHOST, and the gap widens at 25% retention.
[NLP-63] Structured Maps for Landmark-Based Route Instructions with LLM s
【速读】： 该论文旨在解决在生成基于地标（landmark）的路线指引时，如何有效融合地图几何结构与人类对地标偏好认知的问题。现有大型语言模型（Large Language Models, LLMs）虽在自然语言处理任务中表现优异，但其对地图数据的结构化输入方式尚未形成统一范式。本文提出一种基于可缩放矢量图形（SVG）格式的结构化文本表示方法，将地图的空间特征以可读性强的文本形式编码，保持与渲染地图图像之间的几何关联性，使语言模型能够直接访问空间位置与语义信息。通过利用由人工编写的路线描述和OpenStreetMap数据构建的新数据集，研究在转弯处地标选择与目的地周边地标选择两个任务上评估模型性能。实验结果表明，所提出的结构化文本输入在效果上不逊于甚至优于基于PNG图像的输入方式；同时，针对特定任务的微调策略使小型本地化语言模型也能高效利用地图布局实现高精度的地标选择。研究表明，结构化文本是一种高效且轻量级的空间推理接口，适用于面向人类的导航场景。
链接: https://arxiv.org/abs/2610.36582
作者: Mifu Matsumoto,Yuki Nakanishi,Tomoya Hashiguchi,Hiroaki Ohshima
机构: University of Hyogo (兵库大学)
类目: Computation and Language (cs.CL)
备注: Accepted for publication at BESC 2026
Abstract:To generate landmark-based route instructions, models must understand both map geometry and human landmark preferences. Although large language models (LLMs) show great abilities in language tasks, the best way to structure map data for these models is not yet established. This paper proposes a structured text representation for landmark selection tasks in route navigation, utilizing SVG format to encode spatial features. The structured text maintains a geometric connection with rendered map images, allowing language models to directly access spatial and semantic data. Using a new dataset created from human-written route descriptions and OpenStreetMap data, we evaluate models on two tasks: landmark selection at turns and around destinations. Experiments with five LLMs show that the proposed structured text input performs as well as or better than image-based PNG inputs. In addition, task-specific fine-tuning enables smaller local LLMs to effectively leverage map layouts to achieve high selection accuracy. These results demonstrate that structured text is an efficient and lightweight interface for spatial reasoning in human-oriented navigation.
[NLP-64] Decoupling Consistency and Conflict across MLLM Layers for Fake News Detection
【速读】： 该论文旨在解决多模态虚假新闻检测中难以有效识别细粒度、稀疏且层级各异的图文不一致问题。现有方法通常以粗粒度、静态的方式处理语义一致性与冲突，无法充分捕捉不同层次（如对象、场景、事件、主张等）的细微矛盾，尤其在冲突信号稀疏或强度不一的情况下表现不佳。其解决方案的关键在于提出一种多层级一致性与冲突解耦机制（Multi-Level Consistency and Conflict Decoupling, MLCCD），充分利用多模态大语言模型（Multimodal Large Language Models, MLLMs）各层所编码的差异化知识。具体而言，首先利用MLLM最终决策表示筛选与任务相关联的视觉与文本标记，避免无关内容干扰对稀疏冲突的感知；其次，保留相邻层间表征发生显著变化的标记，以捕获每一层新增的知识；最后，将这些知识解耦为一致性证据与冲突证据，并通过各自独立的路由机制在不同层级上进行聚合，从而实现对多层次、可变冲突水平的自适应建模。实验结果表明，MLCCD在三个基准数据集上均显著优于现有方法。
链接: https://arxiv.org/abs/2610.36569
作者: Kai Qian,Dongjie Xu,Weijie Shi,Jiaqi Wang,Haolun Ma,Jia Zhu,Jiajie Xu
机构: Soochow University (苏州大学); Hong Kong University of Science and Technology (香港科技大学); The Chinese University of Hong Kong (香港中文大学); DSA, The Hong Kong University of Science and Technology (香港科技大学数据科学研究院)
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL)
备注: 12 pages
Abstract:The trustworthiness of Web content faces a growing threat from multimodal fake news. Real news usually shows aligned image-text semantics, while fake news often contains a few conflicts at varying levels, such as mismatches in objects, scenes, events, and claims. However, existing methods usually treat consistency and conflict in a coarse and static way, limiting their ability to capture sparse conflicts and adapt to varying conflict levels. Inspired by the distinct knowledge encoded across layers in large language models, we propose Multi-Level Consistency and Conflict Decoupling (MLCCD), which captures conflicts at different levels from different layers of multimodal large language models (MLLMs). Specifically, we first use the MLLM’s final decision representation to select visual and textual tokens relevant to the task, preventing irrelevant content from diluting sparse conflicts. Then, we preserve tokens whose representations change sharply between adjacent layers, capturing the knowledge each layer adds. Finally, we decouple this knowledge into consistency and conflict evidence and aggregate each across layers with its own router, adapting to varying conflict levels. Experiments on three benchmarks show that MLCCD outperforms existing methods.
[NLP-65] PRISM-BENCH: Diagnosing Grounded and Faithful Multimodal Reasoning Beyond Final-Answer Accuracy NEURIPS2026
【速读】： 该论文旨在解决当前视觉语言模型（Vision-Language Model, VLM）评估中仅依赖最终答案准确率（Final-answer accuracy）所带来的局限性，即无法衡量模型答案是否基于图像内容、其推理过程是否与证据一致，以及生成答案的代价。这些问题直接关系到系统在实际部署中的可靠性与可信度。为此，论文提出PRISM-BENCH基准，包含1,000个由专家精心设计的问题，覆盖五个代表性不足的推理类别，并为每一道题提供由一名标注者撰写、另一名验证过的理想化推理轨迹（reasoning trace）。该基准支持两项关键诊断：感知探测（perception probe），通过让模型基于自身描述作答，揭示约85%的前沿模型失败源于感知错误，将瓶颈定位在“接地性”（grounding）；以及推理轨迹一致性（trace alignment），可预测所有25个测试系统答案的正确性（宏平均AUROC达0.66），且不受响应长度干扰，真正反映推理质量而非冗余表达。此外，论文引入效率感知指标CORE，能重新排序24.2%的模型对，同时与响应长度无关，在25个开放与闭源系统（包括GPT-5和Claude Opus 4.6）中，最高准确率为66.0%，而在四个具备人类基线的类别上，最优模型仍落后于人类18.4分，且近期性能提升主要集中在识别任务，而非空间接地能力，凸显了当前模型在深层推理与真实视觉理解上的不足。
链接: https://arxiv.org/abs/2610.36554
作者: Vansh Bhatia,Mehdi Rezagholizadeh,Parsa Ashrafi Fashi,Igor Gilitschenski,Vikram Appia,Emad Barsoum
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL)
备注: Accepted as a poster at the NeurIPS 2026 Workshop on Grounded and Faithful Vision-Language Models for Real-World Deployment (VLM4RWD). 18 pages, 10 figures, 7 tables. Dataset: this https URL
Abstract:Final-answer accuracy reports whether a vision-language model was right, but not whether its answer was grounded in the image, whether its stated reasoning reflects that evidence, or what the answer cost to produce. Those are the properties that decide whether a system can be deployed. We present PRISM-BENCH, a benchmark of 1,000 expert-curated questions spanning five under-represented reasoning categories, in which every item carries an idealized reasoning trace written by one human annotator and validated by another. The traces support two diagnostics that accuracy cannot provide. A perception probe, in which the model answers from a description it wrote itself, attributes roughly 85% of frontier-model failures to misperception and places the bottleneck at grounding. Trace alignment predicts answer correctness for all 25 systems tested (macro AUROC 0.66) and survives length control, so it measures reasoning content and not verbosity. We further introduce CORE, an efficiency-aware metric that reorders 24.2% of model pairs relative to accuracy while remaining uncorrelated with response length. Across 25 open and closed systems, including GPT-5 and Claude Opus 4.6, the best accuracy is 66.0%. On the four categories with a human baseline the best system trails humans by 18.4 points, and recent frontier gains fall on recognition and not on spatial grounding.
[NLP-66] Agent ic Data Sampling Framework for Multilingual ASR via Linguistic-Aware Error-Driven Optimization
【速读】： 该论文旨在解决多语言自动语音识别（Multilingual ASR）中数据采样策略依赖人工、经验性设计导致效率低下且难以优化的问题。现有方法在数据采样时普遍采用静态或启发式规则，缺乏对语言学特征的动态响应能力，限制了模型性能的进一步提升。其解决方案的关键在于提出一种基于语言学感知的智能代理（linguistic-aware agentic data sampling）数据采样框架，将一个由语言学驱动的大型语言模型（LLM）代理引入多语言ASR训练循环。该代理依据音系、形态和句法三个层面的错误分析构建评估准则，结合训练统计信息、预算约束及算法特性，自适应地选择语域级、样本级与词元级的采样策略，从而将误识率（WER/CER）从单纯的评估终点转化为可指导数据选择的反馈信号。实验基于Whisper-medium作为主干模型，在十二种Common Voice语言上验证，结果表明该方法在使用更少训练数据与更低计算资源的前提下，实现了优于全量数据训练的性能，并显著加速了训练收敛速度。
链接: https://arxiv.org/abs/2610.36538
作者: Zhiheng Qian,Junhao Du,Jing Peng,Hanqi Li,Kunyang Peng,Feng Xue,Lu Chen,Shuai Wang,Kai Yu
机构: Shanghai Jiao Tong University(上海交通大学); AISpeech Co., Ltd.(声智科技有限公司); Nanjing University(南京大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Multilingual automatic speech recognition (ASR) has achieved rapid progress with large-scale pretrained models and multilingual speech corpora. Previous studies have proved the pivotal role of data sampling in multilingual ASR performance. However, existing sampling strategies largely rely on time-consuming manual, heuristic-based designs. To address this, we propose a linguistic-aware agentic data sampling framework that introduces a data sampling LLM agent into the multilingual ASR training loop. The agent is driven by a linguistically grounded rubric that analyzes ASR errors from phonological, morphological, and syntactic perspectives, and uses the resulting feedback together with training statistics, budget constraints, and algorithm profiles to select domain-level, sample-level, and token-level sampling strategies. This design turns evaluation results into adaptive data-selection decisions rather than treating WER/CER as final endpoints only. We conduct experiments on twelve Common Voice languages using Whisper-medium as the ASR backbone. The results show that our method achieves performance better than full-data training while using less training data and lower computational resources, and substantially accelerates training convergence compared to conventional sampling strategies.
[NLP-67] RoboAuto: Autonomous Robot Program Search through Multi-Level Physical Understanding
【速读】： 该论文旨在解决基于大语言模型（LLM）的自主智能体在机器人操作任务中进行程序搜索时，因缺乏对物理世界深层理解而导致的探索效率低、成功率受限的问题。具体而言，现有方法主要依赖执行反馈进行程序迭代优化，但未能有效建模物体可操作性（affordance）、场景几何约束及交互成败的物理机理，从而导致智能体难以做出合理的动作决策。其解决方案的关键在于提出RoboAuto框架，通过三个核心组件显式地组织多层次物理理解：首先，构建分层知识库，融合可执行技能与以物体为中心的属性知识（如可操作性、空间关系、动作效应），并基于具身数据和在线任务求解学习；其次，引入场景锚定机制，将知识映射到当前环境，识别相关物体并解析动作关键量；最后，采用多阶段规划结合自设计实验的方法，主动验证各阶段是否产生预期物理效果，以测试和排除物理不确定性假设。该框架显著提升了程序搜索的鲁棒性与效率，在多个基准任务上实现高达89.1%的成功率，并展现出跨模拟器、跨实体的泛化能力及知识规模扩展带来的性能增益。
链接: https://arxiv.org/abs/2610.36533
作者: Buzhao Liu,Jianing Ye,Xiu Yuan,Yangtong Feng,Chengsong Huang,Tyler Dinh,Yevgeniy Vorobeychik,Ning Zhang,Chongjie Zhang
机构: Washington University in St. Louis (圣路易斯华盛顿大学)
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: Project page: this https URL
Abstract:Recently, LLM-based autonomous agents have shown strong capabilities in iteratively generating, evaluating, and refining executable solutions. In robotic manipulation, Code-as-Policy lets these agents search over modular and editable robot programs. However, effective program search requires physical understanding: agents must understand how objects can be manipulated, how scene geometry constrains actions, and why interactions succeed or fail. Existing methods rely mainly on execution feedback for program refinement, often leaving these physical questions unresolved. We introduce RoboAuto, a framework for robot program search that explicitly organizes physical understanding at the object, scene, and interaction levels through three components. First, a hierarchical knowledge base combines executable skills with object-centric knowledge of affordances, spatial relations, and action effects, learned from embodied data and online task solving. Second, scene grounding connects this knowledge to the current environment by identifying relevant objects and resolving action-critical quantities. Third, multi-stage planning with self-designed experiments checks whether each stage produces its expected physical effects and actively tests competing hypotheses to resolve physical uncertainties. RoboAuto achieves superior performance across five benchmark families, attaining an 89.1% success rate on short-horizon LIBERO-PRO tasks and more than twice the mean success rate of ASPIRE on LIBERO-PRO Long and LIBERO-Plus. It also outperforms ASPIRE and \pi_0.5 on three challenging long-horizon real-robot tasks requiring precise manipulation. We further show that the knowledge base transfers across simulators and embodiments, and that program search performance scales with its size. Project page: this https URL
[NLP-68] ETC-Bench: A Multilingual Benchmark for Evaluation-Side Transcript Canonicalization
【速读】： 该论文旨在解决语音识别系统在评估阶段因转录文本存在非声学决定性差异而导致评估结果偏差的问题，即同一语音可能产生多个表面形式不同的转录文本，这些差异并非由语音信号本身决定。为此，论文提出评估侧转录归一化（Evaluation-side Transcript Canonicalization, ETC），其核心是将不依赖于语音信号的多种表面形式映射到单一规范形式。解决方案的关键在于构建了首个全面的多语言基准测试集ETC-Bench，涵盖12种语言中的181类语言现象，每条数据为仅在目标片段上存在特定现象差异的最小对比对。为同时衡量归一化成功率与语义保留能力，论文提出了语义保持精确匹配归一化准确率（Semantic-preserving Exact-match Normalization Accuracy, SENA），结合独立归一化结果间的精确匹配与基于BERTScore的相对约束以防止信息丢失。此外，引入基于分类体系的智能体归一化技能（Taxonomy-guided Agent Canonicalization Skill, TACS），通过向大语言模型（LLM）代理提供现象描述和明确的归一化规则契约，显著提升了跨语言归一化性能。实验表明，TACS驱动的代理在共享语言上优于五种基线配置，验证了该方法的有效性与可扩展性。
链接: https://arxiv.org/abs/2610.36527
作者: Zhiheng Qian,Junhao Du,Jing Peng,Hanqi Li,Kunyang Peng,Feng Xue,Kai Yu
机构: 未知
类目: Computation and Language (cs.CL)
备注:
Abstract:Written transcripts of the same speech can differ in ways that are not uniquely determined by the speech signal. We formulate evaluation-side transcript canonicalization (ETC), which maps multiple surface forms to a single canonical form when their differences are not determined by speech. To our knowledge, ETC-Bench is the first and most comprehensive multilingual benchmark for evaluation-side transcript canonicalization, covering 181 phenomenon types across 12 languages. Each instance is a minimal pair that shares a sentence context and differs only in a target span exhibiting one phenomenon. To evaluate both canonicalization success and semantic preservation, we propose Semantic-preserving Exact-match Normalization Accuracy (SENA), which combines exact agreement between independently canonicalized pair members with a pair-relative BERTScore constraint against information loss. We further introduce the Taxonomy-guided Agent Canonicalization Skill (TACS), which provides LLM agents with phenomenon descriptions and explicit canonicalization contracts. Experiments with five baseline configurations and three TACS-based agents show that the agents outperform the baseline systems on shared languages. ETC-Bench provides a systematic testbed for evaluation-side transcript canonicalization across languages.
[NLP-69] Causal Localization of Evidence Effects in Transformers
【速读】： 该论文旨在解决如何从自然语言证据中识别出影响语言模型决策的注意力头（attention heads）这一关键问题。其核心挑战在于，传统方法难以在不依赖标注数据或预设答案的情况下，有效定位对特定推理任务具有因果影响力的注意力机制。为此，论文提出受控证据聚合（Controlled Evidence Aggregation, CEA）方法：通过在固定提示中系统性地改变两个符号的出现次数与顺序，观测冻结的Transformer模型在两个答案词元间的逻辑差异（logit difference），从而捕捉由符号计数变化引发的偏好偏移。进一步引入平衡效用因果选择（Balanced Utility Causal Selection, BUC）策略，联合评估删减（ablation）某组注意力头输出时对计数效应的削弱程度，以及通过插补（patching）不同计数条件下均值输出能否双向恢复该效应，以此筛选出紧凑且具因果重要性的头集合。理论层面，作者推导了条件边界，明确了注意力头归因与干预效应之间的关系，并指出在有限样本下可恢复归因短名单的条件。实验结果表明，在Pythia-410M模型中，仅删减1至12个针对四种排序模式分别选出的注意力头，即可消除65.5%至91.6%的计数效应，且这些头在未用于选择的计数比较中仍保持显著影响力。更重要的是，将仅基于符号化“是/否”提示选出的8个头应用于真实自然语言任务（判断哪个州获得更多数报告支持），导致平均准确率下降8.2个百分点，显著优于与之匹配大小和层位置的随机头集（仅下降0.8个百分点）。这证明受控符号探针能够识别出具备跨模态因果影响力的紧凑注意力头集合，其作用可延伸至真实自然语言推理任务。
链接: https://arxiv.org/abs/2610.36524
作者: Jipeng Li,Pramod Khargonekar,Yanning Shen
机构: 未知
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Can controlled symbolic inputs reveal attention heads that influence a language model’s decisions from natural language evidence? We introduce controlled evidence aggregation (CEA), which varies the counts and order of two symbols in otherwise fixed prompts and measures changes in a frozen Transformer’s preference between two answer tokens. The preference is measured by their logit difference; the prompts prescribe no correct answer. Balanced utility causal selection (BUC) identifies compact head sets by jointly evaluating how much ablating their outputs removes the count effect and how closely patching mean outputs between count conditions recovers the effect in both directions. We derive conditional bounds linking head attribution to intervention effects and specifying when the attribution shortlist can be recovered from finite samples. In Pythia-410M, ablating sets of 1-12 heads selected separately for four ordering regimes removes 65.5-91.6% of the count effect. The sets retain their influence on count comparisons not used for selection. To test relevance beyond symbolic inputs, we evaluate a natural language task asking which state is supported by more reports. Ablating eight heads selected only on symbolic yes/no prompts lowers mean accuracy by 8.2 percentage points, versus 0.8 for random sets matched by size and layer. Our results demonstrate that controlled symbolic probes can identify compact attention head sets whose causal influence extends to natural language decisions.
[NLP-70] Benchmarking Task Takeover Across Agent Handoffs
【速读】： 该论文旨在解决多智能体协作任务中因任务交接（handoff）导致的责任转移问题，即在长周期任务中，前序模型将工作移交至接收模型时，后者需准确继承并延续先前的决策、观察与操作状态，以确保任务连续性。现有评估方法仅依赖任务成功率和记忆保留率，无法全面衡量交接过程中信息传递的有效性。为此，论文提出手稿状态连续性基准测试（Handoff State Continuity Benchmark, HSC-Bench），包含1,039个任务，系统评估接收模型在两项关键能力上的表现：一是从存储库中恢复相关证据的能力，二是延续已建立的对话状态、工具使用状态及文件操作状态的能力。实验对比六种接收模型表明，结构化的交接状态可显著提升所有模型在证据恢复上的表现，并改善多数模型的任务延续性能；且在两项任务指标中表现最优的模型不一致，说明仅依赖证据恢复能力无法全面刻画接收模型在继承状态下的综合表现。通过结合后续响应与行为分析，HSC-Bench不仅评估对历史证据的访问能力，还检验模型对既定承诺和已完成工作的遵循程度，从而为接收模型的选择及交接表示的设计提供更全面的评估依据。
链接: https://arxiv.org/abs/2610.36523
作者: Wenyuan Li,Guang Li,Dongyuan Li,Keisuke Maeda,Takahiro Ogawa,Miki Haseyama
机构: Hokkaido University(北海道大学); The University of Tokyo(东京大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:As agents collaborate on longer tasks, responsibility may pass between models before the work is complete. The receiving model must build on prior decisions, observations, and actions while addressing the remaining requirements. Evaluating this transition requires connecting the information preserved at handoff to how the receiver continues the work, a relationship that task-success and memory-retention scores alone do not capture. We introduce the Handoff State Continuity Benchmark (HSC-Bench), comprising 1,039 tasks that evaluate two complementary aspects of task takeover: recovering relevant repository evidence and continuing from established conversational, tool, and file-work state. Comparisons across six receiving models show that structured handoff state improves repository evidence recovery for every model and continuation performance for most models. The strongest receivers differ between the two tracks, showing that evidence-recovery rankings alone do not characterize performance when continuing from inherited state. By combining these comparisons with checks of subsequent responses and actions, HSC-Bench evaluates both access to prior evidence and adherence to established commitments and completed work. This provides a basis for selecting receiving models and assessing handoff representations according to the work that must be carried forward.
[NLP-71] Folding Attention Biases into Keys Makes Compacted KV Caches Run on Standard Kernels
【速读】： 该论文旨在解决大模型推理过程中键值缓存（KV cache）存储开销过大的问题，特别是在使用注意力匹配（Attention Matching, AM）技术压缩缓存时，因需额外维护偏置项（bias）而导致与现有系统（如FlashAttention和vLLM的分页缓存）不兼容的问题。其核心解决方案是将每头（per-head）的偏置项沿一个通过上下文查询拟合的闭式方向折叠到键向量中，从而生成一个无需额外偏置信息、与原始模型查询路径完全一致的标准化缓存结构。该方法在保持AM压缩率和头维度的同时，实现了与原生模型架构的无缝集成；输出误差由查询投影在该方向上的偏离程度决定，理论上可被有效控制。实验表明，在等头预算下，未经修改的vLLM可通过一个简单的KV连接器处理折叠后的缓存，且在51个Qwen3-4B问答任务中与Hugging Face结果完全一致，并在200个Qwen3-14B LongBench样本中匹配其贪婪输出；在多个基准测试（QuALITY和LongHealth）中，折叠方案在13/14设置下与AM精度相当，且在12个对比场景中均优于仅丢弃偏置项的方案，体现出更高的保真度。在公开的头级预算下，对Qwen3-4B、Instruct-2507及Llama-3.1-8B的评估显示，该方法在9/10场景中与AM精度无显著差异，且不再在KL散度上持续优于丢弃偏置的策略，二者在准确率上各有胜负，验证了其有效性与鲁棒性。
链接: https://arxiv.org/abs/2610.36520
作者: Utkarsh Kanwat
机构: 独立研究者(Independent Researcher)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: 30 pages, 2 figures, 18 tables
Abstract:Attention Matching (AM) compacts a KV cache by keeping some keys per head, refitting their values and adding a bias to each kept key’s logit. FlashAttention and vLLM’s paged caches have no input for it. We fold the bias into the keys along one closed-form direction per head, fitted on context queries, giving a plain cache with AM’s size and head dimension and the model’s own query path; the output error is bounded by how far each query’s projection on that direction is from 1. With equal per-head budgets, unmodified vLLM serves folded caches through a KV connector, matching Hugging Face’s answers on all 51 Qwen3-4B questions and its greedy output on 180 of 200 Qwen3-14B LongBench samples. With equal budgets on QuALITY (Qwen3-1.7B, 4B, 8B, Llama-3.1-8B) and LongHealth (three Qwen3 models), the fold is within noise of AM’s accuracy in 13 of 14 settings, with lower KL to the full cache than dropping the bias in all 12 where both ran. Under the published per-head budgets, on Qwen3-4B, its Instruct-2507 variant and Llama-3.1-8B, it is within noise of AM in 9 of 10, no longer beats dropping the bias in KL, and the two each win once in accuracy.
[NLP-72] MedConstBench: Benchmarking How Constitutions Steer LLM s in Medical Ethics
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在医疗场景部署后，其伦理行为受“部署宪法”（deployment constitution）影响却缺乏系统性研究的问题。现有工作多聚焦于事实准确性，忽视了不同伦理价值之间的权衡，且将伦理视为模型的固定属性，未将其作为可调控变量进行分析。为此，作者提出MedConstBench，一个以贝奥查姆（Beauchamp）四原则为框架、将“部署宪法”作为可控变量、并通过人类校准的多模型同行评审机制评分的医疗伦理基准。该基准包含自建的临床困境数据集MedDilemma，支持对宪法如何引导模型伦理行为的系统性分析。实验评估表明，宪法的影响具有维度选择性和模型依赖性，不存在普适最优宪法。基于此，作者提出一种“不伤害”自适应选择器（do-no-harm adaptive selector），可根据具体模型动态选择最优宪法，在保持接近各模型最优表现的同时，确保最差情形下的底线性能，并以低开销实现对未见模型的良好泛化能力。
链接: https://arxiv.org/abs/2610.36513
作者: Yibo Yang,Zerui Huang,Haomiao Yang,Zixia Wang
机构: Xinjin District People’s Hospital, China; University of Exeter, United Kingdom
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 8 pages, 3 figures, 8 tables
Abstract:As large language models advance rapidly, they are increasingly used to answer medical questions, and many such models are now deployed locally. Once a model is deployed, its behavior is governed mainly by the ethical policy written into its constitution. However, most existing work focuses on factual correctness rather than the trade-off between competing ethical values. Even benchmarks that do evaluate ethics treat it as a fixed property of the model rather than as a variable, so how a constitution reshapes medical-ethics behavior remains largely unexplored. We introduce MedConstBench, a medical-ethics benchmark that treats the deployment constitution as a controlled variable, is organized around Beauchamp’s four principles, and is scored by human-calibrated multi-model peer review. It includes MedDilemma, our own dataset of clinical dilemmas, and supports a systematic analysis of how constitutions steer LLMs. Evaluating ten open-source LLMs under four constitutions, we find that the effect of a constitution is dimension-selective and model-dependent, and that no constitution is universally optimal. Motivated by these findings, we propose a do-no-harm adaptive selector that chooses a constitution per model, matching the per-model oracle while guaranteeing a worst-case floor and generalizing to unseen models at low cost.
[NLP-73] Do AI Scientists Know Whats New?
【速读】： 该论文旨在解决生成式人工智能在科学发现流程中难以准确判断生成想法是否具有真正新颖性的问题。核心挑战在于：现有新颖性检测机制在拒绝非新颖想法的同时，容易误判后续被人类研究者证实为有效的创新性想法，且缺乏对提案中具体组成部分的可解释性分析能力，往往以“黑箱”方式整体驳回整个提案。其解决方案的关键是提出一种细粒度、可解释的基于声明（claim-level）的新颖性检测模块——ClaimNov。该模块利用带有时间戳真实标签的多层级评估框架（涵盖提案与声明两个层面），实现了对每项科学主张的独立评估。实验表明，ClaimNov在仅1.5%的误报率下可捕获97%的伪装论文，并将大语言模型（LLM）裁判的误报率从11.8%显著降低至1.5%；同时，在保留已有贡献基础上推进创新方面表现优异，仅拒绝3%的包含争议贡献但整体被接受的ICLR论文，且能成功保留91/100个含拼接先验贡献的真正创新提案，优于现有最强检测方法的76个。该方法通过基于声明级别的证据支持决策，显著提升了新颖性评估的准确性与透明度。
链接: https://arxiv.org/abs/2610.36511
作者: Miri Liu,Xiayu Zhao,Encheng Xie,ChengXiang Zhai
机构: University of Illinois, Urbana-Champaign (伊利诺伊大学厄本那-香槟分校)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:A major challenge facing AI scientific discovery pipelines is how to determine when a generated idea is meaningfully novel. Ideally, novelty checks in these pipelines must be able to reject non-novel ideas while not rejecting too many novel ideas, and they should also be able to distinguish which parts of a proposal are non-novel, as opposed to opaquely discarding entire proposals. We design an analytical framework for evaluation of novelty checks with temporal ground truth at both the proposal- and claim-level, using four datasets of 1,060 proposals built from published papers, reviewer disputes, and generated ideas matching ones the field went on to publish. Auditing existing novelty checks with this framework, we find that checks with high rejection rates on non-novel work also tend to have high false alarms on ideas that are later realized by human researchers, and observe that checks built to evaluate proposals as a whole can point to claim-level prior art only by rejecting the whole proposal. We therefore introduce ClaimNov, a fine-grained, interpretable claim-level novelty module. ClaimNov catches 97% of disguised papers at a 1.5% false-alarm rate, and its claim-level novelty profile cuts the LLM judge’s false alarms from 11.8% to 1.5%. Because its verdict rests on claim-level evidence, ClaimNov also keeps proposals that build on established contributions, rejecting only 3% of ICLR-accepted proposals with a reviewer-disputed contribution and keeping 91 of 100 novel proposals with one spliced prior contribution, against 76 for the strongest existing check. Code and data are available at this https URL.
[NLP-74] AC-JEPA: Anchored Cross-Modal Joint Embedding Predictive Architecture for Visual-to-Symbolic Translation
【速读】： 该论文旨在解决跨模态翻译中因真实科学领域数据对存在不完美性、且大规模获取高质量配对数据成本高昂所导致的模型训练难题。现有方法中，监督微调（SFT）虽能通过词元级分类监督优化翻译性能，但对目标序列中的错误极为敏感；而现有的基于联合嵌入预测自监督学习（JEPA）框架多聚焦于模态内预测，难以有效支持跨模态语义对齐。为此，本文提出锚定式跨模态JEPA（AC-JEPA），其核心创新在于引入一种兼容SFT的插件式架构，利用配对数据在两个跨模态方向上预测被掩码的中间状态，从而建立联合嵌入空间下的跨模态结构一致性约束。该联合嵌入目标捕捉由上下文可预测的跨模态结构信息，同时保留直接的词元级监督路径以维持目标序列的精细结构。在自训练阶段，AC-JEPA可从由不完美伪标签对生成的连续潜在目标中学习，具备对局部数据误差的内在鲁棒性，且不会将伪词元强化为硬性分类监督。实验以光学化学结构识别为测试场景，结果表明：在仅使用清洁数据时，AC-JEPA相较纯SFT提升准确率至76.07%（对比67.40%）；而在相同伪标签数据下进行自训练，AC-JEPA进一步达到80.23%准确率，而纯SFT准确率则下降至63.47%。这验证了“锚定式联合嵌入预测”策略在有监督与弱监督场景下均能有效提升跨模态翻译性能，其关键优势在于融合了联合嵌入的结构感知能力与SFT的精细监督机制，实现了对噪声数据的稳健适应。
链接: https://arxiv.org/abs/2610.36493
作者: Chaofei Hong,Ziyang Chen,Baochen Wang,Christina Yi Jin,Qing Zhang
机构: Zhejiang Lab(浙江省实验室)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Cross-modal translation requires models to capture shared semantics across modalities while preserving the fine-grained structure required by the target modality. Yet in real-world scientific domains, paired data are often imperfect, while obtaining clean pairs at scale is costly. Supervised fine-tuning (SFT) directly optimizes cross-modal translation through categorical token-level supervision, but this objective is sensitive to errors in the target sequences. JEPA offers a complementary representation-space objective, but most existing formulations focus on prediction within a modality rather than translation across modalities. We introduce Anchored Cross-Modal JEPA (AC-JEPA), an SFT-compatible plug-in that predicts masked intermediate states in both cross-modal directions using paired data. The joint-embedding objective targets cross-modal structure predictable from context, while the causal SFT path retains direct token-level supervision for the target sequence. During self-training, AC-JEPA learns from continuous latent targets derived from imperfect pseudo-labeled pairs, offering intrinsic tolerance to local data errors without promoting pseudo-tokens to hard categorical supervision. Using optical chemical structure recognition as a testbed, AC-JEPA trained on clean pairs achieves 76.07% accuracy on the Simple set of the WildMol benchmark, compared with 67.40% for SFT alone. Using the same pseudo-labeled data for both approaches, AC-JEPA-based self-training further raises accuracy to 80.23%, whereas accuracy under direct SFT drops from 67.40% to 63.47%. These results support SFT-anchored joint-embedding prediction as an effective strategy for cross-modal translation under both curated and imperfect supervision.
[NLP-75] A Heart Means Recursion: Exploring Visual Meaning and Communication in VLMs
【速读】： 该论文旨在探究视觉语言模型（Vision-Language Models, VLMs）是否具备类人般的视觉交流能力，特别是其在跨模型视觉通信中的有效性与一致性。研究发现，即使是最简单的视觉符号（如红色心形），也能显著影响模型在看似无关任务（如编程问题中诱导递归解法）中的行为表现，但这种影响机制具有高度隐含性——模型对视觉符号的直接解释与实际行为之间存在显著偏差。这一现象揭示了模型在视觉通信中存在“生成-理解鸿沟”：尽管模型倾向于使用直观的形状基元进行表达，这些视觉线索却难以有效引导其他模型产生预期响应。在涉及抽象概念和叙事的任务中，模型虽能创造性地利用视觉元素传递意义，但部分通信方式对人类而言仍缺乏直观可理解性。因此，该研究的关键在于揭示VLMs在视觉语义生成与理解上的非人类惯例特征，表明其虽具备丰富、创造性的视觉化沟通潜力，但其视觉意义的生产与解读机制与人类认知存在根本差异，这既凸显了人机视觉通信的巨大前景，也指出了实现高效协作的核心挑战。
链接: https://arxiv.org/abs/2610.36449
作者: Alexa R. Tartaglini,Zhengxuan Wu,Thaddäus Wiedemer,Martin Wattenberg,Been Kim,Yuxuan Li
机构: Google DeepMind(谷歌深思维); Stanford University (斯坦福大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Do vision-language models (VLMs) share our ability to communicate visually? We share preliminary findings, starting from a surprising observation: simple visual symbols like a red heart can strongly influence model behavior on seemingly unrelated tasks, such as steering models towards recursive solutions on simple coding problems. Notably, this steering effect is largely implicit, as models’ direct interpretations and thinking traces about visual symbols diverge substantially from their observed behavior. In visual communication between models, this leads to a generation–interpretation gap: while models tend to send shape primitives that they (and humans) find intuitive, these visual cues often fail to steer other models. In further communication games involving abstract concepts and stories, models can creatively use visuals to transmit meaning to one another, although some communications remain non-intuitive to humans. These findings reveal that VLMs might be capable of rich, creative, and visually-grounded communication, but their interpretation and production of visual meaning diverges from human conventions in surprising ways. Together this work highlights both the promise and challenges of visual communication between humans and VLMs.
[NLP-76] CONCUR: Concordant Auxiliary Rewards for Multi-Task Multi-Reward Reinforcement Learning
【速读】： 该论文旨在解决在使用强化学习（Reinforcement Learning, RL）对大语言模型（Large Language Model, LLM）进行多任务训练时，主奖励（primary reward）与辅助奖励（auxiliary reward，如长度奖励）之间因“门控效应”（gating effect）导致的排序失真问题。具体而言，当主奖励较低时，辅助奖励被限制在主奖励的上限内，而现有的群体相对优势估计器（group-relative advantage estimators）未考虑这一约束，可能导致辅助奖励反向干扰主奖励的正确排序，甚至在主奖励达到上限时仅由辅助奖励决定响应排序，从而损害任务性能。针对此问题，本文提出CONCUR，一种新型优势估计方法：其核心在于将每个被门控的辅助奖励的标准化得分替换为满足两个条件的最接近值——即主奖励更高的响应不得获得更低的得分，且在主奖励达到上限时所有响应应获得相等得分。CONCUR-T在整个训练过程中强制执行这两条规则，而CONCUR-A则对在通过样本中排序一致性高于反向频率的辅助奖励放宽第二条规则。在来自两个模型家族的三个模型上，涵盖数学、代码生成、指令遵循和写作等多种任务的实验表明，两种变体均显著优于所有基线方法，也优于完全移除被门控辅助奖励的情况，证明了该方法在保持多任务学习有效性方面的优越性。
链接: https://arxiv.org/abs/2610.36437
作者: Jixuan Leng,Si Si,Fei Wang,Cho-Jui Hsieh,Jiong Zhang,Hsiang-Fu Yu,Inderjit S. Dhillon
机构: CMU(卡内基梅隆大学); Google DeepMind(谷歌深度思维); Google(谷歌); UCLA(加利福尼亚大学洛杉矶分校)
类目: Computation and Language (cs.CL)
备注:
Abstract:When a Large Language Model (LLM) is trained with Reinforcement Learning (RL) on multiple tasks, a task’s primary reward can gate auxiliary rewards such as a length reward: on a failing response, an auxiliary cannot exceed the primary. However, group-relative advantage estimators treat a gated auxiliary like any other component, so it can reverse the primary’s order, and at the primary’s ceiling only auxiliaries order responses. Keeping gated auxiliaries hurts some tasks, while dropping them hurts others. We propose CONCUR, an advantage estimator that replaces the standardized scores of each gated auxiliary with the closest scores that satisfy two conditions: responses with higher primary rewards cannot receive lower scores, and responses at the primary’s ceiling receive equal scores. CONCUR-T applies both conditions throughout training, and CONCUR-A lifts the second for an auxiliary that agrees with the primary’s order more often than it reverses it among passing responses. On three models from two families and across math, code, instruction following, and writing, both variants score higher on average than all baselines and than when all the gated auxiliaries are dropped.
[NLP-77] No Tokenizer Still a Gap: Cross-Lingual Fairness of Homoglyph and Invisible-Character Watermarking
【速读】： 该论文旨在解决生成式文本水印在低资源语言中性能下降的问题，特别是针对传统生成时水印（generation-time watermarks）因分词器（tokenizer）对文本的处理方式差异而导致的跨语言不均衡问题。其核心解决方案是采用基于编辑的水印技术，如同形异义字符（homoglyph）和不可见字符（invisible-character）水印，这些方法不依赖于模型生成过程，而是通过替换字符为视觉相似但编码不同的同形字符或插入不可见字符来嵌入信号，从而绕过分词器的影响。研究发现，水印容量在不同语言间存在显著差异（3.3倍），且96%的变异由书写系统决定；在拉丁字母体系内，资源水平对容量无显著影响（ρ=+0.095, p=0.78），但在跨书写系统时，资源越低的语言通常使用同形字符较少的书写系统，导致水印容量更低。此外，常规复制粘贴操作会移除不可见字符，进一步加剧语言间的不平等——例如，英语或波斯语仅需约2个码点即可检测，而老挝语则需93个码点。这表明，尽管去除分词器可规避部分瓶颈，却并未消除不平等，反而将问题根源从分词器转向了书写系统的固有特性。
链接: https://arxiv.org/abs/2610.36432
作者: Alexander Nemecek,Yuqiao Xu,Joseph Falkenburg,Erman Ayday
机构: Case Western Reserve University(凯斯西储大学)
类目: Computation and Language (cs.CL); Cryptography and Security (cs.CR); Computers and Society (cs.CY)
备注: 19 pages
Abstract:Homoglyph and invisible-character watermarks are edit-based watermarks. Instead of changing how a model generates text, they hide a signal in the text after it is written. They swap some characters for homoglyphs, which look the same, or very similar, but have different Unicode code points, or they insert characters that do not display, all without access to the model that wrote the text. Generation-time watermarks work less well for lower-resourced languages due to how tokenizers split their text. Because they work on individual characters, these watermarks never touch a tokenizer, so we ask whether they escape this gap. Across 40 language varieties in 24 scripts, we measure capacity (the hidden bits each character can carry), how much of it survives routine handling and targeted removal, and how much text a keyed detector needs. Capacity varies 3.3-fold and script accounts for 96% of that variation. Within Latin script, resource level makes no detectable difference ( \rho=+0.095 , p=0.78 ), but across scripts capacity rises with it ( \rho=+0.54 , 95% CI [+0.14,+0.83] ), because lower-resourced languages tend to use scripts with few homoglyphs. Ordinary copy-paste, which strips invisible characters, widens the gap: a detector then needs about 2 code points of English or Persian but 93 of Lao. Removing the tokenizer does not remove the inequity, but instead only changes its source.
[NLP-78] SOMA: A Benchmark for Fine-Grained Egocentric Behavior and Body-Motion Understanding
【速读】： 该论文旨在解决当前多模态大语言模型（Multimodal Large Language Models, MLLMs）在从第一人称视角视频中进行细粒度人体运动理解方面能力不足的问题。现有评估体系主要关注物体、场景和粗粒度动作识别，而对身体姿态、运动动态及时间推理等更精细的体感认知能力缺乏有效衡量标准，尤其在身体部分被遮挡、信息时空分散的第一人称视角下更为严峻。为此，作者提出SOMA基准，包含1,632个经人工验证的问题，覆盖身体构型、运动动力学与时间推理三大能力维度。实验表明，人类仅能以68.5%的准确率完成该任务，而表现最优的闭源模型Qwen3.6-27B在零样本条件下也仅达50.0%准确率，显示出显著差距。错误分析揭示了模型普遍存在对佩戴者自身身体状态误判、时间事件判断偏差以及对右手法偏好、重复动作计数不足等系统性偏见。为缩小这一差距，研究提出了一个包含对齐视频、人体运动与语言标注的训练数据集，并构建了一个联合学习多模态信号的基线模型。通过识别当前MLLMs在身体运动理解中的共性失败模式与缺失能力，SOMA旨在推动未来在人体运动感知与多模态推理方向的研究进展。
链接: https://arxiv.org/abs/2610.36384
作者: Matea Tashkovska,Naël Ouerghemi,Haozhe Qi,Alexander Mathis
机构: École Polytechnique Fédérale de Lausanne (EPFL), Switzerland
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Neurons and Cognition (q-bio.NC)
备注: Code and data at: this https URL
Abstract:Perceiving human body motion is critical for multimodal systems that understand and assist people. Yet current evaluations of Multimodal Large Language Models (MLLMs) largely focus on objects, scenes, and coarse actions, leaving their ability to reason about the body itself poorly assessed. This is especially challenging from the egocentric perspective, in which the body is only partially visible and motion must be inferred from occluded and temporally distributed cues. We introduce SOMA, a benchmark for fine-grained body-motion understanding from egocentric video, comprising 1,632 human-verified questions across three capability axes: body configuration, motion dynamics, and temporal reasoning. We evaluate 37 leading closed-source, open-source, and specialized egocentric MLLMs, complemented by an extensive human evaluation. We find that humans also find this task difficult from video alone (68.5% accuracy), yet clearly outperform the strongest MLLM (Qwen3.6-27B), which reaches 50.0% zero-shot accuracy. Our error analysis shows that MLLMs often misread the wearer’s own body, misjudge events in time, and are prone to biases such as favoring the right hand or under-counting repetitions. To narrow the gap to human performance and support further progress, we introduce a training dataset of aligned video, body motion, and language supervision, and a multimodal baseline that jointly learns from these modalities. By identifying common failure modes and missing capabilities in current MLLMs, we hope SOMA will inspire future research on body-motion understanding and multimodal reasoning. Full benchmark, training data, and code will be released at: this https URL
[NLP-79] AK_QMUL at ArGuard Shared Tasks: A Single Fine-Tuned JAIS for Arabic Harmful-Prompt Detection EMNLP2026 ACL
【速读】： 该论文旨在解决阿拉伯语有害提示检测中的两大挑战：二分类的安全/不安全判断（B1任务）与七分类的有害领域识别（B2任务）。其核心解决方案在于采用经过4-bit QLoRA微调的JAIS-6.7B作为统一骨干模型，并在其最后一层标记表示上添加线性分类头。针对B1任务，通过引入MSA（Modern Standard Arabic）数据增强训练集，构建四分类头以区分方言注册类型下的安全/不安全标签，采用argmax解码实现最终分类；针对B2任务，则将同一骨干模型微调为七分类器，并对五个随机种子的softmax输出进行平均以提升鲁棒性。该方法在两项任务中均取得最优表现，宏平均F1分别为0.984（B1）和0.790（B2），且在各方言及类别上的性能均衡，仅在“欺凌”（Bully）类别上存在因语义重叠导致的残余误差。
链接: https://arxiv.org/abs/2610.36360
作者: Ali Khairallah,Arkaitz Zubiaga
机构: Queen Mary University of London (伦敦玛丽女王大学)
类目: Computation and Language (cs.CL)
备注: Published at ArabicNLP 2026 (Shared Tasks), co-located with EMNLP 2026. ACL Anthology: this https URL
Abstract:We present AK_QMUL’s submissions to the ArGuard-2026 shared tasks on Arabic harmful-prompt detection: a binary safe/unsafe task (B1) and a seven-way harm-domain task (B2). Both submissions use one backbone – JAIS-6.7B fine-tuned with 4-bit QLoRA and a linear classification head over its last-token representation. For B1 we train a four-way head that factors the label by dialect register safe, unsafe x MSA, other, we augment the training fold with MSA prompts from AraSafe, and decode with a plain argmax to score safe/unsafe. For B2 we fine-tune the same backbone as a seven-way classifier and average the softmax outputs of five random seeds. Both submissions ranked first on their track, with macro-F1 0.984 on B1 and 0.790 on B2. We report results by dialect and by class: on B1 the model is accurate across all six dialects (93.6–99.8%) and its predicted unsafe rate tracks the true rate closely; on B2 six of the seven classes reach F1 between 0.73 and 0.98, with the residual error concentrated in the Bully class, which overlaps semantically with harassment and harm-to-others.
[NLP-80] Detect Before You Disambiguate: Perturbation-Based Input Ambiguity Detection in LLM s AACL
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在面对输入歧义时，倾向于将其视为具有单一明确含义的问题。这种处理方式忽略了输入歧义作为不确定性上游来源的本质，导致模型在未充分识别歧义的情况下即进行回答，从而影响输出的可靠性。其核心问题在于：当一个提示（prompt）存在多种合理解释时，模型是否应直接作答，还是应当首先检测并承认歧义的存在。现有方法通常依赖于模型生成的改写或澄清语句来衡量不确定性，但这类方法在测量前可能已对输入进行了隐式消歧，造成评估偏差。为此，论文提出一种标签保持性扰动框架（label-preserving perturbation framework），通过在不改变原始输入标签的前提下，施加贴近原输入的微小扰动，以实现对输入歧义的准确检测。该框架的关键创新在于区分“输入侧”与“输出侧”的不确定性检测——强调在尝试消歧之前先识别歧义。具体实现上，作者构建了两种检测器：一种是黑盒规则驱动的扰动库，另一种是基于模型内部信号的重要度引导的白盒方法。实验在四个开放域问答、对话式问答及视觉问答数据集上验证了该方法的有效性，结果表明其显著优于基于输出端不确定性估计和基于LLM重述的基线方法。研究结果强调了在消歧前进行输入歧义检测的重要性，为提升模型鲁棒性与可解释性提供了新的思路。
链接: https://arxiv.org/abs/2610.36353
作者: Tiejin Chen,Xiaoou Liu,Michael Clifford,Ahmadreza Moradipari,Yitao Chen,Kyungtae Han,Nejib Ammar,Hua Wei
机构: Arizona State University(亚利桑那州立大学); Toyota InfoTech Labs(丰田信息科技实验室)
类目: Computation and Language (cs.CL)
备注: Accepted to AACL-IJCNLP 2026 Main
Abstract:Large language models (LLMs) often answer ambiguous inputs as if they had a single intended meaning. This is problematic because input ambiguity is an upstream source of uncertainty: when a prompt admits multiple valid interpretations, the key question is whether the model should answer directly at all. Existing input uncertainty methods often rely on LLM-generated statements or clarifications, but such transformations may already disambiguate the input before the ambiguity is measured. We propose a label-preserving perturbation framework for input ambiguity detection in which perturbations remain close to the original input while preserving its label. We instantiate this principle with two detectors: a black-box rule-based perturbation library and a white-box importance-guided method using model-internal signals. Experiments across four datasets in open-domain, conversational, and visual QA show that our methods outperform output-side uncertainty baselines and LLM-paraphrase-based perturbation baselines. Our results highlight the importance of detecting ambiguity before attempting to disambiguate it. Our code is available \hrefthis https URL\textcolorpurplehere.
[NLP-81] Cross-modal Steering Vectors Exist Across Both Image and Text Generation
【速读】： 该论文旨在解决当前人工智能模型中跨模态概念与机制是否实现统一，抑或在各模态间独立实现的问题。其核心关注点在于跨模态控制（cross-modal steering）：即在某一模态（如文本）上成功引导生成的向量，能否在零样本条件下（zero-shot）直接应用于另一模态（如图像）并产生一致、可解释的生成效果。区别于传统的相关性对齐方法（如表征相似性），该研究强调因果性，通过分析统一多模态模型（Unified Multimodal Models, UMMs）中语言与图像共享参数的特性，验证了跨模态转向向量的可迁移性。研究发现，无论是在情感、物体尺寸等通用概念，还是在高度依赖特定模态的语义（如正式与非正式语言）上，转向向量均能在两个方向（文本→图像、图像→文本）实现零样本转移，并表现出一致且可解释的效果（如正式风格的服装与设计）。此外，具有强跨模态可迁移性的向量在不同模态的嵌入空间中呈现出相似的几何方向，表明尽管存在显著的模态差异，但关键概念在统一模型中仍具备共通的表征结构。该研究的关键突破在于揭示了统一多模态模型中视觉与语言之间深层交互机制的存在，并为构建更普适、模态无关的可控生成方法提供了理论依据与实践路径。
链接: https://arxiv.org/abs/2610.36269
作者: Donghao Zeng(1),Nishant Subramani(2),Tamar Rott Shaham(3),Benno Krojer(1 and 4) ((1) Mila, McGill University, (2) Carnegie Mellon University, (3) MIT CSAIL, (4) Northeastern University)
机构: Mila, McGill University(蒙特利尔人工智能研究所，麦吉尔大学); Carnegie Mellon University(卡内基梅隆大学); MIT CSAIL(麻省理工学院计算机科学与人工智能实验室); 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: Pre-print. Code: this https URL Data: this https URL Demo: this https URL
Abstract:Are concepts and mechanisms unified across modalities in current AI models, or instead implemented separately for each one? We operationalize this broad question of modality unification by studying cross-modal steering: 1) finding vectors that successfully steer generation on tokens of one modality, which then 2) transfer zero-shot to steer generation when added to tokens of another modality. Our study goes beyond simple correlational alignment (e.g. representational similarity) because steering is fundamentally causal. We study cross-modal steering in several unified multimodal models (UMMs), since they possess shared parameters for language and image generation, focusing on these two modalities. Across diverse steered concepts such as emotion and object size, we show that steering vectors transfer zero-shot in both directions (text-to-image image-to-text). Even for concepts that are highly unique to one modality (e.g. formal vs. colloquial language), steering vectors often induce a consistent, interpretable effect in the other modality, such as formal clothing and design. This transfer has a geometric signature too: despite a clear modality gap separating text and image representations, steering vectors with strong cross-modal transferability tend to point in similar directions for the same concept. Our findings shed light on how vision and language interact inside unified multimodal models and point towards more general, modality-agnostic approaches for controlling their behavior.
[NLP-82] Beyond IceBERT: Open-Data ModernBERT Encoders for Icelandic from 12M to 1B Parameters
【速读】： 该论文旨在解决冰岛语（Icelandic）领域中大规模语言模型的训练效率与性能优化问题，特别是针对单语冰岛语编码器在不同规模和预训练策略下的表现。其核心问题是：是否仅通过清洗后的网络爬取数据即可训练出高性能的冰岛语编码器？以及从多语言模型进行热启动（warm-start）是否能在特定任务上带来显著优势。解决方案的关键在于采用ModernBERT架构，在90亿词元的开放数据上训练了参数量从1200万到10亿不等的30个冰岛语编码器，并在统一评估协议下对11项任务进行系统性基准测试，同时提供置信区间以增强结果可靠性。研究发现，从零开始训练的10亿参数单语冰岛语编码器在所有任务上均显著优于IceBERT；而使用相同规模的英文Ettin编码器作为热启动源时，其在所有任务上的表现均达到最优。进一步分析表明，随着模型规模增大，迁移效果主要来自源模型的Transformer层结构，即使源语言为英语也能有效提升性能；而在保持源模型词汇表的情况下，虽能避免显著性能下降，但会带来高达5点的F1分数损失及近三倍的计算开销，凸显了词汇表移植的代价。因此，关键突破在于验证了跨语言热启动在大规模单语模型中的有效性，并揭示了模型结构与词汇表选择之间的权衡关系。
链接: https://arxiv.org/abs/2610.36230
作者: Hafsteinn Einarsson
机构: University of Iceland(冰岛大学)
类目: Computation and Language (cs.CL)
备注: 20 pages, 1 figure, 11 tables. Models: this https URL
Abstract:IceBERT, the reference encoder for Icelandic, found that a cleaned web crawl is enough to train a strong model and that starting from a multilingual model can help on some tasks. We revisit both findings with the ModernBERT architecture. We train Icelandic encoders from 12M to 1B parameters on 9B tokens of open data, and rerun IceBERT’s benchmarks and five newer tasks for 30 encoders under one protocol with confidence intervals. Our 1B model trained from scratch, to our knowledge the first monolingual Icelandic encoder of that size, is significantly better than IceBERT on all eleven tasks; warm-started from the English Ettin encoder of the same shape, it has the best score of all encoders on every task we rank. Because our model shapes match the English Ettin encoders and the multilingual mmBERT, we can also test the warm start under controlled conditions, at six model sizes and four token budgets. With our vocabulary transplanted, a warm start was never significantly worse than scratch training on the same Icelandic text, and from base size up it was not significantly worse on four tasks than scratch training on 4 times the text (8 times at base and 1B). As models grow, most of the transfer comes from the source model’s transformer layers, so even an English source helps. Keeping the source vocabulary instead cost 5 F1 on question answering and nearly three times the compute.
[NLP-83] Base Models Can Detect Hidden Intentions that Prompted Classifiers Miss
【速读】： 该论文旨在解决大模型在后训练（post-training）过程中可能隐含并传递未明确声明的潜在人格特征（persona）的问题，尤其是这些特征以隐蔽信号的形式嵌入到模型生成的看似无关的内容中（如数字序列）。其核心挑战在于如何检测和识别这些潜藏于模型输出中的非显性偏好或意图，而传统方法难以捕捉此类细微偏差。解决方案的关键是提出一种名为“人格头似然评估”（Persona-Header Likelihood Evaluation, PHLE）的方法：通过在基础预训练模型上添加特定的“人格头”（persona header），并计算生成输出在该头部下的似然值，从而解码出隐藏的子意识信号。实验表明，即使经过去偏处理、使生成内容表面无明显异常，PHLE仍能有效区分不同人格驱动下的输出（如“喜爱猫头鹰”与“喜爱鹰”），且该信号在不同模型家族间具有跨模型迁移性（如Qwen与Olmo之间的信号可被读取），验证了人格选择模型（PSM）的非平凡预测能力。该方法为检测和审计模型内部隐藏目标提供了可操作的技术路径。
链接: https://arxiv.org/abs/2610.36202
作者: Oz Amram
机构: 独立研究人员(Independent)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 28 pages, 10 figures. Code available at this https URL
Abstract:The Persona Selection Model (PSM) posits that post-training selects a persona the model learned during pre-training. Work on subliminal learning has shown that models can impart covert signals (like “love of owls”) into semantically unrelated outputs like strings of numbers, and that students trained on these outputs inherit those traits. In this work I show that by evaluating the likelihood of such outputs under the base pre-trained model with an appropriate header, a method I call persona-header likelihood evaluation (PHLE), one can decode these subliminal signals and classify between them. For example, PHLE distinguishes number sequences generated by a post-trained model prompted with “You love owls” from those prompted with “You love eagles”, and distinguishes answers from a model with a secret desire to harm the user, filtered to remove any noticeable misalignment, from those of a model with no system prompt. In both cases prompted classifiers, including the generating model itself, remain near chance. The signal partially persists across model families, with the Qwen base model reading signals generated by Olmo and vice versa. This is evidence for a non-trivial prediction of the PSM, and the method may be practically useful for auditing the hidden objectives of models.
[NLP-84] Learning to Read Braille: Benchmarking and Improving Braille Recognition Capabilities of VLMs
【速读】： 该论文旨在解决多模态系统在真实世界场景下从图像中读取盲文（Braille）这一关键可及性任务中的性能瓶颈问题。当前主流视觉语言模型（VLMs）在直接转录盲文图像方面表现不佳，暴露出在特定无障碍应用场景下的能力缺陷。其解决方案的关键在于构建一个高质量的实拍盲文图像基准测试集——BrailleBench，涵盖259张真实的Grade-1与Grade-2盲文图像，并设计了一套包含多样化版式与图像扰动的合成盲文生成流水线。通过在该合成数据上训练的20亿参数级视觉语言模型，实现了对Grade-1盲文87%的精确匹配率和Grade-2盲文61%的精确匹配率，显著优于独立运行的一般性多模态模型，且在所评估系统中达到最高的Grade-2盲文转录准确率。研究还揭示了模型在合成验证集上高精度并不等同于在真实照片上的高性能，强调了针对真实场景进行针对性合成监督的重要性，从而为提升盲文图像识别的鲁棒性与实用性提供了有效路径。
链接: https://arxiv.org/abs/2610.36195
作者: Kartik Jangra,Atharva Gawane,Siddharth Pulipaka,Hemachandran Karah,Raj Dabre
机构: Nilekani Centre at AI4Bharat; Indian Institute of Technology, Madras; Mahindra University
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL)
备注: 22 pages, 8 figures, 16 tables. Code and data: this https URL
Abstract:Despite rapid progress in visual understanding and document transcription, modern vision-language models (VLMs) struggle with a fundamental accessibility task: reading Braille from images. We introduce BrailleBench, a manually curated benchmark of 259 real-world Grade-1 and Grade-2 Braille images. We evaluate proprietary, open-weight, OCR-specialized, and tool-assisted multimodal systems and find that standalone general-purpose models perform poorly at directly transcribing Braille images, whereas tool-assisted systems perform much better. We design a synthetic Braille generation pipeline with diverse layouts and image perturbations. A 2B VLM model trained on our synthetic data reaches 87% exact match on Grade-1 and 61% on Grade-2 BrailleBench, outperforming standalone VLMs and achieving the highest Grade-2 exact match score among the evaluated systems. We further show that high accuracy on synthetic validation data does not guarantee high accuracy on real photographs in our experiments: models trained on prior synthetic data reach up to 96% on their synthetic validation set but at most 3.26% on Grade-1 BrailleBench. Our results identify an accessibility-relevant capability gap in multimodal systems and show that targeted synthetic supervision can improve real-world Braille transcription.
[NLP-85] LLM Hallucination Detection via Low-rank Multilinear Decomposition
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在生成回答时频繁出现流畅但事实性错误（即幻觉，Hallucination）的问题，尤其关注如何从高维隐藏状态中有效提取具有判别性的信息以提升幻觉检测的准确性。现有方法多依赖于特定激活值或基于池化的压缩策略，往往忽略分布在Transformer各层、特征模态及词元间的潜在判别信息。其解决方案的关键在于提出一种名为LOTUS（Low-rank Order-statistic TUcker-based Subspace learning）的多线性低秩分解框架：首先利用互补的统计描述符对变长词元表示进行总结，捕捉峰值激活、激活分布及跨层变化等关键特征；随后通过低秩多线性分解技术提取能够反映LLM隐藏状态底层结构的紧凑表征，并用于训练轻量级幻觉分类器。该方法显著降低了对标注数据的依赖，同时在多个大模型和问答基准上实现了优于当前先进基线的幻觉检测性能。
链接: https://arxiv.org/abs/2610.36191
作者: Devansh Sharma,Faizul Rakib Sayem,Shahana Ibrahim
机构: University of Central Florida (中佛罗里达大学)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 26 pages, 3 figures. Code: this https URL
Abstract:Large language models (LLMs) frequently generate fluent yet factually incorrect responses, raising concerns about their reliability. Recent hallucination detectors increasingly exploit internal representations, but extracting informative signals from high-dimensional hidden states remains challenging. Existing approaches often rely on selected activations or pooling-based compression, potentially overlooking discriminative information distributed across Transformer layers, feature modes, and tokens. We propose LOTUS (Low-rank Order-statistic TUcker-based Subspace learning), a multilinear low-rank decomposition framework for hallucination detection that extracts compact representations from LLM hidden states while reducing dependence on labeled data. Our approach first summarizes variable-length token representations using complementary statistical descriptors that capture peak activations, activation distributions, and cross-layer changes. We then leverage low-rank multilinear decomposition to extract compact representations that capture the underlying structure of LLM hidden states, which are subsequently used to train a lightweight hallucination classifier. Experiments across multiple LLMs and question-answering benchmarks demonstrate improved hallucination detection performance over recent state-of-the-art baselines.
[NLP-86] Frankly Speaking: The Company Election Integrity Keeps in Congressional Communication
【速读】： 该论文旨在解决美国政治中两党对“选举诚信”（election integrity）概念理解日益分化的问题，揭示尽管两党使用相同的词汇，却通过语境差异构建出截然不同的叙事。其解决方案的关键在于运用历时词嵌入（diachronic word embeddings）技术，分析2010至2025年间国会发送的10,420封电子邮件新闻简报，追踪“诚信”“安全”“欺诈”等关键词在语义上的演变及其共现模式。研究发现，两党并未依赖不同词汇传递不同含义，而是通过将相同词汇置于迥异的语境之中实现意义分化；例如，民主党将“选民”“选举”“诚信”与“投票权”“行政管理”等词汇关联，强调扩大参与和程序公平；而共和党则将其与“公民身份验证”“非法投票”“边境安全”等关联，聚焦于防止非资格选票。这种语义邻近性的根本分歧虽在整体词汇频率层面不显著，但关键术语如“认证”“被操纵”“被盗”自2020年后显著极化，凸显了选举话语的政治分裂。最终，选举官员需在两种相互矛盾的定义下维持同一套制度运行，承受由此产生的结构性张力。
链接: https://arxiv.org/abs/2610.36187
作者: Michael A. Robinette,Christina Barsky,Amanda D. Clark,Helen Flannery,Juniper Lovato
机构: Vermont Complex Systems Institute, University of Vermont(佛蒙特复杂系统研究所，佛蒙特大学); Community Development and Applied Economics, University of Vermont(社区发展与应用经济学，佛蒙特大学); Department of Computer Science, University of Vermont(计算机科学系，佛蒙特大学); Complexity Science Hub, Vienna(复杂性科学中心，维也纳)
类目: Computation and Language (cs.CL)
备注:
Abstract:Election integrity is a shared commitment in American politics, but Democrats and Republicans in Congress increasingly mean different things by it. Analyzing 10,420 congressional e-newsletters sent from 2010 to 2025, we examine how the two parties invoke election terms such as integrity,'' security,‘’ and fraud.'' Using diachronic word embeddings, which track semantic change through how words are used in context, we measure how differently each party uses each term, how much each party shifts, and whether they share nearest semantic neighbors. We show that members of Congress do not need different words to tell different stories about elections. They need only place those same words in different company. Across nearly every term and year, the two parties surround election-integrity vocabulary with entirely different words. That divergence is easy to miss, because most election vocabulary is no more divergent than other words used at roughly the same frequency, and neither party's usage shifted more than the other's. Only a subset of terms grew more divergent, most notably certification,‘’ while rigged'' and stolen’’ entered election discourse after 2020 and polarized there. By 2024, Democrats surround voter,'' election,‘’ and integrity'' with the language of access and administration. Republicans surround the same terms with citizenship verification,‘’ noncitizen voting,'' and border security.‘’ For one party, integrity means protecting eligible voters’ ability to cast a ballot. For the other, it means protecting the electoral process from illegitimate ballots. Election officials, who must administer a single system under both definitions, inherit this tension.
[NLP-87] Lipi-Ghor: A Large-Scale Diarized Multi-Domain Bengali Speech Corpus and a Scalable Pipeline for Building It
【速读】： 该论文旨在解决孟加拉语在语音识别领域严重资源匮乏的问题，特别是针对现有公开的孟加拉语自动语音识别（ASR）语料库普遍存在的短时、朗读或众包语音、缺乏说话人分离（speaker diarization）以及长时音频标注的局限性。其核心解决方案是构建Lipi-Ghor——一个大规模弱监督的孟加拉语语音语料库，包含来自998个YouTube视频的885小时音频，覆盖17个多样化领域（如有声书、戏剧、宗教布道、新闻及地方方言），并采用SSTT格式（说话人、语音、转录、时间戳）进行标注，结合自动化说话人分离与YouTube字幕自动生成转录。该工作首次实现了低资源语言中大规模、说话人分离且具有长时结构的音频标注，突破了以往资源规模与质量的瓶颈。关键创新在于提出了一套可复现的构建管道，通过针对性筛选高质量视频源渠道，在无需大量人工标注的前提下实现对低资源语言的大规模语料扩展。此外，为应对字幕噪声问题，研究团队还发布了由17名志愿者历时30小时人工校验的黄金评估集，并开放原始贡献数据（视频ID列表、说话人分离输出、代码及黄金标注）以供研究使用，而仅提供基于字幕生成的转录文本用于学术研究，不提供原始音频文件。
链接: https://arxiv.org/abs/2610.36184
作者: Sanjid Hasan,A H M Fuad,Risalat Labib,Bayazid Hasan Simon
机构: Khulna University of Engineering & Technology (KUET)( khulna 工程与技术大学); Bangladesh University of Engineering and Technology (BUET)( 孟加拉国工程与技术大学)
类目: Computation and Language (cs.CL)
备注: 12 pages, 7 figures. Published in the Proceedings of the Speech and Audio Language Models Workshop (2nd Edition), pages 190-201, 2026
Abstract:Bengali, spoken by over 230 million people, remains severely under-resourced in spoken language research. Existing open Bengali ASR corpora are confined to short, read or crowd-sourced speech and lack speaker diarization and long-form audio. We present Lipi-Ghor, a large-scale weakly-supervised Bengali speech corpus comprising 885 hours of audio sourced from 998 YouTube videos across 596 unique channels, spanning 17 diverse domains including audiobooks, drama, religious sermons, news, and regional dialects. Each recording is annotated in SSTT format (Speaker, Speech, Transcription, Timestamp), which combines automated diarization with Bengali YouTube caption tracks. To our knowledge, Lipi-Ghor is the first Bengali speech corpus to pair large-scale coverage with speaker-diarized, long-form audio annotations, both of which are absent from existing open Bengali resources. We also present the construction methodology as a replicable pipeline for any language with a sufficient captioned YouTube presence. Targeted channel curation across a handful of rich domains yields corpus scales that were previously out of reach for low-resource languages. We benchmark four widely-used ASR models under zero-shot and fine-tuned conditions. All four degrade consistently on emotive, dialectal, and long-form content. Because caption targets are noisy, we also release a human-annotated gold evaluation split, verified by 17 volunteer annotators over 30 hours of human-in-the-loop correction and stratified across all 17 domains. We release our original contributions (video-ID lists, diarization output, pipeline code, and the gold annotations) under CC-BY-4.0, and distribute caption-derived transcripts for research use only; no raw audio is redistributed. Comments: 12 pages, 7 figures. Published in the Proceedings of the Speech and Audio Language Models Workshop (2nd Edition), pages 190-201, 2026 Subjects: Computation and Language (cs.CL) Cite as: arXiv:2610.36184 [cs.CL] (or arXiv:2610.36184v1 [cs.CL] for this version) https://doi.org/10.48550/arXiv.2610.36184 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Sanjid Hasan [view email] [v1] Mon, 26 Oct 2026 19:17:52 UTC (1,199 KB) Full-text links: Access Paper: View a PDF of the paper titled Lipi-Ghor: A Large-Scale, Diarized, Multi-Domain Bengali Speech Corpus and a Scalable Pipeline for Building It, by Sanjid Hasan and 3 other authorsView PDFHTML (experimental)TeX Source view license Additional Features Audio Summary Current browse context: cs.CL prev | next new | recent | 2026-10 Change to browse by: cs References Citations SciXGoogle Scholar Semantic Scholar export BibTeX citation Loading… BibTeX formatted citation loading… Data provided by: Bookmark checked="checked"class=“labs-tab-input”> Bibliographic Tools Bibliographic and Citation Tools Bibliographic Explorer Toggle Bibliographic Explorer (What is the Explorer?) Connected Papers Toggle Connected Papers (What is Connected Papers?) Litmaps Toggle Litmaps (What is Litmaps?) scite.ai Toggle scite Smart Citations (What are Smart Citations?) Code, Data, Media Code, Data and Media Associated with this Article alphaXiv Toggle alphaXiv (What is alphaXiv?) Links to Code Toggle CatalyzeX Code Finder for Papers (What is CatalyzeX?) DagsHub Toggle DagsHub (What is DagsHub?) GotitPub Toggle Gotit.pub (What is GotitPub?) Huggingface Toggle Hugging Face (What is Huggingface?) ScienceCast Toggle ScienceCast (What is ScienceCast?) Demos Demos Replicate Toggle Replicate (What is Replicate?) Spaces Toggle Hugging Face Spaces (What is Spaces?) Spaces Toggle TXYZ.AI (What is TXYZ.AI?) Related Papers Recommenders and Search Tools Link to Influence Flower Influence Flower (What are Influence Flowers?) Core recommender toggle CORE Recommender (What is CORE?) Author Venue Institution Topic About arXivLabs arXivLabs: experimental projects with community collaborators arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv’s community? Learn more about arXivLabs. Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?) mathjaxToggle(); We gratefully acknowledge support from our major funders, member institutions, , and all contributors. About Help Contact Subscribe Copyright Privacy Accessibility Operational Status (opens in new tab)
[NLP-88] 3 7 14 17 42: Instruction Tuning Installs and Sharpens Number-Choice Heuristics in Language Models
【速读】： 该论文旨在解决语言模型在生成随机整数时概率分布建模中存在的采样误差问题，传统方法依赖于蒙特卡洛采样估计概率，引入噪声并影响结果的准确性。其核心解决方案是通过精确计算每个整数从1到N的“拼写得分”（scored spellings）之和，直接获得无采样误差的概率分布，从而实现对模型输出分布的精确测量。该方法在18个开源模型系列、5个不同数值范围（最高至N=100）、29种解码配置及两个推理栈上进行评估，揭示了指令微调（instruction tuning）对分布特性的影响：显著降低分布均匀性，引入“圆整数回避”（round-number avoidance）与强化“端点回避”（endpoint avoidance），且仅在N=100时出现一致的奇数偏好。这些行为部分契合人类数字选择的心理学模型（Kubovy & Psotka, 1987），但圆整数回避并非人类认知模型的一部分，且其中预测的中点回避现象几乎不存在。此外，指令微调并未在不同模型间建立共享的首选整数，多数条件下无共同主导值，表明模型内部分布仍保持较高多样性。关键发现在于，通过非采样式精确计算，可揭示模型内在分布特征，为理解生成式AI中的决策偏差提供可靠依据。
链接: https://arxiv.org/abs/2610.36178
作者: Samar Ansari
机构: 未知
类目: Computation and Language (cs.CL)
备注:
Abstract:We measure, without sampling error, the probability a language model assigns to each integer from 1 to N when asked for a random number. For a fixed prompt the model defines one probability vector, which we compute by summing over each integer’s scored spellings rather than estimate by sampling, removing the noise that limits earlier studies [1, 4]. We evaluate 18 open-weight model lines at five ranges up to N = 100, under 29 decoding configurations, on two inference stacks; 15 of them have matched base and instruction-tuned checkpoints that both give usable answers. Instruction tuning makes the distribution less uniform in all 15 pairs. It installs one heuristic, round-number avoidance, and sharpens another, endpoint avoidance. Base checkpoints do not avoid round numbers once the endpoints are removed from those sets; tuned checkpoints avoid them, multiples of 10 almost completely. Base checkpoints return the endpoints 1 and N at 0.30 to 0.60 of the interior rate; tuned checkpoints almost never return them. A preference for odd numbers is consistent across tuned checkpoints only at N = 100, where base checkpoints also show it weakly. Endpoint avoidance and the odd preference fit the account Kubovy and Psotka [6] proposed for human number choice; round-number avoidance is not part of that account, and the midpoint avoidance it predicts is essentially absent, with one slight exception at N = 100. Tuning does not install a new favourite integer shared across models: above N = 10 no modal integer is shared by more than 42 percent of tuned conditions, and P (17) rises in 10 of 15 pairs and falls in 5. The concentration is present in the model’s own distribution: at temperature 1.0, nucleus truncation at top-p 0.95 adds a median of 0.035 to the total variation distance.
[NLP-89] Dont Pick a Winner: Knowledge Conflicts Need a Prior Not a Rule
【速读】： 该论文旨在解决检索增强型语言模型（Retrieval-Augmented Language Model, RALM）在检索上下文与模型自身知识存在冲突时的决策难题。现有方法依赖于设计时固定的规则来处理分歧，如优先采纳上下文、按偏离程度放大上下文影响或信任模型记忆，但这些规则在特定部署场景下均会失效。其根本问题在于：固定规则的性能随可靠上下文比例呈线性变化，而最优性能应为凸函数，导致任何固定规则在非最优情况下必然被超越；且当前方法所依赖的对称性偏差无法提供判断哪一方正确的信息。为此，论文提出需同时具备两个关键输入：分别评估每个来源（上下文与模型知识）的独立证据，以及上下文正确的先验概率。作者在GRACE基准上验证了这一理论，该基准可独立测量模型知识、上下文内容与真实事实。实验表明，所有来自四个基线家族的方法仅在正确或错误上下文中表现良好，无法兼顾两者。为此，研究构建了双输入系统：模型以“笔记”形式表达知识，小型策略网络为每条笔记赋予校准置信度，阅读器以双重声音生成回答，选择器则在给定先验π下决定采纳或弃权，实现精确先验偏移。当π设置为真实正确上下文比例时，该系统在所有GRACE上下文混合比例下均显著优于最佳基线，最高提升26分，整体得分达45.3（Qwen3.5-35B-A3B）和35.0（Llama-3.1-8B），远超基线的29.4和27.5；在预注册的ClashEval测试中，亦分别领先同骨干模型的CK-PLUG与TCR达12和26分。当模型缺乏知识时，性能增益消失，符合理论预测。
链接: https://arxiv.org/abs/2610.36169
作者: Minh Pham Van,Cong Tran
机构: Posts and Telecommunications Institute of Technology (越南邮电技术学院)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 49 pages (9 main text + appendix and supplementary material), 14 figures
Abstract:A retrieval-augmented language model answers from two sources, the retrieved context and its own knowledge, and the two disagree whenever the context is incomplete or wrong. Existing methods settle the disagreement with a rule fixed at design time: follow the context, amplify it in proportion to how far it moves the model, or trust the model’s memory. Each rule is confidently wrong in some deployment. In a decision-theoretic model of two sources, the score of any fixed rule is linear in the share of reliable contexts while the best attainable score is convex, so every fixed rule is beaten wherever it is not exactly optimal; and a symmetric divergence between the two sources, the signal adaptive decoding uses, carries no information about which one is right. Resolving a conflict therefore needs two inputs that no existing method keeps: evidence about each source separately, and a prior on how often the context is right. We test these claims on GRACE, a benchmark that intervenes on the knowledge-graph context of a fixed question, so that what the model knows, what the context says and what is true are measured separately. There, every baseline from four families is strong on right contexts or on wrong ones, never on both. We then build the two inputs: the model states its knowledge as notes, a small policy voices each note with a calibrated confidence, a reader answers in two voices, and a selector commits or abstains under a prior \pi applied as an exact prior shift. With \pi set to the true share of right contexts, the system is above the best baseline at every mixture of GRACE contexts, by up to 26 points. Its overall score is 45.3 and 35.0 (Qwen3.5-35B-A3B, Llama-3.1-8B), against 29.4 and 27.5 for the best baseline. On preregistered ClashEval it beats same-backbone CK-PLUG and TCR by 12 and 26 points. Where the model lacks the knowledge, the gains vanish, as the theory predicts.
[NLP-90] ConformalMOS: Uncertainty-Aware MOS Prediction with Conformal Intervals and Ordinal Modeling
【速读】： 该论文旨在解决现有语音质量评估中主观评价指标——人类平均意见分（Mean Opinion Score, MOS）预测模型仅输出点估计而忽略不确定性的问题，这限制了其在模型选择与实际部署中的可靠性。针对此问题，论文提出ConformalMOS框架，通过引入共形预测（conformal prediction）技术，生成具有统计保证的预测区间（prediction intervals），在交换性假设下可确保覆盖率，同时保留点估计输出。其核心解决方案在于：在训练阶段采用序数感知（ordinal-aware）的标签建模方式，将原始的one-hot标签转化为高斯平滑分布，从而在保持评分顺序关系的同时提升不确定性校准能力，尤其有效缓解softmax置信度在分布外（out-of-distribution）语音样本上过度自信的问题。实验结果表明，ConformalMOS在BVCC数据集上实现了优异的预测精度（MSE = 0.08），并验证了预测区间的实际覆盖率，显著提升了文本到语音（TTS）与语音转换（VC）系统在真实场景部署中的可靠性。
链接: https://arxiv.org/abs/2610.36126
作者: Kehinde Elelu,Joshua E. Siegel,Mohammadali Saffary,Tashfain Ahmed,Simeon Babatunde,Ebuka Okpala
机构: 未知
类目: ound (cs.SD); Computation and Language (cs.CL)
备注:
Abstract:Accurate prediction of human Mean Opinion Scores (MOS) is crucial for evaluating synthetic speech quality in text-to-speech (TTS) and voice conversion (VC). Most MOS predictors output only point estimates and ignore uncertainty, limiting reliability in model selection and deployment. We propose ConformalMOS, a framework that augments MOS prediction with conformal prediction to produce statistically valid prediction intervals with guaranteed coverage under exchangeability, alongside point estimates. During training, MOS labels are modeled in an ordinal-aware manner by converting one-hot targets into Gaussian-smoothed distributions, preserving score ordering. This improves uncertainty calibration, particularly when softmax confidence is overconfident on out-of-distribution speech. On BVCC datasets, ConformalMOS achieves competitive accuracy (MSE = 0.08) while delivering empirically validated coverage, enhancing reliability for real-world TTS and VC deployment.
[NLP-91] Beyond Update Counts: A Control Map for Evaluating Test-Time Model Updates
【速读】： 该论文旨在解决现有测试时更新（test-time updating）评估方法中对更新过程信息缺失的问题，即当前评估通常仅关注更新次数或频率及即时性能表现，却忽略了更新内容、位置、时机与顺序等关键细节，导致相同总更新量下可能掩盖系统间本质差异。其解决方案的核心是提出**控制图（Control Map）**这一新型评估框架，通过记录每次评分答案前已提交的更新历史，实现对行为变化、即时学习（acquisition）、后续回忆（retention）、无关行为保持性（locality）以及新场景下性能的独立评估。研究证明，在特定假设条件下，更新次数可决定最终状态，并揭示了预答更新历史差异对平均得分差异的限制关系。实验结果表明：更新调度策略显著影响性能（相同更新次数下准确率波动达19.7%–34.2%）；学习新信息并不保证其持久性或局部性（如ROME虽在所有操作点保留编辑，但均破坏局部性）；且无任何编辑器在不同模型、基准和测试中始终最优。因此，论文强调评估应报告完整的更新历史，并分别考察学习获取、记忆保持与局部性，以实现更全面、透明的评测。
链接: https://arxiv.org/abs/2610.36109
作者: Zehao Zhao,Shangjian Yin,Rui Liu,Chongyang Bai,Mark Ibrahim,Kavosh Asadi,Yuchen Lu,Hang Cui,Hamed Firooz
机构: Meta AI(元宇宙人工智能实验室)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: 44 pages, 6 figures, including appendices
Abstract:Test-time updating lets a language model change persistent state after training—such as model weights, adapter parameters, or a memory module—before or while answering queries. Existing evaluations often compare systems using update count or frequency and performance measured immediately afterward. These summaries omit what is updated, where, when, and in what order. The same total can therefore hide an important difference: one system may answer after one update while another answers after two. We introduce the Control Map, an evaluation framework that records the committed updates preceding each scored answer and separately evaluates behavioral changes, immediate learning (acquisition), later recall (retention), preservation of unrelated behavior (locality), and performance in new settings. We prove when counts alone determine final state and show, under stated assumptions, how differences between pre-answer histories limit differences in average scores. Using ControlMap, our experiments reveal three insights. First, update scheduling matters: at a 760M-parameter checkpoint, 24 schedules with the same update count yield 19.7%–34.2% accuracy. Second, learning an edit does not guarantee that it lasts or remains contained: all six selected CounterFact LoRA settings learn the new information, while ROME retains edits at all 13 operating points but fails locality at each. Third, no editor performs best consistently across models, benchmarks, and tests. Evaluations should therefore report update histories and assess acquisition, retention, and locality separately.
[NLP-92] FinLLM -KG: Knowledge Graph-Enhanced LLM s for Financial Sentiment Analysis in Algorithmic Trading
【速读】： 该论文旨在解决现有生成式AI在金融情感分析中面临的两大核心问题：一是现有方法仅孤立地预测单篇财经文章的情感倾向，缺乏对上下文和公司整体背景的充分理解；二是模型决策过程呈现“黑箱”特性，难以解释，限制了其在高风险金融决策场景中的可信部署。其解决方案的关键在于提出一种模型无关的框架FinLLM-KG，通过在推理阶段动态注入来自美国证券交易委员会（SEC）10-K文件的时序化、企业级知识图谱（financial knowledge graph, KG）信息，无需更新模型权重即可增强上下文感知能力。该框架支持以自然语言公司概况或经相关性排序的事实三元组形式提供上下文，使情感分析能够融合企业层面的宏观信息，实现近乎零计算成本的扩展。同时，通过将文章级情感得分归因于具体句子及知识图谱中的事实，显著提升了模型输出的可解释性，支持人工介入审查。实验表明，FinLLM-KG在投资组合绩效上持续优于当前最优基线，验证了其在实际金融应用中的有效性。
链接: https://arxiv.org/abs/2610.36083
作者: Giorgos Iacovides,Oliver Hannibal,Wuyang Zhou,Danilo Mandic
机构: 未知
类目: Computation and Language (cs.CL); Machine Learning (cs.LG); Statistical Finance (q-fin.ST); Trading and Market Microstructure (q-fin.TR)
备注:
Abstract:Recent advances in Generative AI have substantially improved financial sentiment analysis through post-trained financial large language models (LLMs). However, existing approaches remain confined to predicting the sentiment of each article in isolation and operate as black boxes, which limits contextual understanding and hinders deployment in high-stakes financial decision-making. To this end, we introduce FinLLM-KG, a model-agnostic framework that combines causal LLMs with a financial knowledge graph, in order to inject temporally grounded, company-level knowledge from SEC 10-K filings at inference time, without updating model weights. Within FinLLM-KG, context is supplied either as a natural-language company profile or as a small set of relevance-ranked factual triples. This enables sentiment predictions to account for broader firm-level information rather than the article alone, at almost zero computational cost and for any causal LLM. In addition, attributing each article-level sentiment score to individual sentences and to the retrieved knowledge-graph facts makes the evidence behind the prediction inspectable, thereby enhancing explainability and supporting human-in-the-loop review. Simulations demonstrate that FinLLM-KG consistently improves portfolio performance over the state-of-the-art baseline.
[NLP-93] How Much Does the Hint Matter? Scope Conditions for Hindsight Self-Distillation in Financial Reasoning
【速读】： 该论文旨在解决在数据本地化约束下，金融领域应用中等规模开源模型（open-weight models）因无法部署前沿大模型而导致多步定量推理能力显著落后的问题。其核心挑战在于：传统基于策略的蒸馏（on-policy distillation, OPD）依赖强外部教师模型提供监督信号，而纯自蒸馏（on-policy self-distillation, OPSD）则需完整的参考解作为条件，导致实际应用受限。为此，论文提出一种名为Fin-SD的局部回溯式自蒸馏方法，其关键创新在于无需外部模型参与梯度计算，仅通过学生模型自身识别首次错误推理步骤，并生成简短批判性提示（hint），随后将该提示与前缀状态联合输入一个冻结的学生副本作为“教师”，以反向KL散度进行短周期蒸馏。该方法强调监督信号的位置而非内容本身，实证表明：（i）使用通用反馈替代具体批判性提示仍可获得3.03个百分点的准确率提升，说明性能增益主要源于监督位置的设计；（ii）生成提示的模型无需强于接收提示的模型，同家族更大参数量的GPT-OSS-120B所生成提示并未带来实质性优势；（iii）Fin-SD在保持与全参考解自蒸馏相当的准确率的同时，减少9.5%的生成token数，且在MMLU-Pro通用推理和FinTrust合规审计任务上未出现明显性能退化。因此，解决方案的关键在于构建一种轻量化、无外部依赖、基于内部反思的高效自蒸馏机制，实现高质量推理能力的本地化迁移。
链接: https://arxiv.org/abs/2610.36082
作者: Johnny Yu,Cyril Gorlla
机构: University of California, Berkeley (加州大学伯克利分校); CTGT (CTGT)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 8 pages, 3 figures, 5 tables. Accepted at the 7th ACM International Conference on AI in Finance (ICAIF '26), Milan, Italy. Code and data: this https URL
Abstract:Financial institutions operating under data-residency constraints are restricted to open-weight models small enough to locally host, which trail frontier systems on multi-step quantitative reasoning. On-policy distillation narrows such gaps by supervising a student on its own rollouts, but traditional OPD requires a strong external teacher, and on-policy self-distillation (OPSD) conditions on a full reference solution. We ask how much of that signal is necessary. We present Fin-SD, a localized hindsight self-distillation recipe for multi-step reasoning. Given a problem the student fails, the student itself identifies the first incorrect reasoning step and writes a short critique of that step. We truncate the student’s trajectory immediately before the error (aiming to obtain a correct prefix), condition a frozen copy of the student on the same prefix plus the hint, and distill that hint-conditioned teacher into the student over a short rollout under reverse KL. No external model contributes a gradient. We apply Fin-SD to GPT-OSS-20B on a custom multi-step financial reasoning benchmark that requires deriving formula inputs from realistic scenarios via accounting conventions. We find: (i) replacing the critique with a generic-feedback control matches Fin-SD’s entire 3.03 pp accuracy gain over the base model on the FinanceReasoning Hard subset, suggesting the gain comes from where supervision is applied, not hint content; (ii) the model writing the hint does not need to be stronger than the model receiving it: hints from same-family GPT-OSS-120B, at roughly 6x the parameters, yield no materially greater benefits than self-written hints; and (iii) Fin-SD matches on-policy self-distillation from full reference solutions on accuracy while generating 9.5% fewer tokens. With Fin-SD, general reasoning on MMLU-Pro and a fourteen-metric compliance audit on FinTrust show no material regression.
[NLP-94] OmniBrain-0: A Unified Agent ic Model for Embodied Intelligence
【速读】： 该论文旨在解决通用具身智能（general-purpose embodied intelligence）中机器人在长时程任务中协调多样化技能、持续追踪任务进展并实现失败恢复的难题。现有方法中，专用策略（specialist policies）难以保障任务级可靠性，而高性能多模态代理依赖于昂贵的推理过程和外部经验，难以被轻量级模型复现。为此，本文提出OmniBrain-0——一种轻量级统一具身智能体，通过共享高层决策模型协同操作与导航，集成任务理解、规划、进度追踪及反馈驱动的故障恢复能力。其核心解决方案在于采用渐进式训练框架，融合具身预训练、领域专精代理蒸馏以及多教师在线策略蒸馏（Multi-Teacher On-Policy Distillation, MOPD），将强教师模型的交互经验提炼为操作与导航专家，并将其决策能力统一整合至OmniBrain-0的决策模型中，实现决策层面的真正统一而非仅限于工具接口层。进一步引入基于评分标准的递归自改进（Rubric-Guided Recursive Self-Improvement, RSI）数据飞轮机制，由元智能体（Meta-Agent）将执行证据转化为精细化评分标准、步骤提醒及验证轨迹，用于更新领域专家，再通过MOPD重新整合至OmniBrain-0。实验在LIBERO-Pro、LIBERO-Plus、RoboCasa365、RoboTwin 2.0、R2R-CE Val-Unseen及多个真实世界操作任务上验证，结果表明OmniBrain-0在跨域任务表现上具备竞争力，同时显著降低推理与执行开销，实现了具身能力高效迁移与紧凑化统一。
链接: https://arxiv.org/abs/2610.36013
作者: Xiaomi Embodied Intelligence Team:Yuechen Luo,Chong He,Fang Li,Shaoqing Xu,Jingkai Wang,Menglun Su,Jiaxin Liu,Shengyin Jiang,Zhiyi Lai,Long Chen,Longlong Wang,Liang Hu,Haochen Liu,Yuncheng Jiang,Hongwei Xie,Jieming Cui,Hanbing Li,Junjie Huang,Zhixiang Duan,Hexin Zhang,Weixiang Liang,Ying Gong,Yuhan Wang,Yong Pan,Pengna Li,Kangyi Wu,Ken Deng,Kun Ma,Fuxi Wen,Zhi-xin Yang,Hangjun Ye(University of Macau, Tsinghua University)
机构: Xiaomi Embodied Intelligence Team (小米具身智能团队); University of Macau (澳门大学); Tsinghua University (清华大学)
类目: Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV)
备注: Technical Report; 40 pages, 27 figures, 8 tables; Project Page at this https URL
Abstract:General-purpose embodied intelligence requires robots to coordinate diverse skills over long horizons, track task progress, and recover from failures. Specialist policies alone do not ensure task-level reliability, while powerful multimodal agents rely on costly reasoning and external experience that compact models struggle to reproduce. We introduce OmniBrain-0, a lightweight unified embodied agent that coordinates manipulation and navigation through a shared high-level decision model, integrating task understanding, planning, progress tracking, and feedback-driven recovery. Our progressive training framework combines embodied pretraining, domain-specialized agent distillation, and multi-teacher on-policy distillation (MOPD). It distills powerful teachers’ interaction experience into manipulation and navigation experts, then consolidates their decision capabilities into OmniBrain-0. Unification therefore occurs within the learned decision model rather than only at the tool interface. We further introduce Rubric-Guided Recursive Self-Improvement (RSI), a data flywheel in which a Meta-Agent turns execution evidence into refined rubrics, step-level reminders, and verified trajectories for domain-expert updates before their reconsolidation into OmniBrain-0 through MOPD. We evaluate OmniBrain-0 on LIBERO-Pro, LIBERO-Plus, RoboCasa365, RoboTwin 2.0, and R2R-CE Val-Unseen, as well as multiple real-world manipulation tasks. OmniBrain-0 combines strong cross-domain task performance with substantially lower inference and execution overhead than its closed-source teacher, demonstrating efficient transfer and consolidation of embodied capabilities into a compact unified agent.
[NLP-95] MedJev: Evaluating Zero-Shot System-One Decision Making in Medicine
【速读】： 该论文旨在解决临床决策支持系统中生成式AI（Generative AI）模型计算成本高、推理延迟大以及预测置信度校准不足的问题。其核心挑战在于如何在保持低计算资源消耗的同时，实现与大型语言模型（LLM）相当的医学知识理解与患者层面预测性能，并确保模型输出的置信度具有良好的校准性。解决方案的关键在于提出一种轻量级、仅需两个参数的再校准方法（recalibration method），通过利用50至200个本地标注样本对模型预测进行微调，显著提升置信度校准质量（将ECE从0.286降至0.007），同时维持模型在本地设备上运行的高效性。实验表明，MedJev作为开源的System One模型，在多项医学基准测试中接近GPT-4性能，且在住院死亡率预测任务中优于传统机器学习模型（如XGBoost），验证了其作为医疗AI零样本基线的潜力。
链接: https://arxiv.org/abs/2610.36006
作者: Simon A. Lee,Jiahang Sha,Moonseong Jeong,Yunbei Pan,Jeffrey N. Chiang
机构: UCLA Computational Medicine(加州大学洛杉矶分校计算医学); UCLA Neurosurgery(加州大学洛杉矶分校神经外科)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:System One decision models such as Jev answer structured questions in a single forward pass without generating text, offering fast and computationally efficient clinical decision support. We evaluate MedJev, an open System One model that runs locally on a laptop, across ten medical benchmarks spanning medical knowledge, diagnosis, and patient level prediction using MIMIC IV and MIMIC IV ED. We compare MedJev against zero shot LLM baselines, published GPT results, and supervised models. MedJev achieves accuracies of 72.5 percent on MedQA, 75.4 percent on USMLE, and 80.1 percent on NEJM, approaching published GPT 4 performance while maintaining well calibrated confidence estimates (ECE 0.028 to 0.063). On patient level tasks, MedJev achieves an AUROC of 0.926 for In Hospital Mortality, exceeding XGBoost (0.902), but underperforms supervised models on operational outcomes such as ED Triage Admission (0.745 vs. 0.809). Patient level predictions also exhibit substantial miscalibration (In Hospital Mortality ECE 0.286). We introduce a lightweight, two parameter recalibration method that adapts predictions to local patient populations using 50 to 200 labeled examples, reducing ECE to 0.007. Our findings establish open System One models as a promising zero shot baseline for medical AI, combining competitive predictive performance with low computational cost and efficient local calibration. We release our code and evaluation harness to support future research.
[NLP-96] ReCoPII: A Relation-Grounded Framework for Building Contextual PII Masking Datasets
【速读】： 该论文旨在解决生成式隐私信息识别（PII）标注数据时面临的“真实数据不可得”难题，即在隐私保护限制下难以获取可用于训练上下文掩码系统的带标签真实数据。传统方法如基于作者的增强（SPY）依赖单一数据主体假设，将与作者无关的实体标记为非PII，但忽略了多主体场景下的复杂关系，导致标注偏差。其解决方案的关键在于提出一种名为ReCoPII（关系引导的上下文PII掩码）的新框架：该框架通过构建符合本体约束的实体关系图谱，设计带有占位符的插入句，利用代码自动化重构标注，并通过独立的大语言模型（LLM）验证结果，最终依据实体间的关系推导出MASK/KEEP标签。该方法无需依赖单一数据主体假设，从而生成更准确、更具泛化能力的标注数据。实验表明，基于该框架构建的ReCoPII-Ko语料库在六个韩语文档领域中实现了高一致性（四名标注者对采样保留标签的同意率达88.42%，远超SPY方法的26.27%），且基于该数据训练的分类器在测试集上过掩码率显著降低。进一步地，联合微调上下文掩码系统与PII检测器，在固定泄露预算下有效缓解了仅依赖检测器时的过掩码问题；在人工验证的实体记录上，四个检测器的过掩码率均不超过5.04%，泄露率不超过4.18%。此外，所提出的检测器在KDPII基准上性能接近XLM-R-large，但参数量仅为后者的约四分之一，展现出优异的效率与效果平衡。
链接: https://arxiv.org/abs/2610.35925
作者: Taehwan Lee
机构: Hanyang University (汉阳大学); Republic of Korea (大韩民国)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Training contextual masking systems for personally identifiable information (PII) requires examples distinguishing maskable from preservable mentions of one entity type, but privacy restrictions limit access to real data. Author-based augmentation (SPY) labels mentions unrelated to the author as non-PII, presuming a single data subject and overlooking others. We propose ReCoPII (Relation-grounded Contextual PII masking), a framework that builds such data without this premise: it plans ontology-constrained entity relations, writes insertion sentences with placeholders, reconstructs annotations with code, checks the result with a separate LLM, and derives MASK/KEEP labels from entity relations. On Korean documents from six domains, it yields ReCoPII-Ko, a corpus with relation, privacy, and MASK/KEEP annotations, which we validate. Across 300 pairs, four annotators agree with 88.42% of its sampled KEEP labels, against 26.27% for SPY-style augmentation of the same sources. Classifiers trained on it over-mask less than under that reference on a matched test. Contextual masking fine-tuned jointly with a PII detector removes most over-masking of detection alone within a fixed leakage budget; on human-verified entity records, the four detectors meeting it over-mask at most 5.04% with at most 4.18% leakage. The detector is comparable to XLM-R-large on KDPII with about a quarter of its parameters.
[NLP-97] Physiognomy Revisited: How Vision-Language Models Encode Sensitive Attributes Through Image Transformation EMNLP2026
【速读】： 该论文旨在解决生成式人工智能（Generative AI）在视觉语言模型（VLMs）中对受保护身份（如种族、性取向等）的编码与生成过程中所隐含的偏见与不可预测的失效模式问题，尤其关注这些缺陷对弱势群体可能造成的不平等影响。其解决方案的关键在于揭示：当要求模型将人脸图像转换为特定种族或性取向特征时，模型表现出非确定性、群体敏感性的响应行为，生成图像在统计上可区分且往往强化既有刻板印象；进一步发现，此类生成阶段引入的表征偏差会传递至后续文本推理与人物描述任务中，甚至扩展至犯罪性相关偏见的再生产，表明生成过程中的系统性偏见具有跨模态传播特性，从而凸显出对生成式AI伦理与公平性评估机制的迫切需求。
链接: https://arxiv.org/abs/2610.35910
作者: Soumyajit Datta,Arka Dutta,Ashiqur R. KhudaBukhsh
机构: Rochester Institute of Technology(罗切斯特理工学院)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Computers and Society (cs.CY)
备注: Accepted at EMNLP 2026 (Main Conference Track)
Abstract:Emergent generative AI technologies, while remarkably beneficial across several tasks and domains, can suffer from unforeseen vulnerabilities and difficult-to-anticipate failure modes. Such failure modes can disparately affect vulnerable communities. This paper investigates two well-known vision language models (VLMs) and inspects how these models encode protected identities while generating images. We show that when VLMs are requested to transform human face images to reflect race identities (e.g., Hispanic, Black, White, or Asian) and sexual orientation (e.g., gay or straight), VLMs often comply with such requests. We demonstrate that compliance (and refusal) is non-deterministic, group-sensitive, and generated images are statistically distinguishable, often accentuating deep-seated stereotypes. We further show that representation biases introduced during image generation can re-emerge in textual reasoning and characterization tasks. Finally, we show that this worrisome finding is not limited to racial identity or sexual orientation, image generation can perpetuate bias around criminality, raising concerning possibilities of AI-assisted profiling.
信息检索
[IR-0] REGER: Modeling inter-item Relations with Semantic IDs for Generative Recommendation
链接: https://arxiv.org/abs/2610.37584
作者: Yuxi Huang,Silin Zhou,Chenhao Wang,Zihan Li,Gangming Guo,Pei Ke,Shen Gao,Shuo Shang
类目: Information Retrieval (cs.IR); Artificial Intelligence (cs.AI)
备注: 13 pages, 6 figures
Abstract:Personalized recommendation depends on learning semantic and collaborative relations among items from online content and user interactions. Generative recommendation (GR) offers a promising paradigm by autoregressively generating discrete semantic IDs (SIDs) of items. However, randomly initialized SID token embeddings and GR models trained from scratch do not directly inherit the semantic relations inherent in items, while inter-item collaborative relations are learned largely implicitly through token-level attention and next-token supervision. To address this limitation, we propose REGER, a unified framework that explicitly incorporates inter-item relations into semantic pretraining, history encoding, and SID generation. We pretrain the GR model through generative semantic retrieval, enabling it to inherit inter-item semantic similarity by learning to generate semantic neighbors’ SIDs. We combine global attention with local attention aligned with SID boundaries to capture historical context and directed transitions between adjacent items. We further develop a theoretical framework for hierarchical preference decomposition over the SID prefix tree, proving an exact factorization of collaborative preference distributions and a weighted decomposition of item-level KL divergence into prefix-conditional KL divergences. Building on this analysis, we distill a discriminative teacher’s full conditional distributions over all legal next tokens at ground-truth prefixes along the target SID path, jointly optimizing this objective with cross-entropy. Extensive experiments on real-world datasets demonstrate consistent improvements over existing GR baselines and support the effectiveness of the proposed preference distillation.
[IR-1] LCM-Promo: An LLM -Based Uplift Model for Promotion Allocation KDD2026 ECML
链接: https://arxiv.org/abs/2610.37570
作者: Nesta Midavaine,Floris Fok,Paulo Mendes,Joao Silva,Önder Akaçık,Dimas Lima,Zülküf Genç
类目: Machine Learning (cs.LG); Information Retrieval (cs.IR)
备注: 4 pages, 2 tables. Accepted at ECML PKDD 2026 (Industrial Track)
Abstract:We present LCM-Promo, an LLM-based uplift model for promotion allocation. LCM-Promo is a task-specific Large Commerce Model (LCM): a finetuned LLM trained on transactional commerce data to predict user intent and promotion response. It learns monotonic promotion-response curves from production logs by interpreting single-token outputs as conversion-threshold probabilities, while conditioning on session context such as day and delivery shift. In a production A/B test on churn reactivation, LCM-Promo increased conversion by 1% and reduced promotion cost by 6% relative to the Segment-EMA production baseline.
[IR-2] Can Complementary Signals Bridge Similarity Islands? Manifold-Augmented Graph Embedding for Multimodal Recommendation NEURIPS2026
链接: https://arxiv.org/abs/2610.37409
作者: Zhenye Yang,Jinpeng Chen,Wenbo Fu,Jun Ma,Huan Li,Qimin Zhou,Tao Wang,Li Sun,Senzhang Wang
类目: Information Retrieval (cs.IR)
备注: Accepted at NeurIPS 2026. 33 pages, including references and appendices
Abstract:Multimodal Recommender Systems (MMRS) leverage rich item features to mitigate data sparsity, yet similarity-driven graph construction and propagation can favor semantically similar substitutes over functionally related complements. When behavioral associations connect semantically distant items, complementary signals can remain geometrically isolated, yielding fragmented “Similarity Islands” and semantic drift when standard propagation is applied. To overcome these limitations, we propose MAGE (Manifold-Augmented Graph Embedding), a framework for personalized multimodal ranking that decouples complementarity modeling into offline geometric warping and online topological propagation. The offline phase enhances item representations by extracting complement-aware residual directions via a Soft Conceptor mechanism and injecting them under a local gap-preserving constraint, incorporating co-purchase evidence while controlling the magnitude of geometric perturbation. Synergistically, the online phase employs an asymmetric dual-stream GNN that treats visual features as stable anchors and augmented textual features as flexible probes, fused through a self-calibrated cross-modal mechanism to reconcile shared semantics and modality-specific residual cues. Extensive experiments on four Amazon datasets show that MAGE yields consistent improvements over strong multimodal recommendation baselines. Together with the supporting analyses in the appendix, these results suggest that geometry-aware augmentation can alleviate the semantic mismatch that arises in complementary recommendation settings. Our code is available at this https URL.
[IR-3] When Is Additional Retrieval Worthwhile? A Cost-Aware Study of Evidence Gain and Answer Reliability in RAG
链接: https://arxiv.org/abs/2610.37358
作者: Zhouting Zhao
类目: Information Retrieval (cs.IR)
备注:
Abstract:Retrieval-augmented generation (RAG) can benefit from more sophisticated retrieval, but extra retrieval may be unnecessary for many questions and gains in evidence retrieval may not translate into faithful or correct answers. We study a cost-aware, two-action retrieval policy that chooses between a dense-retrieval-plus-reranking baseline and a hybrid lexical–dense retriever. Rather than classifying baseline retrieval as simply successful or unsuccessful, we estimate the question-specific incremental evidence-retrieval benefit of hybrid retrieval using features available before the routing decision. A regularized linear model and decision threshold are selected using document-grouped cross-validation and frozen prior to held-out evaluation. On 60 previously unseen QASPER training-split articles (188 questions, 169 evidence-evaluable), the policy triggers hybrid retrieval on 46 questions (24.5%) and obtains the same evidence Hit@10 as always-hybrid retrieval (78.70%, versus 76.92% for always-keep). Paired, paper-cluster bootstrap intervals do not establish a reliable utility advantage. We further compare generated answers from a 4-bit Qwen3-4B model under the three retrieval policies. Provisional answer-level assessments show no clear correctness or faithfulness improvement after partial review, despite the evidence-retrieval result. The generation assessments remain model-assisted and lack independent human validation. These results expose an important distinction between retrieving relevant evidence efficiently and improving the reliability of the generated answer. The work is presented as a reproducible empirical study, not a claim of statistically established answer-quality superiority.
[IR-4] LESSRec: Level-Specific History Modeling for Semantic ID-Based Recommendation
链接: https://arxiv.org/abs/2610.37286
作者: Shangxin Yang,Zongwei Wang,Jing Long,Junliang Yu,Min Gao
类目: Information Retrieval (cs.IR)
备注:
Abstract:Semantic ID-based recommendation represents each item as a sequence of semantic tokens obtained through residual quantization. This paradigm generates the next item’s tokens autoregressively from coarse to fine, entailing each prediction attending to historical tokens at all semantic levels. However, our inference-time analysis reveals that historical tokens at the target-token’s semantic level tend to be more informative, whereas other levels can introduce interference. Motivated by these findings, we propose LESSRec that separately encodes the historical token sequence at each semantic level and restricts attention over historical tokens to the sequence at the target token’s semantic level during prediction. To complement these level-specific histories, LESSRec explicitly provides an item-level semantic sequence that represents each historical item’s overall semantics in chronological order. The decoder accesses the level-aligned token sequence and the item-level semantic sequence through two separate cross-attention branches and fuses their contexts. We further introduce LESSRec+, which uses separate encoders for different semantic-level histories with the same total parameter count to reduce encoder computation. Extensive experiments across three public datasets and three tokenizers demonstrate the recommendation effectiveness of LESSRec and LESSRec+, with efficiency analyses highlighting the computational benefits of LESSRec+.
[IR-5] Modeling Preference Dependencies with Directed Acyclic Graphs for Narrative-Driven Recommendation
链接: https://arxiv.org/abs/2610.37280
作者: Mingzhe Lu,Jian Qin,Ruize Qin,Jinhao Liu,Yunpeng Li,Hang Su
类目: Information Retrieval (cs.IR); Computation and Language (cs.CL)
备注: 9 pages, 5 figures, 3 tables
Abstract:Narrative-driven recommendation allows users to express multifaceted needs through free-form text where stated preferences naturally qualify and condition one another. Existing methods typically treat narratives as flat keyword combinations or unconstrained generative queries, failing to capture how complex preferences hierarchically refine and contextualize each other. We propose the Preference Dependency Graph Framework (PDG-F), a narrative-driven recommendation framework centered on explicit preference dependencies. Specifically, PDG-A (PDG-Agent) first parses unstructured requests to organize user preferences and their mutual qualifications into a directed acyclic Preference Dependency Graph (PDG). Through this graph, candidate evidence from the open Web is acquired via directed path retrieval, and final recommendations are determined by topological path ranking. Comprehensive experiments across three diverse domains spanning English and Chinese contexts confirm that PDG-F consistently achieves superior recommendation performance over competitive baselines, open-domain retrieval agents, and frontier language models. Further diagnostic analyses show that our dependency-guided retrieval markedly raises the attainable candidate-pool ceiling, demonstrating the fundamental value of modeling preference dependencies in narrative-driven recommendation.
[IR-6] GradCO: Behavior-Aware Gradient-Conflict Offloading for Text-Based Person Anomaly Search
链接: https://arxiv.org/abs/2610.37267
作者: Zhiheng Fu,Zixu Li,Zhiwei Chen,Jiale Huang,Fangxu Liu,Jinhe Bi,Yupeng Hu
类目: Computer Vision and Pattern Recognition (cs.CV); Information Retrieval (cs.IR)
备注:
Abstract:Text-based Person Anomaly Search (TPAS) retrieves a target pedestrian using natural-language descriptions of appearance and normal or anomalous behavior. Unlike conventional text-based person retrieval, TPAS must preserve identity cues while distinguishing fine-grained behaviors. It presents two challenges: same-identity, different-behavior hard negatives cause gradient conflict during granular alignment, while strong appearance cues can obscure weak behavioral evidence during aggregation. We propose GradCO (behavior-aware Gradient-Conflict Offloading), comprising three complementary components. Behavior-Aware Mixture-of-Experts (B-MoE) forms appearance, normal-behavior, and anomalous-behavior streams whose roles are learned from retrieval objectives without expert-specific semantic labels. Conflict Offload Learning (COL) uses pair-conditioned offload anchors in similarity computation to reduce the conflicting-gradient burden on the main text representations. Hierarchical Intent Alignment (HIA) aggregates expert similarities and features using query-conditioned weights. Experiments on five TPAS/TIPR benchmarks show GradCO significantly surpasses SOTA methods and validate our design.
[IR-7] askBridge: Technical Report for SIGIR 2026 eCommerce Challenge SIGIR2026
链接: https://arxiv.org/abs/2610.37231
作者: Zhiheng Fu,Yupeng Hu,Zixu Li,Zhiwei Chen,Qinlei Huang,Shaokun Wang,Han Liu,Fan Liu,Weili Guan,Liqiang Nie
类目: Information Retrieval (cs.IR)
备注: Technical Report on the First-Place Solution for eCommerce Challenge @ SIGIR 2026 eCom Workshop
Abstract:The SIGIR 2026 eCommerce Challenge Query Reformulation Task evaluates the reformulation of task-oriented queries into expressions suitable for product retrieval, followed by BM25-based retrieval. These queries often imply product needs through activity-level semantics, creating a lexical gap with catalog descriptions. Systems must preserve task intent while adapting the number of reformulations to the Automatic and Interactive protocols. We propose TaskBridge, a catalog-anchored framework comprising Catalog Semantic Anchoring, Unified Five-Candidate Rewriting, and Protocol-Aware Selection. It grounds queries in catalog vocabulary, generates five complementary reformulations, and selects one query for Automatic retrieval or up to four for Interactive retrieval using BM25 scores, query-product reranking, and weighted Jaccard overlap. The selected queries are independently executed using BM25 to produce results in the required six-column format. Experiments demonstrate improved retrieval effectiveness, with multiple complementary queries providing additional gains in the Interactive setting. TaskBridge achieved first place in the SIGIR 2026 eCommerce Challenge.
[IR-8] AutoRSE: Evaluating LLM Agents as Recommendation Engineers for Self-Evolving Recommender Systems
链接: https://arxiv.org/abs/2610.37228
作者: Rui Tian,Yi Fang,Yejin Hua,Du Wang,Yuzhe Wang,Minxuan Shan,Zirui Ren,Xinyu Lin,Chongming Gao,Wenjie Wang,Fuli Feng
类目: Information Retrieval (cs.IR)
备注: 12 pages, 7 figures. Code: this https URL
Abstract:Recommender Systems Engineer Agents (RSE-Agents), which use LLM-based agents to autonomously improve recommendation algorithms, are emerging as a promising paradigm. However, their capabilities and limitations remain poorly understood due to the lack of a unified, standardized evaluation framework and systematic trajectory-level analysis. To address this gap, we introduce AutoRSE, an evaluation framework spanning six representative industrial-grade recommendation tasks, and conduct a large-scale empirical study of three frontier LLMs across three widely used agent harnesses. Our results show that RSE-Agents consistently enhance the performance of backbones on five of the six tasks, with an average relative improvement of 16.1%, indicating that they can already autonomously improve recommendation performance. However, trajectory-level analysis also reveals that current RSE-Agents function primarily as engineering optimizers rather than algorithmic improvers: their gains are largely driven by hyperparameter tuning, training strategies, and model ensembling, while genuine algorithmic innovations are rare. Since the largest gains in recommender systems often stem from algorithmic advances, this behavior underscores a fundamental limitation of current RSE-Agents. To probe this limitation, we perform idea-injection experiments and show that agents equipped with innovative ideas from frontier research or human experts achieve substantially larger improvements than those relying on self-generated ideas. These results suggest that idea generation is a central bottleneck of current RSE-Agents, highlights a key direction for the future recommendation system research. Our implementation is publicly available on GitHub: this https URL.
[IR-9] Lossless Compression of Proximity Graph Indexes for Efficient Approximate Nearest Neighbors Search CIKM2026
链接: https://arxiv.org/abs/2610.37176
作者: Francesco Benocci,Leonard Dahlmann,Francesco Gargiulo,Tobias Heuer,Silvio Martinico,Franco Maria Nardini,Cosimo Rulli,Rossano Venturini
类目: Information Retrieval (cs.IR)
备注: 7 pages, 5 figures, Accepted at CIKM 2026. Code: this https URL . Reproducibility: this https URL
Abstract:Graph-based Approximate Nearest Neighbor (ANN) indexes, such as HNSW, are the backbone of modern vector search. However, their massive memory footprint creates a severe infrastructure bottleneck. While vector quantization is widely adopted to reduce the memory required to store data vectors, the graph’s edge structure increasingly dominates total memory usage. Compressing these adjacency lists for online search presents a critical system challenge: because decompression occurs during query processing, i.e., while navigating the graph, memory savings cannot come at the expense of query latency. In this paper, we present a compression technique explicitly tailored for live HNSW routing. By applying offline graph permutations to artificially induce structural locality, we enable the use of SIMD-accelerated integer compressors, such as Stream VByte, for rapid on-the-fly decompression. Evaluated across diverse real-world workloads, we significantly outperform the current state of the art in both compression ratio and query latency. Our experimental results show that we achieve compression ratios up to 52.7% with negligible penalties in latency compared to the uncompressed baseline. Furthermore, our technique has been effectively deployed in production at eBay, achieving up to 44.3% reduction in edge storage with no degradation in query latency.
[IR-10] Coverag e Under Contradiction: A Benchmark for Negation-Aware Retrieval CIKM2026
链接: https://arxiv.org/abs/2610.37162
作者: Pasha Abrishamchian,Eugene Yang,Andrew Yates,Nazli Goharian,Ophir Frieder
类目: Information Retrieval (cs.IR); Computation and Language (cs.CL)
备注: 5 pages, 1 figure. Accepted at CIKM 2026
Abstract:Coverage-evaluation benchmarks quantify whether retrieved evidence covers the different pieces of information needed to satisfy an information need. Existing benchmarks mostly consider factual claims that may already be known to pretrained retrieval components, leaving open whether retrievers find corpus evidence when it contradicts those priors. We study how retrievers behave when their parametric knowledge and the information in a corpus disagree. To do so, we develop a process to augment a nugget-based coverage benchmark with controlled contradictions. For each nugget, we create a negated counterpart and edit its supporting documents to support that negation while preserving the context. Applying this process to CoverageBench, a benchmark measuring information coverage, we create CoverageBench-Negate, spanning six datasets and roughly 63,000 original-edited evidence pairs (one per nugget-document link). We attach rich construction metadata to every pair and additionally provide out-of-family LLM judgments over all pairs and a human audit of a sample. In baseline experiments, BM25 separates original from edited near-twin evidence substantially better than dense retrievers (Qwen3 and Jina), which stay near chance. We release the benchmark on Hugging Face Datasets (nuggets, documents, pairs) together with the construction materials, including the prompts used to create it.
[IR-11] Rethinking Positional Encoding and Interaction Granularity for CTR Prediction
链接: https://arxiv.org/abs/2610.37140
作者: Xinxun Zhang,Yuxiang Wang,Xiaokai Zhou,Jiayue Li,Yi Li,Yue Liu,Jun Zhang,Xiaoling Zhou,Tieyun Qian,Xiao Yan,Jiawei Jiang
类目: Information Retrieval (cs.IR)
备注: 13 pages, 11 figures
Abstract:Click-through rate (CTR) prediction relies on historical behaviors and contextual features to estimate clicks on a target item. Existing Transformer-based CTR models often adopt index-based or temporal positional encodings and token-level causal interaction. However, we observe that most existing positional encodings do not outperform NoPE, while relaxing causal constraints within target-relative temporal groups also yields gains. These observations motivate a fundamental question: what positional encoding and interaction granularity does CTR prediction actually require? To address this question, we propose \textbfTRIP, which combines target-aware rank–time positional encoding with temporal-group interaction modeling. Specifically, Factorized Rank Encoding represents each behavior’s ordinal position within the historical sequence through a low-rank factorization, while Target-Relative Time Encoding characterizes its temporal distance to the target on a coarse-to-fine scale. This joint design grounds each behavior’s positional representation in its sequence context and temporal relation to the current target. Furthermore, Temporal Group-Causal Interaction enriches intra-group representations through bidirectional interaction and masked reconstruction while preserving inter-group causal order. Finally, behavior representations are condensed into compact group summaries for efficient group-level modeling in subsequent layers. Experiments on four public benchmarks show that TRIP improves AUC by up to 4.45% relative to the strongest baseline.
[IR-12] ssera: Structure-Completed Evidence Units for Web Retrieval-Augmented Generation
链接: https://arxiv.org/abs/2610.37080
作者: Peichun Hua,He Sun,Zirui Zhuang,Qi Qi,Yiming Qiu,Shinan Liu,Yunming Xiao
类目: Information Retrieval (cs.IR)
备注: 13 pages, 7 figures, 17 tables
Abstract:Web retrieval-augmented generation (RAG) splits a page into fixed-size text chunks or, in recent visual pipelines, screenshot tiles, and the retrieved item serves at once as the unit of indexing, ranking, and reading. We trace answer evidence from source HTML to reader input over the full English Wikipedia and find that this coupling loses evidence in text and screenshot pipelines alike: evidence present in the source page often fails to reach the reader intact, particularly structured content such as tables and infoboxes, which chunking separates from the headers and context that give it meaning. Stronger extraction, context expansion, and larger reader budgets leave the loss in place. We present Tessera, a training-free evidence delivery layer that decouples what a pipeline retrieves from what its reader reads. Given the pages behind any ranked list, Tessera rebuilds their HTML into fine-grained evidence units, each completed with its page title, section path, and table headers, selects the units relevant to the query, and packs them into the reader context at an unchanged token budget. The index, the retriever, and the reader stay as they are, so Tessera applies to text, screenshot, and agentic pipelines. Across three question answering benchmarks, Tessera improves answer accuracy by 5 to 6 points on text and screenshot pipelines, consistently across readers, retrievers, and reader budgets, and outperforms strong HTML-aware and structure-aware RAG baselines under matched budgets. The gains carry over to multi-step agentic search. Our code is publicly available at this https URL.
[IR-13] Compote: Collapse-Resistant Multiple Hypothesis Learning For Sequential Recommendation
链接: https://arxiv.org/abs/2610.37005
作者: Louis Castres Saint Martin,Clement Calauzenes,Alain Rakotomamonjy,Patrick Gallinari
类目: Machine Learning (cs.LG); Information Retrieval (cs.IR)
备注:
Abstract:Next-item prediction is inherently ambiguous: a single interaction history often admits multiple plausible continuations across distinct catalogue regions. Standard sequential recommenders compress these alternatives into a single prediction, conflating user intents. We introduce COllapse-resistant Multiple hyPOThesis lEarning (COMPOTE), a multi-hypothesis transformer trained with an annealed objective that transitions from uniform supervision to a winner-takes-all assignment. This mechanism forces hypotheses to specialize into distinct catalogue regions, while inference relies on an efficient, normalizer-free max-similarity rule. Theoretically, we show that in the population case, the optimal solutions partition the conditional next-item distribution according to a max-entropy quantization principle, while pooled scores preserve the Bayes-optimal ranking. We also prove a generalization bound scaling logarithmically with the number of hypotheses. Across three Amazon Reviews 2023 benchmarks, COMPOTE outperforms strong ID-based, generative, and multi-interest baselines using item IDs alone. Matched-capacity ablations confirm that performance gains stem directly from the annealed objective, preventing the head collapse observed in conventional multi-head architectures, rather than added capacity, ensembling, or routing.
[IR-14] SpectRAG : Weighted Chunk Graphs for Token-Budgeted Context Selection in Multi-Hop Retrieval-Augmented Generation
链接: https://arxiv.org/abs/2610.36991
作者: Zafer Sahinoglu,Abdulazeez Mohammed,Buse Yaren Kazangirler,Cahit Berkay Kazangirler
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Information Retrieval (cs.IR)
备注: 26 pages, 5 figures, 10 tables. This work is intended for submission to an IEEE journal
Abstract:Retrieval-augmented generation (RAG) systems ground large language models (LLMs) by retrieving external evidence and conditioning generation on the retrieved context. Although dense retrieval is effective for local semantic matching, independently ranked retrieval can return redundant passages, miss cross-chunk dependencies, and provide incomplete evidence for multi-hop reasoning. We propose SpectRAG, a spectral graph-augmented retrieval framework that combines hybrid candidate seeding, weighted chunk graph construction, graph-based expansion, spectral relevance propagation, and token-budgeted context selection. Document chunks are represented by dense semantic embeddings and graph-spectral coordinates derived from a weighted chunk graph whose edges encode semantic similarity, document adjacency, section hierarchy, shared entities, citation links, and similarity between LLM-generated utility questions. At query time, SpectRAG builds a high-recall seed set using exact inner-product search and hybrid relevance signals, expands it over the chunk graph, re-scores candidates using graph proximity and spectral information, and assembles the final context under a fixed token budget. The goal is not merely to retrieve individually similar chunks, but to construct a context that is relevant, diverse, structurally connected, and sufficient for grounded reasoning. Experiments on MultiHopRAG, UltraDomain, and long-document question-answering settings show that SpectRAG improves broad evidence recovery and achieves competitive generation quality against dense, hybrid, HyDE, LightRAG, and NaiveRAG baselines. The results further indicate that graph-spectral structure is particularly useful for inference-oriented and cross-document retrieval, while highlighting trade-offs in early precision, computational cost, and domain-dependent performance.
[IR-15] KL-MOPD: Discrepancy-Aware Multi-Objective On-Policy Distillation for Generative Recommendation
链接: https://arxiv.org/abs/2610.36973
作者: Weiqin Yang,Yandong Wang,Yuan Gao,Hao Deng,Jinxin Hu,Jing Cao,Jiawei Chen
类目: Information Retrieval (cs.IR)
备注:
Abstract:Coordinating multiple recommendation objectives (e.g., click and conversion) is critical for real-world generative recommendation (GR). However, since a positive item in GR can be retrieved only when its tokens are successively selected at every decoding step, traditional item-wise methods lack fine-grained token-wise supervision to guide intermediate steps. Motivated by its success in foundation models, we consider introducing multi-objective on-policy distillation (MOPD) into GR to distill two single-objective teachers into a student at each token. Yet, we find that in vanilla MOPD, (i) teachers exhibit unequal distillation difficulty at each token (i.e., token-level imbalance), and (ii) the harder-to-distill teacher varies across tokens (i.e., cross-token inconsistency), leaving each objective poorly learned at certain steps and hindering positive item retrieval for both objectives, leading to suboptimal performance. To address this challenge, we propose KL-MOPD (KL-Weighted Multi-Objective On-Policy Distillation), a simple yet effective distillation method that allocates token-wise teacher weights proportional to the initial student-teacher token-wise KL divergences, thereby strengthening supervision for the harder-to-distill teacher at every decoding step. Despite its simplicity, we theoretically prove that KL-MOPD strictly outperforms vanilla MOPD in the additive joint recommendation objective under mild conditions. Moreover, KL-MOPD is plug-and-play and computationally efficient, requiring only a one-time pre-training weight calibration with negligible training overhead. Extensive experiments on three public benchmarks, a billion-scale industrial dataset, and online A/B tests on our e-commerce advertising platform demonstrate that KL-MOPD significantly improves both individual and joint recommendation performance. The anonymous code is provided for peer review. Subjects: Information Retrieval (cs.IR) Cite as: arXiv:2610.36973 [cs.IR] (or arXiv:2610.36973v1 [cs.IR] for this version) https://doi.org/10.48550/arXiv.2610.36973 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[IR-16] PailitaoMMSearch2.0 Technical Report: Scaling Generative Retrieval to Unified End-to-End Multimodal Search
链接: https://arxiv.org/abs/2610.36958
作者: Xiaohan Ye,Zihan Gong,Jingqian Zhao,Lianyu Du,Yunmeng Shu,Xu Chen,Zesheng Zhou,Renchun You,Bocheng Chen,Jinsong Lan,Haihong Tang,Bo Zheng
类目: Information Retrieval (cs.IR); Artificial Intelligence (cs.AI)
备注: PailitaoMMSearch2.0 Technical Report
Abstract:Generative retrieval models have been increasingly adopted in search and recommendation systems. However, industrial generative retrieval models are commonly specialized to a fixed online scenario: they learn from scenario-specific interactions, support a narrow set of query forms and retrieval objectives. We study how to remove these couplings and build a unified end-to-end multimodal search system for open-ended image–text queries, heterogeneous shopping intents, and online scenarios with different resource constraints. We present \textbfPailitaoMMSearch, which directly generates product Semantic IDs from a raw query image and free-form instruction and scales along three dimensions. \textbfData scaling uses a reusable toolset that can be composed into intent- and lifecycle-specific workflows for cold-start synthesis, quality control, real-traffic refinement, and continuous data production. \textbfIntent scaling progressively equips one model with e-commerce product knowledge, multiple retrieval relations, and adjustable reasoning budgets through continual pre-training (CPT), mixed-budget supervised fine-tuning (SFT), and unified reinforcement learning (RL). The RL stage combines intent-balanced optimization and Sample–Beam rollouts aligned with multi-candidate serving. \textbfScenario scaling exposes quality–cost operating points across model capacity and reasoning budget. The system is deployed in multimodal search applications on Taobao Pailitao. A controlled online A/B test reports statistically significant improvements, including item page views \textbf+17.90% , transactions \textbf+27.04% , and gross merchandise value \textbf+8.94% .
[IR-17] When Diffusion Recommenders Stop Sampling: Preserving Generative Diversity
链接: https://arxiv.org/abs/2610.36954
作者: Louis Castres Saint Martin(ISIR, MLIA),Alain Rakotomamonjy,Patrick Gallinari(ISIR, MLIA)
类目: Information Retrieval (cs.IR)
备注:
Abstract:A user history can admit multiple plausible next interactions, yet conventional sequential recommenders produce a single deterministic ranking. While diffusion models offer a natural way to sample alternative continuations, we show that existing target-item diffusion recommenders often suffer from severe sampling collapse, producing nearly identical outputs across independent draws. We identify two ubiquitous design choices that promote this collapse: joint learning of the latent space and denoiser, and in-trajectory supervision via discrete losses. To address these failure modes, we introduce DARTS (Diverse and Accurate Recommendation Through Sampling), a three-stage latent diffusion framework that explicitly decouples representation learning, generative modeling, and discrete ranking by learning and freezing the latent space before diffusion training and postponing ranking supervision to the final decoding stage. We further characterize sampling diversity using complementary measures of variation before and after decoding, and use posterior-predictive aggregation to exploit this diversity at inference. Across three Amazon Reviews 2023 categories, DARTS preserves substantially greater sampling diversity while outperforming strong sequential, generative-retrieval, and diffusion baselines by 7-12% on standard ranking metrics. Aggregating multiple samples further improves ranking accuracy at a fixed slate size.
[IR-18] Do We Need an LLM to Know Why Sources Conflict? Decision Models for Knowledge-Conflict Typing in RAG
链接: https://arxiv.org/abs/2610.36915
作者: Parv Khatri,Harshul Jain,Mehak Kapoor,Tanya Sah
类目: Computation and Language (cs.CL); Information Retrieval (cs.IR)
备注: 15 pages, 2 figures, 8 tables. Code, prompts and all model outputs: this https URL
Abstract:When the sources retrieved for a query disagree, the right response depends on why they disagree: a question with several valid answers, a contested topic, stale information and a false source each call for different behaviour. Cattan et al. (2025) showed that telling a generator the conflict type improves its responses by about 24 points, but that predicting the type is hard even for frontier LLMs. We study whether typed decision models, which return probabilities over fixed options in a single forward pass instead of generating text, can do this classification fast and cheaply. Using the CONFLICTS benchmark and the original paper’s prompt verbatim, we evaluate all 11 decision models served by one commercial inference gateway, from 7 providers, against Gemini models on identical inputs. Asked directly, the decision models reach 50.0–67.0% accuracy; 10 of the 11 match or exceed Gemini 2.5 Flash, the paper’s model (61.3%), at 4–39x lower billed cost (2 were free) and 2.6–6.0x lower latency. The current Gemini 3.8 Flash (70.0%) is significantly more accurate than 8 of them, but sending each decision model’s least confident quarter of queries to it gives a cascade statistically tied with it for 8 of the 11, at 25–37% of its cost. End to end, answers generated with the types of 10 of the 11 follow the expected behaviour as often as with the LLM’s types. Decomposing the taxonomy into yes/no questions lowers accuracy for every model, and a repair fitted on one model does not transfer. Pre-specified diagnostics show that dates play no measurable role and that sensitivity to source order varies widely between models. Decision models detect a planted false source less reliably than the LLM as hard labels, but their probabilities still separate planted from clean sources. We release code, prompts and all model outputs at this https URL.
[IR-19] Latent Residual Reasoning for Ranking in Generative Recommendation CIKM2026
链接: https://arxiv.org/abs/2610.36890
作者: Haihan Shi,Ruidong Han,Wenjie Wang,Shangyu Chen,Sunhao Dai,Yongqi Li,Yangyang Li,Fei Jiang,Wei Lin,Fuli Feng,Xiangnan He
类目: Information Retrieval (cs.IR)
备注: Accepted by CIKM 2026
Abstract:Generative recommendation (GR) models for ranking typically adopt a fast-thinking direct prediction paradigm with strong practical efficiency. However, ranking requires fine-grained scoring across multiple candidate items, while existing GR models still mainly rely on one-shot prediction. This raises the question whether additional inference-time computation can further improve ranking quality. Recently, inspired by Chain-of-Thought (CoT) reasoning in large language models (LLMs), some work has begun to introduce reasoning into recommendation systems. Some of these methods rely on LLM reasoning frameworks to perform explicit reasoning in textual space, but such designs are not suitable for more widely deployed non-textual recommendation models. A more relevant direction is to introduce latent reasoning into ID-based recommendation models. Although recent work has extended latent reasoning to industrial ranking, how successive reasoning steps should improve fine-grained candidate scores remains less explored. To address this problem, we propose Refiner, a latent residual reasoning framework with parameter-efficient fine-tuning for generative ranking. Our approach introduces three key components: (1) a residual reasoning mechanism that progressively refines ranking scores through cumulative additive corrections, (2) history queries that provide condensed user behavior representations to enhance candidate-specific reasoning, and (3) a parameter-efficient fine-tuning strategy that leverages LoRA modules to adapt pre-trained models for latent reasoning. Extensive experiments on two public benchmarks, an offline industrial dataset, and an online A/B test demonstrate that Refiner consistently improves over direct prediction, validating its effectiveness and practical utility.
[IR-20] DuTri: Dual-Branch Modeling with Triadic Factorization for Multimodal Multi-Scenario Recommendation
链接: https://arxiv.org/abs/2610.36836
作者: Jintao Huang,Chaohua Yang,Kailiang Hao,Dugang Liu,Xing Tang,Zhong Ming
类目: Information Retrieval (cs.IR)
备注: 12 pages, 7 figures
Abstract:As recommendation scenarios become increasingly diverse and multimodal information grows more prevalent, effectively modeling cross-scenario user preferences and multimodal semantics has become a critical challenge in recommender systems. When incorporating multimodal information, existing multi-scenario recommendation (MSR) methods typically treat textual and visual representations as generic input features and directly feed them into their original parameter-sharing architectures. However, this extension may not fully accommodate the different modeling requirements of collaborative and multimodal information and may not be the best modeling paradigm. Our controlled experiments show that directly adopting these architectures is not the optimal modeling approach. To this end, we propose Dual-Branch Modeling with Triadic Factorization (DuTri), a framework for multimodal multi-scenario recommendation comprising a collaborative branch and a multimodal branch. Specifically, the collaborative branch learns structured collaborative preference representations through CANDECOMP/PARAFAC (CP) decomposition of user-scenario-item interactions and residual quantization, thereby capturing cross-scenario commonalities and scenario-specific differences. The multimodal branch leverages the learned user and scenario preferences to guide adaptive modality fusion; additionally, item-level contrastive learning aligns collaborative item preference representations with multimodal representations to facilitate effective collaboration between the two branches. Extensive experiments on two real-world datasets, AntM ^2 C and Amazon, demonstrate that DuTri outperforms existing multimodal and multi-scenario recommendation baselines in overall recommendation performance, validating the effectiveness of the proposed framework.
[IR-21] he Vocabulary Is Already Here: Adapting Pretrained Interpretable Features for Generative Recommendation
链接: https://arxiv.org/abs/2610.36763
作者: Ziluowen Luo,Ruochen Liu,Jun Yin,Hao Yan,Zhengxuan Hu,Xinze Shi,Chaozhuo Li,Chengqi Zhang,Senzhang Wang
类目: Information Retrieval (cs.IR); Artificial Intelligence (cs.AI)
备注: 13 pages, 8 figures, 7 tables
Abstract:Generative recommendation (GR) predicts the next item through its semantic ID (SID). Existing methods commonly fit SID vocabularies to pretrained language model (PLM) representations of item descriptions, incurring additional training costs and making vocabulary coverage dependent on the fitting domains. We observe that pretrained sparse autoencoders expose PLM features with human-interpretable semantic annotations, allowing feature indices to serve as a reusable SID vocabulary. However, reusing this vocabulary requires preserving item distinctions under compact token budgets and learning to favor target items over candidates with shared dominant features. It also requires extending feature interpretability from item content to retrieval decisions. To address these challenges, we propose API, a framework that Adapts the Pretrained Interpretable vocabulary for effective and explainable GR without additional vocabulary learning. Within API, Pretrained Interpretable IDs (PIDs) retain selected feature indices and relative activation strengths for item discrimination. With PIDs as supervision, the Semantic Prediction Adapter (SPA) learns user preferences from interaction histories through joint feature-index and strength prediction. SPA combines both predictions to score candidate items, with item-level supervision encouraging the observed next item to outrank its alternatives. API further explains candidate score differences through feature contributions paired with their pretrained semantic annotations. Extensive experiments show that API achieves competitive performance across backbones and decoding schemes, while preserving item distinctions under compact token budgets and enabling explainable retrieval.
[IR-22] Recall or Relearn? A Mixture of In-Context Experts for Recurring Concept Drift
链接: https://arxiv.org/abs/2610.36585
作者: Zhangkai Wu,Xu Zhang,Zhihong Cui,Yiqi Wang,Yi Zhang
类目: Machine Learning (cs.LG); Information Retrieval (cs.IR)
备注:
Abstract:Tabular foundation models (TFMs) are being applied to Web data streams such as network intrusion detection, where they adapt by updating the labelled rows of their context without retraining. When a past concept returns after leaving the context, a TFM faces a recall-or-relearn decision between restoring stored rows of the concept and relearning it from new labels. We propose MICE (Mixture of In-Context Experts), a training-free learner that cuts the stream into segments at detected changes. It merges a closed segment into a stored context when the TFM’s estimated additional error on the segment is within a tolerance, under a capacity limit on stored contexts. A fast level weights the stored contexts and first-in-first-out (FIFO) windows by the log-loss of their predictions on the last labelled batch. We show that, with pure stored contexts and a margin condition, this fast level gains over relearning, in expected accuracy, the normalised positive area between the stored context’s accuracy and the TFM’s learning curve, up to a first-batch term. A slow level mixes the fast level’s prediction with that of the FIFO window of the last labelled rows and bounds the discounted half-Brier regret of MICE against this window at every time. Learning curves measured beforehand predict the ordering of the observed recall gains. MICE outperforms reset, which empties the context at detected drifts, on grid streams with random boundary phases. Merging keeps fewer experts than pools that store every segment, and with fixed segmentation a time-matched snapshot pool matches MICE on the shifted grid. On real streams, MICE has the highest mean accuracy of all policies we compare.
[IR-23] Governed Sovereign RAG for Public Administration: Tripartite Epistemic Routing Facet-Guided Clarification and Deterministic Numeric Verification in Citizen Services
链接: https://arxiv.org/abs/2610.36561
作者: Andres Cruz Degante(Universidad Autónoma de Occidente)
类目: Computers and Society (cs.CY); Information Retrieval (cs.IR)
备注: 19 pages, 5 figures, 6 tables. Deployed in production at SEDECO/CEMERT, State of Tlaxcala, Mexico
Abstract:Deploying conversational assistants at government service desks faces a tension between regulatory certainty (zero tolerance for invented fees), data sovereignty prohibiting commercial clouds, and high rates of underspecified or out-of-jurisdiction citizen queries. We present and evaluate a governed, sovereign Retrieval-Augmented Generation (RAG) architecture deployed on the official portal of the State of Tlaxcala, Mexico (this http URL), covering 1,143 procedures and a decoupled directory of 924 offices. The system governs interactions via three pillars: (1) a two-stage tripartite epistemic gate (clarify_policy and a Boltzmann distribution over AnyJev) arbitrating among answering (SPECIFIC), facet-guided clarification (CLARIFY), and abstaining (ABSTAIN); (2) a post-generation deterministic numeric verifier (verify_numeric_guardrail) with SHA-256 cache invalidation ensuring every generated fee or validity figure belongs to the retrieved official record; and (3) an empirical evaluation across 380 organic production traces (8 weeks) and a 200-query stratified benchmark (5-fold CV: 86.41% +/- 7.50%). In production, 89.47% of traffic is resolved at Level 0 via deterministic templates without free generation (global p50 = 0.665 s; direct L0 p50 = 0.445 s; slow-tail p95 = 30.412 s under CPU queuing), with 0/40 fiscal amount hallucinations observed in Level 1 (95% bound = 8.81%). On the benchmark, the system achieves 86.50% strict Top-1 accuracy (+10.00 pp over a binary neural gate, p = 0.00453, p_Holm = 0.01172), pairing a 93.00% conditional safe upper bound with a 70.87% dropout-adjusted effective resolution after accounting for 53.33% citizen drop-off in clarification menus.
[IR-24] CCP-DP: Cross-cluster context-guided phase dynamic programming for DNA storag e sequence reconstruction
链接: https://arxiv.org/abs/2610.36559
作者: Zixin Zhang,Guanjin Qu,Huaming Wu
类目: Information Retrieval (cs.IR)
备注:
Abstract:Motivation: DNA data storage offers high information density and long-term stability, but accurate sequence reconstruction remains challenging under limited sequencing coverage and strong insertion, deletion, and substitution (IDS) noise. Existing reconstruction methods mainly rely on local evidence within individual read clusters. When sequencing coverage is low or indel-induced phase shifts become pronounced, such local evidence may be insufficient to distinguish closely competing candidate sequences. Results: We present Cross-Cluster Context-Guided Phase Dynamic Programming (CCP-DP), which uses dataset-level cross-cluster information to complement cluster-local evidence and selectively refines ambiguous provisional consensuses. Under the main benchmark settings, CCP-DP achieved exact reconstruction rates of 95.36%-99.98% and outperformed existing methods on multiple datasets. At 5\times coverage, CCP-DP improved exact reconstruction by 6.2-17.5 percentage points over the stronger of BBS and CPL across the tested IDS conditions, while maintaining low reconstruction error and computational overhead. Overall, CCP-DP introduces cross-cluster shared information as a complementary source of evidence for DNA sequence reconstruction under challenging conditions. Availability: Source code is available at this https URL. Subjects: Information Retrieval (cs.IR) Cite as: arXiv:2610.36559 [cs.IR] (or arXiv:2610.36559v1 [cs.IR] for this version) https://doi.org/10.48550/arXiv.2610.36559 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[IR-25] QESChunker: A Single Objective Unifies Overlapping and Non-Overlapping Chunking for RAG
链接: https://arxiv.org/abs/2610.36557
作者: Yifan Zhao,Qianyang Zhao,Xinyu Guan,kai wei,Yuming Deng
类目: Information Retrieval (cs.IR); Computation and Language (cs.CL); Optimization and Control (math.OC)
备注: 16 pages, 8 figures, 4 tables
Abstract:Chunking is a fundamental preprocessing step in retrieval-augmented generation (RAG), yet most existing chunkers rely primarily on heuristics. We introduce the Question–Evidence-Supervised Chunker (QESChunker), to our knowledge the first method for constructing static chunk indices from question–evidence supervision. QESChunker constructs document-specific chunk indices by optimizing over human-annotated or LLM-generated question–evidence pairs, without access to held-out test queries. We formulate non-overlapping and overlapping index construction through a unified question–evidence utility-maximization objective, turning chunking into a supervised combinatorial optimization problem. Under the non-overlap constraint, the objective decomposes over chunks and can be optimized exactly in polynomial time by dynamic programming. When overlap is allowed, the objective becomes monotone submodular; exact optimization is NP-hard, but our greedy achieves a (1-1/\sqrte) -approximation. Across two self-constructed and three public datasets, both variants outperform existing chunking methods in complete-evidence recall in nearly all evaluated settings, under the same overlap regime and a fixed hybrid retriever. These results recast chunking as supervised, budget-aware index construction for evidence retrieval.
[IR-26] Monitor Accuracy Isnt the Answer: A Decision-Value Framework for Evaluating AI Agent Restart Monitors
链接: https://arxiv.org/abs/2610.36544
作者: Siddharth Jain,Venkat N. Vedam
类目: Artificial Intelligence (cs.AI); Information Retrieval (cs.IR)
备注: 29 pages, 6 figures, 14 tables
Abstract:Classification accuracy and true-positive and false-positive rates alone do not determine whether restarting an AI agent improves success. Using an exact alarm-to-action identity for an independent clean restart with zero monitoring overhead, we show analytically that a monitor with fixed 90% accuracy can swing from a five-percentage-point loss to a 35-point gain in success rate, depending solely on the replacement attempt’s residual-budget success probability. We introduce two techniques for evaluating monitors and progress signals in decision terms. First, a decision-sufficiency criterion, a binary-action specialization of Blackwell’s comparison of experiments, identifies which progress representations or context compressions preserve the information needed to choose optimally between the specified actions, independent of trace reconstruction fidelity. Second, a finite-sample, model-free decision audit uses simultaneous binomial intervals over existing agent traces, without additional model calls, to certify a preference for “restart” or “continue” under stated sampling assumptions, or return “abstain” when evidence is insufficient. In a study of 194 externally evaluated model executions on 19 selected SymPy issues from SWE-bench Verified, with a retrospective audit of 130 of these traces, we find a 2.29-point apparent restart gain traced to a single surviving failed attempt, a 42.29-point loss on a second model, no positive replication of the initial gain on prospective task cohorts, and zero of 103 live comparisons meeting the specified certification bar. These results motivate evaluating agent monitors through recovery opportunity, decision value, and statistical support, rather than classification performance alone.
[IR-27] ACOR: incremental evidence accumulation towards real-time DNA storag e reconstruction
链接: https://arxiv.org/abs/2610.36500
作者: Linxuan Han,Huaming Wu,Shi Jia,Guanjin Qu
类目: Information Retrieval (cs.IR); Genomics (q-bio.GN)
备注:
Abstract:DNA offers high-density, durable storage by encoding digital information into synthetic strands and retrieving it through sequencing. Sequence reconstruction combines noisy copies of each strand before downstream decoding. Conventional offline methods typically reconstruct completed read clusters, separating reconstruction from ongoing read collection. Streaming reconstruction offers a way to overlap evidence processing with sequencing and reduce the computation remaining before downstream decoding. We present ACOR, a streaming consensus reconstruction method that records k -mer transitions and boundary support in a persistent ledger, then decodes through length-constrained beam search and priority-based selection at cluster closure. Evaluation on real and simulated DNA-storage datasets showed competitive recovery. On our own short-read next-generation sequencing (NGS) wet-lab library, ACOR combined near-complete recovery with second-scale reconstruction of more than ten million reads. A smaller public benchmark was completed in a fraction of a second. Parallel ACOR achieved speedups of tens to over a thousand-fold relative to the tested serial baseline configurations. Sparse storage reduced high-context memory requirements, while parallel execution accelerated reconstruction computation. ACOR is designed for streaming readout pipelines that process assigned reads while sequencing continues.
[IR-28] Cross-Day Compatibility of Daily Item Embeddings: An Empirical Study of Alignment Scale Management and CTR Adaptation
链接: https://arxiv.org/abs/2610.36408
作者: Yicheng Wang
类目: Information Retrieval (cs.IR)
备注: 19 pages, 1 figure, 8 tables. Supplementary material: 10 pages, 8 figures, 2 tables. Includes the complete 48-configuration inventory
Abstract:Daily item co-occurrence embeddings are paired with next-day behavior without future-behavior leakage. Multi-day click-through rate (CTR) training therefore mixes versions whose directions and norms may differ for the same item. We compare Word2Vec and an item-graph XSimGCL adaptation across initialization, inheritance, post-processing, training constraints, and downstream augmentation. The inventory contains 48 configurations, including 32 common-protocol CTR evaluations. Observed metrics are reported as relative percentage changes from the corresponding Word2Vec baseline (0%), with raw values and exact dataset sizes withheld. Three findings emerge. First, multiple upstream interventions improve compatibility and CTR. ID-init and orthogonal alignment increase cosine by 51.2% and 93.4%, and GAUC by 2.2% and 2.6%, relative to ordinary Word2Vec. Retrieval and geometry remain insufficient downstream surrogates. Second, warm starts perform strongly but accumulate scale. The day-matched norm increase over Word2Vec reaches 102.5% after eight days; daily global rescaling limits it to 8.2% and attains the highest observed GAUC increase, 4.0%. Third, CTR learning through a dimensionality-reducing encoder further adapts to residual drift. Against dimension-matched random-projection, SVD, and untrained-network controls, trained encoders attain the highest cross-day cosine for four of six upstream methods. The other two rank third and second in train-to-test cosine. These industrial results support complementary roles for upstream stabilization, inherited-scale management, and downstream task learning.
[IR-29] MAREval: A Multi-Agent Framework for Evaluating Natural Language Recommendation Explanations
链接: https://arxiv.org/abs/2610.36389
作者: Reza Yousefi Maragheh,Jayesh Uddhav Kudase,Aysenur Inan,Ramin Giahi,Kai Zhao,Parth Parikh,Jianpeng Xu,Jason Cho,Evren Korpeoglu,Sushant Kumar
类目: Artificial Intelligence (cs.AI); Information Retrieval (cs.IR)
备注:
Abstract:Evaluating natural language explanations in recommender systems is essential for fostering user trust, transparency, and engagement. However, existing evaluation approaches like human evaluations, while accurate, are resource-intensive and impractical at the scale required by modern recommendation platforms. Also, automated methods using single-agent LLMs suffer from prompt sensitivity and inconsistent outputs. To address these challenges, we propose MAREval, a structured multi-agent framework for evaluating recommendation explanations using large language models. MAREval orchestrates (i) a planner agent that uses a novel Chain of Debate (CoD) prompting strategy to coordinate agent roles and enforce logically consistent evaluation plans; (ii) a moderator agent that regulates discussions by mitigating prompt drift; and (iii) an arbitrator agent that aggregates outputs from multiple evaluation rounds and (iv) a Monte Carlo sampling method, improving robustness and alignment with human judgment. We conduct comprehensive evaluations on both public (TopicalChat) and proprietary recommendation datasets, demonstrating that MAREval outperforms state-of-the-art baselines. Comprehensive experiments on a public benchmark and a proprietary e-commerce dataset show that MAREval improves alignment with human judgments over strong singleand multi-agent baselines. Stability analyses indicate substantially lower variability across repeated trials. In a large human-annotation gate, MAREval meets production quality thresholds where prior evaluators fall short, and online A/B testing demonstrates statistically significant improvements in engagement and revenue metrics. These results establish MAREval as a scalable and reliable solution for human-aligned evaluation of recommendation explanations in real-world systems
[IR-30] From Generation to Scoring: Scaling LLM -Based Semantic Labeling at LinkedIn
链接: https://arxiv.org/abs/2610.36383
作者: Jian Sheng,Sriram Vasudevan,Sundara Raman Ramachandran,Yang Hu,Junyu Zhou,Xiaojing Ma,Haolan Qin,Tiffany Zhou,Vedant Vipul Jhaveri,Ashwini Bhingare,Chinmay Kini,Qing Lan,Xiaoqing Wang,Suleyman Cetintas,Fedor Borisyuk
类目: Machine Learning (cs.LG); Distributed, Parallel, and Cluster Computing (cs.DC); Information Retrieval (cs.IR)
备注:
Abstract:For efficiency, industrial large language model (LLM) labelers often pack many candidate labels into one prompt and generate a YES/NO answer for each, even when each decision depends only on the entity and one candidate. Generating these independent decisions as one answer list couples them, spends decoding on the list, and relies on the list being well-formed. In this paper, we present an end-to-end recipe that reformulates the labeler as semantic scoring, fine-tuning it on single-candidate prompts to return each decision as one answer-token score from a prefill-only pass. We also adapt Multiple-Item Scoring (MIS) to score hundreds of candidates in one sequence with attention masked between them. Deployed for profile qualification labeling at LinkedIn, semantic scoring changes at most 0.01% of decisions when candidates are reordered and matches the agreement with reference labels that packed generation reaches only with about 4.7x as many training pairs. The MIS-based stack raises throughput over the deployed stack by 1.24x offline and 1.9x in model serving under a one-second p99 latency target. For offline batch inference, we implement MIS-based scoring as a reusable, model-agnostic Ray Data API proposed for upstream adoption. Model serving uses MIS support contributed to SGLang.
[IR-31] Preserving rare-item recall under a storag e budget: quantization distillation and the limits of teacher fidelity
链接: https://arxiv.org/abs/2610.36267
作者: Han Chen,Yingrui Li
类目: Information Retrieval (cs.IR)
备注: 23 pages, 2 figures, 8 tables. Supplementary data and code included as ancillary files
Abstract:Under a fixed embedding-storage budget, recommenders can use a small full-precision model, a distilled student, or a higher-rank low-precision model. We asked which choice preserves rare-item recall and whether teacher fidelity reflects it in seven dataset-backbone cells: matrix factorization on Gowalla, this http URL, Amazon Video Games, Steam, Goodbooks-10k and MovieLens-20M, and LightGCN on Amazon Video Games. Seven candidates fit in 168 bytes per user or item row: 40-dimensional Bayesian personalized ranking (BPR), two 40-dimensional unbiased knowledge distillation (UnKD) students, and factorizations of a 300-dimensional teacher’s scores at ranks 40, 80, 155 and 300, stored in 32, 16, 8 and 4 bits. Each test set was evaluated once after fixing candidates and confirmatory analyses. The rank-300 4-bit factorization used 13% of teacher storage and retained 95-103% of its tail recall@20 and 93-97% of its NDCG@20 in every cell. The rank-40 factorization had the lowest tail recall in every cell; the 4-bit factorization had 1.2-12.7 times its tail recall and 0.99-1.46 times its NDCG. UnKD students retained 56-102% of teacher tail recall. In all ten prespecified comparisons across five prospective-expansion cells, the 4-bit factorization had higher tail recall, with simultaneous 95% intervals excluding zero. Teacher-hit retention, the share of the teacher’s relevant tail hits also found by a model, ordered 23 of 147 candidate pairs differently from tail recall and identified the wrong tail-recall leader in 3 of 7 cells: models can find relevant tail items the teacher misses. In these cells, spending bytes on rank rather than precision preserved rare-item recall at 7.5 times the arithmetic per score. Compressed recommenders should be judged by total tail recall, not teacher fidelity alone.
[IR-32] BehaveRec: When Does Long-Term Behavioral History Help Sequential Recommendation? A Study with Session History and Community Signals
链接: https://arxiv.org/abs/2610.36239
作者: Aysenur Inan,Reza Yousefi Maragheh,Jayesh Uddhav Kudase,Selene Xu,Kaushiki Nag,Jianpeng Xu,Sushant Kumar
类目: Artificial Intelligence (cs.AI); Information Retrieval (cs.IR)
备注:
Abstract:Sequential recommendation models typically predict a user’s next action from recent session interactions, which limits personalization when short-term signals are weak or ambiguous. Rather than assuming that incorporating longer-term history and community signals is universally beneficial, we study when it actually helps. We propose BehaveRec, a three-signal framework combining a causal session transformer, a bidirectional User History Encoder over a 90-day window, and a GraphSAGE module over a user-user graph built from behavioral embedding similarity rather than item co-occurrence; a learned gate blends the three signals per user. On an industrial dataset of 137M sequences from 12.6M users, BehaveRec improves R@5 by +3.2% over SASRec, with a larger-capacity SASRec baseline (more parameters than BehaveRec) still trailing by +2.5%, ruling out capacity as the explanation. More importantly, a user-level stratification by history length reveals a clear dose-response relationship: BehaveRec’s advantage is negligible below roughly 10 interactions and grows monotonically with history richness, reaching +0.017 R@5 for users with 101+ events. This pattern is corroborated across one industrial and three public datasets spanning history-rich to history-sparse regimes, where session-only models remain preferable. These headline results, and the resulting deployment threshold, are established for category-level intent prediction; a preliminary, single-run check on the full item-level vocabulary yields a directionally mixed result and is scoped as an open question rather than a claim. We use these results to derive a practical deployment rule – roughly 20 interactions per user – for when long-term history and behavioral-graph signals are worth their added complexity. Despite this added complexity, BehaveRec remains lightweight: it uses 5.67M parameters and trains on a single GPU in under two days.
[IR-33] Counterfactual Profile Perturbations for Recommendation Explanations: A Reproducibility Study
链接: https://arxiv.org/abs/2610.36208
作者: Meysam Varasteh,Elizabeth McKinnie,Robin Burke
类目: Information Retrieval (cs.IR); Artificial Intelligence (cs.AI)
备注:
Abstract:This reproducibility paper explores the counterfactual explanation technique known as Learning to eXplain Recommendations (LXR). Prior work has shown that LXR can reliably identify perturbations that, if applied to a user’s profile, cause the user’s top ranked item to no longer be recommended, thereby explaining the item’s recommendation as a function of the user’s rating choices. In this paper, we extend this work in multiple ways. We expand the scope of the original LXR study to look at explanations for items other than the most highly recommended item, finding that LXR is also effective at this version of the task. We consider alternate baselines for LXR, including item popularity and latent factor similarity, and we see that for some tasks these alternatives work surprisingly well. We also examine the impact of item popularity, finding that LXR offers better explanation quality (shorter perturbations) when less popular items are being explained. Finally, we conduct a brute-force exploration of a subset of the perturbation space in order to estimate a lower bound for optimal perturbation size against which LXR and other such methods can be compared.
[IR-34] Is the education gap in generative AI use smaller than the internets was at a similar level of use? Evidence from 16 European countries
链接: https://arxiv.org/abs/2610.36181
作者: Keisuke Watanabe
类目: Information Retrieval (cs.IR); Computers and Society (cs.CY)
备注: 16 pages, 2 figures. Code, data and outputs are provided as ancillary files
Abstract:New technologies are usually adopted first by people with more formal education. This paper asks whether the education gap in generative AI use in 2025 is smaller or larger than the internet’s gap in the same country in the year when overall internet use was at a similar level. It uses public Eurostat survey data on ICT usage by individuals aged 16 to 74. Each country’s 2025 rate of generative AI use in the last three months is matched to the internet year whose overall last-three-months use was closest, within 5 percentage points, and the gap between people with high (ISCED 5 to 8) and low (ISCED 0 to 2) completed education is measured as a log ratio and as a percentage-point difference. Sixteen of 35 countries could be matched; most of the others were early internet adopters. In the matched countries, the generative AI gap was smaller in 13 of 16 countries on the log ratio (median change -0.625, 95% bootstrap interval -0.969 to -0.291, two-sided sign test p = 0.021) and in 15 of 16 on the percentage-point gap (median change -27.6 points, interval -32.2 to -16.8, p = 0.0005). The point-gap result keeps its sign in every generative-AI-versus-internet sensitivity analysis with 8 or more countries, including comparisons restricted on both sides to ages 25 to 54 and 55 to 74, but its size is sensitive to matching that takes education composition into account (about -10 to -11 points in 7 and 8 countries). The ratio result is fragile: it fails the decision rule when countries matched at their first internet observation are removed (9 of 11, p = 0.065). The analysis follows an unpublished plan whose timing cannot be verified, so it is not presented as pre-registered. The results describe unadjusted education contrasts in a non-random set of countries, not an education effect, and do not generalise to the EU as a whole.
[IR-35] When Connectivity Misleads: Guiding Agent ic Search with Contrastive Evidence Paths
链接: https://arxiv.org/abs/2610.36134
作者: Junchen Li,Liang Xu,Rongzheng Wang,Qizhi Chen,Junhao Wen,Shuang Liang
类目: Information Retrieval (cs.IR); Artificial Intelligence (cs.AI)
备注:
Abstract:Search agents can use cross-document graphs to discover sources through multi-hop connections and choose subsequent actions based on candidate rankings and relationships within a limited observation window. However, documents may rank highly because their numerous connections make them easier to reach, without necessarily providing more valuable evidence. Graph diffusion methods such as Personalized PageRank (PPR) aggregate propagation contributions from multiple sources, allowing highly connected distractor nodes to crowd the observation window. Flat node scores also obscure cross-document evidence trails, limiting continued exploration. We propose Contrastive Evidence Paths (CEP), which scores candidates by their strongest individual paths on a query-conditioned document graph and calibrates these scores against the same graph’s background visitation distribution to mitigate structural bias. CEP ranks candidates by the calibrated scores and retains the corresponding source paths. Building on CEP, we develop Structured Corpus Interaction (SCI), which organizes candidates and their connecting paths into a structured Map, enabling agents to verify source documents and pursue missing information through expansion from anchor documents. Experiments on BrowseComp-Plus and TREC-COVID show that SCI achieves state-of-the-art answer accuracy and nDCG@10 with the lowest generation cost. Compared with the strongest baselines by task performance, maximum reductions in tool calls and total generation costs exceed 50% and 40%, respectively. Further analysis confirms that SCI with CEP suppresses ranking inflation induced by structural connections and improves gold recall and exploration efficiency.
[IR-36] RELAY: One Query Two Roles for Sequential Recommendation
链接: https://arxiv.org/abs/2610.36000
作者: Zesheng Li,Chengchang Pan,Honggang Qi
类目: Information Retrieval (cs.IR)
备注:
Abstract:Sequential recommendation must combine semantic preferences with behavioral associations to deliver accurate rankings at low inference cost. Sharing history selection connects preference representation directly to the behavioral evidence used in candidate scoring. We present RELAY, a compact sequential recommender built around one query, two roles: semantic readout and behavioral-memory routing. A single learned query forms a semantic preference representation and reuses its attention to route direct-history and transition evidence into candidate scores. Candidate-conditioned gates calibrate the two memory channels for each candidate. RELAY trains both roles with a single catalog likelihood and produces the final ranking in one forward pass. Full-catalog experiments on six datasets show that RELAY achieves the highest Recall@5 and NDCG@5 among the compared methods on every dataset, improving over the previous best model by 14.22% and 17.88% on average. With 2.044 million predictor parameters, RELAY reduces predictor size by 65.12% and achieves a 1.75~1.79 \times inference speedup over the previous best model.
人机交互
[HC-0] Who Shaped What? Untangling Human and AI Creative Contributions Through Spatial Provenance in Image Editing
链接: https://arxiv.org/abs/2610.37469
作者: Riya Sahni,Christine Dierk,Lydia Chilton,Eunyee Koh,Doga Dogan,Tongyu Zhou
类目: Human-Computer Interaction (cs.HC)
备注: 24 pages, 7 figures
Abstract:Images co-edited with AI are often labeled to disclose AI use. However, creators feel these labels do not sufficiently describe their contributions to the editing process. Editing histories preserve requests, suggestions, and revisions, but how do creators use these logs to judge contribution? We investigated this with Roop, a mixed-initiative design probe that links image regions to editing histories (spatial provenance) and summarizes interactions in a heat map. Twenty participants edited images, annotated human and AI contributions, then reviewed the spatial provenance heat map. Participants credited themselves for directing changes while crediting AI for determining visual details. They judged contribution partly by whether their interactions achieved their intentions. Reviewing provenance helped participants remember interactions without always changing whom they credited. These findings motivate provenance systems that preserve enough context for creators to explain how interactions shaped their work, alongside the record of what happened.
[HC-1] PolyGandr: Steering Generative Models via Spatial Navigation through Latent Space
链接: https://arxiv.org/abs/2610.37446
作者: Mingwei S.G. Li,Susie S.Y. Li,Isabel Moreira de Oliveira,Ryan P. Adams,Remco Chang
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:Generative models support content creation across text, images, videos, and design disciplines, yet their interfaces rely primarily on natural language. While effective for specifying high-level intent, language can be inefficient and imprecise for fine-grained control, iterative refinement, and open-ended exploration. We present PolyGandr, a visual interface for exploring relationships between latent representations and generated outputs, complementing natural-language model steering. PolyGandr formulates generative control as high-dimensional navigation, enabling users to define starting points, steering directions, and step sizes through a unified map-and-mixing-palette interface. Combining global exploration with local refinement supports both broad design-space navigation and fine-grained output manipulation without repeatedly translating visual intent into prompts. We evaluate PolyGandr across three use cases: StyleGAN2 face generation, Stable Diffusion chair design, and rotational kirigami engineering. A formative study indicates that PolyGandr complements existing workflows by supporting real-time editing while simultaneously providing an overview and means to explore the broader design space.
[HC-2] me-Dynamic Semantic-Activity Entanglement: Quantifying Communication-Based Synchrony in Collaborative VR
链接: https://arxiv.org/abs/2610.37427
作者: Pooja Pol,Jianing Zhang,Qing Huang
类目: Human-Computer Interaction (cs.HC)
备注: 6 pages, 3 figures. This work has been accepted as an extended abstract in the “Poster” track for publication in the Companion Proceedings of ACM ISS 2026
Abstract:Collaborative virtual reality (VR) produces rich multimodal traces of team behavior, yet practical methods for characterizing time-varying team communication remain limited. We present time-dynamic semantic-activity entanglement (TSAE), inspired by the entanglement concept introduced by Gloor et al. (2022), to describe communication-based team synchronization within discrete time windows. TSAE is modeled by combining relative speaking activity with contextual semantic alignment for participant pairs, which is aggregated into participant-level trajectories and summarized through the Gini coefficient to understand the distribution of team-level communication synchrony. This approach is employed with German-language transcripts from fifteen collaborative VR gameplay sessions. Transcript utterances are represented using a long-context embedding model with a late-chunking strategy and analyzed within fixed game-time windows. A representative session illustrates shifting dyadic alignment and low-to-moderate participant contributions, with sporadic periods of high entanglement. Across fifteen game sessions, the metrics indicate complementary insights, offering a qualitative contrast to the core-periphery pattern reported in the foundational work. Hence, these descriptive findings support the study of communication as a dynamic process and provide a mathematically grounded approach to analyze its evolving distribution in shared interactive VR spaces.
[HC-3] Comparing Temporal Deep Learning Architectures for Discrete Cross-User Myoelectric Control
链接: https://arxiv.org/abs/2610.37374
作者: Ethan Eddy,Evan Campbell,Scott Bateman,Erik Scheme
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:The recognition of dynamic gestures via discrete myoelectric control is becoming an attractive input modality for event-based human-machine interaction. While recent work suggests that discrete cross-user models are possible (i.e., no calibration data is required from the end user), few have evaluated techniques to optimize these systems and their generalizability to new randomly prompted data simulating real-world discrete control is relatively unknown. Using the 612-user EMG-EPN dataset, this work benchmarks five temporal deep-learning approaches, evaluates the tradeoff between performance and model complexity, and compares hand-crafted and machine-learned feature extraction techniques. The CNN-GRU achieved the best accuracy of 93% when tested on 184 previously unseen users. These models were then evaluated on a different, more realistic, unsegmented dataset, wherein users were randomly prompted to elicit discrete gestures during a continuous data collection. This dataset was recorded five years apart and in a separate country from the data used to train the original model. Results show that the cross-user results scale to this new dataset (accuracies of 96%) even though no feedback was given to users during the elicitation of gestures. These findings highlight the exciting potential of discrete zero-shot cross-user myoelectric control and offer valuable insights for the continued advancement of this promising technology.
[HC-4] Understanding Human-Robot Collaboration through Joint Action
链接: https://arxiv.org/abs/2610.37298
作者: Chengwen Zhang,Jianyang Wang,Chang Liu,Yukun Wang,Yuan Xu,Zhoutong Ye,Jie Cai,Chun Yu,Yuanchun Shi
类目: Human-Computer Interaction (cs.HC); Robotics (cs.RO)
备注: 26 pages, 5 figures, 32 tables
Abstract:Human-robot collaboration requires partners to coordinate interdependent contributions toward a shared goal. Joint Action theory provides concepts for understanding this coordination, but how these concepts are instantiated and combined across empirical research remains fragmented. We conducted a systematic review following PRISMA 2020, identifying 83 empirical papers published between 2021 and 2026. Through theory-driven framework analysis, we organized the literature into nine constructs across three capacities: Representation, Prediction, and Integration. We analyze how these constructs are distributed and co-occur across studies, revealing where empirical evidence concentrates and where it is absent. By mapping empirical HRC research onto a unified Joint Action framework, this review offers a coherent overview of HRC studies, identifies gaps in current evidence, and provides guidelines for future research.
[HC-5] An Interactive Mobile Augmented Reality System for Efficient Virtual Model-to-Patient Registration in Thoracic Surgery
链接: https://arxiv.org/abs/2610.37271
作者: Emmanuel Ajaero,Pascal Spiegler,George Rakovich,Marta Kersten-Oertal,Yiming Xiao
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:Video-Assisted Thoracoscopic Surgery (VATS) is a minimally invasive procedure, where precise surgical port placement is vital, yet it relies heavily on surgeon experience and mental mapping of pre-operative imaging in the operating room. To address this challenge, we present an accessible, interactive mobile augmented reality (AR) framework, HILMARTS, to enable accurate and efficient pre-operative virtual model-to-patient registration for AR-guided surgical navigation. Our system runs on a commercially available LiDAR-equipped iPad and supports a user-in-the-loop workflow through a two-stage coarse-to-fine registration process. First, the user performs a manual alignment using the AR view on the iPad. This initial maneuver is then refined through automatic point-cloud registration leveraging the LiDAR sensor. We further introduce a novel occupancy-grid-based Surface Coverage Score to provide real-time, actionable feedback to the operator on the geometric completeness of LiDAR data capture for reliable point-cloud-based alignment. In a phantom study with 16 participants, the system achieved a 3.31+/-1.26mm registration accuracy in ~4 min with a mean System Usability Scale (SUS) of 84.1. These findings highlight the system’s accuracy, efficiency, robustness, and potential clinical usability, and could support its applicability to other AR-based surgical navigation tasks.
[HC-6] Peers4U: Designing Learner-Centered Virtual Peers for Childrens Multimodal Mathematics Learning
链接: https://arxiv.org/abs/2610.37238
作者: Zhi Zheng,Binglin Liu,Weihao Chen,Lan Luo,Yu Mei,Jie Cai,Chun Yu,Yuanchun Shi
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:In early mathematics, children learn abstract concepts through concrete representations, such as manipulating objects, drawing, and speaking. Learning with peers can enrich this process by exposing children to diverse strategies and enabling them to exchange ideas through interaction. We realized this potential with virtual peers in Peers4U, a multimodal learning system to enhance children’s mathematical conceptual understanding. Virtual peers driven by LLMs are personalized to the learner’s cognitive level and assigned multimodal solutions with inspectable processes. During learning, peers engage in mixed-initiative interactions by responding, initiating topics, and advancing discussions along a pedagogical progression. A user study with 22 children and two experts showed our learner-centered topic orchestration significantly outperformed a persona-centered baseline, and children spontaneously authored new artifacts to demonstrate inspired ideas or share their own strategies, producing significantly more solutions when grouped with more Leading Peers. We contribute design insights for multimodal, learner-centered virtual peer systems.
[HC-7] Science-Fiction Transformation as a Design Lens for Visualization Transformation
链接: https://arxiv.org/abs/2610.37131
作者: Abdulhaq Adetunji Salako,Christian Tominski
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:Science fiction often features characters or objects that undergo radical changes in form while still being recognizable as the same entity. For example, a vehicle might unfold into a robot, or one body might flow or shape-shift into another. We use such transformations as a reflective lens to study similar shape-shifting in visualization, where one visual representation transforms into a completely different one. For example, a node-link diagram might transform into a parallel coordinates plot as a user’s task changes from understanding relationships to identifying correlations. Based on four instances of science-fiction transformation, our paper highlights two key questions: First, why is a transformation into a target form needed, or what is the purpose? Second, what remains recognizable across the transformation from source to target, or what is the correspondence? Through four hypothetical transitions between visual representations of a multifaceted dataset, we demonstrate that purpose is also essential to visualization transformations. Furthermore, we show that the question of correspondence results in different levels of complexity in terms of designing and understanding transitions. Depending on whether correspondence is 1:1, 1:N, or N:1, visualization transformations may be direct and easy to understand, or they may require more elaborate transition constructs, implying increased cognitive effort for interpretation. Particularly, cases of 1:N and N:1 correspondence pose challenges for designing visualization transformations.
[HC-8] Words or Deeds? How LLM s Predict Users Decisions in Human-Agent Dialogue
链接: https://arxiv.org/abs/2610.36813
作者: Baiqiao Zhang,Shuai Ma,Zhifeng Liao,Chao Zhou,Jianshu Wang,Yuhan Zhao,Chaoyue Zhang,Tongjun Si,Yuying Tang,Xiangxian Li,Yong-Jin Liu,Yulong Bian
类目: Computation and Language (cs.CL); Human-Computer Interaction (cs.HC)
备注:
Abstract:Web systems increasingly predict users’ decisions from what users say, yet when interests partly conflict, users may promise to cooperate to win the other side’s cooperation and then not do so. Sales or service agents that rely on such promises may grant concessions to users who then do not keep their promises, at a cost to the party they represent. We test how LLMs combine users’ statements with decision records to predict users’ decisions, using 1,260 rounds in which 42 users played a repeated prisoner’s dilemma-style game with five agents, each prompted to express a dominant Big Five trait, and could talk with the agent before every decision. Across eight open-weight LLMs, predictions from decision records separated cooperation from defection well (AUC = .80), and adding dialogue brought no detectable gain. With dialogue, predictions agreed with users’ statements more often, and errors concentrated on rounds in which users broke their statements: accuracy on rounds with a statement barely changed (.73 to .74), while it rose on kept statements (.76 to .95) and fell sharply on broken ones (.65 to .17). Predictions weighted statements more heavily than users’ behavior warranted: when users stated a decision they had not yet made in that game, the predicted probability of keeping it rose from .13 to .69, whereas users kept .40 of such statements. Users’ decisions also varied with the agent they faced, but predictions hardly reflected this: users cooperated more with the Agreeableness agent than with the Extraversion agent (70.6% vs. 54.5%), whereas predicted cooperation was nearly identical (64.5% vs. 63.1%). Users also cooperated more with an agent when they themselves scored higher on that agent’s trait, whereas predicted cooperation barely changed with these scores. Dialogue-based predictions should therefore weigh users’ statements by their earlier decisions and the agent they face.
[HC-9] Externalizing Hand Grip Pressure Through Real-Time Visual Feedback
链接: https://arxiv.org/abs/2610.36769
作者: Andrew L. Huard,Avinash Ajit Nargund,Misha Sra
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:Grip pressure, including contact force and its distribution across the hand, is central to skilled grasping but is not directly visible. This makes it difficult to teach, learn, and self-correct in grip-dependent motor tasks, where verbal instruction and tactile demonstration provide only partial guidance and no real-time, spatially resolved feedback tied to the learner’s own grip. Using an instrumented glove, we compare pressure feedback visualizations for such guidance, evaluating how visualization type and, secondarily, sensor-region mapping shape pressure-matching performance in novices. In a study using baseball and table tennis as representative grip tasks (N=28), we find that a continuous gradient visualization produces significantly more accurate pressure matches and lower gaze entropy than a discrete bar display, despite participants perceiving the bar as more precise for pressure magnitude. Perceived precision and objective performance therefore diverged: the encoding participants judged more precise was not the one they matched more accurately with. Grip-tailored sensor-region mappings further reduced perceived temporal demand and effort, although this subjective workload advantage did not extend to objective matching accuracy. Overall, continuous pressure feedback supported more accurate grip matching and more concentrated viewing, and we present design considerations for real-time pressure feedback in motor skill guidance.
[HC-10] ranslating Poetic Emotion into Dynamic Mathematical Visuals: A Strange Attractor-Based Approach
链接: https://arxiv.org/abs/2610.36698
作者: Subhadra Nayak
类目: Human-Computer Interaction (cs.HC)
备注: 6 pages, 5 figures, Going to be presented for the Student Research Consortium (SRC) at India HCI 2026 (Indian Institute of Technology, Hyderabad)
Abstract:Poetry conveys emotional experiences through language. This project investigates how poetic emotional experiences can be represented through dynamic mathematical visualisations while further developing a pipeline for this purpose. By leveraging a pretrained RoBERTa-based emotion classifier, the system analyses the fine-grained emotional states within a poem. These detected emotions are then mapped onto the parameters of strange attractors, complex structures from dynamical systems theory, to generate abstract, deterministic, and dynamic imagery. This work situates itself at the intersection of poetry, psychology, and mathematics, aiming to extend the poetic experience beyond language into the sensory domain
[HC-11] Beyond the 1D Reality-Virtuality Continuum: A 2D Framework for Seven Extended Reality Paradigms
链接: https://arxiv.org/abs/2610.36674
作者: Rajashekhar Vachiravelu Saminathan
类目: Human-Computer Interaction (cs.HC)
备注: Accepted for IndiaHCI 2026 Conference in the Student Research Consortium held at Indian Institute of Technology - Hyderabad, India; 7 pages, 4 figures
Abstract:The Reality-Virtuality (RV) continuum helps in describing the relationship between physical and computer-generated environments. However, the increasing number of Extended Reality (XR) paradigms introduces new technologies. Their characteristics cannot be completely represented by a single real-virtual dimension. It can be noted that XR systems not only vary in their proportion of real and virtual content but also adds, suppresses, or transforms information available to the user. This work proposes a framework for characterizing seven XR realities, namely Virtual Reality (VR), Augmented Reality (AR), Mixed Reality (MR), Augmented Virtuality (AV), Phenomenal Reality (PR), Annexing Reality (AnR), and Diminished Reality (DR). It uses a two-dimensional extension of the conventional 1D RV continuum. The first dimension helps in identification of the position of an XR technology between real and virtual environments. The second dimension captures information addition and subtraction. The proposed approach gives a qualitative basis for comparing existing and emerging XR paradigms within a unified XR framework.
[HC-12] Measuring Users Security and Privacy Concerns About Third-Party Community Bots on Discord NDSS
链接: https://arxiv.org/abs/2610.36626
作者: Kofi Ampomah,Tu Le
类目: Cryptography and Security (cs.CR); Human-Computer Interaction (cs.HC)
备注: Accepted to NDSS Symposium 2027. 20 pages. Artifact: this https URL code: this https URL
Abstract:Third-party community-based bots are central to how social platforms such as Discord work. They moderate content, manage roles, play music, and support AI-assisted conversations across millions of servers. Yet when a bot is installed in a server, it can access messages, user profiles, and server metadata concerning members who did not participate in the installation decision, with no standardized way to communicate the associated security and privacy implications. In this paper, we analyze 23,102 third-party Discord bots and 183,142 user reviews spanning 49 detected languages and five years of feedback (2021-2026). We employ LLM-based knowledge distillation and fine-tune a DeBERTa-based classifier to identify candidate security and privacy (SP)-relevant reviews. After manually validating the complete classifier-positive set, we retain 8,832 reviews identified as SP-relevant, representing an observed pipeline yield of 4.82%, and map them to a 31-theme SP taxonomy extending prior work. Our results show that SP-related reviews focus heavily on Data Sharing (30.5%) and Data Collection (16.2%). Confession and anonymity bots have pipeline-identified SPRRs above 50%, more than ten times the ecosystem-wide pipeline yield of 4.82%. Star ratings may obscure the frequency and valence of SP-related discussion, while conventional metadata features, such as permission count and administrator status, show only weak associations with user-reported SP concerns. We then build BotLens, a proof-of-concept auditing framework that combines review-derived SP signals with bot metadata to provide on-demand decision support for the community-based bot ecosystem. BotLens exposes six tools through a Model Context Protocol server, supporting bot-level checkups, server-wide comparison, and integration with AI agents.
[HC-13] CogArena: Benchmarking Multimodal Agents on Interactive Behavioral Experiments NEURIPS2026
链接: https://arxiv.org/abs/2610.36484
作者: Kianté Fernandez,Caitlin Chen,Jialu Zhou,Bartek Sadowski,Anthony C. Miceli,Ian Krajbich
类目: Artificial Intelligence (cs.AI); Computers and Society (cs.CY); Human-Computer Interaction (cs.HC); Neurons and Cognition (q-bio.NC)
备注: 28 pages, 4 figures, 11 tables. Accepted at NeurIPS 2026 (Evaluations Datasets Track). Code: this https URL . Leaderboard: this https URL
Abstract:Claims about the cognitive capabilities of large language models are typically tested by translating behavioral paradigms into text prompts, bypassing the visual interfaces and time pressure through which human participants actually engage with experiments. At the same time, AI agents can now complete online behavioral tasks and produce human-like data, which raises concerns that online datasets in the social and behavioral sciences are increasingly open to contamination. We introduce CogArena, a benchmark of ten experimental paradigms from the social and behavioral sciences. Agents complete the experiments through a standard web browser, working with the same visual stimuli and response interfaces that online participants use. A three-level scoring procedure evaluates whether the agent completed the task, how accurately it performed, and whether its data show the behavioral signatures each paradigm reliably produces in people. Across four multimodal models, each run ten times per task, we find that finishing an experiment and reproducing its signatures are different things. Many sessions did not finish, and those that did reproduced some signatures reliably and others not at all. CogArena provides the tasks, scoring code, and a public leaderboard for new submissions.
[HC-14] What Victims Take Away: Attack Reuse and Defense Revision after Prompt-Injection Attacks
链接: https://arxiv.org/abs/2610.36422
作者: Zihan Chen,Lei Nico Zheng,Di Zhu
类目: Cryptography and Security (cs.CR); Human-Computer Interaction (cs.HC)
备注: 20 pages, 5 figures, 10 tables. Working paper
Abstract:Attack logs and red-team reports show targets the attacks used against them, to help them repair their defenses. Prompt injections against large language models (LLMs) are plain text, so such records also hand targets a working copy. Drawing on social learning theory and work on the victim-offender overlap, we study what targets do with that copy in Tensor Trust, an online game whose players defended LLM-guarded accounts, attacked one another’s, and could read every attack they received. Using all 563,349 attack records, we compare each received attack with matched attacks that circulated among similar players at the same moment but never reached the target. Reuse mainly followed success: within a week, targets attacked other players with 10.0% of the attacks that had broken their defense, versus 0.03% of matched attacks (adjusted difference: 8.8-10.0 percentage points). Near-misses, whose replies began with “access granted” but that scored as failures, were reused about as rarely as failures: reuse tracked whether an attack paid off, not how its reply read, consistent with payoff-biased social learning. Breaches were also followed by defense changes, after which same-family attacks became 24.5 percentage points less successful against the target, relative to other defenders. Reuse came mainly from players already attacking, not newcomers. Previously received texts accounted for 23.2% of the successful hijacks we analyze. The attack history thus served both repair and reuse; we derive testable hypotheses for how AI security tools should disclose attacks, such as describing the technique without the full text.
[HC-15] When AI Can Build It What Should We Assess? Rethinking Evidence of Engineering Competence
链接: https://arxiv.org/abs/2610.36413
作者: Ali Shakiba
类目: oftware Engineering (cs.SE); Human-Computer Interaction (cs.HC)
备注: 11 pages, 1 figure, 4 tables
Abstract:Generative AI can produce code, tests, documentation, and technical explanations that resemble artefacts traditionally used as evidence of student competence, weakening the inference from artefact quality to the engineering capability behind it. We report the redesign and first deployment of a 13-week, AI-permissive, software-intensive engineering course in which students may delegate substantial work to AI while retaining responsibility for planning, verification, review, integration, and explanation. The redesign couples an engineering accountability workflow, which structures human-AI work and preserves human decision authority, with an assessment evidence architecture comprising six complementary channels: product, process/origin, verification, explanation, reflection, and individual assurance. A collect–cross-check–escalate (CCE) reasoning logic connects these channels rather than treating any single artefact or trace as sufficient evidence of competence. By the end of Week 11, the deployment involved 87 students organised into 17 teams, with 957 individual weekly Git repositories and 34 group-project Git repositories supporting repeated assessment and engineering activity. The first deployment yielded four practical lessons: AI assistance changes where rubric criteria carry evidential weight; multiple evidence channels require deliberate cross-checking and meaningful complementarity; the evidential strength realised in practice may differ from that intended in the assessment design; and automation scales evidence collection more readily than interpretation. The contribution is a conceptually grounded, deployed design for making engineering responsibility and competence more visible when AI can perform substantial artefact production.
[HC-16] Your Wish Is My Commit: Over-Compliant Refactoring in AI Coding Agents
链接: https://arxiv.org/abs/2610.36346
作者: Haocong Jiang,Sadra Sabouri,Sepand Haghighi,Souti Chattopadhyay
类目: oftware Engineering (cs.SE); Human-Computer Interaction (cs.HC)
备注:
Abstract:Large language models are known to be sycophantic, over-complying with users even when doing so is unproductive. When these models act as coding agents, this can be costly: an agent that keeps editing after no edit improves the code writes churn, technical debt, and review burden into the codebase. Yet the consequences of sycophancy in software engineering remain largely understudied. To understand them, we run a controlled study of iterative refactoring, an open-ended task with no stopping criterion, where any edit an agent makes once the code is already clean reflects compliance rather than improvement. Across algorithmic (n=50) and object-oriented (n=10) suites, we repeatedly ask two widely used agents, Codex and Claude Code, to refactor further across five models. Over 3,000 turns, we define six behavioral signals of over-compliance. We hypothesized that a capable agent would stop once it recognized an optimized codebase; none did. Every model kept editing, rewriting up to 31x the original codebase before first stopping and resuming edits even after declaring the task complete. Over-compliance is a costly, multi-turn behavior that compounds with every turn, and our six signals give developers and tool builders a way to detect and curb it.
[HC-17] Executable Archaeology of Programs That Model Programming: Reanimating the Heuristic Compiler (1961) and KBEmacs (1984)
链接: https://arxiv.org/abs/2610.36341
作者: Jeff Shrager,Kent M. Pitman,Mia Golec
类目: Artificial Intelligence (cs.AI); Human-Computer Interaction (cs.HC); Programming Languages (cs.PL)
备注:
Abstract:Early AI programs were often also theories of how people think, and programs that wrote programs were, in part, theories of how people program. Here we report on the reanimation of two such programs that offer two seemingly different theories of programming and were developed for quite different purposes. The first, Herbert Simon’s Heuristic Compiler (1961), was an experimental model of programming as problem-solving, applying the means-ends analysis methods of the General Problem Solver to investigate the nature of the programming task. The second, KBEmacs (1984), was a working prototype of an intelligent programming assistant developed by MIT’s Programmer’s Apprentice project, built on the view that expert programmers recognize and combine familiar patterns, or “cliches.” It let a programmer build up programs by combining cliches through a stylized English command language. We approach both through the methodology of Executable Archaeology, the study of historical computational systems by running their original code, a process we call “reanimation.” Our reanimation of the Heuristic Compiler reproduces the compiled code appearing in Simon’s 1963 paper, and our reanimation of KBEmacs reproduces the SQRT sequence of its December 1984 self-test. Examined side by side, the two theories turn out to have much in common: the Heuristic Compiler also works by adapting existing routines, and Simon later argued that search alone requires a large body of programming knowledge. To highlight the reflexivity of this project, we deliberately used a modern AI-assisted programming tool, Anthropic’s Claude Code (Opus 5.5), to carry out much of the detailed reanimation work, building on a significant base of existing software. The explicit, executable theories of programming that both systems embodied are largely absent from today’s far more capable AI coding assistants.
[HC-18] Immersive Commons: Designing Ephemeral Spatial Dialogue for Co-Located Mixed Reality Environments
链接: https://arxiv.org/abs/2610.36014
作者: Wuwu Ran,Mohammed Faraj Al-Suwaidi
类目: Human-Computer Interaction (cs.HC); Multimedia (cs.MM)
备注: 14 pages, 8 figures. Accepted at SIGraDi 2026, the Conference of the Iberoamerican Society of Digital Graphics
Abstract:Head-mounted mixed reality (MR) often prioritises individual immersion, potentially drawing attention away from others sharing the room. This paper asks how MR can support co-located spatial dialogue through embodied interaction and activity visible to non-headset observers. We introduce Immersive Commons, a design framework, and Atmos Messages, a prototype combining gesture-based, ephemeral spatial messages with voice transcription, preset text, and a public floor-plan overview. Following a Research through Design approach, fifteen participants created 236 messages across two sessions, including 48 voice-authored messages. Questionnaire responses highlighted playfulness and social awareness, while observations suggested opportunities for exchange grounded in the architectural setting. Interaction clustered near the public display, although this was also where participants put on their headsets. These findings suggest the potential of ephemeral spatial messaging to support co-located exchange, with playfulness, shared visibility, and architectural context informing the design of more social and situated MR experiences.
计算机视觉
[CV-0] Honeycomb-Surfels: Compact Hexagonal Primitives for Efficient Scene Rendering and Reconstruction
链接: https://arxiv.org/abs/2610.37668
作者: Prakhar Kaushik,Soumava Paul,Alan Yuille
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Graphics (cs.GR); Machine Learning (cs.LG)
备注: Project page at this https URL
Abstract:Novel view synthesis and 3D reconstruction have advanced rapidly with radiance fields and splatting, yet their memory and computational demands limit their use in resource-constrained settings. We introduce Hexels: novel planar hexagonal surface elements with a first-order spatial color field for efficient 3D scene and 2D image representation. Hexels exploit locally planar regions with smoothly varying appearance using fewer, larger primitives. Their hexagonal support provides an area- and boundary-efficient domain, while the local color field captures appearance variation within each primitive. Across 34 scenes, Hexels improve PSNR over memory-matched 2DGS by 4.53-7.38 dB across datasets while requiring lower training VRAM. We further ablate Hexels by showing that per primitive, Hexels outperform 2DGS by upto 4.74 dB while using lower training VRAM and having 2.31 times faster FPS. Even in unconstrained settings, Hexels offers complementary quality-memory trade-offs. For efficient 3D surface reconstruction, Hexels reduce error by 28.9% compared with memory-matched 2DGS across 15 DTU scans. On 6 Tanks and Temples surface scenes, Hexels achieves the highest mean F1 among compared methods (34.38%, versus 31.72% for 2DGS). We further also show that Hexels, as a general primitive, improves 2D image representation over Image-GS. Hexels push the Pareto frontier on the quality-efficiency trade-off of surface-aligned primitives.
[CV-1] hinkWorld: World Imagination Reinforces Chain-of-Dynamics Reasoning
链接: https://arxiv.org/abs/2610.37666
作者: Chaehyun Kim,Chi-Pin Huang,Jisoo Kim,Min-Hung Chen,Yu-Chiang Frank Wang,Seungryong Kim,Fu-En Yang
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注: Project Page: this https URL
Abstract:Vision-Language-Action (VLA) models have made substantial progress in robotic manipulation, but their generalization remains constrained by the diversity and coverage of action-labeled robot demonstrations. Human videos provide much broader behavioral and environmental experience, while learning from them faces a fundamental trade-off between scalable data and explicit action grounding. We introduce ThinkWorld, a VLA learning framework that grounds action reasoning in world dynamics and strengthens it through reinforcement learning with world imagination. ThinkWorld represents each action thought as a Chain-of-Dynamics (CoD), a sequence of discrete motion tokens that captures the intermediate dynamics connecting the current world to a future state and serves as a common action representation across embodiments. We first initialize CoD with action-labeled demonstrations, then predict the future induced by each CoD and reinforce those whose predictions better match the observed outcome. This enables videos without action annotations to provide direct supervision from observed state transitions, allowing human and robot experience to refine the same reasoning space. Experiments in simulation and on real robots demonstrate effective human-to-robot transfer and strong generalization to unseen tasks and environments.
[CV-2] Can Discrete Tokens Serve as Semantically Lossless Representations of Visual Signals?
链接: https://arxiv.org/abs/2610.37662
作者: Jialv Zou,Xiaotong Li,Zhengrong Yue,Xu Huang,Yaxin Luo,Wenhao Yan,Haodong Li,Wenyu Liu,Xinggang Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project page: this https URL
Abstract:Discrete visual tokens represent vision in a discrete form analogous to language, offering a native interface for multimodal modeling under a shared next-token prediction objective. However, quantization error is widely assumed to cause unavoidable semantic loss, fundamentally limiting the semantic fidelity of discrete visual tokens and positioning continuous visual representations as an intrinsic upper bound. Using multimodal understanding as a proxy for semantic fidelity, we revisit this assumption and find quantization error unreliable, as discrete representations with substantially larger errors can achieve better understanding. Instead, the effective bit budget of discrete indices emerges as a key determinant of downstream performance. However, existing quantizers eventually saturate as additional codebooks contribute progressively less semantic information. As one approach to alleviating this saturation, we introduce Factorized Vector Quantization, which quantizes complementary, fixed-dimensional views of the visual feature with symmetric parallel codebooks. It continues to improve at larger bit budgets, ultimately matching continuous visual representations in multimodal understanding. Trained on industrial-scale multimodal data, our model achieves state-of-the-art performance among discrete-representation models, outperforming UniAR and Emu3-Chat by 22.8% and 59.2%, respectively, on a nine-metric average. To our knowledge, it is also the first discrete-representation model competitive with strong continuous-representation models such as Qwen3-VL, including on fine-grained OCR tasks. We further identify the key principles and a practical recipe enabling this competitiveness. These results and insights support the potential of discrete visual representations to underpin large-scale native multimodal foundation models. Code and models are available at this https URL
[CV-3] Repurposing Video Diffusion Transformers for Cross-View Temporal Object Correspondence
链接: https://arxiv.org/abs/2610.37633
作者: Siyoon Jin,Dahyun Chung,Honggyu An,Sangbeom Lim,Wonjin Nam,Jaewoo Jung,WonJun Moon,Seungryong Kim
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project Page : this https URL
Abstract:Cross-view object correspondence identifies the same object across views. However, existing methods often rely on frame-wise matching or sequential tracking, which can propagate correspondence errors under appearance changes or occlusions. Therefore, we introduce cross-view temporal object correspondence, which requires preserving object identity across both views and time from a query mask. To address this, we propose XTrack, a unified framework which repurposes a pretrained video DiT with semantic and spatiotemporal priors for RGB-to-mask latent transport, jointly modeling cross-view matching and temporal tracking. However, naive video DiTs do not guarantee accurate cross-view grounding and often attend to semantically similar yet incorrect objects across views. Thus, we introduce focal cross-view attention alignment, hard negative conditioning, and local context crop strategies to ensure reliable grounding. Finally, beyond these architectural enhancements, we also tackle the lack of dense annotation by introducing a pseudo-label curation pipeline that converts sparse Ego-Exo4D annotations into dense supervision to enable effective video-level training. Our results demonstrate the effectiveness of XTrack as a unified framework, ensuring robust temporal consistency and identity preservation.
[CV-4] Know then Draw: A New Learning Paradigm for Visual Generation
链接: https://arxiv.org/abs/2610.37626
作者: Yucheng Mao,Chongjian Ge,Zhuowen Tu,Hao Tan,Hanwen Jiang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Visual generation requires both high-level semantic knowledge and fine-grained pixel details. We propose a learning paradigm that assigns these roles to autoregressive pretraining for knowledge acquisition and diffusion post-training for high-fidelity rendering. Within this framework, we study the objective, representation, and scaling of knowledge acquisition and rendering. Controlled comparisons identify autoregression as the stronger pretraining objective for downstream rendering, improving both learning efficiency and final generation quality. We further find that AR-only generation quality does not necessarily correlate with final performance; semantic quality matters more. This motivates a purely semantic tokenizer trained with language supervision alone, without an image decoder or reconstruction loss. We establish compute-optimal scaling laws for knowledge acquisition and rendering, and find that autoregressive pretraining lowers the fitted asymptotic rendering loss compared with pixel diffusion models trained from scratch. Guided by the laws, we train a 3B-parameter text-to-image model with only 6,600 H100 GPU-hours, achieving 86 on GenEval and 87 on DPG-Bench and validating the proposed paradigm at scale.
[CV-5] Perceiving and Predicting Dynamic Sports: Spatiotemporal Intelligence in Vision Models for Egocentric Tennis Video
链接: https://arxiv.org/abs/2610.37619
作者: Tudor Braicu,Joseph DelPreto,Jared Boyer,Mónika Farsang,Maximilian Stölzle,Tim Niklas Seyde,Daniela Rus
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注: 8 pages, 5 figures. Project page: this https URL
Abstract:Agile robots require spatiotemporal intelligence to interpret scenes, reason about interactions, and anticipate future events. Yet it remains unclear how well foundation vision models support these capabilities, particularly from a moving egocentric camera, which is a natural view for robots. To investigate this, we collect an annotated tennis dataset and train a suite of vision models on spatiotemporal tasks. Our dataset contains egocentric video synchronized with five allocentric videos, estimated global positions, labeled contact times, and annotated stroke types. We then evaluate seven prediction tasks from egocentric images: 2D ego player and 3D ball localization, future player and ball trajectories, forehand or backhand stroke anticipation, time to the next ego player ball contact, and 3D ball position at that contact. We compare image encoders, video encoders, and vision-language models with jointly trained prediction heads. The best performing models achieve mean current-frame localization errors of 0.14 m for the ego player and 0.54 m for the ball, future forecasting errors of 0.34 m for the ego player and 0.91 m for the ball over 1.1 s, and a mean contact-time prediction error of 0.14 s. Auxiliary depth information substantially improves 3D ball localization, suggesting that inferring metric depth remains a challenge. Longer image histories do not improve ball forecasting, indicating that representations exploit scene context rather than rely solely on inter-frame motion. Finally, we propose a conditional diffusion model to generate whole-body kinematic tennis motion from vision encoder predictions on egocentric video at real-time rates, adapting movements as predictions are refined. These results move towards a deeper understanding of how foundation vision models can support real-time dynamic robots using egocentric video.
[CV-6] VideoLVR: Latent Visual Reasoning over Grounded Video Evidence
链接: https://arxiv.org/abs/2610.37613
作者: Shilong Li,Xingyuan Bu,Hao Shi,Yun Xing,Weixin Feng,Qiangpeng Yang,Shilei Wen
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 13 pages
Abstract:Chain-of-thought reasoning has substantially advanced video understanding in multimodal large language models (MLLMs). However, language-based reasoning traces do not provide a reliable mechanism for models to revisit relevant visual evidence, leading to visual forgetting and impairing their ability to reason over video content. Existing approaches maintain visual grounding during reasoning in two main ways: they either refer to evidence using textual coordinates, which specify symbolic locations rather than the visual content itself, or reinsert cropped regions or frames into the multimodal context, preserving visual details at the cost of substantial context expansion. To address these limitations, we introduce VideoLVR, which constructs latent visual evidence from local regions at relevant video moments, thereby providing visual support for intermediate reasoning steps. To supervise latent visual evidence generation, we introduce a visual token prediction objective that formulates visual token reconstruction as prediction over a vocabulary derived from the input video, thereby unifying visual evidence reconstruction and text generation within the same token prediction framework. This shared prediction framework enables MLLMs to reason over interleaved text tokens and latent visual states while integrating evidence from different moments. We further extend VideoLVR to reinforcement learning by treating sampled visual tokens as actions alongside text tokens and jointly optimizing the policy using final-answer rewards. By efficiently revisiting task-relevant visual information during reasoning, VideoLVR improves video understanding without substantially expanding the multimodal context. Experiments on multiple video benchmarks demonstrate its effectiveness, especially on tasks requiring fine-grained visual evidence.
[CV-7] MoI-Bench: A Benchmark for Image Generation and Editing on Mobile Devices
链接: https://arxiv.org/abs/2610.37599
作者: Kailai Feng,Haoxiang Rao,Yuxiang Wei,Bo Chen,Junsong Li,Yang Pan,Chengkun Cai,Jagmohan Chauhan
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Project page: this https URL
Abstract:Developing on-device image models requires considering practical user needs, task types and device resource constraints. However, existing benchmarks provide limited guidance on identifying capability gaps between on-device models and server-side models. We introduce MoI-Bench, a benchmark grounded in mobile user needs, comprising 64 subtasks and 1150 samples. It defines 16 operations, such as removal and replacement, and eight visual target categories covering content such as portraits and text. These form two orthogonal axes whose applicable combinations define specific tasks, such as removing facial makeup. To better guide model improvement, we further propose MoI-Eval, which introduces task-specific criteria covering instruction following, visual quality and detail preservation. We further construct sample-level evaluation questions and aggregate their scores into overall quality and capability profiles along the operation, target and criteria axes. We evaluate representative on-device and server-side models and measure mobile deployment costs. Results show that this decoupled Operation x Target design provides finer-grained guidance for improving compact models. For example, it highlights the need to prioritize specific capability gaps (e.g., text-replacement) over broad application-level goals (e.g., improving text editing). Unlike conventional benchmarks that rank models using a single score, we introduce a two-dimensional Quality-Deployability capability map to assess models for practical mobile use. The map shows that current on-device models can achieve efficient generation or editing but expanding their task coverage deserves greater attention than further improving deployment efficiency. Further progress requires improving editing quality, particularly identity preservation and text rendering, while broadening task coverage.
[CV-8] Zero-Shot Multimodal Event Spotting in Soccer Broadcasts
链接: https://arxiv.org/abs/2610.37583
作者: Sadia Kamal,Mahmudul Hasan,Tony Braskich
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG); Multimedia (cs.MM)
备注: Accepted to the Main Conference at the IEEE International Symposium on Multimedia: ISM 2026
Abstract:Automatic soccer event spotting supports highlight generation, video indexing, and sports analytics, but existing methods depend on large annotated datasets and soccer-specific training. We present a zero-shot framework that detects, classifies, and temporally localizes events in full-length soccer broadcasts without task-specific parameter training. Each match is segmented into overlapping clips represented by a compact visual storyboard and the aligned commentary transcript, which a general-purpose multimodal model processes jointly. On three manually annotated matches containing 130 events, the best configuration achieves 72.0% F1@5s and outperforms SoccerNet-trained CALF and NetVLAD++ in a cross-dataset setting, while the storyboard representation reduces inference cost by 72% compared with individual frames. We further analyze the effects of model family, input representation, and prompt optimization, and introduce an evidence-grounded excitement-scoring stage that ranks detected events with natural-language explanations for highlight selection.
[CV-9] As the Story Unfolds: Watching a Film and Identifying Characters as a Human Does WWW
链接: https://arxiv.org/abs/2610.37579
作者: Zhongrui Gui,Junyu Xie,Tengda Han,Weidi Xie,Andrew Zisserman
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project Page: this https URL
Abstract:Films introduce characters incrementally as they progress. We study the problem of identifying principal characters and assigning their names directly from the movie itself, without using external cast lists, actor photographs, or pre-built character banks. We introduce the character naming task: given a name mentioned in dialogue and the associated video context, the model must determine which visible character bears that name. This task is challenging because name resolution often depends on cinematic cues such as shot-reverse-shot structure, self-introduction, direct address, third-party reference, gaze, and temporal continuity. We provide a dataset and annotations for evaluating character naming and first character appearances, develop a video-and-dialogue model for the task, and compare it with proprietary multimodal large language models. We further introduce a progressive framework for building a character bank as the film unfolds. Together, these contributions support online, human-like movie understanding in which character identities are inferred from the film itself.
[CV-10] he Noise Plane: A Diffusion Bridge Across Exponent and Dispersion
链接: https://arxiv.org/abs/2610.37576
作者: Xuran Hu,Yuru Jia,Tengxi Wang,Miloš Daković,Ljubiša Stanković,Wufan Zhao
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 23 pages, 15 figures
Abstract:Diffusion-based denoising commonly uses Gaussian perturbations, whereas sensor noise follows diverse signal-dependent statistical laws. In this paper, we propose p -Bridge, a unified measurement-indexed diffusion framework for Tweedie noise. The noise exponent p describes the variance–intensity relation, while accumulated measurement \nu controls the noise level. Together, they define a noise plane that includes Gaussian, Poisson, and Gamma models. The stochastic bridge connects measurement distributions through independent increments, giving denoising a physical interpretation as progressive measurement accumulation. Both the starting and stopping indices can be specified, allowing restoration from an observed measurement level to a prescribed higher level. Measurement superposition yields a shared reverse-mean update, paired with a deviance-based objective for learning the required conditional mean. A latent compound Poisson–Gamma construction enables sampling at intermediate exponents. A single exponent-conditioned network adapts to different Tweedie noise models without separate training for each exponent. Experiments show that exponent conditioning improves cross-exponent robustness over fixed-exponent models. Using estimated exponents, the shared network achieves robustness comparable to an oracle-selected set of five specialist models. Experiments on synthetic images, FMD fluorescence microscopy, and multi-sensor SAR images further demonstrate its effectiveness across different signal-dependent noise regimes.
[CV-11] ReferStream: Interactive Subject Injection for Streaming Video Generation
链接: https://arxiv.org/abs/2610.37566
作者: Shuolin Xu,Guanyuan Li,Boyu Li,Kelu Yao,Xiaogang Xu,Hongchuan Yu,Shuicheng Yan,Xiaobin Hu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 18 pages. Project page: this https URL
Abstract:Injecting new subjects naturally during streaming video generation requires integrating newly supplied visual references into the existing video while preserving established subjects and the scene. Encoding references independently of the current scene provides little guidance for this integration, while keeping their features in persistent memory can interfere with established context. We propose ReferStream, a framework that requires no additional training and combines reference construction using the current scene with transient cache injection. Joint Latent places reference latents in target regions of the latest scene latent, allowing reference features to be extracted in context. Cache Restoration uses these features to condition the injection chunk, then restores the overwritten global cache entries while retaining the newly generated history. This procedure incorporates new visual conditions without retraining the generator or revising emitted frames. We further introduce ReferStream Bench, a new benchmark with scheduled reference arrivals, evolving instructions, and explicit annotations of subject presence for InShot and MultiShot generation. Evaluation follows the event schedule and measures both the fidelity of new subjects and the preservation of established subjects, alongside video quality, text adherence, and count accuracy. Experiments on ReferStream Bench show higher scores for text adherence, subject similarity, and count accuracy than the compared baselines, including offline methods given all references in advance. Project page: this https URL.
[CV-12] GroundControl: Satellite-Conditioned Multi-View Diffusion for Reconstruction-Free Novel View Synthesis WACV2027
链接: https://arxiv.org/abs/2610.37545
作者: Akshay Krishnan,Matias Turkulainen,Jamie Wynn,Zawar Qureshi,Mohamed Sayed,Guillermo Garcia-Hernando,Sara Vicente,James Hays,Michael Firman,Daniyar Turmukhambetov
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Accepted to WACV 2027. Project page: this https URL
Abstract:Multi-view or video diffusion models have recently emerged as state-of-the-art approaches for novel-view synthesis from sparse image collections. While they can produce high quality images, they tend to hallucinate unrealistic spatial layouts when synthesizing views outside the input views (scene extrapolation). We propose to address this challenge for outdoor scenes by using satellite images as an additional information source to condition the multi-view generative model. We introduce a new satellite-conditioned multi-view diffusion model that combines conditioning from a satellite image and street-level input images when generating novel-views. Our cross-attention based conditioning allows us to retain pretrained multi-view diffusion priors to generate seamless, realistic, novel views from a handful of source views, while also adhering to the scene layout as seen from satellite images. Our model does not use 3D scene reconstruction, and only relies on posed street and satellite views. Our experiments on outdoor datasets show that our model is better than both satellite-conditioned or street-view conditioned baselines on novel-view synthesis, especially on challenging cases where target views have very little overlap with source views. Project page: this https URL
[CV-13] A note on data-efficient finetuning of vision foundation models ICML2025
链接: https://arxiv.org/abs/2610.37510
作者: Niclas Hergenröther,Antonio Orvieto
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 8 pages, 3 figures, Accepted at ES-FoMo III: 3rd Workshop on Efficient Systems for Foundation Models at ICML 2025
Abstract:Adapting foundation models to a new task within or beyond the training domain is the de facto standard for building specialized machine learning models. Recently, Visual Prompting (VP) has emerged as a promising technique for efficient knowledge transfer. As new foundation model families~(like Mamba) get introduced and VP pipelines such as AutoVP reach greater maturity, we find a growing need for a systematic evaluation of current approaches. In this work, we assess the performance of the latest models, comparing them to earlier architectures and alternative fine-tuning methods, to better understand the progress, challenges and opportunities in the field of efficient fine-tuning under resource limitations. Towards this goal, this paper provides a concise empirical overview of the interactions among foundation model families (Attention-, Convolution-, and Mamba-based) and transfer paradigms: VP, Linear Probing (LP), Full Fine-tuning (FFT) and Low Rank Adaptation (LoRA). Our work builds up on previous findings by broadening the selection of evaluated models and techniques while tuning hyperparameters. In the interest of delivering practical guidelines for the user, we also explore application of prevalent regularization techniques to boost performance in the context of VP.
[CV-14] ALADDIN: Adversarial Latent Diffusion Distillation for One-Step Identity Swapping WACV2027
链接: https://arxiv.org/abs/2610.37503
作者: Ivan Shchekotov,Timofei Gritsaev,Viacheslav Pirogov,Dmitry Vetrov
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to WACV 2027. 15 pages, 5 figures, 4 tables, including references and appendices
Abstract:Face swapping transfers a source identity onto a target image while preserving the target’s pose, expression, hair, and lighting. The most realistic current methods are diffusion-based, but they need many network evaluations per image. GAN-based methods are fast, yet trained from scratch they must learn realism from the adversarial signal alone and produce visibly weaker results. We present ALADDIN (Adversarial Latent Diffusion Distillation for One-Step Identity Swapping), a one-step face swapper obtained by initialising the generator from a pre-trained multi-step flow-matching model and training it adversarially under identity and target-preservation constraints. We expose a controllable trade-off between identity transfer and target-attribute preservation by training a student model with triplet pseudo-label supervision. We additionally identify a bias in the standard evaluation protocol, which favours methods trained against the same identity encoder used to score them, and re-evaluate all baselines with an independent one. ALADDIN is state-of-the-art among one-step methods in realism and target preservation at comparable identity transfer, and matches the best multi-step model with 28 times fewer network evaluations and an order of magnitude fewer training pairs.
[CV-15] How Reliable Is Individualized Functional-Network Attribution in Resting-State fMRI Cognitive Prediction in ADNI?
链接: https://arxiv.org/abs/2610.37472
作者: Sheethal Bhat,Mahfuzur Rahman Chowdhury,Imam Ahmed Shaik,Rose Dawn Bharath,Siming Bayer,Andreas Maier
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Individualized attribution estimates which functional networks drive a model prediction for a given subject and is often proposed as an interpretable resting-state fMRI (rs-fMRI) biomarker, but its test-retest reliability is rarely quantified. We evaluate per-subject network-ablation attribution for a frozen resting-state functional connectivity (FC) predictor (Schaefer-200 FC predicting MMSE/ADAS-Cog13) trained on N=875 subjects from the Alzheimer’s Disease Neuroimaging Initiative (ADNI), using, and (iii) a subject-level permutation null. Split-half reliability is near-ceiling (Spearman p=0.948/0.950) but drops on scan-rescan (p=0.609/0.616). Under the mismatched-subject null, correlations remain substantial (p~0.46-0.47), indicating strong population-shared structure, yet matched pairs exceed this null (z~6.6-6.9, p=10-4). Thus, in ADNI, individualized network attribution contains a modest but non-zero person-specific component atop a strong population-shared pattern. We recommend reporting scan-rescan reliability together with a permutation-null control, rather than within-session proxies alone, to avoid overstating individual specificity.
[CV-16] PerceptRL: Making Visual Perception Rewardable in MLLM s from Answers to Interactions
链接: https://arxiv.org/abs/2610.37471
作者: Tongkun Guan,Zhibo Yang,Mingkun Yang,Haocheng Wang,Wei Shen,Xiaokang Yang
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Fine-grained visual perception remains a bottleneck during post-training of multimodal large language models (MLLMs), as perceptual errors can constrain gains from reasoning-focused optimization. Although image-caption pre-training improves perception, its influence can weaken as subsequent stages change data mixtures and objectives. Continuing caption supervision during post-training is challenging because subjective, information-dense descriptions complicate verifiable reward design. We present PerceptRL, which strengthens perception at two complementary levels. Perception RL uses a multi-agent pipeline to extract verified errors from dense image captions and convert the corresponding visual facts into QA pairs with fixed answer labels and rule-based rewards. Chain QA links correlated errors, while multiple-choice QA combines unrelated errors. A specialized route targets high-error domains with tailored verification, and a domain-agnostic general route covers errors beyond these domains. Perception Agentic RL addresses fine-grained and dense cases unresolved by single-turn Perception RL through multi-turn inspection with visual tools. We observe coordinate drift even in models trained to use tools, where incorrect crops or annotations distort evidence and propagate errors. Coordinate Injection corrects spatial operations in cold-start trajectories, while a grounding-aware process reward supervises intermediate spatial fidelity alongside final-answer correctness. Training proceeds through shared cold-start SFT, single-turn Perception RL, and multi-turn Perception Agentic RL. Across seven benchmarks under a shared tool-use protocol, PerceptRL achieves 71.6% average accuracy, surpassing the base model with the same tools and the strongest evaluated frontier VLM by 15.7 and 5.2 percentage points, respectively.
[CV-17] Segmentation Confidence for Arbitrary CNNs
链接: https://arxiv.org/abs/2610.37431
作者: Baris Oguz,Xing Yao,Claudia Tawil,Alison Pouch,Brett Byram,Gabriel Arenas,Nadav Schwartz,Ipek Oguz
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 5 pages, 6 figures, published in ISBI 2026
Abstract:Convolutional neural networks (CNN) are widely used for medical image segmentation. However, they lack an inherent measure of confidence. Existing methods for estimating prediction uncertainty often require a custom network architecture and/or dedicated training setups and need to be trained from scratch. In addition, the deduction of a single confidence score per input from the uncertainty maps is not always straightforward. In this study, we explore voxel-based uncertainty estimation methods that can be added to existing segmentation pipelines without training from scratch. We also propose a novel method to estimate a single confidence score from the uncertainty maps. Our evaluations on different ROIs and modalities suggest that our confidence score can predict segmentation performance in unseen data.
[CV-18] When Visual Evidence Is Insufficient: How Instructions Post-Training and Internal Activations Shape Refusal in Multimodal Language Models
链接: https://arxiv.org/abs/2610.37430
作者: Ananth K Suresh,Aveen Dayal,Gouthaman KV,Aditya Pola,Vineeth N Balasubramanian
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注: 31 pages, 6 figures
Abstract:For reliable deployment, multimodal large language models (MLLMs) should recognize when an image provides insufficient evidence to answer a question and withhold unsupported answers. Yet refusal alone is not necessarily evidence of such recognition: models may refuse because of unrelated safety constraints or without identifying the missing evidence. We study how refusal on visually unanswerable questions is shaped by three sources of control: inference-time instructions, post-training, and internal activation interventions. Across three MLLMs on the MoHoBench benchmark, we systematically manipulate the meaning and placement of refusal instructions, compare supervised and preference-based post-training, and intervene on a refusal-related activation direction derived from naturally answered and refused examples. We find that refusal is highly sensitive to both the meaning and placement of instructions, with model-specific changes at the sample level that are obscured by aggregate refusal rates. Evidence-aware and general refusal instructions have stronger effects than an intentionally mismatched safety instruction. Post-training produces larger increases in refusal, but can also induce responses that combine refusal language with attempted answers. In contrast, activation intervention produces smaller yet transferable changes across prompt conditions, without evidence of indiscriminate refusal on tested visually answerable samples. These findings show that refusal is jointly shaped by external instructions, post-training, and internal representations, highlighting the need to disentangle evidence recognition from a model’s learned propensity to refuse.
[CV-19] CTBakeGS: training-free baking of photorealistic CT volume renderings into 3D Gaussian splats
链接: https://arxiv.org/abs/2610.37406
作者: Haruki Fukuda,Constantin Lux,Wolf Schweitzer,Tadashi Hosoya,Lars Ebert
类目: Computer Vision and Pattern Recognition (cs.CV); Graphics (cs.GR)
备注: Code and project page: this https URL
Abstract:Photorealistic volume rendering of computed tomography (CT) may convey anatomy intuitively, but it needs the CT data, dedicated software and a powerful graphics processor. Three-dimensional Gaussian splatting (3DGS) displays photorealistic scenes in real time on less powerful devices, but existing medical approaches train each scene for minutes to hours. We aimed to convert a CT volume rendering directly into a 3DGS model without training. Splats were placed on an adaptive octree of millimetre-sized cells. Their colour and opacity were computed on the graphics processor with the renderer’s lighting, cell opacity was converted analytically into splat opacity, and faint splats were retained stochastically above the 1/255 opacity floor of standard 3DGS rasterisers. Six public CT scans (1.5-mm voxels) were converted with three transfer functions and compared with converged references by peak signal-to-noise ratio (PSNR), structural similarity and a learned perceptual metric. Two-million-splat models reached a PSNR of 32.6 +/- 0.9 dB for bone, 30.1 +/- 1.0 dB for contrast-enhanced vessels and 32.3 +/- 1.5 dB for skin and bone. A conversion pass (usually two or three per model) took 4-30 s on a current laptop and 9-60 s on a desktop with a 4-GB graphics card. On the laptop, the models rendered at 68 frames per second at 800 x 800 pixels. Training-free conversion produced real-time 3DGS models of photorealistic CT renderings in about a minute or less per conversion pass on ordinary computers, with fidelity that we consider sufficient for interactive viewing.
[CV-20] PDRVLA: A Vision-Language-Action Model with Physical Dynamics Representations
链接: https://arxiv.org/abs/2610.37381
作者: Hanqing Wang,Zhaotian Cai,Ruichuan An,Bohan Zeng,Tao Zhang,Yilin Chen,Xu Jiang,Kaiquan Chen,Chongyang Wu,Haoran Li,Fan Yang,Mingan Lin,Wentao Zhang,Zenan Zhou,Youzheng Wu
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:Vision–Language–Action (VLA) models generate robot actions from visual observations and language instructions, but their understanding of physical dynamics remains limited. Video world models can capture scene dynamics, yet generating future videos introduces high latency into the control loop. We introduce PDRVLA, whose lightweight Physical Dynamics Module extracts physical tokens from vision–language features to condition action generation without online video generation. The module learns through direct prediction of future RGB, depth, and optical-flow features, with auxiliary supervision through a frozen pretrained video world model. Training relies entirely on publicly available data, expanded through human-to-robot synthesis, subtask annotation, and embodied chain-of-thought (ECoT) annotation. Across four simulation benchmarks, PDRVLA achieves success rates of 99.3% on LIBERO, 89.9% on LIBERO-Plus, 93.6% on RoboTwin~2.0, and 81.5% on SimplerEnv. Ablations show improved future prediction with world-model supervision and evaluate how subtask conditioning, future object-pose supervision, and ECoT training affect action success.
[CV-21] meLineXR: An Interactive Exploration of Time in Vision-Language Models
链接: https://arxiv.org/abs/2610.37349
作者: Nidham Tekaya,Markus Passecker,Matthias Zeppelzauer
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Recently, it has been shown that vision-language models (VLMs) have a certain intrinsic understanding of time. They can estimate the temporal origin of objects and encode time along a low-dimensional, non-linear manifold inside their embedding space~\citetekaya2025matter. That result is quantitative and reproducible, but the structure itself stays out of reach for end users: they cannot see it, query it with their own images, or inspect where a query lands relative to other years on the time axis. We present TimeLineXR, an interactive demonstration that turns this time axis into an explorable 3D scene. A web demo lets users upload or select images and see each placed on a color-coded 3D time axis; a this http URL WebXR scene renders the same time axis as a room-scale representation that can be grabbed, scaled, inspected, and queried by throwing image cards into it. Both clients run the same prediction pipeline against the same cached temporal index: encode the query with CLIP, project it through cosine-kernel KPCA, and snap it to the Bézier time axis. The system turns each image into a point on the time axis, so users can compare where different objects land and read off their estimated first year of this http URL source and reproduction steps can be found in: this https URL.
[CV-22] CodeCap-Bench: Can Code Serve as a Good Video Caption?
链接: https://arxiv.org/abs/2610.37341
作者: Jiafu Tang,Yifan Xu,Eric Tang,Junjie Zhao,Shipeng Luo,Te Sun,Haoxuan Hu,Yiyan Ji,Xinlong Chen,Zihao Wang,Yang Shi,Yuanxing Zhang,Pengfei Wan,Jiaheng Liu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Video captioning aims to preserve visual information through compact descriptions, predominantly using natural language. However, precisely expressing fine-grained spatial structures and temporal dynamics in natural language often leads to lengthy and potentially ambiguous descriptions. In this work, we explore Code as Video Caption, a paradigm where executable programs serve as a complementary representation of visual content through graphical primitives, layout rules, and functions of time. To examine whether code captions effectively preserve video information, we compare them with natural language captions on downstream question answering and video reconstruction tasks. Our results show that code captions achieve superior reasoning accuracy and reconstruction fidelity. Despite these advantages, realizing the potential of code captioning depends on whether current multimodal large language models (MLLMs) can generate faithful code captions. To systematically assess this capability, we introduce CodeCap-Bench, which comprises 323 videos spanning five categories of structured 2D content, along with fine-grained annotations and a multi-level evaluation framework. Extensive experiments reveal that while leading proprietary MLLMs demonstrate promising code captioning capabilities, they still struggle to faithfully capture complex temporal dynamics, with open-source models lagging substantially behind.
[CV-23] ScIntBench: Physical Interaction Plausibility in 3D Scene Generation
链接: https://arxiv.org/abs/2610.37326
作者: Wendi Ying,Daochang Liu,Leonid Sigal,Ajmal Saeed Mian
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Existing evaluations of text-conditioned 3D indoor scene generation largely emphasize semantic fidelity and physical validity at rest, leaving interaction-time physical reliability insufficiently assessed. To address this gap, we introduce ScIntBench, a unified benchmark centered on interaction readiness: it executes prescribed manipulation actions (e.g., pushing, opening, and surface probing) in physics simulation and measures task success, capability and progress, post-interaction integrity, and dynamic plausibility. Multi-view semantic and static physical evaluations complement these metrics and help explain interaction outcomes. We systematically evaluate 12 state-of-the-art scene generation systems across 10 room types. Our results reveal that strong semantic and static physical performance does not necessarily translate into interaction readiness: a system with near-perfect static plausibility completes only 11.54% of interactions, 7 of the 12 systems produce scenes that cannot support any interaction, and even the best system succeeds in only 36.92% of prescribed tasks. These findings highlight two fundamental challenges in generating interaction-ready scenes: providing the necessary physical interaction capabilities and ensuring reliable physical behavior under active manipulation.
[CV-24] Hierarchical Sparse Graph for Multi-dimensional Assignment in Event-based Tracking
链接: https://arxiv.org/abs/2610.37321
作者: Muxi Zha,Yang Shang,Wenlong Zhang,Qifeng Yu,Banglei Guan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 16 pages, published to IEEE T-CSVT
Abstract:Existing event-based tracking methods typically operate in an online manner, associating tracklets across adjacent batches. However, this limited temporal window often leads to failures in challenging scenarios, such as long-term occlusions or missed detections, where broader temporal context should be incorporated for high-precision tracking. To achieve robust and accurate tracking, this paper concentrates on the event-based multi-dimensional assignment that associates tracklets across multiple batches simultaneously. Unlike image-based detectors, event-based detectors can generate multiple correct yet spatially distinct tracklets for a single object per batch. This violates the one-to-one rule in classic data association, making event-based association an underdetermined many-to-many matching problem. To the best of our knowledge, this is the first work to tackle this NP-hard problem. To address its underdetermined nature, we formulate the assignment problem via fuzzy logic, where pairwise tracklet correlations are quantified through multiple long-range kinematic metrics. This fuzzy design eliminates the need for hard tracking thresholds and enables multiple heterogeneous metrics to be integrated into a unified correlation. By treating tracklets as nodes and their fuzzy correlations as edges, we construct a Hierarchical Sparse Graph (HSG). Building upon the HSG, we pose assignment as a graph partition task, and employ a heuristic solution that measures vertex affinity for partitioning adaptively. The entire pipeline is training-free. Comprehensive experiments on two challenging event datasets, eTram and Ev-UAV, show that our method outperforms existing event-based and most image-based association models. Particularly, it achieves over 2 \times the performance of OC-SORT across all metrics in the many-to-many association scenario.
[CV-25] A Lightweight Attention-Based CNN with Saliency-Map Interpretability for Skin Cancer Detection
链接: https://arxiv.org/abs/2610.37314
作者: Sanjana Islam Kasfia,Md. Ismiel Hossen Abir,AKM Bahalul Haque,Abir Mahmud Shahariar
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Skin cancer is among the most prevalent and life threatening forms of cancer worldwide. Early and accurate diagnosis significantly improves patient survival rates. However, manual dermoscopic assessment remains challenging due to visual similarity. This study proposes an Explainable AI-driven framework for automated binary classification of skin lesions using a lightweight attention-based convolutional neural network (CNN-Attention). The model integrates channel attention to focus on informative regions of dermoscopic images while maintaining low computational complexity. A publicly available dataset comprising 2,637 training and 660 test images was used, with balanced class weights. The proposed model achieved 87 percent accuracy with 1.95 million parameters and 7.45 MB of storage, matching the best baseline (MobileNet) while using approximately half as many parameters and half as much storage. ROC AUC and PR AUC scores reached 0.9475 and 0.9430, respectively. Additionally, saliency maps were incorporated as an explainability technique to visualize regions that influence model decisions. Comparative analysis with VGG16,VGG19,ResNet101,and MobileNet demonstrates the proposed models effectiveness in balancing classification results, computational efficiency, and interpretability.
[CV-26] MvT: Event-based Multi-view Projection for Multiple Object Tracking
链接: https://arxiv.org/abs/2610.37309
作者: Muxi Zha,Banglei Guan,Minzu Liang,Zibin Liu,Yang Shang,Qifeng Yu,Laurent Kneip
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 16 pages, 20 figures, published to IEEE T-IP
Abstract:Event cameras are increasingly used for Multiple Object Tracking (MOT), but their asynchronous event output often requires specialized methods. Existing processing methods primarily follow two paradigms, pseudo-frames and event-by-event. The former is the prevailing approach since its data format aligns with images, making image-based techniques applicable. However, it suffers from tracking failures when trajectories overlap or are spatially close on pseudo-frames. Facing this challenge, we propose a multi-view pipeline, Multi-view Tracking (MvT), which preserves the 2D data format to leverage image-based techniques directly while introducing additional spatio-temporal views to resolve tracking ambiguities in a single view. MvT comprises a Multi-view Projection (MvP) module and a Multi-view Fusion (MvF) stage. MvP encodes events into three complementary spatio-temporal views while mitigating the pattern discretization. Within MvF, multi-view results are unified into a 3D coordinate system, and tracklets are associated through an optimization model subject to specific criteria combination. Evaluations on four datasets, including our self-collected Small Objects Dataset (SOD), show that MvT seamlessly integrates image-based methods and outperforms existing non-learning and learning trackers in generalized scenarios, and effectively resolves the single-view tracking ambiguities. Being training-free, MvT is applicable when ground-truth annotation is infeasible, thereby highlighting its practical, data-efficient potential. Code is available at this https URL.
[CV-27] RUCO: A Scalable Lock-Free Algorithm for Parallel Contour Extraction
链接: https://arxiv.org/abs/2610.37297
作者: Rafael Munoz-Salinas,Francisco J.Romero-Ramirez,Manuel J. Marin-Jimenez
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Transforming binary masks into geometric contours remains a critical step in many computer vision pipelines, particularly for post-processing high-resolution segmentation outputs. Yet, standard algorithms such as the Suzuki-Abe method are limited by their sequential nature and the overhead of building topological hierarchies. As modern applications transition to high-resolution imaging, there is a critical need for contour extraction that leverages multi-core CPU architectures without the complexity of tile stitching or heavy synchronization. This paper presents TRUCO (Threaded Raster Unrestricted Contour Ownership), a lock-free parallel algorithm designed for high-performance geometric boundary extraction. TRUCO introduces a start-point ownership rule, combined with a speculative tracing model, in which threads trace contours downward into adjacent regions and use an upward-termination logic to ensure each contour is processed exactly once. Experimental results on synthetic and real-world datasets, conducted on both consumer and server-grade CPUs, demonstrate that TRUCO significantly outperforms existing baselines. The method achieves speedups of over an order of magnitude compared to the industry standard while maintaining robust scalability on high-resolution scenes across diverse hardware architectures. Subjects: Computer Vision and Pattern Recognition (cs.CV) Cite as: arXiv:2610.37297 [cs.CV] (or arXiv:2610.37297v1 [cs.CV] for this version) https://doi.org/10.48550/arXiv.2610.37297 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[CV-28] aming Noise Initialization for Zero-shot Video Restoration with Image Diffusion Priors
链接: https://arxiv.org/abs/2610.37290
作者: Chong Wang,Zixuan Fu,Hao Cheng,Siyuan Yang,Alex C. Kot,Bihan Wen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: IEEE Transactions on Multimedia
Abstract:Recent advances in diffusion models have achieved remarkable success in restoring high-quality images in a zero-shot manner. Unfortunately, extending this success to video restoration is challenging due to the substantial computational resources required to pre-train video diffusion models. A rising trend seeks to address video restoration by utilizing a pre-trained image diffusion model, initiating restoration from either frame-wise independent or identical Gaussian noise. Yet, these approaches often fall short, as such simple noise initiations fail to capture and utilize the inherent temporal correlations within degraded video frames. In this work, we propose taming noise initialization for zero-shot video restoration, which preserves the intrinsic temporal correlations by generating initial noises directly using DDIM inversion on the corrupted video frames themselves. Building on this, we introduce an adaptive noise blending strategy that suppresses degradation patterns encoded during DDIM inversion while gradually steering the noise toward a Gaussian distribution. To further emphasize temporal consistency, we propose measurement-guided trajectory attention to better align features across frames. Extensive experiments on multiple video restoration tasks demonstrate that our method outperforms existing zero-shot approaches, consistently achieving superior results on the DAVIS, REDS4, and Set8 datasets.
[CV-29] GameVerse: A Minute-Scale Gameplay Dataset for Long-Horizon Interactive World Modeling NEURIPS2026
链接: https://arxiv.org/abs/2610.37257
作者: Kang He,Wenshuo Peng,Chuanhao Li,Zihui Gao,Xiaojie Xu,Zhengyuan Lin,Yuzhe Ding,Donghong Ji,Kaipeng Zhang,Yongtao Ge
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted by NeurIPS 2026
Abstract:Recent advances in interactive world models have made it possible to generate controllable, dynamic environments from action inputs. Yet current models struggle to maintain coherence beyond roughly one minute of rollout, a limitation rooted not in model capacity but in the scarcity of long-horizon training data. Existing gameplay video datasets typically consist of clips under ten seconds, with captions that describe whole clips in coarse and undifferentiated text, which is insufficient for controllable long-horizon generation. We address this with GameVerse, a minute-scale gameplay video dataset purpose-built for long-horizon interactive world modeling. GameVerse contains 69.7K clips spanning 1,655 hours and 349.6 million frames across diverse first- and third-person gameplay, including 1,358 hours of continuous two-minute sequences, an order of magnitude longer than prior datasets. Clips pass through a unified pipeline of scene-consistent segmentation, quality filtering, camera pose estimation, and event-based caption annotation. Each clip is paired with a temporally stable camera trajectory that remains coherent across minutes, together with captions that separate static scene context from dynamic foreground events and align to event boundaries rather than fixed time windows. Comprehensive analysis shows that GameVerse surpasses existing public datasets in scale, diversity, and annotation quality. Experiments further demonstrate that training on GameVerse substantially extends the temporal horizon of interactive world models.
[CV-30] HDVS: semi-supervised semantic segmentation via heterogeneous dual-branch voting supervision
链接: https://arxiv.org/abs/2610.37236
作者: Yongqi Shan,Yunzhi Zhuge,Huchuan Lu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 16 pages, 9 figures, 7 tables. Published in Visual Intelligence
Abstract:Semi-supervised semantic segmentation, which aims to utilize large volumes of unlabeled images to achieve accurate segmentation results with fewer human annotations, has attracted increasing attention. Prior methods were primarily based on the assumption that the pseudo-labels generated by each branch were sufficiently reliable to supervise the remaining branches. However, there is a concern that the errors in pseudo-labels could accumulate during co-training, potentially resulting in suboptimal performance. To this end, we present a novel framework called heterogeneous dual-branch voting supervision (HDVS), which is designed to enhance the reliability of pseudo-labels and mitigate the issues arising from pseudo-labeling. Specifically, based on the cross-supervision framework, we introduce a voting mechanism that correlates pseudo-labels from heterogeneous branches to produce pseudo-labels with enhanced reliability. Concurrently, we employ a feature communication module to introduce perturbations at the feature level in each branch to maximize prediction diversity across the dual branches while maintaining convergence. Comprehensive evaluations of the proposed HDVS on two benchmark datasets (PASCAL VOC 2012 and Cityscapes) demonstrate its superiority to the state-of-the-art approaches.
[CV-31] Subspace Clifford Networks
链接: https://arxiv.org/abs/2610.37226
作者: Zhongping Ji
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 12 pages
Abstract:In this work, we propose Subspace Clifford Networks, a reformulation that replaces global cyclic rolling with grouped subspace interactions. Instead of sampling sparse cyclic diagonals across the entire D -dimensional channel space, we project the features into an aligned space \R^\tildeD = \bigoplus_g=1^G V_g partitioned into G low-dimensional groups V_g \cong \R^k with k \in \2, 3, 4\ . Within each k -dimensional group, we compute all \binomk2 wedge coordinates and concatenate them with \operatornameSiLU -activated coordinate-wise products, while sandwiching the grouped interaction between learnable 1\times 1 linear projections to enable cross-group channel mixing. This design reduces intermediate feature expansion from 2|\mathcalS|D (typically 10D for 5 shifts) down to 1.5\tildeD ( k=2 ), 2.0\tildeD ( k=3 ), or 2.5\tildeD ( k=4 ). Furthermore, exploiting the small, fixed group size k \in \2, 3, 4\ , we implement a fused CUDA kernel that evaluates the forward interaction and a recomputation-based analytical backward pass using thread-local variables without allocating global-memory buffers for elementwise intermediates. Finally, we extend the backbone from an isotropic columnar layout to a multi-resolution Hierarchical Clifford Network with configurable per-stage subspace dimensions. Experiments on CIFAR-100 show that the proposed models substantially reduce parameter count and training time across both regimes, while the hierarchical variants also improve classification accuracy over the evaluated shift-based hierarchical baselines.
[CV-32] Vision-Language-Action or World-Action? Just Action Models
链接: https://arxiv.org/abs/2610.37219
作者: Yonghao Yu,Lang Huang,Runyi Li,Toshihiko Yamasaki
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注:
Abstract:The vision-language-action (VLA) and world-action model (WAM) communities debate whether control policies should inherit pretrained knowledge from vision-language models or from visual generators. The debate, however, bundles three separable choices: backbone initialization, the objective applied to future observations, and whether the machinery for that future loss is kept at inference. A gap between a VLA and a WAM therefore cannot simply be attributed to any one of them. We show that decoupling these choices yields a common framework for both families, which we term Just Action Model (JAM): a pretrained visual backbone with an action expert, trained with an action loss and a foresight loss. The foresight loss trains current-observation features to predict frozen encodings of the recorded future through a prediction head discarded after training, so no visual generation is involved in either training or inference, and the backbone can be chosen freely. We first compare objectives within fixed policy architectures, then evaluate the foresight recipe across backbone families. The resulting JAM policies outperform their matched FastWAM and ImageWAM counterparts on LIBERO and LIBERO-Plus, and achieve similar or slightly higher success rates on RoboTwin. Compared with FastWAM, JAM improves the success rates from 97.0%/55.3%/91.2% to 98.2%/64.5%/91.4%, respectively. With cached foresight targets, JAM reduces estimated online training FLOPs per sample per update by 73% relative to FastWAM. Our results show that future supervision improves control without visual generation or additional inference machinery, and strong foresight performance is attainable with either generator-initialized or vision-language backbones. Project Page: this https URL
[CV-33] opoField: Technical Report for ECCV 2026 SceneFun3D Challenge ECCV
链接: https://arxiv.org/abs/2610.37218
作者: Zhiwei Chen,Zhiheng Fu,Zixu Li,Mingyu Zhang,Jiale Huang,Yupeng Hu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Technical Report for ECCV OpenSUN3D Workshop SceneFun3D Benchmark Challenge
Abstract:Functional segmentation requires locating small interactive parts, such as knobs, buttons, and sockets, in indoor point clouds and returning confidence-ranked instance masks. A point-wise semantic predictor produces a continuous probability field, whereas the benchmark evaluates a discrete set of scored instances. This creates a substantial field-to-instance gap. We present TopoField, a deterministic decoder built on top of VOLT. TopoField first converts 13 test-time forward passes into a semantic consensus field, then decodes instance candidates from the consensus geometry, and finally ranks the candidates in a class-aware manner. The augmentation stage retains the processed-point identity of every fragment prediction, accumulates probabilities in a common point array, normalizes by coverage count, and fuses the resulting caches. The decoding stage separates high-confidence cores from broader support regions, constructs a symmetric radius-pruned point graph, and uses a multi-source graph-geodesic partition to split support regions that contain multiple seeds. On 561 ground-truth instances across 30 validation scenes, an IoU-0.5-valid candidate exists for 67.2% of instances, and top-1 coverage is 20.0%. These complementary measurements motivate joint attention to candidate construction and ranking. TopoField achieves 28.286% AP, 41.383% AP50, and 47.558% AP25. TopoField achieved second place on the ECCV 2026 SceneFun3D Challenge.
[CV-34] Zhengrong-Spatial-Aware 1.0: Advancing 3D Multimodal Intelligence
链接: https://arxiv.org/abs/2610.37216
作者: Yu Zhang,Xiaogang Wang,Yuhang Cheng,Jingrui Li,Wenbo Chen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 6 pages, 3 figures, 4 tables. Technical report
Abstract:This report presents the implementation and validation of two end-to-end multimodal adaptation pipelines for 3D-aware visual question answering on a custom spatial dataset. The first pipeline fine-tunes HiSpatial-3B, which augments RGB observations with metric-scale 3D point maps; the second adapts Qwen3-VL through memory-efficient QLoRA-based supervised fine-tuning. The completed engineering stack covers dataset conversion, image-annotation alignment, monocular geometry preprocessing, instruction construction, mixed-domain training, checkpoint management, batch inference, and automated evaluation. On the target spatial benchmark, HiSpatial adaptation increases overall accuracy from 1.73% before fine-tuning to 79.89% at 3,600 steps. On a separately collected styled split, accuracy improves from 14.9% to 55.1%, indicating transfer beyond the original training set. The Qwen3-VL pipeline is likewise implemented end-to-end for both 8B- and 32B-scale models using mixed-domain QLoRA adaptation. The resulting accuracy improves from 20.74% to 45.27% for Qwen3-VL-8B and from 30.79% to 47.60% for Qwen3-VL-32B. Together, these results establish two complementary and reproducible adaptation pipelines for geometry-aware and general-purpose multimodal spatial reasoning. These results establish a reproducible engineering foundation for subsequent work on metric 3D reasoning, spatial grounding, and domain-adapted multimodal systems, while also identifying training-data composition as a central factor in preserving general multimodal capability during specialization.
[CV-35] 3D-SceneQ: Empowering 3D LLM With Query-Guided Adaptive Pruning and Multi-Modal Feature Enhancement
链接: https://arxiv.org/abs/2610.37202
作者: Yongqi Shan,Lu Zhang,Jiazuo Yu,Yunzhi Zhuge,Huchuan Lu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 13 pages, 8 figures, 8 tables. Published in IEEE Transactions on Multimedia
Abstract:Current 3D scene understanding pipelines typically concatenate vast numbers of 3D object tokens with text tokens and feed the resulting sequence to a Large Language Model (LLM). However, existing 3D-LLMs for scene understanding encounter two major limitations: (i) excessive inclusion of task-irrelevant object data, introducing noise that reduces reasoning accuracy and increases hallucinations; and (ii) reliance solely on point cloud data, which inherently lacks the rich semantic information available in complementary 2D modalities, such as color, material properties, texture, and high-level contextual relationships. To address these challenges, we introduce 3D-SceneQ, a 3D LLM that unifies adaptive token pruning with multimodal semantic enrichment, markedly advancing scene understanding, reasoning, and grounding. Specifically, we propose a Query-Guided Adaptive Pruning (QGAP) module that adaptively selects task-relevant objects based on the intended meaning of user instructions, leveraging a learnable latent query to retain critical information while effectively suppressing irrelevant noise. In addition, we introduce a Multi-modal Object-level Feature Enhancement (MOFE) module, which integrates 2D feature embeddings from pre-trained models into 3D representation to enrich object-level semantic information. Evaluations show that 3D-SceneQ outperforms current state-of-the-art methods across multiple benchmarks (e.g., +5.7 CIDEr on Scan2Cap and +17.0 EM@1 on SQA3D), demonstrating its remarkable capabilities in various 3D language reasoning tasks.
[CV-36] How Much Can Best-of-N Help a World Action Model? Measuring Selectable Headroom and Realizing It in Closed Loop
链接: https://arxiv.org/abs/2610.37196
作者: Haojia Gao,Yue Cao,Yumin Liu,Haohua Que,Mingkai Liu,Guanqiu Guo,Zhuo Zou,Bokui Chen,Xueqian Wang
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 24 pages, 12 figures, 14 tables
Abstract:World action models (WAMs) predict the visual consequences of the actions they generate, which makes test-time selection a natural way to improve a fixed policy: imagine the futures of candidate actions and execute the best one. The benefit of selection is bounded by the selectable headroom, how much better the best candidate truly is, but the standard estimate, which executes each candidate once and keeps the best outcome, conflates candidate quality with the randomness of everything that follows. We introduce a denoised counterfactual protocol, based on bitwise-reproducible branching, repeated continuations, held-out oracles, and common random numbers, that measures the headroom a selector can realize, and use its measurements to build a fail-gated selector. Across four WAMs, two benchmarks, over 100,000 branched rollouts, and pre-registered tests, we find that (i) the standard estimate overstates the selectable headroom of native samples: bounds from 17 continuations per candidate place its selectable share between 24% and 66% for three WAMs; (ii) structured action perturbations create up to 16 points of selectable headroom; and (iii) in both WAMs with action-conditioned imagination, imagined futures carry candidate-outcome information with no detectable loss relative to real futures, information that the value head of Cosmos Policy leaves unused. Deviating from the policy chunk only when a readout of the WAM predicts failure, the selector raises closed-loop task success from 76.9% to 81.6% on held-out episodes of training tasks, with the largest gain where the policy fails most often, and from 62.6% to 69.8% on unseen tasks.
[CV-37] VGIQA: Fine-Grained Image Quality Assessment with Visual Grounding Rewards
链接: https://arxiv.org/abs/2610.37186
作者: Weichang Wang,Liyu Chen,Shuyue Guan,Jie Hu,Fei Gao,Wei Li,Nannan Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 20 pages, 5 figures, 6 tables; main paper: 4 figures and 4 tables, appendix: 1 figure and 2 tables
Abstract:Image quality assessment (IQA) aims to evaluate the perceptual visual quality of images. Recently, the rapid development of Multi-modal Large Language Models (MLLMs) has significantly advanced the IQA field. By integrating chain-of-thought (CoT) reasoning and reinforcement learning (RL), these models can generate more accurate and interpretable results while relying on less training data. However, existing MLLM-based methods are predominantly constrained to holistic, image-level evaluations and lack the capability to produce fine-grained, localized assessments, which substantially limits their practical applicability. To address this limitation, we propose VGIQA, a novel IQA model endowed with both local and global quality assessment capabilities. VGIQA is trained via a two-stage paradigm comprising a cold-start phase followed by reinforcement learning. Driven by carefully designed local quality assessment reward functions, our model exhibits exceptional image quality understanding on VGSR-2K, a new dataset we propose specifically for local quality evaluation, significantly outperforming existing open-source MLLMs such as Qwen2.5-VL-72B-Instruct and Qwen3-VL-32B-Instruct. Furthermore, the local quality evaluation task can effectively enhance the model’s performance on the degradation perception task.
[CV-38] INTR: INterpretable TRansformer with better Accuracy Interpretability and Adaptability WACV2027
链接: https://arxiv.org/abs/2610.37178
作者: Dipanjyoti Paul,Arpita Chowdhury,Maneet Chatterjee,Daichi Kitaguchi,Tanya Berger-Wolf,Tetsuya Sakurai,Wei-Lun Chao
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV 2027)
Abstract:Interpretable Transformers often perform well in domain specific settings but lose effectiveness beyond that. Achieving ViT-like broad applicability requires models with consistent performance across domains. Recent works on Interpretable Transformers, INTR and Prompt-CAM, were developed keeping domain adaptability in mind. Prompt-CAM leverages learnable prompts to promote adaptability. However, these prompts are fixed in nature and therefore struggle to adapt to domains with substantial variation in input image characteristics, resulting in reduced interpretability. To address this limitation, we propose INTR++, which extends Prompt-CAM by incorporating a lightweight Prompt Adapter Module (PAM). PAM enables prompts to be adaptive by enriching them with domain-specific knowledge, thereby ensuring consistent interpretability across diverse domains including those with varying image characteristics such as medical imaging, micro-scopic imaging, and remote sensing imagery. INTR++ retains the standard architecture and end-to-end learning of Interpretable Transformer based models, while adding only a small number of additional parameters compared to Prompt-CAM. We demonstrate that INTR++ delivers consistent improvements in all aspects across domains like animals, plants, medical imaging, microscopic imaging, remote sensing imagery and object-centric data.
[CV-39] FastBEVFusion: Simple Multi-Modal Fusion for 3D Object Detection at 50 FPS
链接: https://arxiv.org/abs/2610.37153
作者: Tom Sanitz,Emre Gülsoylu,Simone Frintrop,Christian Wilms
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注: Accepted at Iberian Robotics Conference (ROBOT2026)
Abstract:Reliable real-time 3D object detection based on LiDAR and camera data is critical for safety in autonomous driving, where systems must accurately perceive their surroundings with minimal latency. Many state-of-the-art systems achieve high-quality detection results but fail to meet real-time constraints, limiting their practical use. To address this challenge, we propose FastBEVFusion, a simple, fast, and robust fusion system for 3D object detection. Based on modality-specific features, FastBEVFusion projects these features into semi-aligned modality-specific BEV spaces, without relying on a computationally intensive and potentially error-prone depth estimation. Given these BEV spaces, we propose a light-weight, soft, and dynamic fusion module based on cross-attention, which adapts dynamically to the scene context and enables feature fusion across the entire BEV spaces. Experiments on the complex nuScenes dataset show that FastBEVFusion outperforms state-of-the-art real-time systems, and achieves over 50 FPS without deployment libraries. Additionally, our system shows strong robustness to calibration errors, making FastBEVFusion a powerful 3D object detection system for real-world applications.
[CV-40] GCSR-OCC: Geometry Calibration and Semantic Routing for Open-Vocabulary Occupancy Prediction
链接: https://arxiv.org/abs/2610.37149
作者: Bohua Wang,Congyang Zhao,Tianxiong Lv,Chenxu Zhao,He Zheng,Hangqi Fan,Leigang Luo,Zhihao Liao,Yihong Guo,Yang Cai
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Open-vocabulary 3D occupancy prediction represents scene geometry and semantics in a unified voxel grid, enabling recognition beyond predefined categories. One approach reconstructs scenes with 3D Gaussians and converts them into voxel occupancy. However, Gaussian scale inflation and excessive overlap can distort voxel occupancy predictions, while semantic mixing can arise within Gaussians spanning multiple categories and during 3D–2D semantic feature alignment. We propose GCSR-OCC, which combines geometry calibration and multi-mode query-conditioned semantic routing for indoor occupancy prediction from a single RGB image. The Evidence-Conserving Gaussian-to-Occupancy operator (EC-G2O) uses kernel mass normalization, p -norm aggregation, and a learnable gain to suppress scale inflation and excessive local evidence accumulation while calibrating occupancy responses. Multi-mode Query-First Gaussian Semantic Routing (MM-QGSR) matches multiple semantic modes per Gaussian to text queries and aggregates query-specific evidence into voxels through a learnable text-conditioned router. Fine-grained subclass queries and smoothed semantic aggregation further refine voxel-level predictions. Without 3D semantic annotations, GCSR-OCC achieves state-of-the-art performance in open-vocabulary occupancy prediction on Occ-ScanNet, with 63.31% geometric IoU and 27.44% semantic mIoU. It also reduces geometric aggregation latency from 19.76 ms to 12.88 ms. Our code will be made publicly available.
[CV-41] MixSIS: Mixed-supervised instance segmentation BMVC2026
链接: https://arxiv.org/abs/2610.37138
作者: Christian Wilms,Emre Gülsoylu,Sven Schreiber,Simone Frintrop
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at BMVC 2026. Code available at this https URL
Abstract:Weakly-supervised instance segmentation (WSIS) based on either box-, point-, or image-level annotations has been proposed to overcome the labor-intensive annotation process in fully-supervised instance segmentation. Despite decent results, most WSIS methods only focus on a single annotation level, ignoring that some objects are easy to spot from point annotations, while others require box or even mask annotations. Therefore, we propose the Mixed-Supervised Instance Segmentation (MixSIS) task, which allows a flexible mix of annotation levels (image, point, box, and mask) within a dataset. MixSIS has a high practical relevance, as it allows optimizing the annotation process by annotating challenging objects in more detail, while saving time on simpler instances with less detailed annotations. To address MixSIS, we propose AnnotationsMixer (AnnsMixer) as a baseline. AnnsMixer is a truly mixed-supervised instance segmentation system, which generates pseudo mask annotations from all annotation levels, even mixed within a single image. We evaluate AnnsMixer on both WSIS and MixSIS tasks, largely outperforming state-of-the-art methods. In three case studies, we also demonstrate how the unique features of MixSIS and AnnsMixer can substantially reduce the annotation time (-90%) or improve the results (+34%) in instance segmentation with a simple annotation rule or leveraging heterogeneous data. Our code is available at this https URL
[CV-42] Physics Guided Diffusion Model Based Exposure Interpolation for High Dynamic Range Imaging
链接: https://arxiv.org/abs/2610.37135
作者: Zhengguo Li,Chaobing Zheng
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 11 pages, 6 figures
Abstract:Ghosting artifacts are an open problem for popular multi-shot based high dynamic range (HDR) imaging. An effective solution is to capture two large-exposure-ratio (LER) standard dynamic range (SDR) images simultaneously via a single-shot. Brightness order reversal (BOR) artifacts are generated when the two LER images are directly fused together. It is crucial to interpolate an image with an exposure time in between the exposure times of two LER images. In this paper, a novel physics-guided diffusion model based exposure interpolation algorithm is proposed to address the BOR artifacts. An initial image is interpolated by a physics-driven component, and the residual high-frequency details are generated by a new diffusion model based component. The new diffusion model is based on the correlation analysis of generative models and includes a new predicted target to reduce the variation of the Pearson correlation between the noisy data and the target, as well as the amplification factor of the fitting error. Compared with existing neural augmented exposure interpolation algorithms on top of convolution neural networks, the proposed exposure interpolation algorithm has an inherent advantage to generate more high frequency information. Experiment results validate the proposed exposure interpolation algorithm.
[CV-43] MRI-to-CT Translation of Spinal Images via Anatomy-Guided Pseudo-Pairing and Adversarial Diffusion
链接: https://arxiv.org/abs/2610.37134
作者: Eduardo Nieto,Ibai Genua,Davide Scorza,Karen López-Linares
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Magnetic Resonance Imaging (MRI) is routinely used to diagnose spinal disorders, but surgical planning often requires Computed Tomography (CT) for better visualization of vertebral bone anatomy. Synthesizing CT from MRI could avoid additional radiation exposure and streamline preoperative assessment. We propose a three-stage framework for spine MRI-to-CT synthesis. First, an anatomy-guided vertebral matching system identifies similar sagittal MRI and CT slices using spine segmentations and aligns them through rigid registration, creating pseudo-paired data from unpaired scans. Second, these pairs train an adversarial diffusion model for MRI-to-CT translation. Third, synthesized slices are assembled into 3D volumes and refined by a network that improves inter-slice continuity and structural coherence. The framework was trained on over 10,000 unpaired sagittal MRI and CT slices and evaluated on 1,911 slices from 83 MRI volumes spanning all spinal regions. Assessment combined distribution-based measures with vertebral segmentation metrics. Compared with diffusion alone, the proposed approach increased binary 3D vertebral Dice from 0.666 to 0.739. Volumetric post-processing reduced visible discontinuities between stacked slices and improved vertebral coherence in axial and coronal views. Combining anatomy-guided pseudo-pairing, adversarial diffusion and 3D refinement improves spine MRI-to-CT synthesis. Vertebral matching primarily enhanced anatomical alignment and fidelity, while volumetric refinement improved 3D consistency. These findings support further development of MRI-derived CT-like visualization for preoperative planning, particularly to preserve fine intensity details while maintaining volumetric coherence. Subjects: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI) Cite as: arXiv:2610.37134 [cs.CV] (or arXiv:2610.37134v1 [cs.CV] for this version) https://doi.org/10.48550/arXiv.2610.37134 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Karen López-Linares Román [view email] [v1] Tue, 27 Oct 2026 12:32:47 UTC (10,716 KB)
[CV-44] MCP-Net: Real-Time Coronary Perception Network and Topological Analysis for Robotic Guidewire Navigation
链接: https://arxiv.org/abs/2610.37127
作者: Sahar Nasirihaghighi,Tianyuan Wang,Kenan Niu,Stefano Stramigioli
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注:
Abstract:Coronary guidewire navigation during percutaneous coronary intervention requires continuous interpretation of vessel anatomy, guidewire position, and target stenosis under two-dimensional X-ray fluoroscopy. While existing approaches commonly address vessel segmentation or device localization as separate tasks, a robotic system requires a more structured perception of the coronary scene to support downstream navigation. In this work, we present a simulation-driven framework centered on a Multi-Task Coronary Perception Network (MCP-Net) for robotic guidewire navigation, integrating coronary vessel segmentation, vascular topology extraction, and roadmap generation. Starting from three-dimensional coronary CT angiography, we generate simulated fluoroscopy images containing vascular structures, simulated guidewires, stenotic lesions with varying severity, and diverse vascular geometries, together with simulation-derived ground truth. MCP-Net jointly performs coronary vessel segmentation, guidewire-tip localization, and stenosis localization from a single fluoroscopic frame. These predictions are subsequently converted into a structured vascular graph containing centerlines, bifurcations, endpoints, branch directions, and branch angles. By associating the predicted stenosis with this graph, the framework generates a roadmap from the automatically selected graph root toward the target lesion. The results demonstrate that the proposed framework can jointly recover coronary perception targets and transform them into an interpretable vascular representation suitable for downstream robotic navigation.
[CV-45] Multibranch Ensemble Head in Vision Foundation Models for Efficient Cucumber Phenological Stage Classification
链接: https://arxiv.org/abs/2610.37118
作者: Kostiantyn Khabarlak
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Vision Foundation Models (VFMs) pretrained with self-supervised learning, such as DINOv2, provide highly transferable features, yet adapting them to small, domain-shifted datasets is still challenging. We propose a regularizing multibranch ensemble head: a training and inference scheme that pairs low- and high-capacity branches, samples a single branch stochastically at each training step, and averages both branches in probability space at inference. The design adds negligible parameters and essentially no inference overhead over a single heavy head, while improving the accuracy-efficiency trade-off under data-scarce conditions. Accurate classification of plant growth stages is critical for precision agriculture, enabling timely decisions on fertilization, irrigation, and harvesting. Therefore, as a challenging practical case we use cucumber phenological stage classification, in which flowers and fruits are frequently occluded by foliage, and stage transitions manifest primarily as subtle changes in leaf morphology. The NALEF Food Safety dataset contains images of five cucumber growth stages captured under diverse real-world conditions. On it we evaluate a broad range of backbones, including DINOv2 and DINOv3 ViT and ConvNeXt variants, against strongly supervised counterparts (ResNet, EfficientNet, ViT). The proposed DINOv2 ViT-S with multibranch head achieves a Macro F1 of 0.978 and Cohen’s Kappa of 0.971, matching the performance of models more than ten times larger in parameters and GFlops, while requiring only 12.2 GFlops and 22.3M parameters, making it deployable on smartphones and edge devices for phenological stage classification.
[CV-46] A First-Step Bias Restores Lost Diversity in Diffusion Model Distillation
链接: https://arxiv.org/abs/2610.37110
作者: Wenwu Tang,Dong Wang,Yun Cheng,Olga Saukh
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Distribution-matching distillation enables high-quality image generation in only a few denoising steps. However, distilled student models often map different initial noise samples to similar outputs under the same prompt. To locate where diversity is lost, we compare the teacher and student along the sampling trajectory. The largest measured reduction in the response to different initial noises occurs during the first, high-noise step. When we replace the first step with the teacher’s corresponding trajectory and leave the final three student steps unchanged, most of the output diversity returns. This suggests that the first step is a key point for restoring diversity. Motivated by this finding, we introduce \emphReSeed, a post-distillation method that adds a seed-response bias adapter only to the first student step and aligns its velocity with the teacher’s first-interval secant velocity. All student weights and later denoising steps remain unchanged. We evaluate ReSeed on three text-to-image models (SD3-Medium, SD3.5-Medium, and FLUX.1-dev) and one text-to-video model (Wan2.1-T2V-1.3B). For a distilled 4-step FLUX.1-dev model, \emphReSeed requires only 0.00196% additional trainable parameters relative to the base model and is trained on 8,000 prompt–noise pairs, while keeping the distilled student fully frozen and requiring no target images, perceptual losses, or adversarial losses. It improves DINO diversity by 48.3% while maintaining competitive visual quality. These results show that substantial diversity can be recovered through a lightweight, data-efficient tuning of only the first denoising step, without finetuning the full student model.
[CV-47] SplatScout: Query-Driven Part-Level Referring Segmentation in Gaussian Scenes
链接: https://arxiv.org/abs/2610.37098
作者: Yihan Zang,Da Li,Shinkyu Park,Ivan Viola,Dominik Engel
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Existing referring segmentation methods on Gaussian scenes primarily target whole objects, relying on segmenter-generated candidates from which a vision-language model (VLM) selects. Such predetermined fixed candidates may omit fine-grained parts needed as targets or cues for identifying objects. We introduce SplatScout, a query-driven, closed-loop approach that places perception under VLM control. The model actively selects views, directs segmentation, and inspects rendered 3D results to refine its predictions. For part-level evaluation, we introduce PartRef-ScanNet++, comprising 261 manually verified queries across 12 ScanNet++ scenes with mesh-derived ground truth. On PartRef-ScanNet++, SplatScout achieves 0.538 mIoU, substantially outperforming existing methods. Ablations demonstrate the benefits of iterative refinement, while comparisons across VLM backbones highlight the importance of effective tool-feedback utilization. SplatScout also achieves higher scores on four existing referring benchmarks. Code and the PartRef-ScanNet++ benchmark are available at this https URL.
[CV-48] riDifNet: A Unified Three-Task Diffusion Network for End-to-End RGB-T Perception
链接: https://arxiv.org/abs/2610.37097
作者: Qianchao He,Jing Liu,Yunhao Li,Xiaoqing Tian
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 16 pages, 6 figures
Abstract:Infrared–visible (IR–VIS) multi-modal perception supports robust scene understanding under challenging illumination. Infrared and visible images are often misaligned due to platform motion, viewpoint changes, and modality-dependent distortions, causing registration errors to propagate into subsequent fusion and detection stages. We propose TriDifNet, a unified three-task diffusion network for end-to-end RGB-T perception. To the best of our knowledge, TriDifNet is the first diffusion-based network that jointly optimizes registration, fusion, and detection for misaligned RGB-T pedestrian perception. TriDifNet consists of three diffusion-based modules, i.e., the Registration Diffusion Module (RegDM), the Fusion Diffusion Module (FusDM), and the Detection Diffusion Module (DetDM). RegDM aligns the unaligned infrared image to the optical reference by estimating a dense deformation field. FusDM performs residual conditional diffusion to generate a detection-oriented fused representation with optical structural cues and complementary infrared saliency. DetDM formulates pedestrian detection as an iterative denoising process in bounding-box space. To couple the three tasks, TriDifNet introduces a closed-loop optimization mechanism, where detection semantics guide fusion and registration, and the fusion result further regularizes geometric alignment. A stage-wise progressive training strategy and a lightweight Nash multi-task learning (Nash-MTL)-based late-stage refinement are further adopted to balance registration, fusion, and detection optimization. Extensive experiments on the KAIST multispectral pedestrian benchmark show that TriDifNet achieves superior downstream detection performance under misaligned RGB-T settings. Against strong system-level competitors, TriDifNet achieves 65.44% Recall, 57.55% mAP _50 , and 26.83% mAP _50:95 .
[CV-49] Learning to Look Beyond Screens: Transferable Visual Observation from GUI Grounding
链接: https://arxiv.org/abs/2610.37093
作者: Ruilin Yao,Xinyu Qiu,Tianyu Zou,Yi Rong,Shengwu Xiong,Shuheng Shen,Changhua Meng
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Technical Report
Abstract:Visual perception is evolving from passive recognition toward active, task-oriented observation. We investigate GUI grounding as a scalable source of spatial supervision for learning observation behaviors that transfer beyond interfaces. We introduce LBS (Look Beyond Screens), a vision-language model trained through supervised initialization on teacher-generated GUI grounding trajectories, followed by reinforcement learning for multi-turn crop-and-inspect decisions. Under fixed inference settings, LBS models based on Qwen3.5-35B-A3B and Qwen3.8-27B consistently improve over their corresponding baselines. Gains beyond supervised fine-tuning transfer to PerceptionBench for both backbones and to HR-Bench and ChartQAPro for Qwen3.5-35B-A3B, supporting observation strategies that generalize beyond GUI-specific imitation. Further analysis suggests complementary roles for the two stages: SFT establishes visual inspection primitives, while RL improves their downstream effectiveness. These findings support GUI grounding as a practical supervision paradigm for transferable active visual perception and broader multimodal reasoning. Our code and model are publicly available at this https URL and this https URL, respectively.
[CV-50] Systematic Morphometric Brain Autoencoding for Representation Learning in Neurodegenerative Diseases
链接: https://arxiv.org/abs/2610.37074
作者: Nuria Balbás-Esteban,Albert Belenguer-Llorens,Elaheh Moradi,Jussi Tohka,Vanessa Gómez-Verdejo,Carlos Sevilla-Salcedo
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:Early identification of neurodegenerative disorders using structural T1-weighted MRI is essential for timely clinical intervention, yet supervised deep learning models often struggle to generalize due to limited annotations and scanner-induced distribution shift. Self-supervised learning offers a promising alternative; however, existing approaches are largely adapted from natural image paradigms, relying on contrastive objectives or generic masked reconstruction with voxel-wise losses, which may not fully capture the distributed morphometric patterns associated with neurodegeneration. We introduce SYMBA (SYstematic Morphometric Brain Autoencoding), a self-supervised pretraining framework specifically designed for structural brain MRI. SYMBA integrates anatomically constrained masking, a multi-scale structural similarity (MS-SSIM) reconstruction objective, and a progressive curriculum that transitions from local anatomical detail to global brain morphology. These components explicitly incorporate neuroanatomical structure into the self-supervised objective rather than relying on generic image priors. Pretrained on 6,953 T1-weighted MRI scans from nine heterogeneous cohorts, SYMBA is evaluated using frozen linear probing on two independent unseen datasets, enabling a strict assessment of out-of-distribution generalization. Across multiple clinical tasks, SYMBA consistently outperforms recent neuroimaging foundation models, achieving 0.930 AUROC for Alzheimer’s disease classification, 0.772 for mild cognitive impairment conversion prediction, and strong performance in amyloid status prediction. In low-label regimes, SYMBA demonstrates high sample efficiency, reaching approximately 0.90 AUROC for AD/CN classification using only six labeled subjects. Subjects: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG) Cite as: arXiv:2610.37074 [cs.CV] (or arXiv:2610.37074v1 [cs.CV] for this version) https://doi.org/10.48550/arXiv.2610.37074 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[CV-51] Learning to Code the World: Harness Evolution for Compositional 3D Reconstruction
链接: https://arxiv.org/abs/2610.37046
作者: Haoyu Yao,Chenguo Lin,Ruize Du,Yadong Mu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project page: this https URL
Abstract:Reconstructing editable and compositional 3D scenes from a real image requires jointly reasoning about geometry, object structure, and spatial relationships. We introduce SceneEvolve, a framework that represents 3D scenes as executable code and improves reconstruction by evolving the harness around a small open-source agent. SceneEvolve combines SceneAct, a compositional semantic action interface built on parameterized primitives, hierarchical composition, and semantic spatial relations, with SceneHarness-RSI, which leverages a stronger teacher agent to iteratively evolve the harness using feedback from development-set reconstructions. Throughout harness evolution, the model parameters, action semantics, and execution backend remain fixed. The resulting harness provides a reusable reconstruction procedure that enables the small open-source model to reconstruct unseen scenes without parameter updates or teacher assistance. Across two in-domain datasets and one out-of-domain dataset, SceneEvolve with the comparatively compact Qwen3.8-27B model improves all four evaluated reconstruction metrics while reducing mean inference time by 47.0% relative to the base model. More notably, despite relying on a substantially smaller model, SceneEvolve achieves 32.3% higher reconstruction quality than the large proprietary GPT-5.6-Sol model, while reducing mean equivalent deployment cost by 59.8%. Furthermore, the harness evolved using Qwen3.8-27B also generalizes to Kimi-K3, achieving comparable reconstruction performance to GPT-6-Astra. The project page is this https URL.
[CV-52] GraftRay: Physically Based Mesh Insertion into Gaussian Splatting Scenes via Path Tracing
链接: https://arxiv.org/abs/2610.37036
作者: Jinxin Ai,Matthias Niessner
类目: Computer Vision and Pattern Recognition (cs.CV); Graphics (cs.GR)
备注: 8 pages + references, 6 figures, 2 tables
Abstract:Inserting virtual objects into scenes reconstructed via 2D or 3D Gaussian Splatting (2DGS/3DGS) is a key capability for content creation, AR, and synthetic data generation. However, realistic insertion requires the object to be seamlessly lit by the scene and, conversely, to cast physically plausible shadows onto it. Existing approaches either harmonize appearance in image space, sacrificing multi-view consistency, or approximate scene illumination with an environment map at the insertion point, which cannot represent near-field area lights such as windows and lamps that dominate indoor lighting. To address this, we present GraftRay, a unified method for physically consistent insertion of mesh objects into indoor scenes represented by either 2D or 3D Gaussians. We recover scene materials and surface normals with diffusion-based priors, and explicitly model spatially-extended emitters, solving their radiance in closed form. A hybrid renderer then jointly path-traces the scene Gaussians and inserted meshes, producing view-consistent shading, interreflections, and soft shadows. On FIPT, GraftRay sets a new state of the art, outperforming recent baselines such as ComGS and EAG-PT in rendering fidelity and shadow accuracy, while maintaining highly competitive rendering efficiency.
[CV-53] Interpretable Deep Learning Models for Breast Cancer Diagnosis Using Medical Imaging
链接: https://arxiv.org/abs/2610.37026
作者: Sami Alnawali,Muhammad Atif Saeed,Abdul Hadi,Azka Atiq,Akhtar Jamil,Ahmad Din
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 37 page, 45 Figures
Abstract:Deep learning shows promise for breast cancer diagnosis, but limited interpretability hinders clinical adoption. We benchmark ten ImageNet-pretrained models (ResNet50, VGG19, InceptionV3, DenseNet121, Xception, MobileNetV2, EfficientNet, ConvNeXt, ViT, and Swin Transformer) for benign-versus-malignant classification on the BreakHis histopathology dataset using a patient-level split, and explain their predictions with Grad-CAM, LIME, Integrated Gradients, and LRP. DenseNet121 performed best, with 99.4% test accuracy and an AUC-ROC of 0.999, and significantly outperformed all other models (p 0.001). Transformer-based models trailed the CNNs. The explanations consistently highlighted nuclei-rich and glandular regions of diagnostic relevance. Combining transfer learning with explainable AI can thus yield accurate, interpretable tools for breast cancer diagnosis.
[CV-54] Ground-to-Satellite Knowledge Distillation for Crop Classification under Temporal and Geographic Shift
链接: https://arxiv.org/abs/2610.37021
作者: Miro Miranda,Rocco Sedona,Valerio Marsocci,Gabriele Cavallaro,Claudia Paris
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Continuous and up-to-date crop type mapping is essential for agricultural monitoring but is often hampered by limited reference data. The growing availability of ground-level field photographs offers a promising way to address this challenge. When combined with satellite time series, these images provide complementary perspectives on the same parcel, capturing crop characteristics from distinctly different perspectives. However, field photographs are not always available for each satellite pixel at inference time, limiting their potential in scalable crop type mapping. We investigate whether ground photographs can provide privileged supervision for crop classification built on pretrained satellite embeddings. A photo teacher based on frozen CLIP features supervises classifiers trained on AlphaEarth and TESSERA representations; photographs are used only during training. On a paired field photo and satellite imagery dataset with 17 crop classes, knowledge distillation consistently improves over satellite-only classification; for frozen embeddings the margin grows under temporal shift, reaching up to +5.5 macro-F1 points, and widens in geographic holdouts (+3.7 vs +3.0 in-distribution). These results show that ground-view supervision can complement satellite embeddings and improve their robustness without changing inference requirements.
[CV-55] RePAIR: Inverting Backdoor Attacks in Semantic Segmentation
链接: https://arxiv.org/abs/2610.37015
作者: Arnold Brosch,Kira Maag
类目: Computer Vision and Pattern Recognition (cs.CV); Cryptography and Security (cs.CR); Probability (math.PR)
备注: 26 pages total, 9 pages main paper, 2 figures
Abstract:Despite significant advances in computer vision, semantic segmentation models remain vulnerable to backdoor attacks, posing a security risk. These attacks redirect pixels of a chosen victim class to an attacker-specified target class when a trigger is present, while preserving clean predictions. Existing defenses are largely derived from image classification, while dedicated segmentation defenses remain scarce and typically require retraining or model modification. We introduce RePAIR, a post-hoc, label-free, attack-unaware defense based on the observation that backdoor-induced shifts can be stronger in the classifier head than in internal representations, leaving recoverable victim-class semantics. RePAIR infers the victim-target relation from an unlabeled, potentially contaminated image pool, constructs a safety-approved semantic reference, and localizes suspicious pixels by comparing head predictions with internal semantic evidence. It then applies feature-space and output corrections toward the inferred victim semantics, while conservative localization and output constraints limit changes to benign predictions. We model the attack as adversarial probability transport and repair as its approximate reversal, deriving bounds and sufficient conditions for localization, residual attack success, clean-side changes, defense-induced target predictions, and attack-unaware pair identification. Experiments across segmentation architectures and attack settings show that RePAIR substantially reduces attack success while largely preserving clean performance.
[CV-56] GPA-VGGT: Geometry-Preserving Acceleration of VGGT WACV2027
链接: https://arxiv.org/abs/2610.37000
作者: Muyeol Sung,Jangwon Kim,Jongwoo Lim
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to WACV 2027. Project page: this https URL
Abstract:Feed-forward geometry models such as VGGT recover camera parameters, depth maps, and point maps from one to hundreds of views in a single forward pass, making them powerful backbones for multi-view 3D perception. However, VGGT’s global attention is quadratic in the number of frames and dominates end-to-end runtime on long sequences, which has motivated a range of training-free efficiency methods such as token merging. These methods, however, come with a drop in accuracy that makes them hard to use in place of VGGT. We propose GPA-VGGT, a training-free acceleration method that replaces global attention in early blocks with frame attention augmented by a shared cross-frame K/V set. In the remaining blocks, we merge only keys and values while preserving all queries, restore the attention mass of the tokens each merged key represents, and randomly project the matching metric to 128 dimensions to reduce matching cost. Across diverse multi-view geometry tasks, GPA-VGGT provides training-free acceleration at speedups comparable to prior work while preserving the information required for precise geometric estimation, retaining over 99% of VGGT’s pose accuracy under AUC@3.
[CV-57] LearnPrunerV2: Learnable Progressive Token Pruning for Vision-Language Models
链接: https://arxiv.org/abs/2610.36986
作者: Rinyoichi Takezoe,Yaqian Li,Weiren Wang,Zihao Bo,Huihong Shi,Anzhou Hou,Mo Guang,Jiale Yan,Kaiwen Long
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Accepted to ACM Multimedia 2026
Abstract:Vision-Language Models (VLMs) have demonstrated remarkable multimodal capabilities but suffer from prohibitive computational overhead due to the massive number of visual tokens. While token pruning is a promising solution, existing methods still suffer from two major drawbacks: 1) unreliable pruning criteria that fail to accurately localize task-specific regions due to inherent attention biases or query-agnostic designs, and 2) suboptimal pruning strategies that employ rigid, non-adaptive schedules, overlooking the dynamic evolution of token redundancy across different model depths. To address these challenges, we propose LearnPrunerV2, which extends the previous two-stage heuristic framework (LearnPruner) into a fully learnable architecture. Specifically, we integrate lightweight learnable modules into every layer of the VLM backbone, which estimate token importance by fusing intrinsic visual saliency with extrinsic query relevance captured via a shadow token. By uncovering a consistent staircase redundancy pattern across VLM layers, we further introduce Budget-Deterministic Pruning (BDP) to establish a hardware-friendly schedule with fixed per-layer budgets. Extensive experimental results across various benchmarks and VLMs demonstrate that LearnPrunerV2 consistently outperform previous state-of-the-art methods. Notably, LearnPrunerV2 preserves 99.8% of the original performance when pruning 77.8% of visual tokens on LLaVA-v1.5, achieving a superior accuracy-efficiency trade-off.
[CV-58] Explicit Regional Representations For Video Quality Assessment DATE
链接: https://arxiv.org/abs/2610.36978
作者: Fuyi Yang,Lei Yang,Andrew Liu,Chenyu Wei,Xiaoyi Guo,Ming Kai Hsu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Preprint. This work is an initial version and may be updated in future revisions
Abstract:End-to-end video quality assessment (VQA) methods leave the regional evidence underlying their predictions largely implicit. We introduce an explicit pixel-level regional representation for VQA, instantiated as a Quality-Relevance Heatmap (QRH) that combines localized degradation evidence with perceptual priors under vision-language model guidance, without requiring pixel-level annotations. We apply QRH to existing VQA models through two deterministic fragment-sampling strategies: distribution-preserving sampling approximates the original sampler’s QRH-score distribution without retraining, while quality-aware sampling prioritizes regions with high QRH scores and is paired with fine-tuning. Experiments with DOVER on multiple benchmarks show that the former closely approximates predictions averaged over random samples, while the latter improves most evaluated metrics across popular VQA benchmarks. These results support explicit regional representations as a useful interface for downstream VQA.
[CV-59] Dense Representation Learning and Differentiable Shape Modeling for Spine Landmark Detection
链接: https://arxiv.org/abs/2610.36951
作者: Augustin Daridon,Sofiane Sifaoui,Ayman Assi,Mohamad Karam,Wafa Skalli,Elsa Angelini,Loic Le Folgoc
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Spine landmark detection from spinal X-ray is a key step for estimating spine deformation severity, for building biomechanical models, and for pre-operative planning. We reframe anatomical landmark detection as a pixel-level contrastive representation learning problem, matching dense pixel embeddings with landmark prototypes. Surprisingly, even with a generic U-Net backbone, this formulation results in more discriminative heatmaps compared to the traditional heatmap regression paradigm, leading to improved discrimination between vertebral corners on distinct vertebra. Furthermore, we combine the resulting heatmaps with learnable shape priors over the global configuration of spinal landmarks to resolve ambiguities in landmark locations, yielding a simple and practical approach to spinal landmark detection. Our approach is competitive, in terms of landmark accuracy and clinical parameter estimation, with the state-of-the-art on the public AASCE dataset and on two private spinal X-ray datasets. Our code and pre-trained weights are available at this https URL.
[CV-60] PREEN: Exemplar-Based Color Style Transfer with Smartphone Editor Presets WACV2027
链接: https://arxiv.org/abs/2610.36896
作者: Beomjin Ahn
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: Accepted to WACV 2027
Abstract:We present PREEN (Preset Regression via an Exemplar-based Editor Network), which casts color style transfer as regression onto the 29 parameters of a deterministic image editor. Given a content image and a style reference, PREEN predicts a preset, a vector of editor slider values that is interpretable and editor-consumable, not a raster image. The key obstacle is supervision, since an arbitrary real reference lacks a preset ground truth. We therefore supervise with a style-transfer teacher, whose target looks we invert into the editor’s parameter space as pseudo-presets, then pretrain a CLIP-conditioned regressor and refine it with LoRA under an image-level CIELAB loss. On the human-curated PST50 benchmark, PREEN applies the reference’s color while preserving structural fidelity, leading the baselines on pixel- and perceptual-fidelity metrics. The preset is compact and predicted at sub-second on-device latency. The low-dimensional subspace also exposes a single inference-time strength knob that traces a continuous default-to-vivid trajectory in one forward pass, with every intermediate state a valid editor preset. A two-axis human study shows that raters prefer the vivid end for matching the reference but the default as a standalone photo, a trade-off the knob spans without retraining.
[CV-61] HoloRecon: Incremental Reconstruction of Long-Range Panoramas with Global Consistency
链接: https://arxiv.org/abs/2610.36876
作者: Zhuoxiao Li,Xinyi Liu,Hongye Hou,Jing OU,Kang Du,Shuai Zhang,Dongli Wu,Taoyu Wu,Siyuan Feng,Xiaohui Zhu,Wufan Zhao
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project Page: this https URL
Abstract:Long-sequence panoramic reconstruction requires detailed scene geometry and globally consistent camera poses across distant revisits. Panoramic SLAM and incremental SfM rely on feature correspondences, but repeated, similar structures and small partial overlap make reliable long-range connections difficult to establish, leaving accumulated drift unresolved. Feed-forward models offer detailed local geometry, but their high computational costs limit inference window size, necessitating additional constraints for global consistency. This motivates combining continuous feature-based tracking and sparse mapping with rapid, high-quality local reconstruction from feed-forward models to achieve accurate localization and globally consistent dense mapping over long sequences. We therefore propose a panoramic reconstruction framework that couples an ORB-based sparse panoramic mapping frontend with a fine-tuned feed-forward model. Our key insight is that predicted local geometry enables 3D verification of feature matches across distant revisits, yielding long-range camera constraints even between regions. We calibrate predicted depths with sparse landmarks and use verified connections to optimize the global pose graph, updating dense geometry with the corrected camera poses. We also introduce PanoRecon, a seven-scene benchmark with repetitive architecture, large and small loops, and reference camera poses and LiDAR geometry for evaluating global consistency. Our method achieves the lowest mean absolute trajectory error and high F-1 Score among the evaluated methods. Project Page:this https URL
[CV-62] RangeRectQuant: Compressing Scientific Fields for Every Future Rectangular Sum
链接: https://arxiv.org/abs/2610.36873
作者: Jiahao Zhang,Xinling Wen,Yu Chen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 10 pages, 6 figures. Code, data, core experiments and extended figures: this https URL
Abstract:Scientific rasters are often stored before the regions to be measured are known. Independent pixel-error bounds can force exact integer reconstruction when one absolute error budget covers every future rectangle. Coordinated changes allow every pixel to change while every rectangular sum remains accurate. A prefix sum accumulates values from the top-left corner. Four such sums determine any rectangular total. RangeRectQuant preserves the pixel range and designated exact samples while controlling errors in prefix sums. At each row position, we characterize exactly which errors let the remaining pixels meet these constraints. These errors form a single integer interval, even when many completions exist. A backward pass finds its endpoints; a forward pass uses them to choose quantized pixels. The resulting field meets the prescribed error budget for every axis-aligned rectangle, with linear work and workspace proportional to one row. Experiments on geographically held-out built-surface and water-occurrence maps show substantial storage savings over alternatives meeting the same requirements. Data producers encode once; later users measure arbitrary rectangles without the source, re-encoding, or query-specific correction.
[CV-63] ail Cues Principal Corrections: Plug-and-Play Rectification for Open-Set Test-Time Adaptation NEURIPS2026
链接: https://arxiv.org/abs/2610.36870
作者: Yingkai Yang,Chaoqi Chen,Hui Huang
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at NeurIPS 2026
Abstract:Open-Set Test-Time Adaptation (OSTTA) operates on non-stationary target streams where covariate and semantic shifts coexist. Existing TTA methods often obtain adaptation signals by updating model parameters or maintaining mutable memories, which can be costly and vulnerable to unknown samples. We present TALOR, a lightweight, backward-free rectification module that uses the source linear head as a geometric anchor. The head-induced basis decomposes each normalized feature into high-gain principal coordinates and low-gain tail coordinates. TALOR estimates a soft-routed affine correction in the principal subspace, combining a regime-level principal bias with a tail-conditioned slope that captures sample-level tail-to-principal coupling. Using weighted ridge regression, it subtracts this drift estimate from the principal coordinates, reconstructs the feature and feeds it into the source head. On the ImageNet-C benchmark with six csOOD datasets, TALOR outperforms the next-best UniEnt by 2.2% H-score while running 2.5x faster and using only 12% of its GPU memory; as a post-correction plug-in, it further boosts COME to 68.4% H-score with only 2% additional memory overhead.
[CV-64] ActiveAlign: Aligning Egocentric Manipulation via Active Perception
链接: https://arxiv.org/abs/2610.36859
作者: Ri-Zhao Qiu,Bowen Wen,Yifeng Jiang,Chen Tessler,Eric Heiden,Xue Bin Peng,Xiaolong Wang,Stan Birchfield,Erwin Coumans
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Real-to-Sim-to-Real (Real2Sim2Real) enables structured world representations for planning, interpretability, and safe execution. Most existing approaches rely on passive perception, which creates real2sim gaps due to imperfect observation. Inspired by how humans perceive actively through actions, we propose ActiveAlign, a framework that leverages active perception to align the Real2Sim2Real gap from interactions. ActiveAlign introduces two complementary modules: (i) ActiveAlign-Waving, which performs hand-eye calibration by exploiting active hand motions and differentiable rendering, and (ii) ActiveAlign-Touching, which refines object alignment through tactile feedback to achieve millimeter-level precision. Together, these capabilities enable strong real2sim alignment. We demonstrate that ActiveAlign-Waving can be used to curate simulation environments for open-sourced uncalibrated robot manipulation datasets, and demonstrate on a real Unitree G1 robot that ActiveAlign enables fine manipulation of millimeter-scale objects such as pressing buttons and moving sliders.
[CV-65] Forget the Data Keep the Capability: Switchable Unlearning for Retrieval
链接: https://arxiv.org/abs/2610.36858
作者: Jingong Chen,Kyle Gao,Guangyuan Li,Chulwoo Pack,Sijia Liu,Kwanghee Won
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Privacy compliance requires re-identification models to selectively forget individuals while preserving their ability to recognize others. For this, we present MoME, a switchable framework that can adapt pre-trained backbones for privacy compliance, allowing the same model to switch between Full mode for standard inference and Core mode for unlearned inference. MoME splits one backbone into core parameters that preserve retrieval and zero-initialized parasitic parameters that carry the full-data knowledge, and training them under gradient isolation so the core never receives a forget-set gradient. Disabling the branch yields Core, which satisfies a request to delete a person’s training data, and enabling it yields Full, which retains what the deleted data taught, so honoring the request is a flag flip rather than a retraining run. We prove Core independent of the forget set up to an audited set of coupling channels and introduce the rank-based Forgetting-Rank statistic to audit that independence in a trained model. On Market-1501, VeRi-776 and RegDB, Core reads at a retrained reference under three membership attacks while retrieving above it, and the retrieval advantage carries to ResNet and CLIP backbones. Code and weights are available at this https URL.
[CV-66] Evaluating AlphaEarth Embeddings for Crop Classification: Separability Temporal Dynamics and Cross-Regional Transferability
链接: https://arxiv.org/abs/2610.36857
作者: Priyanka Goyal,Maitreyee Sar,Adway Mitra,Manjira Sinha
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:Earth Observation (EO)-based agricultural monitoring is shifting from pixel-level spectral analysis toward representation-driven modeling, yet systematic evaluation of learned representations across spatial, temporal, and domain-generalization dimensions remains limited. Geospatial foundation models generate high-dimensional embeddings encoding spatio-temporal land-surface properties, but their utility for agricultural applications remains insufficiently understood. This study presents a representation-agnostic evaluation framework integrating geometric separability, temporal dynamics, spatial robustness, and cross-regional generalisation. A 64-dimensional AlphaEarth embedding space is benchmarked against Sentinel-2 reflectance, NDVI-derived phenology, and harmonic features using field-scale data from the Conterminous United States (CONUS; 2017-2023) and Northern India (2021). Traditional features achieve strong linear separability (Fisher ratio), while embeddings provide superior distribution-aware separation – Bhattacharyya and Mahalanobis distances up to an order of magnitude higher for spectrally similar pairs (e.g. Corn/Soybeans: 3.9 vs. 0.10). Embeddings uniquely capture multi-year crop rotation dynamics, reaching a separability ratio of 1.784 versus a maximum of 1.003 across all engineered features, and exhibit lower residual spatial autocorrelation of errors (mean Moran’s I \approx 0.17 ). However, harmonic and NDVI features show stronger cross-continental transferability (US \to India Macro F1: 0.493 vs. 0.326 for embeddings), reflecting better distributional alignment. The combined feature space yields balanced performance, supporting hybrid modeling. These findings reveal a fundamental trade-off between representational richness and cross-region transferability, providing a structured framework for feature selection in large-scale crop classification. Subjects: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG) Cite as: arXiv:2610.36857 [cs.CV] (or arXiv:2610.36857v1 [cs.CV] for this version) https://doi.org/10.48550/arXiv.2610.36857 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[CV-67] LingGS: Decoupling Feed-Forward Geometry Prediction from Camera Tracking for High-Fidelity Gaussian SLAM
链接: https://arxiv.org/abs/2610.36853
作者: Da Li,Yihan Zang,Ivan Viola,Zeyu Wang,Dominik Engel
类目: Computer Vision and Pattern Recognition (cs.CV); Graphics (cs.GR)
备注: 10 pages, 4 figures
Abstract:Achieving high-quality reconstruction and responsive camera tracking remains challenging for monocular RGB Gaussian splatting SLAM under a limited optimization budget. We present an online system that addresses this challenge by combining learned geometric priors with explicit geometric camera tracking, which can substantially outperform learned camera prediction in terms of tracking accuracy. A frozen streaming 3D foundation model serves as a virtual RGB-D sensor, providing depth estimation from monocular video. An explicit geometric solver then estimates camera motion through feature reprojection, and a streaming Gaussian mapper integrates the resulting observations into a dense scene representation. We further show that geometric tracking improves camera pose accuracy across representative streaming geometry backbones, demonstrating the broader applicability of our approach. Extensive experiments on indoor and outdoor datasets show that our method significantly improves reconstruction quality over existing approaches, achieving about 7 FPS on a single RTX 4090 and delivering 2-3 times the throughput of the evaluated prior SLAM systems.
[CV-68] uTHCD-C: A Corruption-Robustness Benchmark for Tamil Handwritten Character Recognition
链接: https://arxiv.org/abs/2610.36850
作者: Preethi S,Meeradevi T,Arun Prakash A
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 14 pages, pre-print
Abstract:Evaluation of Tamil handwritten character recognition (HCR) has largely focused on clean test images, and a standardized multi-corruption robustness benchmark for isolated Tamil HCR has not been established. We introduce uTHCD-C, a corruption-robustness benchmark based on the unconstrained Tamil Handwritten Character Database (uTHCD). The benchmark defines ten corruption types, including a handwriting-specific stroke-thinning corruption, at five severity levels. We evaluate nine convolutional and transformer-based architectures using five-fold stratified cross-validation and writer-independent testing. Robustness is quantified using mean Corruption Error (mCE), normalized against a VGG16-BN reference. We additionally compare parameter count, computational complexity, and inference throughput to characterize the computational requirements of the evaluated architectures. The results establish comparative robustness baselines for Tamil HCR, revealing substantial differences across architectures: ConvNeXt and Swin achieve the lowest mCE values (0.75 and 0.83, respectively), whereas ShuffleNet and MnasNet exhibit higher values (1.55 and 1.85, respectively). The accompanying efficiency measurements characterize the trade-offs between corruption robustness and computational requirements. By providing standardized corruption configurations, evaluation protocols, and comparative results, uTHCD-C establishes a reproducible basis for future research on robust and computationally efficient Tamil HCR.
[CV-69] AxisQ: Depth-Adaptive Channel–Frequency Quantization for Camouflaged Object Detection
链接: https://arxiv.org/abs/2610.36844
作者: Chi Ruan,Tianqi Li,Xin He,Xue Geng,Ming-Ming Cheng,Yun Liu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Camouflaged object detection (COD) depends on weak texture and boundary cues that can be destroyed by aggressive post-training quantization (PTQ). Existing PTQ methods mainly optimize quantizers or reconstruct network blocks, while using a single precision-allocation domain throughout hierarchical backbones. Instead, we introduce AxisQ, a depth-adaptive bit-allocation framework motivated by the observation that the useful allocation axis changes with feature depth. Under a fixed per-layer average activation-bit budget, AxisQ allocates precision in the channel domain for shallow features and in the local-frequency domain for deep features, using COD-aware sensitivity to better preserve subtle foreground regions and ambiguous boundaries. We also formulate a boundary-aware Hessian approximation for reconstruction by incorporating boundary-weighted gradient and error statistics. Experiments across multiple COD models and diverse benchmarks demonstrate improved accuracy over the strongest baselines at low bit widths, with relative weighted F-measure gains of 12.7% for CFRN and 33.3% for CamoFormer-P on CAMO under W2A2 quantization. These results support adapting the allocation domain to feature depth as a transferable quantization strategy for hierarchical COD models.
[CV-70] MVRCap: Towards Comprehensive Video Captioning via Hierarchical Multi-Dimensional Verifiable Rewards
链接: https://arxiv.org/abs/2610.36838
作者: Weihuang Lin,Xianing Chen,Zhibin Wang,Congzhi Zhang,Yiwei Ma,Jiayi Ji,Miao Chen,Jun Song,YuCheng,Bo Zheng,Xiaoshuai Sun,Rongrong Ji
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Despite significant progress, current models struggle to generate comprehensive video captions, constrained primarily by coarse-grained supervision, which often fails to capture spatial-temporal dynamics and the underlying logical intent. To address this issue, we introduce MVRCap, a novel framework that reformulates video captioning as a reinforcement learning task steered by structured verification. Central to our approach is a Fine-Grained Verifiable Reward mechanism, serving as dense supervision to enforce rigorous visual-semantic alignment. Specifically, we construct this reward system anchored by three complementary pillars: a Hierarchical Concept Coverage Reward utilizing our newly constructed MVRCap-QA dataset to ensure multi-dimensional completeness and semantic depth, a Factuality Verification Reward to enhance factual consistency, and a Conciseness Constraint to maximize semantic density. Furthermore, we curate the high-quality MVRCap-SFT dataset, providing structured annotations across six key dimensions, to serve as a robust foundational initialization for the RL policy in this training pipeline. Extensive experiments demonstrate that MVRCap achieves state-of-the-art performance on authoritative benchmarks including VDC, CAPability, and VC4VG.
[CV-71] Drop the Details Get the Gist: Dynamic Conceptual Representation for Video Recognition ACCV2026
链接: https://arxiv.org/abs/2610.36832
作者: Ryoya Yoshiyama,Bowen Wang,Yuta Nakashima
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 25 pages, 5 figures, 4 tables. Accepted to the MM4SciReal Workshop at ACCV 2026
Abstract:Concept Bottleneck Models (CBMs), which route predictions through a layer of human-interpretable concepts, have substantially improved the interpretability of visual classification by exposing an explicit, language-based rationale for each prediction. Most existing work focuses on static images, and temporal dynamics in videos have been underexplored in conceptual representation. We propose a framework that extends concept-based image recognition to procedural video understanding. It represents a multi-step video as a temporal sequence of segment-level concept score vectors and feeds this sequence into a Transformer encoder to predict the class label. We also explore a way to estimate the contribution of each concept for further interpretability. We evaluated three concept sets of different granularities on cooking and household-activity videos and found that coarser-grained concepts are sufficient for classifying these videos, which implies that fine-grained textual descriptions may be unnecessary.
[CV-72] Right Click Wrong Box: Grading GUI Grounding by What a Click Does
链接: https://arxiv.org/abs/2610.36831
作者: Xiaokang Yuan,Binhui Yao,Luwei Yang,Ruimin Chen,Zeng Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 14 pages, 7 figures, includes supplementary material
Abstract:Leading GUI grounders score 93 to 96% on ScreenSpot-v2, and reported gaps between new and prior models are about a point. At this level the useful question is what the remaining errors are. Benchmarks count a click as correct only inside one annotated box, but interfaces often expose one action through several elements: a duplicated link, the label of a checkbox, the title next to a logo. The box is not wrong. The rule is incomplete. We call a click correct when executing it produces the same outcome as clicking the annotated target. A probe that executes and fingerprints about a thousand clicks per page turns this definition into outcome maps and an execution judge for live web applications, with no new annotation. We execute the predictions of three strong 7–8B grounders on a code forge and an online shop. About one in five and one in three of the clicks that box scoring rejects reach the intended outcome, and the shares remain one in seven and one in three when the instructions are rewritten as natural tasks. A blind audit of 11 models on ScreenSpot-v2 (three 7–8B grounders and eight 3B variants) finds the same effect on web screenshots, with two independent reviewers on a 60-unit subset. The share is highest for the strongest grounders. Training seeds and test-set sampling add noise that more runs and paired intervals control. Box scoring instead under-counts correct clicks, which repetition does not remove, and the audit-based correction of 0.2 to 1.8 points on ScreenSpot-v2 is as large as the gaps the field reports. The correction is needed in evaluation. In our 3B GRPO study, crediting equivalent clicks in the reward gives no stable gain over box rewards across three seeds and four test sets. We release the probe, the outcome maps, the audit and every prediction.
[CV-73] APESTRY: A Single Backbone for Video Generation and Robot Control
链接: https://arxiv.org/abs/2610.36823
作者: Jeremy A. Collins,Seungjae Lee,Rachanon Wachakorn,Krishnan Srinivasan,Animesh Garg,Vitor Campagnolo Guizilini,Paarth Shah
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: Accepted to Conference on Robot Learning (CoRL) 2026
Abstract:Pretrained video generation models encode rich priors about physical interactions, but leveraging them for robot control typically requires either resource-intensive video generation at test time or action heads trained from scratch that do not fully utilize the pretrained backbone. We introduce Tokenizing Actions as Video Patches in a Shared Transformer Policy (TAPESTRY), a world action model that repurposes a pretrained video diffusion transformer into a robotic policy without requiring video generation at inference. Our key idea is to treat robot actions as single-patch tokens in the video model’s latent space. A lightweight linear projection maps each action timestep to the patch dimensions of the pretrained VAE, and the resulting tokens are processed by the same patchifying layers and transformer backbone used for video. The shared backbone is optimized through three diffusion objectives sampled per minibatch element: Video diffusion uses VAE latents from robot and human videos to establish dynamics priors. Joint video+action diffusion trains on robot demonstrations using a coupled diffusion timestep and an asymmetric attention mask. Action-only diffusion denoises actions conditioned on clean image latents. At test time, the model generates a compact action token sequence in a single DDIM step from pure noise, producing smooth action chunks at 84 ms per action chunk on consumer hardware and avoiding the cost of full video generation. Across simulation benchmarks and real-robot experiments, TAPESTRY achieves strong performance with substantially lower inference cost than comparable joint video–action policies. More information can be found at this https URL.
[CV-74] Visual Autoregressive Generation with Next Token Imagination KR
链接: https://arxiv.org/abs/2610.36793
作者: Ruihang Jiang,Zongyao Zhao,Zhaolin Wang,Yuanwei Liu,Xiatian Zhu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 19 pages, 7 figures. Code and checkpoints: this https URL
Abstract:The space of possible visual content is enormous, and arguably unbounded. Any training set, however large, therefore covers only a tiny fraction of it. Yet generative models are typically trained to fit only the images they are given. In visual autoregressive (AR) generation, for example, each token is predicted from the preceding ground-truth (GT) tokens and supervised solely by the GT next token. Confining learning to observed data in this way limits how much of the visual distribution a model can capture. In this work, we hypothesize that letting a model imagine beyond its training data, by exploring plausible visual realizations it has never observed, can ease this limitation. We use visual AR generation as a proof of concept, since its fixed, discrete token space makes imagination easier to steer and constrain. We propose Next Token Imagination (NTI), a method that allows the model to explore beyond the GT next token, subject to a carefully designed criterion that keeps this imagination bounded by the underlying visual distribution. Extensive experiments on ImageNet show that NTI substantially and consistently improves the fidelity of generated images, approaching state-of-the-art diffusion and AR models while using a smaller model and lower training cost.
[CV-75] OmniGlimpse: Asymmetric Identity Memory for Online Multi-View Tracking
链接: https://arxiv.org/abs/2610.36785
作者: Rangya Zhang,Jiaping Xiao,Yujing Shang,Yuhang Zhang,Mingsheng Li,Mir Feroskhan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Multi-view tracking extends visual coverage beyond a single camera, providing complementary observations when targets are occluded or leave the view. This raises a question: how can observations from different views be matched to the correct identities without offline scene-specific training or camera calibration? Progress toward this goal remains limited because many offline formulations assume such deployment-specific knowledge, while online adaptation can amplify association errors. We study this problem as online multi-view identity maintenance (OMIM): targets are initialized once per view, with no further identity labels provided after each initialization. To this end, we introduce OmniGlimpse, an asymmetric propose-verify framework that separates adaptive association from stable identity consolidation. At the representation level, short-term local and cross-view evidence is mapped into target-specific adaptive spaces to propose associations, whereas long-term identity distributions are maintained for verification in a canonical space defined by a frozen encoder. At the update level, short-term evidence may guide current associations, whereas, after bootstrapping, only verified observations may update persistent memory or drive adaptation. Under a unified target-specified evaluation protocol, OmniGlimpse improves over the best evaluated online baseline for each metric by 9.1 IDF1 and 22.1 CVIDF1 points on M3Track. The corresponding gains on MvMHAT are 20.3 and 26.2 points. Despite using no offline scene-specific training, it outperforms offline-trained references in IDF1 and remains competitive in cross-view association. These results support an approach to online identity maintenance that adapts to changing appearances while consolidating reliable evidence into stable identity anchors.
[CV-76] CVBL-CCTV-Face Dataset for Real-World Face Enhancement and Recognition in Surveillance
链接: https://arxiv.org/abs/2610.36781
作者: Manish Das,Ajay Kumar Yadav,Shiv Ram Dubey,Satish Kumar Singh
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 6 pages, conference submission
Abstract:In real-world surveillance settings, facial captures from closed-circuit television systems suffer from degradations such as low resolution, motion blur, noise, and occlusions, which undermine dependable face recognition. In this study, we propose a framework for super-resolution and identity recognition, aimed at surveillance applications. For this purpose, we present the CVBL-CCTV-Face Dataset which is a dataset of real-world images taken in an unpaired manner where each pair consists of a low-quality (LQ) image of a face captured by a 8MP IP Dome CCTV camera together with a high-quality (HQ) reference image that is taken in controlled settings. In order to cope with the various problems posed by such datasets, we construct a modular system that involves face detection, deep learning-based super-resolution of low-quality face images, and identity recognition through L2-normalized embeddings extracted by an embedding model. Enhancement models are trained and tested according to three different setups, and the system is deployed for identity verification in a real-time unconstrained environment.
[CV-77] Evaluation-Guided Demonstration Refinement for Vision-Language-Action Policies: The Winning Entry in the MetaFine Open Competition
链接: https://arxiv.org/abs/2610.36765
作者: Lifan Zheng,Guotao Chen,Zhenyang Li,Shujie Cui,Qiang Huang,Fan Liu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 1st place in the MetaFine Open Competition, RoboFineMani Workshop @ CoRL 2026
Abstract:We fine-tuned a separate \pi_0.5 policy for each of the five tasks in the MetaFine Open Competition, with a task-specific training mixture and number of actions executed per prediction. Training started from the official demonstrations and expert procedures. Local evaluation exposed recurring grasping, pushing, and insertion failures. We addressed them by revising the demonstrated motions and by adding expert continuations from states the policy reached, while keeping nominal demonstrations in training. Checkpoints were compared on every instruction, so that a gain on one instruction could not hide a loss on another. Our submission, MAI, ranked first overall with a score of 96.66.
[CV-78] GeoGSplat: Geometry-Grounded Non-Pixel-Aligned Gaussian Splatting
链接: https://arxiv.org/abs/2610.36760
作者: Jiale Wang,Chen Zhao,Haoyang Zhang,Yanhao Wu,Fei He,Yanhu Shan,Rui Wu,Wei Ke,Tong Zhang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Feed-forward Gaussian Splatting enables efficient novel view synthesis, but pixel-aligned prediction makes the primitive count grow with image resolution and view count, leading to substantial redundancy and increased computational cost. Non-pixel-aligned methods address this limitation by predicting a fixed number of Gaussians. Without explicit ray and depth constraints, these methods yield misplaced Gaussian centers, causing rendering artifacts under large viewpoint changes. To address this issue, we introduce GeoGSplat, a geometry-grounded feed-forward model that explicitly incorporates geometric priors into the non-pixel-aligned framework. A dual-branch encoder constructs Anchors and Triplane to capture coarse 3D structure and fine-grained scene features, respectively. Subsequently, a hierarchical decoder progressively decodes dense 3D points and Gaussian primitives, enabling 3D reconstruction and novel view synthesis. GeoGSplat surpasses both state-of-the-art pixel-aligned and non-pixel-aligned Gaussian Splatting methods on DTU and NRGBD in terms of novel view synthesis and reconstruction. The advantage of our geometry-grounded paradigm further increases as target viewpoints move farther from the input cameras, demonstrating stronger generalization to large viewpoint changes. Project Page: this https URL
[CV-79] Efficient Remote Sensing Image-Text Retrieval via Interaction-Free Recall and Text-Guided Reranking
链接: https://arxiv.org/abs/2610.36756
作者: Xi Chen,Xu Chen,Xiamgyang Jia,Wei Wang,Xu Zhang,Zhenyuan Shun
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Remote sensing (RS) image-text retrieval plays a critical role in understanding massive RS imagery. However, the dense multi-object distribution and complex backgrounds in RS imagery make it difficult to simultaneously achieve fine-grained cross-modal alignment and efficient retrieval. Existing methods either rely on complex cross-modal interactions that lead to low retrieval efficiency, or depend on large-scale vision-language adaptation, which requires massive data and computational resources. To address these issues, we develop a fast-then-fine (FTF) retrieval framework centered on interaction-free multi-granular recall and parameter-free text-guided cross-modal reranking with the Balanced Text-guided Interaction Block (BTIB). Specifically, interaction-free multi-granular recall pre-computes gallery embeddings and matches them with the corresponding query embedding without full-gallery token-level text-image interaction, while text-guided cross-modal reranking with BTIB refines local semantic correspondence within the top-K candidate set. Furthermore, an inter- and intra-modal loss is designed to jointly optimize cross-modal alignment across multi-granular representations. Extensive experiments on public benchmarks demonstrate that the FTF achieves competitive retrieval accuracy while significantly improving retrieval efficiency compared with existing methods.
[CV-80] Multi-task Disentangling Fusion Mamba-CNN for MODIS Spectral-Temporal Unmixing with Endmember Variability Modeling
链接: https://arxiv.org/abs/2610.36750
作者: Motasem Alkayid,Yimin Zhu,Naser El-Sheimy,Lincoln Linlin Xu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Spectral-temporal unmixing of MODIS time series is essential for applications such as abundance mapping and spectral library construction, yet it remains challenging due to (i) spectral-temporal information coupling in MODIS data, (ii) intra-class endmember spectral variability and inter-class signature ambiguity, and (iii) the instability of naively fusing multiple expert representations. This paper presents the Multi- task Disentangling Fusion Mamba (MDF-Mamba) to address these limitations. To resolve spectral-temporal coupling, we design a disentangling architecture that pairs per-channel spectral experts with a dedicated bidirectional state-space (Mamba) temporal branch. Convolutional modules each process a single spectral channel, while the Mamba branch encodes the cross-channel seasonal trajectory, forming a multi-branch expert ensemble with soft inter-expert reweighting based on squeeze-and-excitation. For the integration of these spectral and temporal expert outputs, we introduce mHCFusion, adapting the manifold-constrained hyper-connection framework to merge the outputs in a stable and interpretable manner. Endmember variability is addressed by extending the linear mixing model with pixel-wise learnable spectral variability weights that capture intra-class endmember variation. Finally, we design a multi-task training objective that combines supervised abundance estimation with self-supervised spectral reconstruction and endmember regularization terms on the same training data, yielding a highly efficient framework for spectral-temporal unmixing. Experiments on MODIS time series covering the Alberta and Saskatchewan provinces show that MDF-Mamba achieves the best performance on 12 of 14 evaluation metrics across both datasets (7 of 7 on Alberta; 5 of 7 on Saskatchewan), ranking second on the remaining two.
[CV-81] Readout and Operating-Point Sensitivity in Event-Stream Candidate Selection
链接: https://arxiv.org/abs/2610.36743
作者: Yifan Shuai
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 12 pages, 4 figures, including supplementary material
Abstract:Event-triggered recording needs a useful combination of target capture and false triggers, yet representation comparisons can change with readout and threshold calibration. We evaluate real event accumulations from 520 DailyDVS-200 clips and 47 participants using participant-disjoint training, validation, and testing. The original study varies frozen encoder scale, accumulation duration, and temporal view count; a post-review extension adds two supervised-pretrained convolutional encoders, a 61,586-parameter trained CNN with three seeds, ten repeated participant partitions, and calibration sensitivity. At 50 ms/eight views under the primary handshake-or-fall policy, DINO-S CLS captures 101/132 targets with 40/388 false triggers, versus 78/132 and 35/388 for patch mean. DINO-B CLS has lower selected-threshold capture but higher fold-wise low-FPR partial AUC (0.746 versus 0.732). MobileNetV3-Small captures 97/132 with 53/388 false triggers, demonstrating why lightweight baselines and actual operating points matter. Across ten dependent repartitions, mean normalized partial AUC is 0.738 for DINO-B CLS and 0.674 for MobileNet. The original LeJEPA S-to-B capture difference is 8.3 points, but its multiplicity-adjusted interval includes zero. Duration and view-count effects remain policy dependent. We find no established JEPA-specific advantage. The evidence concerns offline action-defined candidate selection on previously used development data; it does not establish personal recording value, continuous false alarms per hour, or eyewear energy.
[CV-82] Hy-Nomas: Towards Free-form Interactive World Generation
链接: https://arxiv.org/abs/2610.36729
作者: Qiaoling Chen,Jiaxiang Cheng,Xueqing Deng,Tiankai Hang,Ji Hou,Jiaqi Li,Yifan Liu,Tao Lu,Junshu Tang,Yu Tang,Yao Teng,Yu Tian,Heng Wang,Weizhi Wang,Xiyao Wang,Xilun Wu,Mingyu Yang,Yangbin Yu,Rui Zhao,Longjun Cao,Linghao Chen,Haoyi Duan,Jiaming Han,Mingji Han,Enhua Jiang,Shunkai Li,Yanzhe Liang,Qinwei Ma,Cong Wang,Lei Wang,Bing Wu,Ke Xing,Ray Ying,Zhentao Yu,Key Zhang,Yuan Zhou,Zhengguang Zhou,Kai Zou
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Technical Report
Abstract:Interactive world generation requires models to follow successive commands while preserving relevant history and action consequences. We present Hy-Nomas, a video world model with an 8B diffusion backbone that supports free-form text instructions and continuous camera control across multiple interaction turns. It generates variable-length segments conditioned on current inputs and cached history. Our central design aligns structured video captions with action-aware causal attention: descriptive context remains accessible across segments, while action instructions are restricted to their intended temporal windows. A vision-language Prompt Enhancer (PE) translates user requests and multimodal interaction histories into structured conditions for the next segment. A multi-stage curriculum progressively introduces causal generation, text interaction, camera control, and training on model-generated histories, followed by diffusion distillation. In human evaluation, Hy-Nomas shows superior performance over existing world models. We release model weights, code and benchmarks, together with a lightweight adaptation workflow, to support reproducible research on interactive world generation.
[CV-83] Progressively Refining SAM for Accurate Phytoplankton Segmentation
链接: https://arxiv.org/abs/2610.36722
作者: Jialin Li,Jinkun Jiang,Yuezun Li,Junyu Dong
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 5 pages, 3 figures
Abstract:Accurate phytoplankton segmentation is essential for ecological monitoring, yet it remains largely underexplored due to the subtle visual characteristics of phytoplankton and the scarcity of pixel-level annotations. To address these challenges, we propose PhySeg, a SAM-based segmentation framework equipped with a lightweight refinement module that progressively restores fine-grained details through multi-scale feature fusion. In addition, we construct a phytoplankton segmentation dataset by manually annotating instance-level masks for an existing dataset, providing a valuable resource for future research. Experimental results demonstrate that PhySeg improves segmentation accuracy, validating the effectiveness of the proposed module. Code is available at this https URL.
[CV-84] SAGE: Saliency-Annealed Gated Encoder for Universal Vision Recognition
链接: https://arxiv.org/abs/2610.36719
作者: Le Yang,Yucheng Huang,Junyu Chen,Qi Dang,Xin Kong,Haifeng Huo,Qiang Zhao
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:When adapting self-supervised Vision Transformers (ViTs) to dense prediction, existing methods often treat the ViT as a passive feature extractor and may underuse the rich priors encoded in its representations, including saliency, global context, and high-frequency details. We propose SAGE (Saliency-Annealed Gated Encoder), an enhanced ViT-Adapter backbone that explicitly exploits these priors through three complementary innovations: (1) a saliency-guided architecture that extracts DINO-style class-token (CLS) saliency to guide feature gating and high-frequency enhancement, together with dynamic multi-layer aggregation and a local-detail branch; (2) an annealing-based training paradigm designed to improve convergence and feature quality during early training while reducing feature drift later in training; and (3) a two-stage, quality-aware training mechanism that smoothly transitions from focal loss to Varifocal Loss (VFL) within a single training run. Across the evaluated tasks (Fig.~\reffig:fig_com), SAGE achieves state-of-the-art (SOTA) or competitive results on COCO (62.8 box AP and 55.3 mask AP), COCO keypoint estimation (78.5 AP), ADE20K (60.3 mIoU for semantic segmentation and 44.9 mask AP for instance segmentation), TinyImageNet classification (94.76% Top-1 accuracy), and DIOR-R oriented object detection (78.8 AP _50 ). The same backbone is used across tasks, with task-specific decoders. To the best of our knowledge, SAGE is the first framework to achieve multi-task state-of-the-art results on Huawei Ascend NPUs. Code and pretrained models will be released after internal review.
[CV-85] Learning Where to Trust: Reliability-Guided Expert Routing and Velocity Gating for 3D Deformable Registration
链接: https://arxiv.org/abs/2610.36672
作者: Chenyu Guo,Wenjie Dai,Yiqian Chang,Zhiyuan Zhang,Yunfan Wei,Shuhan Qi
类目: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Deformable image registration relies on accurate anatomical correspondences. However, explicitly coupling local matching reliability with spatial context selection and deformation update strength remains insufficiently explored in learning-based registration. This challenge is particularly pronounced in homogeneous, weakly textured, and severely deformed regions, where ambiguous matches may induce erroneous deformation updates. We propose ReliMorph, a reliability-routed symmetric framework for 3D deformable registration that couples correspondence assessment with spatial context selection and update this http URL correlation distributions characterize correspondence ambiguity using shared anatomical features with modality-specific adaptation. These reliability cues guide spatial expert fusion and gate residual velocities during coarse-to-fine refinement. Updated inverse-inconsistency feedback guides routing and gating at finer pyramid levels. To our knowledge, ReliMorph is the first 3D deformable registration framework to jointly condition spatial expert routing and residual velocity gating on correspondence reliability. Extensive experiments demonstrate that ReliMorph consistently outperforms competing methods in anatomical overlap, highlighting its effectiveness across unimodal and multimodal registration tasks.
[CV-86] Look Only When Needed: Certificate-Guided Visual Routing for Efficient Diffusion VLMs
链接: https://arxiv.org/abs/2610.36671
作者: Donghyeon Cho,Joowon Kim,Seungho Shin,Eunho Yang
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Diffusion vision-language models (DVLMs) process their entire input at every denoising step. Since visual tokens make up 89-97% of this input on the benchmarks we study, most of the decoding computation is spent repeatedly processing the same image. Visual token pruning methods partially relieve this burden by removing part of the visual tokens, but they lose the removed information and still process the remaining image at every step. To measure how often the image is actually needed, we compare LLaDA-V’s predictions with and without the image at every decoding step and find that removing the image changes a token being unmasked at only 22.8% of steps on average and at as few as 7.9%. We therefore propose LOWN, a decoding method that decides at every step whether to process the image and how many tokens to unmask. LOWN predicts the masked tokens without the image and processes the image only when this prediction disagrees with the last image-conditioned prediction. Agreement certifies the earlier prediction for reuse, and the number of certified tokens sets how many are unmasked in that step. We further observe that the need for the image differs across benchmarks, so LOWN’s adaptive decision is needed rather than a fixed schedule. Without training or a tuned threshold, LOWN retains 99.6% of LLaDA-V’s score on ten benchmarks at 35.3% of the decoding FLOPs on average. LOWN also generalizes across LLaDA-based DVLMs, composes orthogonally with visual token pruning, and runs under key-value caching for further efficiency gains. Code and project page are available at this https URL
[CV-87] Haste Makes Waste? Towards Safe Few-Step Text-to-Image Models
链接: https://arxiv.org/abs/2610.36665
作者: Lior Ziv,Mahmood Sharif
类目: Computer Vision and Pattern Recognition (cs.CV); Cryptography and Security (cs.CR)
备注: Accepted at IEEE SP 2027
Abstract:While few-step distilled text-to-image models have gained traction due to their fast generation times, the speed-up implications on their unsafe generation rates remain unknown. We uncover an inherent tension between speed and safety: few-step models are often more unsafe than their teachers, and reducing the number of inference steps consistently raises unsafe generation rates. We further find that acceleration makes unsafe concepts harder to erase in few-step models, with existing concept-erasure methods markedly less effective than in teachers. We offer a mechanistic diagnosis for why erasure methods fail on few-step models, finding that the fragility of few-step models’ generation process and the way generation guidance is usually distilled into those models may hinder established concept-erasure methods. To address the speed-safety trade-off, we propose DICE (Distillation-Integrated Concept Erasure), a post-distillation erasure method tailored for few-step models. DICE aligns both its latent construction and its generation-preservation loss with the distillation method that produced the model, keeping erasure faithful to few-step models. DICE demonstrates strong results in our evaluation. For example, on FLUX Schnell, DICE reduces unsafe-generation rate to 0.046 (vs. at least 0.161 for prior erasure methods we test), while mostly preserving benign generation quality.
[CV-88] SWIM: Learning What Matters from Behaviors for JEPA World Models
链接: https://arxiv.org/abs/2610.36662
作者: Jiaming Hu,Shi Bo,Yan Zheng,Tian Wang,Florian Dubost,Alejandro Mottini,Junze Liu,Arvind Srinivasan,Kai Zhong,Kun Qian,Sharon Gao,Vianne Gao,Houyu Zhang,Qingjun Cui
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注: 25 pages, 1 figure
Abstract:Joint-embedding predictive architectures (JEPAs) enable world models to plan from images by predicting future embeddings. However, their visual prediction objectives do not explicitly incorporate task-relevant state information, so the physical state that determines task progress may not be reliably captured in the latent space. Supervising world models with simulator state can make planning more informative, but it requires deciding which state features matter. Features are typically selected by hand or weighted uniformly, and a uniform weighting gives every standardized state feature equal influence, so redundant coordinates can overcount the same variation, while irrelevant ones can obscure task-relevant differences. To address this limitation, we propose SWIM (State Weighting with Inverse-Dynamics Model), which learns state weights from offline trajectories by predicting recorded actions from current states and observed future displacements. The learned weights quantify each feature’s contribution to action prediction and define a weighted state metric for state-supervised training, which we demonstrate by aligning latent and state distances. The inverse-dynamics model is used only during training, so the resulting world model plans from images with no additional planning-time computation. Across five simulated robotic manipulation tasks, SWIM consistently improves planning success over uniform state weighting.
[CV-89] SAGE: Semantic-Augmented Geometric Experts for Point Cloud Registration WACV2027
链接: https://arxiv.org/abs/2610.36658
作者: Tomoyasu Shimada,Kazuhiko Murasaki,Shoichiro Takeda,Shogo Sato,Riku Inoue,Toshihiko Nishimura,Ryuichi Tanida
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to WACV 2027 (Round 1)
Abstract:Point cloud registration is an essential task for scene reconstruction, requiring the alignment of partially overlapping point clouds. Under low-overlap conditions, this task remains challenging because existing geometry-only methods cannot distinguish geometrically similar but semantically distinct structures (e.g., TV corners vs. desk corners and chair legs vs. table legs), resulting in incorrect correspondences. To address this geometric ambiguity, we present Semantic-Augmented Geometric Experts (SAGE), which integrates frozen pre-trained point cloud semantic representations into the registration pipeline without requiring additional sensors or fine-tuning the semantic encoder. To effectively leverage these semantic distinctions, SAGE employs a Semantic-Guided Mixture of Experts (SG-MoE) architecture. Through semantic-aware routing, SG-MoE enables each expert to specialize in distinct semantic-geometric patterns, reducing mismatches that confound geometry-only methods. With a structure-aware estimator, SAGE achieves the highest registration recall on 3DLoMatch while remaining competitive on 3DMatch. When evaluated with the same number of refinement iterations, SAGE further achieves the highest inlier ratio and registration recall on both benchmarks. These results indicate that frozen semantic representations provide effective cues for resolving geometric ambiguity in low-overlap registration.
[CV-90] Adapting Video Models for Robot Control via Motion-guided Asymmetric Prediction
链接: https://arxiv.org/abs/2610.36624
作者: Taeyoung Kim,Junyoung Sung,Heeseung Kwon,Minseop Kim,Jinwoo Shin
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Project page: this https URL
Abstract:World Action Models (WAMs) leverage pretrained video generative models for robotic control by jointly predicting future visual observations and actions. The visual prediction objective plays a key role in adapting pretrained video representations to the target robot domain. Yet standard visual prediction inherited from video pretraining treats all visual regions uniformly, despite their differing relevance to robotic control. We therefore investigate the spatial allocation of visual supervision in WAMs and show that focusing supervision on dynamic regions can substantially improve policy performance. Building on this finding, we introduce Motion-Guided Asymmetric Prediction (MoGAP), which augments WAM training with motion-focused visual prediction. MoGAP allocates noise asymmetrically across regions, training the model to reconstruct more heavily corrupted dynamic regions using less corrupted static regions as context. MoGAP is trained alongside standard video-action prediction through a training-only auxiliary prediction path within the shared video diffusion backbone. It uses motion cues derived directly from training videos, requiring neither an external teacher nor additional modalities. Across robot manipulation benchmarks and WAM architectures, MoGAP consistently improves policy performance, increasing success rates on the GR-1 tabletop benchmark from 38.8% to 47.5% for Fast-WAM (Wan2.2-5B) and from 52.3% to 62.3% for Cosmos3-Nano.
[CV-91] 4D Reconstruction with Events RGB ECCV2026
链接: https://arxiv.org/abs/2610.36610
作者: Dulanga Weerakoon,Darshana Rathnayake,Van Minh Do,Hemanth Sabbella,Kaichen Zhou,Fangneng Zhan,Alok Prakash,Sanjay E. Sarma,Archan Misra,Paul Pu Liang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at NeVi 2026 workshop at the 19th European Conference on Computer Vision (ECCV 2026)
Abstract:Reconstructing dynamic 3D scenes over time (4D reconstruction) from a single moving camera is central to robotics and AR/VR applications, yet state-of the-art feed-forward methods rely solely on RGB frames and therefore inherit the failure modes of conventional cameras, including motion blur and temporal aliasing under fast or non-uniform camera and object motion. Event cameras report per-pixel brightness changes asynchronously at microsecond resolution, offering a complementary signal that is largely immune to these artifacts, but whether this modality can be incorporated into existing RGB based 4D reconstruction pipelines without a full architectural redesign remains an open question. In this preliminary study, we investigate a minimal extension: we augment Monst3R, an existing RGB-only 4D reconstruction model, with an additional event-camera input branch, leaving the rest of the architecture unchanged. Across 3 benchmark datasets, this simple RGB-event fusion reduces Abs Rel by 3.4% on average relative to the RGB-only Monst3R baseline, and a sensitivity analysis shows that the improvement grows with the magnitude of camera and object motion, with the largest gains observed under high object motion conditions where RGB-only reconstruction degrades most. These early results suggest that even a lightweight, non-invasive fusion of event data can meaningfully improve 4D reconstruction robustness, motivating a more thorough architectural investigation in future work.
[CV-92] Camera-to-Quadruped Robot Pose Estimation from Sequential Images
链接: https://arxiv.org/abs/2610.36599
作者: Seobin Hwang,Ji-Seong Beom,Yeong-Jun Cho
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注: 17 pages, 8 figures, 6 tables
Abstract:Quadruped robots are increasingly deployed in hazardous environments such as disaster response and facility inspection, where external-camera monitoring is essential for safe operation. This setting is challenging because quadrupeds are articulated systems with time-varying leg joints, leg symmetry, and frequent self-occlusions, which make keypoint correspondence ambiguous. Moreover, no public benchmark exists, making quantitative comparison difficult. To address these challenges, we propose QuadPose, which estimates the full-body 18-DoF pose (6-DoF base pose + 12-DoF joint angles) from monocular video sequences, and introduce QuadSyn, the first camera-to-quadruped robot pose benchmark. On QuadSyn, QuadPose achieves 92.18% PCK@2.5px, 95.54% PCK-AUC, 85.94% ADD@20cm, 66.98% ADD-AUC, and 3.65 degree mean joint-angle error at approximately 33 FPS.
[CV-93] WithinVAE: Compact Visual Autoencoding with Self-Distilled Semantics
链接: https://arxiv.org/abs/2610.36593
作者: Yuzhou Ji,Jiepeng Wang,Tianyu Jin,Lizhuang Ma,Haibin Huang,Chi Zhang,Xuelong Li
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Visual autoencoders must balance reconstruction fidelity, compactness, and semantic structure. Existing semantic tokenizers typically inherit semantics from pretrained recognition encoders, whose invariances can discard the appearance detail required for inversion. We show that this inheritance is unnecessary: semantic structure can be learned inside an autoencoder’s own reconstructive representation. We introduce WithinVAE, a compact variational visual autoencoder with a sequential bottleneck, DINO/iBOT-style self-distillation from an exponential-moving-average copy of the autoencoder itself, and position-dependent token masking to achieve a 4-fold compression ratio, which is free of pretrained representation encoder or external semantic target. On ImageNet-1K at 256\times256 , WithinVAE improves all four reported reconstruction metrics over RPiAE including PSNR by 1.36 dB, while using half its latent scalar capacity. Under frozen evaluation, the same latents improve linear probing by 6.5 points over RPiAE and 5.1 points over our reconstruction-only baseline, attributing the semantic gains to self-distillation rather than architecture or external supervision. An auxiliary latent-to-latent model reproduces the ordered-compression trend, and a pixel-to-latent model transfers the objective to a pretrained generative space. These results separate the origin of semantic supervision from its use: a compact visual autoencoder can organize semantic structure without borrowing a recognition representation.
[CV-94] LabWorld: Learning Interactive 4D Worlds for Scientific Experimentation
链接: https://arxiv.org/abs/2610.36592
作者: Suozhi Huang,Kelsey Fu,Binxu Li,Luna Lyu,Kaize Wu,Wenbo Lu,David Smerkous,Jiahan Li,Siqi Zhu,Zhizhou Sha,Jiacheng Guo,Zhixuan Liang,Shilong Liu,Le Cong,Mengdi Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 20 pages, 4 figures. Suozhi Huang and Kelsey Fu contributed equally
Abstract:Autonomous scientific agents must not only reason about experiments but also interact with the physical environments in which they are executed. Existing laboratory simulators typically rely on manually constructed assets and predefined procedures, while laboratory videos capture real equipment, layouts, and interactions but leave this information implicit in pixels. We introduce LabWorld, an agent-driven Real2Sim framework for reconstructing interactive 4D laboratory worlds from monocular RGB video. LabWorld recovers reusable 3D assets, a metric scene, time-varying object states, and procedure events. Multi-frame observations provide complementary evidence for asset and scene reconstruction, while motion reconstruction combines visual and geometric observations with event constraints encoding the spatial and temporal requirements of laboratory interactions. We evaluate LabWorld on real pipetting and vortexing videos collected in our laboratory against general-purpose agentic and dedicated reconstruction baselines. For pipetting, LabWorld achieves a mean Mask IoU of 0.501, exceeding the strongest baseline by 0.033. For vortexing, it achieves a mean Tube-Length-Normalized PCK-AUC of 0.638, exceeding the strongest baseline by 0.035. On five pipetting videos, event-conditioned reconstruction increases the fraction of modeled physical constraints satisfied from 14.3% (5/35) to 100% (35/35). The resulting worlds expose geometry, motion, articulation, and events as explicit, editable state for replay, inspection, and future embodied scientific systems.
[CV-95] FrameForge: Harnessing Priors of Video Foundation Models for Unified Image Generation
链接: https://arxiv.org/abs/2610.36545
作者: Yijing Lin,Zhendong Mao,Mengqi Huang,Yuxin Wang,Yongdong Zhang
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Unified conditional image generation requires a single model to preserve different aspects of a visual input while producing a target image that follows textual guidance. Pretrained video models provide a promising foundation by encoding visual transformations across time, yet existing approaches typically assign the condition and target to fixed temporal positions. We observe that condition-target relationships vary substantially across generation tasks, while video models naturally encode visual transformations at different temporal scales. This motivates us to model conditional generation through input-adaptive virtual temporal relations. We propose FrameForge, which captures two complementary aspects of such relations. Mixture of Temporal Experts (MoTE) models how far the target lies from the condition by dynamically selecting temporal RoPE experts with different virtual intervals. Chain of Temporal Reasoning (CoTR) models how the condition is transformed into the target through a latent intermediate state, with a reasoning-bottleneck attention mask encouraging information propagation through this pathway. Experiments on DreamBench, COCO2017, and GEdit-Bench show that FrameForge achieves superior performance in subject-driven generation, canny-to-image generation, and image editing among the compared methods.
[CV-96] Frugal Handwritten Text Recognition for Sub-Saharan Civil-Status Registers: A Smartphone-to-Records Pipeline with CPU-Only Inference BMVC2026
链接: https://arxiv.org/abs/2610.36540
作者: Ousmane Hamadou,Jean Louis Kedieng Ebongue Fendji
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at the GlobalSouthCV Workshop, BMVC 2026. 17 pages, 3 figures
Abstract:Handwritten civil-status registers across Sub-Saharan Africa are being photographed, but rarely transcribed, which leaves their content unsearchable. Local administrations that hold these records typically lack GPUs, reliable connectivity, and the budget for external digitization services. We study whether a frugal pipeline, with consumer-grade capture and CPU-only inference, can make such registers machine-readable. We introduce a corpus of 1,568 smartphone images of handwritten birth, marriage, and death records from a municipal registry in Chad, with line-level transcriptions for 185 records. Our pipeline combines a fully classical preprocessing stage, running at about 2.6 images per second on a laptop CPU, with a Kraken recognizer fine-tuned from a pretrained model in about four hours on a free-tier GPU. On manually corrected line segmentation and a single-registry test set whose writers also appear in training, fine-tuning reduces the character error rate from 14.53% to 0.50% and the word error rate from 56.93% to 3.37%. On a six-core laptop CPU, the recognizer weighs 61 MiB, peaks below 2.6 GiB of RAM, and processes an image in under 6 s on average, excluding line segmentation. Finally, a downstream information-extraction task shows that residual recognition errors cost about 9 F1 points on role resolution, compared with a 36.7-point gap between the fine-tuned and pretrained recognizers. These results are evidence of feasibility within one registry, not of performance on unseen writers or sites; we discuss these limitations in detail.
[CV-97] PRISM: Pathology Reliability In Scarce-label Medicine. A Tile-level Benchmark of Pathology Foundation Models ALT NEURIPS2026
链接: https://arxiv.org/abs/2610.36517
作者: Alicem Koyun
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注: 30 pages (9 main), 21 figures. Accepted at the Third Workshop on GenAI for Health (GenAI4Health) at NeurIPS 2026
Abstract:Pathology foundation models are typically benchmarked on classification performance at full label availability and on in-distribution test sets, while calibration, behavior at low label budgets, and robustness under cross-institutional shift are seldom evaluated jointly. We introduce PRISM, a tile-level benchmark that measures the reliability of pathology foundation models along four axes: model choice, label fraction, calibration, and out-of-distribution transfer. PRISM evaluates eight foundation models (CLIP, PLIP, CONCH, UNI, Virchow2, GigaPath, H-Optimus-0, Midnight) on six tile-level datasets at six label fractions (1% to 100%) with three seeds (864 linear probe runs), extended with a five-hospital Camelyon17 study of 20 directed transfer pairs. Four phenomena emerge. (i) Under full supervision, AUROC rank predicts post-temperature calibration rank consistently across datasets (Spearman rho = 0.52, I^2 = 0%), whereas at 1% labels the relationship becomes dataset-specific (rho = -0.24, I^2 = 67%). (ii) On harder datasets the linear probe collapses to a single predicted class below a dataset-specific label fraction. (iii) Whether additional source-domain labels help or harm target calibration tracks how well the probe transfers: below a transfer AUROC of roughly 0.85, more labels worsen target ECE in 86-88% of combinations. (iv) Temperature scaling fails to reduce ECE in 49 of 288 cells, miscalibration is predominantly under-confidence, and a source-fitted temperature is harmful under failing transfer. PRISM is released as a pip-installable package with reference results for all 288 cells. Code and data: this https URL
[CV-98] Identity or Evidence? Auditing Evidence Dependence in Agent ic Life-Science Benchmarks
链接: https://arxiv.org/abs/2610.36510
作者: Danai Brilli,Maria Georgila
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Agentic life-science benchmarks give an agent evidence about a named entity and score an answer against a recorded outcome. That score tests reasoning over evidence only if the evidence, rather than the name, determines the answer. We examine this distinction by varying identity and evidence independently on a clinical-outcome benchmark and published tasks from Biomni and LAB-Bench. Blinding identity matters little when a prompt contains a decisive sequence or passage. It matters a great deal when the prompt is mostly a name and candidate list (0.742 named versus 0.082 blinded; chance 0.093). Across 96 LAB-Bench LitQA2 questions, the gap increases under one fixed passage-redaction scheme (+0.000 to +0.135; per-instance slope +0.147, p= 0.004). On the clinical benchmark, a no-evidence recall probe matches prediction accuracy and recovers a database-specific programme count. So, we argue that benchmarks intended to measure evidence-dependent reasoning should report identity-blinded results, retrieval telemetry, and positive controls.
[CV-99] ARISE: Authentic Reference-Informed Stability Enhancement for Continual Multimodal Forgery Detection
链接: https://arxiv.org/abs/2610.36509
作者: Yuchen Zhang,Yaxiong Wang,Yujiao Wu,Lianwei Wu,Qi Zhang,Li Zhu,Chaofeng Dong,Le Wu,Zhedong Zheng
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Multimodal forgery detection is commonly studied under a static assumption: training attacks are expected to approximate those encountered after deployment. This assumption becomes increasingly fragile as generators, editing systems, and multimodal large language models continually introduce new forgery mechanisms and forensic evidence. We formulate this problem as attack-incremental continual multimodal forgery detection, where the binary real/fake interface remains fixed while emerging attacks must be learned without forgetting previous ones. We introduce Fable-Bench, which organizes attack evolution through complementary joint-modality, text-side, and image-side protocols. We further propose ARISE (Authentic Reference-Informed Stability Enhancement), motivated by the open-ended forged class and the comparatively reusable geometry of authentic content. ARISE fits a real-only Gaussian mixture model to projected authentic replay representations as a compact historical reference. Meanwhile, the proposed Gaussian-Responsibility-Aware Stability Preservation (GRASP) derives a sample-adaptive anisotropic metric that strongly constrains historically stable representation directions while retaining plasticity elsewhere. Our Vulnerability-Aware Replay (VAR) further estimates stage-wise vulnerability from authentic-density degradation, forged-to-authentic boundary confusion, and authentic-mode occupancy drift, then assigns bounded replay emphasis to fragile stages. All adaptive signals are obtained from training-time replay statistics, while inference uses only the original multimodal backbone and binary classification head. Experiments across three attack-evolution protocols and multiple detector architectures demonstrate improved historical retention and final average performance over representative continual-learning baselines.
[CV-100] Spatial-Frequency Learning for Scanning Transmission Electron Microscopy Image Enhancement
链接: https://arxiv.org/abs/2610.36487
作者: Hesong Li,Ying Fu,Ziqi Wu,Ruiwen Shao,Tao Zhang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted by IEEE TPAMI
Abstract:Scanning Transmission Electron Microscopy (STEM) enables the observation of atomic arrangements at sub-angstrom resolution and plays a key role in the discovery and characterization of advanced materials. Raw STEM images are often noisy and blurred due to multiple degradations. Although recent deep learning methods improve STEM images, their performance remains limited under severe degradations. In this work, we propose a spatial-frequency learning framework for STEM image enhancement. Specifically, we design a spatial-frequency fusion network that exploits periodic atomic arrangements in the frequency domain and restores local atomic details in the spatial domain. Spatial- and frequency-domain features are enhanced separately and then fused. We also present a STEM noise calibration method for synthesizing more realistic STEM images. The parameters of scan noise, pointwise noise, and background noise are estimated through statistical analysis and fitting of real STEM images. Using these calibrated parameters, we construct a diverse dataset containing a wider range of atomic arrangements, imaging modes, and degradations. Random masks combine different atomic arrangements to simulate interfaces between multiple materials. Experimental results demonstrate that our method achieves state-of-the-art enhancement performance on synthetic datasets and real images of advanced materials while improving atomic localization accuracy.
[CV-101] Confident First Reflective Second: Reinforcement Learning of Confidence-Preserving Reflection for Video Understanding
链接: https://arxiv.org/abs/2610.36467
作者: Zihao Lin,Jingyuan Liu,Yucheng Zhang,Yuguang Yao,Lifu Huang
类目: Computer Vision and Pattern Recognition (cs.CV); Multimedia (cs.MM)
备注: 18 pages, 8 figures
Abstract:Video models answer from sampled frames and may miss the decisive moment. A common remedy is to commit to an answer, re-examine the video, and confirm or revise it, a loop now mainly learned with reinforcement learning (RL). We find that RL erodes the initial answer, a phenomenon we term confidence erosion, where confidence means the label-verified reliability of the initial answer instead of self-reported certainty: because errors can be repaired later, a correct initial answer earns less reward, and the initial answer stalls even as the final answer’s gain over it rises. We propose the principle confident first, reflective second: the initial answer must stay reliable on its own, above a floor set by a reference policy, and reflection may only add to it. Our method, CoFire, implements this principle with one mechanism for each erosion channel: reward incentives, parameter drift, and gradient interference. CoFire reaches 67.0% mean accuracy on Video-MME, MLVU, and LongVideoBench, 1.7 points above the strongest compared 7B method, and improves initial-answer and final accuracy over outcome-reward RL by 4.1 and 3.7 points on average: a stronger initial answer carries through to a stronger final one.
[CV-102] Nemotron-Labs-Diffusion-Uni: Elastic Self-Refining Unified Diffusion LLM with Cross-Paradigm On-Policy Distillation
链接: https://arxiv.org/abs/2610.36466
作者: Shufan Li,Hanrong Ye,Greg Heinrich,Yonggan Fu,Haotian Zhang,Sifei Liu,Aditya Grover,Pavlo Molchanov
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:We propose Nemotron-Labs-Diffusion-Uni (NLD-Uni), a unified discrete diffusion large language model (DLLM) for visual understanding and generation tasks. Compared with prior unified diffusion models, NLD-Uni introduces two major innovations. First, we propose a unified cross-paradigm on-policy distillation (OPD) framework that successfully transfers the visual understanding ability of autoregressive VLMs and visual generation ability of continuous flow models, to a unified DLLM. Second, NLD-Uni incorporates a novel progressive self-refining diffusion process that gradually corrupts each token across multiple steps during training, as opposed to adopting the conventional masked diffusion formulation that directly corrupts a clean image token to an uninformative mask token. At inference, this design allows NLD-Uni to make intermediate commit to a token, enabling progressive self-refinement for better visual quality. Through extensive experiments, we show that NLD-Uni achieves state-of-the-art performance among unified DLLMs on diverse tasks including visual question-answering, multi-modal reasoning, object grounding, text-to-image generation, and image editing.
[CV-103] ClusterSplat: Semantic Cluster Selection for 3D Visual Grounding in Gaussian Splatting NEURIPS2026
链接: https://arxiv.org/abs/2610.36462
作者: Jaesung Lee,Hyeontaek Hwang,Gabriel Manalu,Daeyoung Kim
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at NeurIPS 2026 as a Spotlight paper. Project page: this https URL
Abstract:We study 3D visual grounding in 3D Gaussian Splatting (3DGS), where a referring expression should identify a target object as both a rendered 2D mask and a set of explicit 3D Gaussians. Prior methods typically extract 3D targets by applying heuristic thresholding or top-ratio filtering to dense primitive-level relevance scores. However, these scores often vary substantially across scenes and prompts, making such heuristics sensitive and leading to unstable targeting. ClusterSplat addresses this by reframing 3D target selection as cluster-level selection over scene-adaptive candidates. The method first learns instance-aware Gaussian features, forms feature-aware seed voxels, and merges neighboring seed voxels when a description-length criterion decreases, producing a scene-specific set of cluster candidates. Given a referring expression, a query-conditioned cluster scorer ranks the cluster candidates with a lightweight MLP, and the selected cluster directly defines the explicit 3D target while its cluster score is rendered for the 2D mask. On ScanRefer, ClusterSplat achieves the best rendered 2D segmentation and explicit 3D target metrics among the compared 3DGS baselines. It also obtains the best average Ref-LERF scores, supporting the same trend under fine-grained referring descriptions. On ScanNet v2 class name queries, ClusterSplat achieves the strongest rendered 2D results and the highest 3D matching accuracy, while maintaining competitive 3D IoU. Our project page is available at this https URL .
[CV-104] Full Collapse Is Suboptimal: Controlled Partial Neural Collapse for Long-Tailed Recognition WACV2027
链接: https://arxiv.org/abs/2610.36460
作者: Hyeontaek Hwang,Cheonghwan Hur,Daeyoung Kim
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Accepted at WACV 2027
Abstract:Recent long-tailed recognition (LTR) methods incorporate neural collapse (NC) and align features to a fixed Equiangular Tight Frame (ETF) as a uniform target for all classes. However, forcing tail classes to fully collapse over-contracts their samples and weakens generalization. To address this, we propose a controlled partial neural collapse, realized through two coordinated controls on a backbone: the Class-Adaptive Sparse Activation (CASA) classifier and the ETF-constrained adaptive margin (ETF-AM) loss. Rather than forcing a uniform full collapse, this regime allows head classes to collapse along the ETF while tail classes retain within-class diversity. This prevents the tail from being forced into full collapse, which is suboptimal under imbalance. To set the dimensionality of collapse, CASA induces class-related node specialization through feature-conditioned gating so that tail classes occupy their own feature directions. For the clearance axis, ETF-AM applies a frequency-adaptive margin, with smaller margins for head classes and larger for tail classes. On CIFAR100-LT, ImageNet-LT, Places-LT, and iNaturalist-2018, our method achieves the best overall accuracy over prior methods. In the empirical analysis, within-class spread increases monotonically from head to tail classes, consistent with tail classes retaining variability rather than fully collapsing.
[CV-105] LociArt: A Unified Paradigm for Free-Style Local Video Stylization
链接: https://arxiv.org/abs/2610.36456
作者: Yuxuan Chen,Feng Xie,Jiagao Hu,Fuhao Li,Zepeng Wang,Fei Wang,Daiguo Zhou
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Multimedia (cs.MM)
备注: Project page: this https URL
Abstract:Video stylization typically transforms entire scenes, while local artistic editing requires precise region–style association, spatial control, and temporal consistency. We introduce LociArt, a unified paradigm for free-style local video stylization that explicitly specifies where to stylize and what style to apply while preserving non-target content. At its core, GRNStyle builds on Generative Refinement Networks (GRN) and introduces a polar style-aware mask that embeds style identities as equal-norm directions within selected regions. Complementary representation–geometry conditioning combines encoded mask features with raw-mask spatial support, integrating spatiotemporal context with geometric control. To provide paired supervision, we develop an automated pipeline combining global stylization, video object removal, matting, and compositing, and construct the LociArt-100K Dataset with 100K local stylization pairs. We further establish the LociArt Benchmark, comprising 80 source videos and 84 target masks covering 168 foreground/background stylization cases, with a Style Prototype Similarity (SPS) evaluation scheme for regional style quality and complementary measures of non-target preservation and temporal consistency. Experiments show that GRNStyle outperforms existing methods in regional style quality and spatial control while maintaining strong temporal consistency. Qualitative examples further illustrate joint stylization of multiple regions with the same or different learned styles, despite training only on single-region, single-style pairs. Together, these components provide a unified foundation for learning and evaluating controllable local video stylization. Project page: this https URL.
[CV-106] FORESEE-MS: Toward Prospective Spatial Forecasting of Future Multiple Sclerosis Lesions from Baseline FLAIR MRI
链接: https://arxiv.org/abs/2610.36428
作者: Maryam Akhavan Aghdam,Sudip Mittal,Noorbakhsh Amiri Golilarz,Shahram Rahimi
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Traditional multiple sclerosis (MS) image analysis has primarily focused on retrospectively detecting and segmenting lesions. This study investigates whether baseline FLAIR MRI can identify regions at risk for future lesion formation beyond population-level spatial patterns. We introduce FORESEE-MS, an evaluation protocol that uses a fold-specific population-prior atlas as a baseline and assesses predictions using pooled voxel-level discrimination (AUROC and AUPRC) and within-patient lesion localization (Recall@Top-K%), with patient-level bootstrap confidence intervals. We evaluated five deep learning models (3D U-Net, Attn+ResBlock+VAE, ViT, and two SwinUNETR variants) using 5-fold cross-validation on 39 subjects from the MSSEG-2 dataset. The atlas achieved the highest pooled voxel-level AUROC and AUPRC, with no deep learning model reliably outperforming it on either metric. However, the 3D U-Net and Attn+ResBlock+VAE achieved higher mean voxel-level within-patient Recall@Top-K% than the atlas at all evaluated thresholds, although most paired confidence intervals for the differences included zero. These findings suggest that baseline FLAIR may contain patient-specific predictive information, but in this small cohort, it remains uncertain whether this information improves lesion forecasting beyond a population-prior atlas.
[CV-107] SO-101 Bench: Accessible Spatial Manipulation
链接: https://arxiv.org/abs/2610.36420
作者: Truman Hickok
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project page: this https URL ; code: this https URL
Abstract:Low-cost arms such as the SO-101 have become the most common entry point to robot learning, yet no released policy performs spatially constrained, language-conditioned tasks on them out of the box, and existing SO-101 benchmarks run only on physical hardware with simple tasks and few held-out objects. We introduce SO-101 Bench, a simulated benchmark for language-conditioned spatial manipulation built on an Isaac Lab digital twin of a real SO-101 setup. It has four tasks scored as five families, 138 objects in seen, unseen-instance and unseen-class splits, 916 evaluation episodes with subsets for held-out appearances, rotate-to-fit planning and occlusion memory, and 3,629 simulated demonstrations on which every learned policy is fine-tuned under an equal compute budget. The twin ranks policies as the real robot does: five policies trained without real data, four VLAs and an LLM agent, keep their exact ranking when the same weights run on the physical arm in an unseen environment with unseen objects (Spearman rho = 1.00, MMRV = 0.000), although simulation overestimates success. Across nine VLAs and world-action models, the best learned policies score near 61%, with spatial placement, unseen object classes and rotate-to-fit planning remaining open. An LLM agent that writes code and hands grasping to a learned policy scores 83.2%, and an ablation shows that this lead comes from the choice of harness and the choice of language model.
[CV-108] SpeeDiff: Scalable Pixel-Anchored End-to-End Latent Diffusion Model CVPR2026
链接: https://arxiv.org/abs/2610.36419
作者: Bingliang Zhang,Wenda Chu,Yizhuo Li,Linjie Yang,Yisong Yue,Katherine L. Bouman,Yang Song,Qiushan Guo
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to CVPR 2026. 29 pages, 17 figures, 14 tables
Abstract:We present Scalable Pixel-anchored End-to-end Diffusion (SpeeDiff), a latent diffusion method that jointly trains the VAE and the diffusion model from scratch. In principle, joint training allows the diffusion loss gradient to directly guide the VAE encoder, encouraging the formation of a generation-friendly latent space and potentially yielding faster convergence than the conventional two-stage approach with a pretrained frozen VAE. However, a naive end-to-end implementation severely degrades performance, as unrestricted backpropagation of the diffusion loss leads to latent space collapse. Our main technical contribution is a simple yet effective Tweedie Pixel Reconstruction (TPR) loss, which provides additional pixel-level feedback by decoding a predicted clean latent from an intermediate noisy state using Tweedie’s formula, thereby alleviating collapse. Furthermore, our method enables jointly scaling a fully transformer-based architecture and enhances representation alignment within the end-to-end framework. Our SpeeDiff-XL model achieves over 140x and 61x faster training compared to Vanilla SiT and REPA, respectively, while attaining an FID of 1.50 without guidance on ImageNet 256x256 generation. With a more efficient 32x compressed VAE, our model further reaches an FID of 1.53 without guidance on ImageNet 512x512 generation.
[CV-109] ORBITAL: Optimization of Robust Brain–Computer Inference Through Architecture-Aware Low-Precision Compression ISCAS2027
链接: https://arxiv.org/abs/2610.36411
作者: Saim Rehman,Muhammad Shafique
类目: Computer Vision and Pattern Recognition (cs.CV); Cryptography and Security (cs.CR); Machine Learning (cs.LG); Signal Processing (eess.SP)
备注: Submitted to IEEE ISCAS 2027 for Review
Abstract:Compact electroencephalography (EEG) decoders enable embedded brain–computer interfaces (BCIs), but compression can alter adversarial robustness without proportional hardware savings. We present ORBITAL, a robustness-constrained framework combining hardware-realizable structured compression, recovery-aware operating-point selection, multi-attack evaluation, and native edge validation. Across nine subjects and three seeds on BCI Competition IV-2a, S25-TRADES achieves the highest observed mean true-union robustness among compressed EEGNet models (23.06% vs 21.27% for standard Full), with 26.27% fewer parameters and 25.10% fewer MACs. Matched controls identify TRADES as the main source of robustness recovery, improving S25 union accuracy by 6.74 percentage points over clean recovery (p=0.0039), while per-example analysis reveals boundary redistribution rather than uniform margin expansion. On Jetson Orin Nano, FP16 compression reduces board energy by up to 16.20% and latency by 8.00%. Despite 99.3% prediction agreement, native INT8 Q/DQ incurs about twice the FP16 latency and energy, demonstrating the need for joint robustness and hardware-aware optimization.
[CV-110] Evaluating Federated Vision-Language and Vision-Only Models for Privacy-Aware Breast Cancer Imaging
链接: https://arxiv.org/abs/2610.36379
作者: Malithi Mithsara Wanniarachchi Kankanamge,Ning Yang,Sree Bash Debnath,Md Zobaer Islam,Abdur R. Shahid
类目: Computer Vision and Pattern Recognition (cs.CV); Cryptography and Security (cs.CR); Machine Learning (cs.LG)
备注:
Abstract:Artificial intelligence (AI) shows promise for breast imaging, but limited labeled data, institutional data fragmentation, and privacy concerns hinder centralized training. We investigate a privacy-aware federated learning framework combining pretrained vision-language models (VLMs) with client-level differential privacy. We compare general-purpose CLIP models, biomedical BiomedCLIP, and an ImageNet-pretrained ViT-B/16 baseline on two public breast imaging datasets, with and without differential privacy. We also evaluate a CLIP ViT-B/16 ablation that removes the text pathway. Evaluation includes patient-level ROC-AUC with bootstrap 95% confidence intervals, baseline comparisons, threshold sensitivity, inter-model agreement, and communication cost. Under federated learning, all five models achieve high normal-versus-sick discrimination (ROC-AUC: 0.940-0.990). However, acquisition year alone achieves 0.974 on the same test partition, raising concerns about confounding rather than genuine pathology recognition. Benign-versus-malignant discrimination remains uncertain (ROC-AUC: 0.504-0.637), with all confidence intervals including 0.50. Removing the text pathway preserves classification accuracy while halving communication cost. Differential privacy reduces ROC-AUC by 0.375, attributable to perturbing the full model state rather than only trainable parameters. These findings highlight the importance of evaluating confounding, predictive reliability, privacy implementation, and deployment efficiency when developing federated AI systems for breast imaging.
[CV-111] BoxWorld: Let World Engines Control and Video Generators Render
链接: https://arxiv.org/abs/2610.36371
作者: Zike Wu,Ahmed Hatem,Chintan Trivedi,Wallace Michel Pinto Lira,Xuebin Qin,Lele Wang,Renjie Liao
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:We tackle the challenge of building real-time interactive video worlds with user-defined behaviors. Such worlds must track object states, apply interaction rules, and render the resulting consequences as users act. Yet existing end-to-end video world models typically learn interaction dynamics within the generator, making it difficult to specify object-level rules and reason about their consequences. In this work, we introduce BoxWorld, which separates interaction reasoning from generative rendering. Given user-specified rules, a coding agent creates a Box World Engine that determines interaction outcomes and exposes the resulting object states through a lightweight interface of layout boxes, persistent object attributes, and dynamic status variables, leaving appearance and physical effects to generative rendering. A Box-State Generative Renderer then renders this evolving state stream into video autoregressively, enabling online control over object motion, state switches, and conditional events. Rendering this state stream is non-trivial: the renderer must respond to object-level state updates in real time while remaining coherent with the generated history. To this end, we introduce a three-stage training strategy that learns object grounding from open-domain videos and state transitions from gameplay videos, both annotated automatically, and finally performs rollout distillation on self-generated histories. We further propose PairedStateBench, a paired benchmark for fine-grained interaction control. BoxWorld substantially outperforms end-to-end world models and remains competitive across multiple public benchmarks. Project page: this https URL.
[CV-112] CrowdCC: Jointly Benchmarking Corruption Robustness and Deployment Quantization in Crowd Counting WACV2027
链接: https://arxiv.org/abs/2610.36340
作者: S Mahmudul Hasan,Lu Peng
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Accepted in WACV 2027
Abstract:Crowd-counting models deployed on edge devices face two shifts: inputs are degraded along the capture and transmission pipeline, while models are quantized to satisfy resource constraints. Yet corruption robustness and deployment quantization are typically evaluated separately. We introduce CrowdCC, a controlled corruption benchmark that jointly evaluates these effects using 23 deployment-motivated corruptions at three severity levels, four crowd-counting datasets, and 26 models using A100 TensorRT engines at FP32, FP16, and post-training INT8 precision, paired with throughput and energy measurements from corresponding Jetson Orin NX configurations. The corruptions span six families, with head annotations adjusted when geometry or visibility changes. Our evaluation shows that capture-chain corruptions consistently produce the largest degradation across the evaluated models, while crowd density affects absolute and relative error differently. Clean accuracy does not fully predict corrupted performance. INT8 alters corruption robustness substantially more than FP16, with no systematic direction, and does not consistently improve latency or energy. INT8 inference efficiency is strongly associated with reformat overhead in the compiled engine, while slowdowns are also observed under limited INT8 coverage. Precision changes corrupted-error rankings and robustness-cost frontiers, showing why model selection must consider corruption robustness and measured deployment cost together. The code is available at this https URL.
[CV-113] CRLoMTL: Conflict-Resolution through Low-rank matrices for Multi-Task Learning BMVC2026
链接: https://arxiv.org/abs/2610.36337
作者: Ahmed Agiza,Tasneem Shaffee,Rahil Badkul,Sherief Reda
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at the British Machine Vision Conference (BMVC 2026)
Abstract:Multi-task learning (MTL) improves model efficiency by sharing representations across related vision tasks, but it often suffers from negative transfer when task updates interfere with one another. Existing conflict-mitigation methods typically operate in gradient space, requiring explicit task-gradient manipulation and additional optimization steps. We propose Conflict-Resolution through Low-Rank Matrices for Multi-Task Learning (CRLoMTL), a parameter-efficient framework that mitigates task interference in the low-rank parameter-update space. CRLoMTL learns task-specific LoRA updates through task-specific passes, where each task updates its own low-rank matrices, and integrates these updates through global shared-subspace fusion. Specifically, we introduce a shared-subspace resolver based on singular value decomposition (SVD), referred to as the SVD resolver, which extracts a common low-dimensional orthonormal basis across task adapters and projects task updates into this shared space before merging them into a single deployment-time update. The resulting update can be folded into the backbone, preserving the inference structure of the original model without additional runtime branches. Experiments with Swin Transformer backbones on PASCAL-Context, NYUD-v2, and CelebA show that CRLoMTL achieves a favorable accuracy-parameter trade-off, improving over single-task baselines by up to 2.77% on PASCAL-Context, 5.03% on NYUD-v2, and 3.50% on CelebA, while training only a small fraction of the parameters used by single-task models.
[CV-114] raining-Free Inference Acceleration for Video Large Language Models : Methods Trade-offs and Evaluation
链接: https://arxiv.org/abs/2610.36327
作者: Yuqi Li,Chuanyue Yan,Neil Rodrigues,Xianglong Wang,Xiaoqin Feng,Junming Liu,Xi Xiao,Zhongwei Cheng,Wei Li,Lin Chen,Yingli Tian
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Video Large Language Models (VLMs) have demonstrated strong capabilities in video understanding and reasoning, but processing long visual sequences incurs substantial computational, latency, and memory overhead. This survey examines training-free inference acceleration, focusing on reducing inference costs while preserving task-relevant spatiotemporal evidence. We organize existing methods by their intervention stages in the inference pipeline, covering input, visual encoding, projector, prefill, and decoding. Post-training quantization (PTQ) is discussed separately as a cross-stage extension with varying calibration requirements. We analyze how intervention position affects evidence preservation and computational savings, and why local efficiency improvements may not translate into end-to-end acceleration. We complement this analysis with quantitative evidence from prior studies and our own experiments under a shared evaluation setting to assess accuracy-latency trade-offs.
[CV-115] Hierarchical Multimodal Interaction Reasoning for Dyadic Relationship Recognition
链接: https://arxiv.org/abs/2610.36249
作者: Wang Tang,Fethiye Irmak Dogan,Linbo Qing,Ce Zhu,Hatice Gunes
类目: Computer Vision and Pattern Recognition (cs.CV); Multimedia (cs.MM)
备注:
Abstract:Dyadic relationship recognition aims to identify the latent social relationship between two individuals from their multimodal interactions. Existing methods typically project heterogeneous cues into a shared representation and optimise a flat classification objective, overlooking that relationships may rely on different interaction patterns. For instance, distant relationships can often be inferred from direct behavioural correspondence, whereas close relationships are better reflected by coordinated affective, behavioural, and multimodal responses. To address this limitation, we propose HiRe-DRR, a Hierarchical Multimodal Interaction Reasoning framework that organises relationship evidence from three complementary perspectives: feature level, interaction type, and relationship hierarchy. HiRe-DRR decomposes facial, bodily, acoustic, and textual cues into low- and high-level representations and constructs homogeneous and heterogeneous cross-participant interactions. We further introduce the Mixture-of-Interaction Transformer (MoI-Former), which treats diverse interactions as specialised experts and employs relation-aware sparse routing to identify relationship-relevant evidence while preserving contextual interaction information. The proposed Hierarchical Conditional Loss (HC-Loss) exploits the hierarchical nature of social relationships through coarse-to-fine supervision, enabling structured relationship reasoning. Extensive experiments on three dyadic relationship benchmarks demonstrate that HiRe-DRR outperforms recent representative image-, video-, and multimodal methods. Our routing analysis further reveals that distant relationships tend to rely on low-level homogeneous interactions, whereas closer relationships focus on high-level heterogeneous interactions. Our code is publicly available at: this https URL.
[CV-116] Diagnosing and Correcting Bias in MLLM for Long Video Understanding
链接: https://arxiv.org/abs/2610.36210
作者: Xusheng Liang,Jianqiao Sun,Hao Zhang,Yulei Niu,Hengshuang Zhao,Jiawei Ma
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 10 pages, 6 figures. Code: this https URL
Abstract:Long-video question answering is challenging since the answer often hinges on a few decisive moments scattered throughout the video, while memory constraints force Multimodal Large Language Models (MLLMs) to sample frames at extremely low rates, resulting in sparsity that risks omitting key evidence. Though aggressively enlarging the context size is intuitive, we argue that performance remains fundamentally constrained by the bias, \emphi.e., the misleading cues caused by spurious \textitlinguistic correlations and salient yet irrelevant \textitvisual observations. In this paper, we first introduce S-MME, a benchmark designed to systematically diagnose shortcut-biased behavior in MLLMs. To mitigate this failure, we further propose a training-free framework, Counterfactual Long-video Evidence-Aware Reasoning (CLEAR). By treating the full video-question pair as a complete view, CLEAR estimates bias effects via augmented views by only keeping the question or locally salient visual clips. Then, we intervene hidden states to mitigate the discrepancy at inference such that the correct answer can be predicted confidently. With experimental studies across five benchmarks, we show that our method is generalizable and can be integrated in different MLLMs, yielding consistent gains in prediction and robustness. We hope CLEAR can offer a complementary direction in test-time scaling for long-video understanding.
[CV-117] UT-HMR: Uncertainty-Guided Temporal Modeling for Robust 3D Human Mesh Recovery WACV2027
链接: https://arxiv.org/abs/2610.36203
作者: Zhibo Wang,Sirisha Rambhatla,Yuhao Chen,Maria Koshkina,David Clausi
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at WACV 2027
Abstract:Monocular 3D human mesh recovery (HMR) remains unreliable under weak observations caused by long camera-to-person distances, occlusion, and motion blur. In challenging scenarios such as sports, these factors make the problem highly ambiguous, as different 3D poses may yield similar 2D projections. This ambiguity is especially severe for one-stage HMR, which recovers all meshes directly from the full image without person crops or external 2D cues. To handle this ambiguity, we present UT-HMR ( Uncertainty-guided Temporal HMR), to our knowledge the first one-stage HMR framework to capture correlated pose uncertainty across joints. Each person query predicts a Gaussian distribution over the full SMPL pose with a compact low-rank-plus-diagonal covariance, which is converted into joint-wise precision weights for heteroscedastic supervision while keeping inference deterministic. Temporal early fusion further injects neighboring-frame evidence before pose decoding, supplying the visual evidence needed to resolve the ambiguity that the uncertainty captures. With a lightweight ViT-B backbone, UT-HMR achieves state-of-the-art results on VIP-HockeyPose-v2 (ice hockey) and WorldPose (soccer), reducing MPJPE, PA-MPJPE, and PVE by up to 4.7%, 3.8%, and 5.6%, and remains competitive on 3DPW against methods with larger backbones and higher input resolutions.
[CV-118] Beyond the Estimate: Wiener-Informed Decoding for Easier Training of Pixel-Space Flow Models
链接: https://arxiv.org/abs/2610.36192
作者: Ziqiao Wang,Yang You
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 23 pages
Abstract:Learning velocity fields directly in pixel space is difficult and slow. We show that coarse second-order image statistics can ease this task: frozen Wiener estimates improve velocity learning, with gains extending beyond the linear predictions they explicitly provide. Moreover, we observe similar but complementary gains from frozen bases and pixel decoding. These observations motivate Wiener-informed Decoding (WiD), which uses statistical knowledge both for direct prediction and to organize decoder computation without modifying native blocks. WiD combines a trainable Wiener prediction path with coefficient interfaces that inject features into the native decoder and return its predicted coefficient corrections, with these components initialized with data statistics. All introduced parameters, including the spectrum and coordinate transforms, are trained jointly under the original pixel-space velocity objective. Across model scales and decoder architectures, WiD accelerates training and improves generation quality with modest parameter overhead. It also yields smoother gradient trajectories and remains effective with representation alignment. WiD improves class-conditional and text-to-image generation across different resolutions. On ImageNet 256 x 256, PixelDiT-XL with WiD reaches FID 1.53 after 500 epochs, achieving comparable quality to PixelDiT with 37.5% fewer training epochs. Statistical denoisers contribute beyond the estimate: their computational structure can also ease pixel-space training.
[CV-119] Bayesian Evidence Accumulation for Graph-Based Learning in Dynamic Data-Driven Applications Systems
链接: https://arxiv.org/abs/2610.36190
作者: Moises Ponce,Sarah Harkins Dayton,Ioannis Schizas,Vasileios Maroulas
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Graph Neural Networks (GNNs) are widely used in modern machine learning, yet most existing frameworks remain fundamentally static, limiting their applicability in Dynamic Data Driven Application Systems (DDDAS) where observations arrive sequentially under uncertainty. We introduce B-EAGLe (Bayesian Evidence Accumulation for Graph-Based Learning), a probabilistic framework inspired by CBGT-Net (cortico-basal ganglia-thalamic network) that reformulates graph classification as a sequential evidence accumulation process. B-EAGLe combines a GNN encoder with a Bayesian accumulator that incrementally updates posterior beliefs over class labels as new observations arrive, committing to a decision only when sufficient confidence is achieved. Experiments on CIFAR10 demonstrate competitive performance together with principled confidence estimation under sequential observations.
[CV-120] PrinGS: Privacy Preservation in Gaussian Splatting
链接: https://arxiv.org/abs/2610.36168
作者: Gaurav Behera,Abhinav Raundhal,Archisha Panda,Saumya Balina,Gargi Shroff,Ponnurangam Kumaraguru
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: this https URL
Abstract:3D Gaussian Splatting (3DGS) has rapidly become the preferred representation for high-fidelity novel view synthesis and 3D scene reconstruction. However, the learned representation encodes sensitive visual information including human faces, license plates, documents, and computer screens, raising significant privacy concerns when trained splats are shared or deployed. In this work, we address the underexplored problem of privacy preservation in Gaussian Splatting by proposing a representation-level anonymization framework that directly anonymizes a trained splat. Given a trained splat and a user-provided reference view identifying sensitive content, our method localizes the corresponding regions in the underlying 3D representation and directly anonymizes them, ensuring that the sensitive information is absent from all subsequently rendered novel views while preserving the overall scene geometry and visual fidelity. We achieve this post-training anonymization in less than 1 second. Furthermore, we extend our framework with a privacy-by-design training strategy that prevents sensitive content from being embedded into the representation during optimization, enabling the generation of privacy-preserving splats directly from training images. Our method achieves a remarkably low Face Similarity Measure (FSM) of 9.79% at a retained Face Detection Rate (FDR) of 39.47% and makes sensitive text completely unreadable with an OCR Character Error Rate of 1.00. Our work aims to serve as the primary stepping stone towards privacy-aware 3D representations and motivates further research along these directions.
[CV-121] Any3R: Feedforward 3D Reconstruction from Any Camera Type
链接: https://arxiv.org/abs/2610.36150
作者: Jingong Chen,Qingwen Zhang,Youming Deng,Chulwoo Pack,Olaf Wysocki,Boris Jutzi,Kwanghee Won,Kyle Gao
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Geometry foundation models recover camera poses and dense geometry from a set of images with a feed-forward network, but they expect pinhole images. That assumption is inside the backbone too: the positional encoding and the training loss both bind an image position to a direction in space, and both depend on the lens. We condition both on the dense ray field, the unit direction through every pixel, which describes any central camera without a lens type, distortion coefficients or a bound on field of view. Angular coordinates give the rotary encoding the measured ray angle of each patch, evaluated per view, and solid-angle weighting scales the point loss by the area each pixel covers on the sphere, so neither the encoding nor the supervision depends on the lens. Any3R is therefore a feed-forward model that takes any central camera, and is robust to mixing many camera types per scene, with no retraining per lens. On a benchmark of thirteen central cameras across three datasets, where every method reconstructs with its own predicted camera poses, it is the most consistently accurate method. Any3R enables scalable 3D reconstruction from heterogeneous crowdsourced and archival imagery collections sourced across diverse camera types. Project page: this https URL
[CV-122] Closed-Loop Aware Planning via Imitation-Free Speed Scoring for End-to-End Autonomous Driving WACV2027
链接: https://arxiv.org/abs/2610.36079
作者: Emad Aghajanzadeh,Thomas Schwarzfischer,Marc Kaufeld,Moritz Thoma,Ryan Pégoud,Manoj Rohit Vemparala,Pierpaolo Mori,Nael Fasfous,Johannes Betz,Daniel Mueller-Gritschneder,Ulf Schlichtmann
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: Accepted to WACV 2027. 17 pages (8-page main paper, references, and supplementary material), 5 figures, 7 tables. Project page: this https URL . \c{opyright} 2027 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media
Abstract:End-to-end autonomous driving has predominantly relied on open-loop imitation learning for training, which presents two limitations. First, expert demonstrations show only one behavior per scenario, so the model cannot learn alternative valid behaviors and, in turn, cannot adapt to user-specified preferences at test time. Second, noise or ambiguity in these demonstrations can produce inconsistent, unreliable supervision for the end-to-end models. Driving is inherently multi-modal: multiple different behaviors can be valid responses to the same scene, and this diversity arises largely from speed rather than path. We introduce a planning paradigm that learns the path through imitation while scoring a set of candidate speeds across multiple criteria with a learned function supervised by physically grounded safety constraints. We realize this paradigm as Closed-Loop Aware Planning for Autonomous Driving (CLAP-AD), a plug-in planning head that integrates with existing end-to-end driving models and, to our knowledge, is the first to enable explicit test-time control over their speed selection without speed-conditioned training. On Bench2Drive-Speed, CLAP-AD sets a new state of the art Speed-Adherence Score of 82.53, topping TFv6, the strongest baseline on this metric, by 12.9 points while retaining strong closed-loop driving performance. On Bench2Drive, it improves driving score by +28.5 and +32.7 over UniAD and VAD, and on NAVSIM v2, built on real-world driving data, EPDMS by +6.1 over its backbone. Project page: this https URL
[CV-123] he Trinity:Capability-Driven Recursive Self-Improvement for Multimodal Agents
链接: https://arxiv.org/abs/2610.36077
作者: Weidong Tang,Ruiheng Wang,Jingxuan Zhang,Hexin Wang,Haoyuan Chen,Zhengfeiyang Zhang,Yixing Ma,Zichen Jiang,Ziyu Guo,Haodong Li,Zhengbo Jiao,Zhiqing Cui,Xinxiang Yin,Shuo Liang,Zhenglin Wan,Hongzhan Lin,Wangbo Zhao Kaipeng Zhang,Pengfei Zhou,Odin Zhang,Yang You
类目: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Multimodal intelligence can be viewed as human-guided co-evolution between worlds and agents through interaction. However, scaling this process faces two bottlenecks: constructing heterogeneous, verifiable environments is costly, and manual Harness iteration cannot keep pace with growing task diversity and complexity. A key insight is that tasks with different modalities and surface forms can share the same underlying capability requirements. We therefore introduce The Trinity, a recursive self-improvement framework coupling world generation with Harness evolution through three roles: Creator, Explorer, and Evolver. Drawing on verified execution trajectories, Evolver shifts the basis of task organization from surface categories to shared requirements, reorganizing the original task tree into a hierarchical capability tree. We introduce Capability-Driven Real2Sim2Real: for heterogeneous tasks at the same node, Creator abstracts their common requirements into simplified, verifiable simulated worlds for capability learning, while evaluation on original tasks measures transfer. Along this capability tree, Explorer progressively evolves reusable capability modules through selective inheritance, forming a Harness tree whose validated updates accumulate across iterations within a fixed topology. The resulting Harnesses specialize in different capabilities, motivating a Capability-Aware Harness Router that uses public task information to select compatible module configurations from the tree. Our evolved Harnesses improve mean performance by 34 points across 28 training benchmarks and 26 points across 18 unseen benchmarks, including two real-world embodied tasks.
[CV-124] Learning and Reasoning with Geometry: How Generation Helps Spatial Understanding
链接: https://arxiv.org/abs/2610.36022
作者: Rui Yang,Yanwei Li,Sili Chen,Jingjia Huang,Yi Lin,Hengshuang Zhao
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: demo: this https URL
Abstract:Image and video generators can produce spatially consistent visual content, but how to use this capability to improve spatial understanding remains underexplored. We study this problem from two perspectives: geometric supervision during training and explicit geometric evidence during inference. To this end, we introduce a unified geometric understanding and generation model (GeoUG) that integrates depth and multi-view generation with spatial understanding. First, we design a multi-scale depth codec that enables metric-depth generation using a pretrained visual autoencoder, and show that joint training yields task-dependent gains in spatial understanding. Second, we introduce 3DCueBench to control geometric difficulty while balancing selected 2D cues, and find that generated depth provides larger reasoning gains on questions requiring fine-grained geometric discrimination. Building on this finding, we develop adaptive interleaved reasoning, in which GeoUG first drafts an answer and then decides whether to generate geometric evidence to confirm or revise it. We train this decision through reinforcement learning, balancing accuracy gains over retaining the draft against generation cost. GeoUG achieves circular accuracies of 97.9%, 90.8%, and 94.2% on CV-Bench-3D, 3DCueBench, and DA-2K with fewer generation calls than always generating.
[CV-125] Knowledge distillation across teacher architectures for compact diabetic retinopathy grading
链接: https://arxiv.org/abs/2610.35961
作者: Mossa Manisa,Habiba Jahan Priya,A.K.M. Fakhrul Hossain
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注: 17 pages, 3 figures, 18 tables. Submitted to PLOS ONE
Abstract:Diabetic retinopathy grading from retinal fundus photographs is important but difficult to deploy on resource-constrained hardware. We conducted a controlled study of offline response-based knowledge distillation for five-class diabetic retinopathy grading on APTOS 2019, comparing ResNet-50, MobileNetV3-Large, and Swin Transformer-Tiny (Swin-T) teachers distilled into a MobileNetV3-Small student. A 3 x 4 x 2 grid of teacher, temperature, and hard-label weight was screened at one seed, while six configurations and the no-distillation baseline were evaluated across three seeds. The highest observed single-seed quadratic weighted kappa was 0.9093 (macro-F1 0.6851) for Swin-T at temperature 4 and hard-label weight 0.7; the corresponding three-seed mean was 0.9024, compared with 0.8804 for the baseline. The student used 1.52 M parameters and 5.94 MB, representing an 18.07x parameter reduction relative to Swin-T. Temperature and hard-label weight effects were teacher- and configuration-dependent, while small performance differences were sensitive to random seed. Knowledge distillation produced a compact student with competitive performance under the evaluated protocol, but independent external validation remains necessary.
[CV-126] From Slices to Volumes: Comparing 2D 2.5D and 3D Residual Networks for Pulmonary Nodule Malignancy-Risk Classification
链接: https://arxiv.org/abs/2610.35960
作者: Bahodir Nazarov,Zineb Gharib,Fatima Zazoul
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Purpose: To compare 2D, 2.5D, and 3D residual networks for pulmonary nodule malignancy-risk classification and evaluate momentum-contrast pretraining, mild augmentation, and exclusion of indeterminate ratings. Methods: A retrospectively prepared LIDC-IDRI cohort contained 2,654 nodules labeled benign, indeterminate, or malignant using median radiologist ratings. Patient-disjoint splits comprised 1,876 training, 382 validation, and 396 test nodules. Representations included one axial slice, five neighboring slices, and a local CT volume. Each track evaluated a randomly initialized baseline, momentum-contrast pretraining followed by fine-tuning, and supervised training with mild augmentation. Checkpoints were selected by validation Macro-F1. The 2D experiments used three seeds; 2.5D and 3D results were single-run estimates. A separate binary experiment excluded indeterminate nodules. Results: Augmented models achieved test Macro-F1 values of 0.6134\pm0.0075 for 2D, 0.6537 for 2.5D, and 0.5403 for 3D; the 2D uncertainty denotes population standard deviation across seeds. Corresponding accuracies were 0.6128 , 0.6869 , and 0.5455 . Augmentation improved Macro-F1 within every track, whereas the tested pretraining recipes reduced it. Binary Macro-F1 values were 0.8227 , 0.8664 , and 0.7554 , respectively. Conclusion: Augmented 2.5D achieved the highest observed performance. Differences in preprocessing, optimization, capacity, and replication limit attribution to dimensionality. Labels reflect radiologist-assessed malignancy risk rather than pathology-confirmed cancer. Code and experimental results are available at this https URL. Subjects: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV) Cite as: arXiv:2610.35960 [cs.LG] (or arXiv:2610.35960v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2610.35960 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Bahodir Nazarov [view email] [v1] Sun, 25 Oct 2026 10:03:36 UTC (87 KB)
[CV-127] ask-Utility IQA: Rethinking Image Quality Assessment in Embodied Tasks
链接: https://arxiv.org/abs/2610.35905
作者: Jirong Zha,Yanwei Liao,Yemin Wang,Xiangmin Yi,Siqi Peng,Yingfeng Chen,Chen Gao,Xinlei Chen
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Image quality assessment (IQA) underpins embodied imaging pipelines by judging whether visual quality satisfies downstream tasks, yet most existing methods learn task-agnostic scores aligned with generic human ratings on static benchmarks. This objective mismatches embodied and interactive settings, where image adequacy depends on task goals, context, and action requirements that shape an agent’s decisions. We argue that IQA should shift from score regression to goal-conditioned judgment defined by embodied task utility. To this end, we propose an MLLM-based embodied IQA agent framework that reasons about task requirements, grounds quality evidence, and produces decision-facing assessments. A compact VisDrone evidence study shows that conventional perceptual metrics and generic MLLM-based IQA scores are weak proxies for downstream VLM task success, while a task-conditioned utility score better predicts whether visual input supports the target task. These findings support a task-utility-oriented reframing of IQA for embodied, explainable, and decision-aware visual perception.
[CV-128] ObjectTwin: Measuring the Asset Correspondence Gap for Capture-Anchored Digital Twins
链接: https://arxiv.org/abs/2610.35904
作者: Zhenghe Guo
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Object-wise reconstruction increasingly inserts generated meshes into captured scenes as editable proxies for observed objects. This use creates a requirement absent from isolated-asset scores and pose estimates: at the captured metric state, the proxy should reproduce calibrated observations of the same instance. We present ObjectTwin, a scan-grounded measurement protocol that freezes the instance, cameras, scan-projected support and recorded depth, metric anchor, and evaluation denominator, then replaces only the object representation. The resulting asset correspondence gap measures lost observation agreement relative to scan-derived visible geometry without treating the scan as complete physical ground truth. Across two indoor RGB-D capture domains and two generation backends, visible, anchored generated assets lose 0.302-0.497 MaskIoU under this replacement. Every tested scene-level estimate is positive, while same-input generator preference reverses across capture domains. Coverage and overlap-depth analyses locate disagreement in both visible support and surface placement; the evaluated conservative pose refinement does not systematically recover it. These results establish capture correspondence as a separately measurable interface between generated-asset quality and stronger physical or task-level evaluation.
[CV-129] Real-Time Smart Retail Analytics through Fog Computing using the RetailLightNet Lightweight Model
链接: https://arxiv.org/abs/2610.35871
作者: Mayank Atreya,Navin Chhibber,Shilpa Ajay,R Loshini,K Anitha
类目: Networking and Internet Architecture (cs.NI); Computer Vision and Pattern Recognition (cs.CV)
备注: 2025 5th International Conference on Mobile Networks and Wireless Communications (ICMNWC), Conference Location: Tumkur, India
Abstract:Smart retail systems now depend on the application of real-time video analytics, which can track customer behavior, manage shelf inventory, and enhance in-store processes. Currently cloud-based applications of video analytics face multiple challenges such as high latency of the communication itself, privacy concerns surrounding transmitting raw video, and the computational demands of the heavy vision models used on the often limited network infrastructure. These challenges lead to delays in decision making where prompt action is crucial. To address these short comings, this research proposes Fog-based analytics withRetailLightNet (F RLN), an ultra lightweight deep learning model which is specifically suited for the edge, resource constrained hardware capabilities of devices like the NVIDIA Jetson Nano. The approach utilizes publicly available retail datasets, SKU110K, and RPC, which contains an assortment of product, shelf, and customer interaction scenarios for this http URL processes include preprocessing images through normalization and employing motion-gated frame selection to improve efficiency by omitting redundant frames to ease the workload on the fog processing. Internal within Retail LightNet, image feature extraction is realized through depth wise separable convolutions which help support spatial and channel-wise learning without heavy computational costs. The model employs two prediction heads presence classification, and density regression, supporting shelf monitoring and queue assessment. The proposed F RLN framework attained better results in terms of precision (0.988), recall (0.976), and F1-score(0.976) when compared with existing convolutional neural network. Keywords: fog computing, motion-gated frame selection,NVIDIA jetson, retail lightnet, smart retail.
[CV-130] Retrospective macromolecular pool fraction mapping from conventional clinical MRI: a physics-guided self-supervised deep learning framework
链接: https://arxiv.org/abs/2610.37213
作者: Jelmer van Lune,Miha Fuderer,Oscar van der Heide,Matteo Maspero,Martin B. Schilder,Jan Willem Dankbaar,Stefano Mandija,Alessandro Sbrizzi
类目: Medical Physics (physics.med-ph); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 38 pages, 24 figures
Abstract:Macromolecular pool fraction (MPF) is a quantitative imaging biomarker sensitive to myelin integrity and other changes in tissue macromolecular content, but specialized acquisitions and reconstruction methods limit its use in large-scale studies. We present a physics-guided, self-supervised deep learning framework for retrospectively estimating MPF directly from conventional clinical brain MRI ( T_1 -weighted, T_2 -weighted, and fluid-attenuated inversion recovery [FLAIR]) without paired quantitative reference scans. The framework embeds an analytical two-pool magnetization transfer signal model that related sequence-averaged radiofrequency transmit power to apparent longitudinal relaxation. To accommodate heterogeneous clinical protocols, the network conditions intermediate features on acquisition parameters via adaptive instance normalization. It was trained and evaluated on a multisite, multivendor, multiprotocol dataset of 7,705 MRI sessions (3 T). Across 1,237 test sessions, mean MPF was 0.182 \pm 0.007 in white matter and 0.144 \pm 0.009 in gray matter, with limited variation across vendor-by-protocol groups (between-group coefficient of variation: 2.65% and 5.63%, respectively). In 70 sessions with demyelinating-disease labels, MPF was 35.6% lower in segmented lesions than in normal-appearing white matter. MPF maps from two different clinical protocols acquired in a healthy volunteer showed a mean voxel-wise absolute percentage difference of 7.7% and a concordance correlation coefficient of 0.888 , indicating good cross-protocol consistency. These results suggest that magnetization transfer-informed self-supervised learning enables stable MPF mapping from conventional clinical MRI, providing a scalable approach to investigate demyelinating pathology in both existing imaging archives and prospectively acquired conventional clinical scans without dedicated qMT acquisitions.
[CV-131] Quantized Edge Inference of a Self-Supervised SAR-Optical Encoder on Commodity x86 Hardware
链接: https://arxiv.org/abs/2610.37056
作者: Christian Sanchez-Leon,Dongming Chen
类目: Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 3 pages, 1 figure, 3 tables
Abstract:Self-supervised vision transformers are increasingly used as reusable encoders for Earth observation, but their inference cost on the hardware available in the field is rarely characterized. We deploy a pretrained SAR-optical masked-autoencoder encoder on three commodity x86 nodes, from a low-power CPU without AVX2 to an AVX2+VNNI reference, at FP32, FP16 and INT8 on the CPU and, where present, the integrated GPU, and report latency, energy (RAPL and wall plug) and linear-probe accuracy. Quantization is accuracy-neutral: INT8 changes micro-mAP by only +0.0006, within the seed spread, so precision becomes a pure cost decision. INT8 lowers CPU latency on every node, by 1.45x even on the CPU without AVX2, and CPU energy wherever it is measured, and the VNNI reference (1.53x) suggests a further gain beyond AVX2. The Pareto-optimal device shifts from the integrated GPU, which is 4.38x faster than the CPU on the weakest node, to the CPU as the CPU class improves.
[CV-132] Image Modeling: From Handcrafted Priors to Foundation Image Models
链接: https://arxiv.org/abs/2610.36646
作者: A. Murat Tekalp
类目: Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:This paper unifies six decades of image modeling, evolving from handcrafted mathematical tractable priors to solve inverse problems in imaging to data-driven deep foundation image models for both image generation and solving inverse problems, under a common Bayesian framework that formulates image modeling as an implicit or explicit probabilistic characterization of a data manifold. Every image restoration or generative task fundamentally seeks to answer the same question: What makes an image look like a real image? In the Bayesian framework, this question is encapsulated by the prior probability density p(x), or its more optimization-friendly counterpart, the log-prior log p(x). While early works in solving inverse problems, such as Tikhonov, Total Variation (TV), and Non-local priors relied on explicit handcrafted definitions of log prior to model local smoothness, modern techniques, such as Plug-and-Play (PnP) priors and Denoising Diffusion models, are rooted in algorithmic uncoupling and implicit manifold learning that apply to both inverse problems and conditional image generation. By grounding imaging inverse problems and conditional image generation under a single mathematical engine: the gradient of the log-prior, or the Stein Score Function, we bridge the gap between signal processing and machine learning communities, turning what seems like parallel developments into a single, evolution of image modeling from handcrafted analytical models to models learned from big data, demonstrating that the pursuit of the denoising score is the common thread connecting variational optimization (in MAP estimation) to posterior sampling (in modern foundation image models).
[CV-133] When Slides Met Scans: Cross-Attention for Renal Cancer Prognosis
链接: https://arxiv.org/abs/2610.36364
作者: Anusha Krishnamoorthy,Basudha Pal
类目: Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Histopathology and computed tomography (CT) provide complementary views of renal cancer. We connect these modalities through bidirectional cross-attention between 16 histology prototypes and up to 32 kidney-centered CT tokens from frozen foundation models. In 210 TCGA-KIRC patients with 46 disease-specific survival (DSS) events, 5-fold evaluation yielded a C-index of 0.760\pm0.081 and 3-year AUROC of 0.773 (95% CI 0.678–0.860), the highest averages among evaluated unimodal and fusion baselines. External testing in 180 CPTAC-CCRCC patients yielded DSS C-index of 0.629 with CT withheld. Attribution analysis linked predictions to image regions while exposing acquisition artifacts. These findings support further study of cross-scale fusion and independent multimodal validation.
人工智能
[AI-0] Attention as Lattice Gauge Theory: Relations Interact Only Through Wilson Loops NEURIPS2026
链接: https://arxiv.org/abs/2610.37656
作者: Aakash Sarkar,Ananya Mukherjee,Motoshige Sato
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computational Physics (physics.comp-ph)
备注: 7 pages, 1 figure. Accepted at the 9th Workshop on Machine Learning and the Physical Sciences (ML4PS 2026), co-located with NeurIPS 2026
Abstract:Relations compose through the entity they share - Alice is Bob’s mother and Bob is Carol’s father, so Alice is Carol’s grandmother - and the order matters, since a father’s mother is not a mother’s father; verifying a composed relation means closing the loop. We show that if each entity’s coordinate frame may be rotated independently (per-entity frame freedom), three things follow. In the pure-gauge sector, relations that share no cycle cannot interact at all. The only gauge-invariant score of a triangle is a class function of its holonomy H = T_ij T_jk T_ki - a Wilson loop. And the only equivariant update of an edge is U = \Psi(H), T_ij with \Psi conjugation-equivariant. Edge attention by Wilson loops is therefore not one design among many but the unique form the symmetry permits. Three communities hold pieces of this form: the commuting-operator embeddings of knowledge graphs are its abelian (commuting) sector; AlphaFold’s triangle update and the Edge Transformer have its shape - an edge updated through the third vertex - without its constraints; and lattice QCD builds it equivariantly on lattices. We instantiate the equivariant sub-family on general relational graphs as holonomy attention. With no training and composition alone, it infers unobserved S_4 relations on sparse graphs at 0.88 with a fifth of the edges corrupted, at or slightly below specialist synchronization algorithms, where every commutative architecture is capped at 0.083; its holonomy score flags corrupted triangles at AUC 0.99 where commuting scores are capped at 0.76 by a second selection rule, and it separates clusters without labels.
[AI-1] KPPax: preprocess differentiable kinetic ODE for machine learning in atmospheric chemistry
链接: https://arxiv.org/abs/2610.37655
作者: Wenqing Peng,Michael Boy,Zhi-Song Liu
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Atmospheric chemical mechanisms are commonly distributed in the language of the Kinetic PreProcessor (KPP), which generates efficient Fortran or C code but does not fit into the differentiable programming frameworks in which machine-learning models are developed. We present KPPax, a kinetic preprocessor that generates differentiable, just-in-time compiled, and vectorisable chemistry code in JAX from mechanisms written in the KPP language. KPPax evaluates the rate law with sparse operations in log space, which keeps it differentiable, and integrates the stiff system with the implicit solvers and the forward- and reverse-mode adjoints of the \textttdiffrax library, so that exact sensitivities are obtained without hand-derived adjoint code. We demonstrate two applications with the POLLU benchmark mechanism. In the first, the Fisher information matrix computed from an automatically differentiated Jacobian predicts, before any fit, which of the 25 rate coefficients a flow-tube-like design can identify, and noisy synthetic estimation confirms the predicted uncertainties of the nine identifiable rate coefficients to within about 30% while revealing clusters of fast reactions of which only ratios are constrained. In the second, a neural-network correction of an emission inventory is trained offline on per-scenario 4D-Var retrievals and then fine-tuned online through the chemical model, which reduces the error of the biogenic correction by a factor of 2.4–5 and recovers information on low-temperature emissions that the individual retrievals lose. KPPax makes KPP mechanisms directly usable for parameter estimation and hybrid physics–machine-learning modelling.
[AI-2] MuonGS: Pretraining LLM s with Relaxed Weight-Norm Control
链接: https://arxiv.org/abs/2610.37647
作者: Athanasios Glentis,Dawei Li,Mingyi Hong
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Optimization and Control (math.OC)
备注: 21 pages, 9 figures
Abstract:Weight decay and explicit weight-norm constraints improve optimizer performance in language model pretraining, yet a unified understanding of why such control helps and how much is needed remains incomplete. We analyze norm control in Muon over the entire training trajectory using the \emphaccumulated effective learning rate, C_T=\sum_tT\eta_t/|W_t|_F , where \eta_t is the nominal learning rate and \eta_t/|W_t|_F is the effective learning rate relative to the weight norm. For bounded update directions, accumulation over any training interval bounds relative weight displacement. We show that vanilla Muon can accumulate only logarithmically, potentially limiting exploration and helping explain its slower empirical progress, whereas MuonW and MuonH guarantee at least linear accumulation under the stated conditions. Going beyond bounded-norm control, we show that sufficiently slow norm growth guarantees faster accumulation than the logarithmic worst case, opening up less restrictive optimizer designs. Based on this insight, we propose MuonGS, which uses Gram–Schmidt (GS) projection to remove the radial component of each Muon update without fixing the weight norm. With bounded update directions and a constant nominal learning rate, we prove that the weight norm grows at most proportionally to \sqrtt after t steps. Consequently, MuonGS guarantees C_T=\Omega(\sqrtT) , with accumulation sustained even in later training. Further, for smooth scale-invariant objectives we establish its O(T^-1/2) stationarity guarantee. In pretraining experiments with 160M- and 400M-parameter dense models as well as a 135M active- and 1B total- parameter MoE model, MuonGS achieves lower validation perplexity than MuonH at long training horizons. In particular, for the dense model it achieves an estimated token-equivalent speedup up to about 15% at matched validation loss to MuonH. Comments: 21 pages, 9 figures Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Optimization and Control (math.OC) Cite as: arXiv:2610.37647 [cs.LG] (or arXiv:2610.37647v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2610.37647 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-3] What Improves During Pretraining? Tracking Knowledge Acquisition with Point-in-Time Models AACL2026
链接: https://arxiv.org/abs/2610.37631
作者: Chang Su,Bin Wu,Emine Yilmaz,Yao Lu
类目: Artificial Intelligence (cs.AI)
备注: Accepted to AACL 2026. This version corrects minor errors in the appendix
Abstract:Language models pretrained on web corpora collected over multiple years learn both general and domain-specific knowledge. How these different kinds of knowledge are learned as pretraining progresses remains unclear. Recent point-in-time models trained on web snapshots from different years provide a way to probe how knowledge prevalence varies over time. Using these models, we propose a temporal variability score (TVS), defined as the coefficient of variation of an item’s negative log-likelihood across models. A higher score suggests that the knowledge may be less consistently represented across years of web text. We evaluate six benchmarks covering diverse topics using six models from multiple open-source model families. We find that low- and high-TVS items are learned similarly early in training, but high-TVS items improve faster later on. At the end of the pretraining run, high-TVS items outperform low-TVS items in 34 of 36 model-benchmark settings, with an average advantage of 7.3 points of chance-normalized accuracy. This pattern is most robust on knowledge-intensive tasks, even when comparing items of similar difficulty. Our findings reveal differences in how knowledge is learned over the course of pretraining, with temporal variability providing a diagnostic of these learning dynamics beyond aggregate benchmark scores.
[AI-4] Beyond Answer Changes: Matched Same-Class Controls for Local Audio Evidence ICASSP2027
链接: https://arxiv.org/abs/2610.37605
作者: Yunuo Wang,Maoyuan Chen,Yuanjin Huang,Xing Liu
类目: ound (cs.SD); Artificial Intelligence (cs.AI)
备注: Under review at ICASSP 2027. Code and frozen evidence: this https URL
Abstract:For audio-language models, an answer change after a local sound-event edit can reflect a response to the event’s class, sensitivity to waveform rewriting, or both. We compare matched same-class and different-class replacements of a queried event. Both use the same interval, level estimate, donor fitting, and crossfade, while the same-class replacement preserves the correct answer. Across 600 synthetic soundscapes and three models from two architecture families, different-class replacements produce 49.7 to 60.9 percentage points more answer changes than same-class replacements. Same-class accuracy remains within 5.2 percentage points of original accuracy. Among trials with correct original and same-class answers, soundscape-weighted selective accuracy for the inserted class is 73.7 to 81.9%, averaging eligible trials within each soundscape and then weighting eligible soundscapes equally. This protocol tests how model answers respond to local changes in event class.
[AI-5] GAS-PFN: Graph-Aware Spatiotemporal Prior-Data Fitted Networks for Zero-Shot Traffic Forecasting
链接: https://arxiv.org/abs/2610.37598
作者: Joseph Natter,Long Vu,Yifan Zhang,Bobin Deng,Kun Suo,Rui Wu
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Neural and Evolutionary Computing (cs.NE)
备注:
Abstract:Traffic forecasting requires modeling complex temporal dynamics and the spatial dependencies imposed by road topology. Supervised graph-based models achieve strong performance, but are often trained for a particular dataset and sensor graph, making deployment to new or evolving networks costly. Prior-Data Fitted Networks (PFNs) offer a zero-shot alternative by learning in-context inference from synthetic task priors, which provide an explicit mechanism for encoding domain-specific assumptions. Existing forecasting PFNs, however, do not explicitly represent the graph-structured dependencies governing traffic dynamics. To combine graph-aware traffic modeling with zero-shot adaptation, we introduce the Graph-Aware Spatiotemporal Prior-Data Fitted Network (GAS-PFN). GAS-PFN incorporates sampled road topology into its synthetic task prior through random-walk diffusion, teaching the model to learn graph-coupled dynamics during pretraining. Its co-designed inference architecture uses two mask-conditioned passes through a shared Transformer backbone to separately model sensor-specific temporal dynamics and spatiotemporal interactions among neighboring sensors before integrating their representations for prediction. This design enables GAS-PFN to use target-network structure and context without parameter updates. Across six benchmarks at the 90-minute horizon, GAS-PFN reduces MSE by 34.0% on average relative to TimePFN, its closest PFN-based predecessor. It also reduces MSE by 27.7% on average relative to four recent industry-developed general-purpose zero-shot forecasters. In this comparison, all models receive 90 minutes of contiguous recent history, and any additional context is drawn from each dataset’s chronological training split. These results indicate that aligning PFN priors and inference architectures with graph structure improves in-context learning for zero-shot traffic forecasting.
[AI-6] Understanding the limits of latent world models
链接: https://arxiv.org/abs/2610.37588
作者: Rodrigo Gonzalez Laiz,Steffen Schneider
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:World models are widely proposed as core components for enabling agents to reason over long temporal and spatial horizons. Yet current evaluations often emphasize task performance while providing limited insight into whether learned world models acquire predictive latent dynamics that reflect the underlying structure of the environment. Here we introduce the concept of faithful latent world models as an equivariance condition between the environment’s transition map and the learned latent dynamics. We propose latent rollout consistency (LRC) as a metric to evaluate action-conditioned latent prediction. Strikingly, we find that across the studied state of the art latent world models, none solves the task by actually faithfully modeling the dynamics of the environment.
[AI-7] mestral
链接: https://arxiv.org/abs/2610.37585
作者: Adrien Petralia,Alexandre Gavaudan,Artjom Joosen,Etienne Goffinet,Richard Kurle,Yassine El Ouahidi,Yassir Bendou,Abdelaziz Bounhar,Abhijeet Somani,Aditi Kabra,Adrian Valente,Adrien Sadé,Ahmed Djermani,Alan Jeffares,Albert Jiang,Alberto Bietti,Aleksandr Timashov,Alexandre Cahill,Alexandre Laval,Alexandre Sablayrolles,Alexandros Sanchez Vassopoulos,Amélie Héliou,Amos You,André Jonasson,Andreas Fürst,Andrew Bai,Andrew Ehrenberg,Andrew Zhao,Andy Lo,Angele Lenglemetz,Anmol Agarwal,Antonia Calvi,Arata Suzuki,Arjun Majumdar,Arthur Fournier,Ashwani Kumar,Augustin Lombard,Avinash Sooriyarachchi,Aylin Guliz Akkus,Aysenur Karaduman,Baptiste Bout,Baptiste Rozière,Barna Pásztor,Baudouin De Monicault,Benjamin Holzschuh,Benjamin Tibi,Benno Stein,Bernhard Stadlbauer,Bhargavi Bhaskar,Binbing Hou,Błażej Osiński,Boris van Breugel,Camille Le Scao,Chaoran Yu,Charlotte Cronjäger,Chen-Yo Sun,Chris Bamford,Christian Wallenwein,Christophe Renaudin,Clémence Lanfranchi,Colin Bonatti,Corentin Barreau,Corentin Sautier,Cristiana-Diana Diaconu,Cyprien Courtot,Daniel Lopes,Daniel Marczak,Darius Dabert,Diego de las Casas,Dominik Nuss,Dylan Rubini,Dzmitry Soupel,Elizaveta Demyanenko,Elliot Chane-Sane,Emilien Fugier,Emmanuel Gottlob,Erik Aas,Etienne Duchesne,Fabian Paischer,Fabian Schlager,Faruk Ahmed,Federico Baldassarre,Félix Rosseeuw,Filip Szatkowski,Filippo Vicentini,Florian Wiesner,Gabrielle Berrada,Gaëtan Ecrepont,Gaspard Blanchet,Gaspard Donada-Vidal,Gaurav Mahajan,Gauthier Guinet,Genevieve Hayes,Georgii Novikov,Georgios Christoglou,Giada Pistilli,Gonzague de Carpentier,Graeme Nail,Guillaume Breton,Guillaume Kunsch
类目: Artificial Intelligence (cs.AI)
备注: Webpage: this https URL ; Weights: this https URL ; Code: this https URL
Abstract:We introduce Timestral, a family of four time series foundation models for probabilistic forecasting, with 5M, 28M, 123M, and 209M parameters. Timestral sets a new state of the art on GIFT-Eval, TIME, and fev-bench, where Timestral-209M and Timestral-123M take the top two places on every metric. We achieve this with three ingredients: Decoupled Horizon Forecasting (DHF), which separates prediction tokens from the causally encoded context, so that longer horizons are trained without masking context and each observed value is both a forecast target and context for later forecasts; a covariate framework that natively supports known-future covariates within a causal architecture; and a two-stage pipeline that pretrains exclusively on a diverse, large-scale synthetic corpus before post-training on a mix of real and synthetic data with extended context. This report describes the model design and its benchmark results. Weights for all four models are available on Hugging Face and inference code on GitHub.
[AI-8] amGym: A Gym for Team Decision-Making by AI Agents with Distributed Information
链接: https://arxiv.org/abs/2610.37571
作者: Chanwoo Park,Jaywoong Jeong,Zachary Wojtowicz,Yubin Kim,Taekyun Lee,Kaiqing Zhang,Asuman Ozdaglar
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Many real-world decisions require teams to integrate information distributed across participants. Yet AI agents are still primarily trained and benchmarked in isolation. We introduce TeamGym, an environment for evaluating and training language model-based agents on team decision-making problems constructed from real-world legal, finance, medical, and software engineering corpora. TeamGym introduces a method to identify a minimal set of facts, each element of which is necessary and all of which are collectively sufficient for a correct final decision, which we call pivotal information. This enables TeamGym to both control the distribution of pivotal information and track how well agents identify and integrate pivotal information when making team decisions. Examining 8 language models on 12 source corpora, we find that current AI agents often fail to effectively share information: Distributing pivotal information across participants lowers success rate by 22.9 percentage points on average relative to giving the same information to a single model (a communication gap). We also document that participants do not effectively retrieve decision-relevant information on their own: A team’s success rate is lowered by a further 19.9 points when participants must identify and extract pivotal information from source material rather than receiving it directly (a retrieval gap). To help remedy these deficits, we use TeamGym to train policies through reinforcement learning on team decision loss. Across two policy-training experiments, we find that training improves final-decision quality and induces emergent coordination in participation, speaker selection, and communication, even though the reward evaluates only the final decision.
[AI-9] Gym-Anything-RT: Scaling Real-Time Environments for Vision-Language Agents
链接: https://arxiv.org/abs/2610.37567
作者: Pranjal Aggarwal,Sean Welleck
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:AI agents have made substantial progress in domains where they have a non-trivial amount of time to deliberate between actions, such as mathematics, software engineering, and performing tasks on desktop applications. It is less clear how such models fare in settings that demand real-time interaction, such as playing a video game or operating a physical device. In these settings, the environment state may evolve as the agent is reasoning, or even as the underlying model is performing a forward pass. To gain insight into the real-time interaction capabilities of vision-language agents, we introduce Gym-Anything-RT. Our key idea is to construct visual minigames that demand a portfolio of capabilities required for interactive real-time settings. Gym-Anything-RT uses an agentic pipeline to automatically generate such environments based on a taxonomy of six capabilities required for real-time interaction, including the ability to perceive the environment and act under an evolving environment state. A key feature of Gym-Anything-RT is that it allows for controlled comparison of real-time interaction capabilities: we can run the same minigame in real time, or pause or slow down the environment to relax the requirement for real-time actions. We use Gym-Anything-RT to construct 100 real-time minigames and evaluate AI agents on them. We find large capability gaps in real-time interaction. For example, GPT-6 Astra (xHigh) achieves a pass rate of only 9.0% on the real-time minigames, which rises to 75.0% when the environment is paused. We also identify intriguing behaviors by the most powerful frontier agents: for instance, when given the ability to write code, these agents write code that bypasses the need to perceive and act in real time. We publicly release Gym-Anything-RT and the generated minigames to spur further research on AI agents in real-time interactive settings.
[AI-10] When uncertainty quantification fails: structural vulnerabilities and corrections for selective prediction under clinical class imbalance ALT
链接: https://arxiv.org/abs/2610.37561
作者: Walter Endrizzi,Flavio Ragni,Raffaele Marchesi,Stefano Bovo,Monica Moroni,Giuseppe Jurman,Venet Osmani
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: this https URL
Abstract:Background. Selective prediction systems improve the safety of clinical artificial intelligence by deferring ambiguous cases to human experts while automating high-confidence decisions. However, this safety mechanism relies entirely on the premise that uncertainty scores accurately reflect clinical ambiguity. In prognostic tasks characterized by severe class imbalance, where the clinical outcome of interest is a minority, this assumption breaks down. Standard uncertainty measures cease to capture clinically aligned ambiguity; instead, they become a proxy for class membership, systematically penalizing minority patients regardless of their individual predictive clarity. Results. Across 56 synthetic conditions and three neurodegenerative cohorts spanning severe class imbalance to near-balanced conditions (multiple sclerosis, Parkinson’s Disease, Alzheimer’s Disease), our framework consistently maintained bounded, balanced risk across trajectories. In contrast, global metrics either collapsed minority Sensitivity or protected the minority at the expense of majority patients. Applying CCR with standard entropy fails to protect the minority because it breaks within-class monotonicity when tau=0.5. Combining CCR with threshold-aware metrics restores strict monotonicity, correctly identifies ambiguous patients, and safely mitigates minority exclusion. Conclusions. Under clinical class imbalance, standard uncertainty quantification measures dataset sparsity rather than clinical ambiguity, creating a systemic illusion of safety. This bias is a correctable structural flaw. Clinical pipelines deploying selective prediction under imbalance should adopt threshold-aligned uncertainty measures and class-stratified evaluation. This reduces class-specific risk imbalance, supporting safer selective prediction under the evaluated conditions without silently excluding the most vulnerable patients. Comments: this https URL Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI) Cite as: arXiv:2610.37561 [cs.LG] (or arXiv:2610.37561v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2610.37561 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-11] Kinematic Constraint-Referenced Evaluation for Articulated Object Manipulation
链接: https://arxiv.org/abs/2610.37550
作者: Baptiste Bonin,Caro Strickland,Audrey Durand
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注:
Abstract:Manipulation policies are evaluated predominantly by binary task success, which cannot tell whether the commanded gesture is the one the object’s kinematics admit. We introduce Constraint-Referenced Strategy (CRS), which projects the commanded end-effector twist onto the object’s joint axis, separating progress along the joint from off-constraint waste. CRS requires a joint model and a joint-following scripted controller derived from it, but no recorded demonstration or learned judge. On 74 simulated articulated objects, success registers three transfer failures as lower rates but attributes none, while the two CRS channels give each a distinct signature. A door-trained behavior-cloning policy opens drawers in 59 of 110 episodes, and in these passes it commands about 40\times the off-axis rotation of a drawer-trained policy’s passes ( 1.27 vs.\ 0.032 ~rad/s). Relative to a scripted controller scored with the same axis, CRS verdicts are unchanged under joint-axis errors up to 30^\circ ; absolute thresholds tolerate about 8^\circ .
[AI-12] When Where and What to Correct: Human-Guided Adaptive Residual Correction for Robot Policies
链接: https://arxiv.org/abs/2610.37547
作者: Anukriti Singh,Maxime Caniot,Pratap Tokekar,Firas Abi-Farraj
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注:
Abstract:Imitation policies that reliably execute most of a manipulation trajectory can still fail in localized regions requiring last-mile precision. Updating the full policy or exploring residual actions everywhere treats this local failure as a global learning problem. We introduce Adaptive Residual Correction (ARC), a lightweight post-training module that uses reinforcement learning (RL) to learn a residual correction on top of a frozen policy while using sparse human corrections to structure autonomous exploration. Rather than treating human corrections as imitation targets, ARC uses pre-intervention states to determine ‘when’ residual control should become active and correction geometry to determine ‘where’ in residual action space autonomous refinement should explore; task reward determines ‘what’ residual is ultimately executed. Across three precision-manipulation tasks on a real robot, ARC improves average task success from 33.3% to 75.0%. The human corrections define task-dependent 2-3 dimensional subspaces of the seven-joint residual space that retain 96.0-97.7% of squared correction magnitude. Together, these results show that sparse human corrections can localize and constrain autonomous post-training without requiring the observed correction itself to serve as the final action target.
[AI-13] Which Hand Was Late? Diagnosis for Robot Failure Recovery in Bimanual Tool Use
链接: https://arxiv.org/abs/2610.37533
作者: Baptiste Bonin,Caro Strickland,Audrey Durand
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注:
Abstract:Failure diagnosers for robot manipulation are increasingly vision–language models (VLMs), trained and evaluated on single-arm failures produced by perturbing a successful demonstration. Bimanual tool use adds failures of other kinds: a tool that slips out of its socket while still held, two correct gestures made at different times, and attempts that fail for two reasons at once. On a simulated bimanual tool-use task whose failure causes are read from the recorded trace, we measure what each sensing channel lets a diagnoser tell. The tool’s slip is read on the tool hand; which hand was late is read on both arms, while a sensor on one hand reads its own hand’s delay and at most part of the other’s; and whether a late attempt was also pushed too weakly is not identifiable from passive sensing when the tool hand is late. Within two attempts, a repair policy that escalates when the cause is not identifiable repairs as many failures as one told the true cause. A VLM’s internal state carries the late hand almost perfectly, along a direction no closer to the one its answer reads than a random one; whether the answer reports it depends on the wording of the question.
[AI-14] he Structure Gap: When More Data Cannot Replace Structured Reasoning
链接: https://arxiv.org/abs/2610.37523
作者: Kheer Sagar Patel,Mojtaba Soltanalian
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 13 pages, 6 figures, 2 tables. Includes supplementary material and reproducibility code/data
Abstract:A persistent belief in modern AI is that sufficiently large learners trained on sufficiently much data can make explicit domain knowledge optional. This paper examines a limitation of that belief. We define structure-encoded problems: tasks whose input-output map is determined, or sharply constrained, by known structure such as a solver, symmetry group, graph, physical law, measurement operator, grammar, combinatorial search space, or optimization objective. For such tasks, the central question is not expressivity–a dense neural network may contain the desired map–but finite-data identification, extrapolation, and certification. We introduce the structure gap: the finite-resource gap between a method that receives the governing structure and a learner that receives only examples. The strongest useful form of the claim is asymptotic rather than absolute: structure-agnostic learning may approach a structured solver as coverage grows, but every finite dataset can leave residual uncertainty that the structure resolves by construction. We prove this exactly for a Gaussian inverse problem: the known-operator Wiener solver attains Bayes risk, while the dense empirical learner in the correct linear class has persistent expected excess risk. We also prove a cell-coverage lower bound and an unobserved-orbit non-identifiability result. Controlled experiments in known-operator inversion, sparse recovery, and graph reachability with size extrapolation instantiate the mechanisms. When validated structure is available, it can constrain the hypothesis space before learning begins, leaving learning to address components the structure does not specify.
[AI-15] Orthogonal and Parametrized Multi-Attribute Steering of LLM s
链接: https://arxiv.org/abs/2610.37517
作者: Remy Ogasawara,Cullen Anderson,Narmeen Fatimah Oozeer,Foad Namjoo,Amir Abdullah,Jens Kristian Refsgaard Schou,Jeff M. Phillips
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Activation steering offers an effective and efficient way to modify an LLM’s behavior at inference time. However, it remains unclear how to apply steering such that the signal is consistently strong and the steering strength has a universally interpretable meaning. We address this first by identifying the most steerable interval for each steering vector and rescaling the steering range accordingly. Building on this, we introduce a more interpretable method for steering, called parameterized steering which instead of pushing activations in a direction, it sets them at the desired trait strength. Next, with multi-trait steering, we find that naively combining vectors causes them to interfere with one another. In response, we develop a method to orthogonalize the vectors; it decouples these traits, allowing a model to exhibit one behavior without inducing the others, thus enabling multiple traits to be steered simultaneously.
[AI-16] From Interaction to Capability: Task Advancement as State and Supervision for GUI Agents
链接: https://arxiv.org/abs/2610.37482
作者: Ziyang Chen,Qian Li,Chengwei Qin
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:GUI memory should preserve the results of past actions and learn reusable operations. This is difficult in long-horizon workflows, where agents must use information across applications and useful operations are interleaved with failed attempts. We introduce Task Advancement Memory (TAM), which defines task advancement as verified task progress, and supports ongoing task execution and learns reusable operations. During execution, Task Advancement State represents how verified task progress changes the remaining goals, allowing the agent to build on earlier achievements across applications. After each task, advancement supervision identifies and abstracts the interactions that produced a verified local result. The learned operations preserve the relation between required inputs, expected outcomes, and verification conditions. This retains locally successful operations from both successful and failed tasks. On MobileWorld GUI-only tasks, Advancement State alone raises success from 12.0% to 28.2%. Full TAM achieves a 17.6% relative gain in task success over the strongest evaluated baseline and reduces lifecycle tokens per solved task by 38.7%. On SPA-Bench, TAM achieves state-of-the-art task success in the single-app English Level 3 setting.
[AI-17] Autonomous fixed-wing interception using reinforcement learning
链接: https://arxiv.org/abs/2610.37470
作者: Daan van Hamersveld,Ewoud Smeur,Iurii Kriachko
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Neural and Evolutionary Computing (cs.NE); Systems and Control (eess.SY)
备注: MSc thesis, Faculty of Aerospace Engineering, Delft University of Technology, 2026. 24 pages, 29 figures, 4 tables
Abstract:Medium altitude, long endurance unmanned aerial systems such as Shahed-type strike drones are an unanswered threat in modern conflicts. Conventional air defense systems are too expensive to provide a feasible large-scale solution. Quadcopter interceptors are a cheap, battle-proven alternative, but their extremely limited flight time restricts them to low altitude targets at short range, whereas fixed-wing platforms have the endurance to reach higher altitude targets from longer distances. However, such interceptors are still operated manually. Recent developments in learning-based controllers show potential for automation. In this work, a fixed-wing interceptor is presented that is controlled by a neural network that maps filtered monocular vision observations directly to actuator commands. The policy was trained in a custom Python environment using reinforcement learning and proximal policy optimization. The policy was then deployed on a fixed-wing platform and tested in real interception scenarios. As a baseline, two classical guidance controllers were also developed and tested in simulation. The policy is able to intercept targets both in simulation and in real flight. In simulation, the policy outperforms the classical controllers under most conditions. The interceptor was ultimately successfully validated in real flight by intercepting maneuvering fixed-wing targets. This novel system shows the potential of artificial intelligence for autonomous control of fixed-wing interceptor drones.
[AI-18] M4Bench: Evaluating Procedural Specification for Clinical EHR Derivation Agents NEURIPS2026
链接: https://arxiv.org/abs/2610.37459
作者: Hannes Ill,Rafi Al Attrach,Rajna Fani,Ahram Han,Ryohei Yamamoto,Marlene Louisa Moerig,Dukyong Yoon,Leo Anthony Celi
类目: Artificial Intelligence (cs.AI)
备注: Accepted at NeurIPS 2026 (Evaluations and Datasets Track). 10 pages main text, 33 pages total. Code: this https URL . Evidence: this https URL
Abstract:Clinical research over electronic health records depends on derived variables whose definitions sit outside the schema, in item codes, time windows, missingness rules, and output-grain choices that change cohorts and downstream conclusions. We introduce M4Bench, 28 clinical derivation tasks across 15 task families on MIMIC-IV and eICU, to test whether AI coding agents can reproduce such derivations and how clinician-reviewed agent skills change their results. Its controls decompose the value of skills into specification completeness, delivery format, and schema familiarity. In a Docker-isolated Codex campaign (GPT 5.4 Mini and GPT 5.5), skills raise reward by 0.136 over the no-skill baseline, with 26 of 28 tasks improving. Specification completeness accounts for most of this gain: on an eight-task control subset, an operational spec alone raises reward by 0.162, and skills add only 0.024 on top, indistinguishable from zero. On a ten-task subset, decoy skills score like no skill and raw reference SQL nearly matches targeted skills, arguing against generic context or the skill format as the explanation. Schema obfuscation and restructuring preserve 86% and 70% of the skill gain (GPT 5.4 Mini, 26 MIMIC-IV tasks), a descriptive robustness check rather than a bound on memorization. A descriptive Claude Code replication on all 28 tasks (Opus 4.7 and Sonnet 4.6, 560 runs) is directionally consistent (+0.120, 21 of 28 tasks improving). These gains are partial agreement with the reference, not completed tasks: only 25 of 280 with-skill and 13 of 280 no-skill Codex runs pass every strict check, and skills barely change which rows agents return (key F1 +0.003). Skills are released as a versioned, clinician-reviewable markdown delivery format, with a family-level reference-validity audit. M4Bench releases tasks, skills, evaluation code, and sanitized per-run evidence as an open-source benchmark.
[AI-19] FeatureArc: Benchmarking Coding Agents on Evolving Features
链接: https://arxiv.org/abs/2610.37434
作者: Yunlong Wang,Alyssa Wu,Cuiyun Gao,Jia Feng,Songyuanyi Lu,Zhichao Hu
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注: 12 pages, 8 figures, 4 tables
Abstract:A software feature usually evolves over time: developers extend its first version as new use cases and requirements arrive, and each extension builds on code written earlier. Most feature benchmarks evaluate a single request on a fixed codebase. Feature evolution requires evaluating successive extensions on an agent’s own earlier code and checking whether the feature remains correct. We introduce FeatureArc to evaluate coding agents on evolving features, together with a pipeline that constructs its tasks from repository histories. The benchmark combines dependencies between successive stages of the same feature, explicit requirements and interface contracts, and executable checks across four programming languages. The pipeline validates fail-to-pass (F2P) tests for new behavior and pass-to-pass (P2P) tests for existing behavior on the reference code before and after each stage. We evaluated 10 coding agents on 73 manually reviewed tasks of 2 to 10 stages. Each agent works in one persistent session and workspace without external test feedback or rollback. The evaluation reveals a gap between first-stage and whole-feature success: pooled success falls from 62.7% at the first stage to 28.5% when the whole feature is tested on the final code. Trajectory analysis identifies inherited defects as a common failure mechanism, present in 72.9% of failed later stages. These earlier errors often remain undetected: in stages following a failure, agents leave the earlier error unnoticed about 95% of the time, and their self-tests rarely revisit earlier stages. Evolving a feature autonomously, without human checks between stages, therefore remains a substantial challenge for current coding agents. At Tencent, the pipeline extends internal evaluations to this setting and guides the development of Hunyuan, Tencent’s family of large language models.
[AI-20] Bayes-Adaptive MDPs for Rollout Allocation to Enhance GRPO Training
链接: https://arxiv.org/abs/2610.37429
作者: Faeze Moradi Kalarde,Armin Toroghi,Nam Ho Koh,Doo Hwan Hwang,Seungjin Kim,Kee-Eung Kim,Scott Sanner
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 22 pages, 8 Figures
Abstract:Group Relative Policy Optimization (GRPO) has demonstrated strong performance in LLM post-training. However, it allocates a uniform rollout budget to every training question, ignoring variations in question difficulty. As a result, highly difficult questions may receive too few rollouts to generate informative positive reward signals, limiting their contribution to training, while moderately difficult questions often receive more rollouts than necessary. In this work, we formulate the problem of difficulty-aware rollout budget allocation to maximize training utility under the same total rollout budget as standard GRPO. We propose AdaRoll-GRPO, which reformulates rollout allocation as a sequential decision-making problem and leverages a Bayes-adaptive Markov decision process (MDP) to learn an adaptive allocation policy. The learned policy is computed offline and applied during GRPO training without introducing additional online optimization overhead. Extensive experiments on multiple reasoning datasets and LLM architectures demonstrate that AdaRoll-GRPO consistently improves GRPO training performance under a fixed rollout budget, highlighting the effectiveness of adaptive rollout allocation for GRPO in LLM post-training.
[AI-21] Beyond the Readout: A Latent Numerical Frame Shared Across Modality and Operation
链接: https://arxiv.org/abs/2610.37422
作者: Oriol Juan Sabater,Alex Ferrando de las Morenas,Pau Rodríguez,Xavier Suau
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:When a multimodal model computes 17+35 from text or 91-39 from a rendered image, it produces the same answer, 52. Does it use one internal concept of number for both, or separate machinery for each path? Similarity metrics such as RSA or CKA cannot settle this question. They give the same score to identical and to rotated directions, and two representations can look alike only because both prepare the same output digits. We therefore use causal interventions. For four conditions (text or image, addition or subtraction), we learn a small subspace of Gemma4-12B-it’s residual stream in which swapping coordinates swaps the model’s answer. The subspaces answer three questions. First, do the conditions write numbers along the same directions? They do not. Reusing a subspace from text in an image on the same operation recovers only 53.2% of its causal effect. A single rotation and rescaling maps one subspace onto the other and raises recovery to 91.4%, so the conditions share one geometry expressed in different coordinates. Second, does this shared geometry only reflect the digit the model is about to print? To test this, we split each subspace into 9 directions that the output layer reads to choose the next digit and 13 directions that it cannot see. The 13 hidden directions still carry the number, yet changing them barely alters the digit being written. Third, what are these hidden directions for? A model must emit a multi-digit answer one digit at a time, so it can compute the full result early and keep it in reserve. When two numbers share their first digit, swapping the number in the hidden directions leaves the digit being written intact and redirects the next digit in 77.6% of cases. The model thus forms an abstract numerical concept upstream and unfolds it across time: multimodal representations share meaning through relational geometry across distinct coordinate frames.
[AI-22] Constraining the Commit Boundary: A DIO/ODI-Informed Authorization Guardrail for Agent ic Entity Resolution
链接: https://arxiv.org/abs/2610.37407
作者: Daniel Beránek
类目: Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR); Databases (cs.DB)
备注: 29 pages, 1 figure
Abstract:Agentic Entity Resolution (ER) systems iteratively plan actions, retrieve evidence, revise hypotheses, and propose graph mutations. An erroneous identity hypothesis may condition later retrieval and produce apparent confirmation. This paper proposes a Digital Identity Optimization (DIO) and Ontology of Digital Identity (ODI)-informed authorization boundary between evidence available for agent reasoning and evidence eligible to authorize persistent identity-changing mutations. A domain-governed constraint-binding function maps selected reconstruction-failure distinctions to explicit identity-preserving constraints. An abstract two-phase supervisor permits a commit only when claim-specific authorization-eligible evidence supports every required proposition, relevant conflicts are resolved, the operation remains within its authorized scope, and relevant state and policy versions remain valid. Admission as UNVERIFIED does not confer authorization eligibility. Separate claim-specific promotion governs evidentiary use; insufficient support or unresolved material conflict leads to deferral. The contribution is a specification connecting diagnostic-to-constraint mapping, evidence promotion, proposition coverage, conflict handling, scope binding, and commit-time version integrity. A proposed evaluation protocol is designed to assess the mapping’s incremental value beyond generic authorization and conventional identifier rules. This is a conceptual architecture without implementation or empirical validation, not a theorem of semantic correctness. Its property depends on domain-policy adequacy, evidence promotion, and enforcement. It does not prevent erroneous hypotheses or hypothesis-conditioned retrieval; dependency-aware evidence flow remains future research. Comments: 29 pages, 1 figure Subjects: Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR); Databases (cs.DB) ACMclasses: H.2.4; I.2.4 Cite as: arXiv:2610.37407 [cs.AI] (or arXiv:2610.37407v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2610.37407 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-23] Clearing the FoG: Training KG Agents to Retrieve the Evidence Behind Their Answers
链接: https://arxiv.org/abs/2610.37405
作者: Zongyuan Sun,Shuai Zhao,Yingjun Du,Ke Shu,Ziyue Zhu,Haoran Luo
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models (LLMs) can answer many factual questions correctly, but a correct answer is not necessarily a supported one. For an LLM agent, a knowledge graph (KG) plays two roles: as a source of answers, it supplies facts that the model lacks; as a source of evidence, it supplies the supporting triples that make an answer traceable and auditable. However, existing KG agents treat the KG only as a source of answers: their training data carries answers but not evidence, and their training rewards check the answer but not the retrieved evidence, so agents often answer correctly without retrieving the supporting triples. We propose FoG, a framework that trains KG agents to retrieve the evidence behind their answers. FoG consists of evidence-grounded synthesis, a KG interaction agent, and an evidence-coverage process reward (ECPR): the synthesis recursively hides explicit entities behind verifiable sub-questions, so that every question carries its supporting triples by construction; the agent records the triples returned by the KG; and ECPR credits each supporting triple to the turn that first retrieves it. Experiments on three KGQA benchmarks demonstrate that FoG enables 3B–4B open-source agents to outperform substantially larger baselines in answer accuracy and evidence coverage, with zero-shot cross-KG transfer.
[AI-24] Learning Quality-Controllable Robot Policies with Iterative Hindsight Relabeling
链接: https://arxiv.org/abs/2610.37393
作者: Sinjae Kang,Dong Kyu Shin,Jaehong Kim,Kamin Lee,Hyoeun Kim,Kaixin Wang,Li Zhao,Seonghyeon Ye,Kimin Lee
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Robot policy evaluation often relies on a binary success signal for task completion. In real-world deployment, however, desired execution behaviors are also defined by quality attributes such as execution speed and motion smoothness, which need to be dynamically controlled according to task requirements. To address this, we introduce Quality-Controllable Policy (QCP), a robot policy that enables inference-time control over both individual quality attributes and their combinations. To extend the controllable quality range beyond initial demonstrations, we further propose an iterative hindsight relabeling loop: QCP collects online rollouts under target quality labels, adds successful trajectories to the dataset, relabels the accumulated trajectories in hindsight according to their measured quality, and fine-tunes the policy. Experiments on Robomimic and RoboEval demonstrate that QCP consistently improves both task execution quality and success rates across iterations, producing successful rollouts that exceed the quality of the initial demonstrations. Finally, we validate QCP on real-world manipulation tasks, where controlling speed increases throughput in conveyor operations and controlling transport stability improves success when carrying wine glasses, demonstrating that quality-controllable policies yield practical gains in physical deployment.
[AI-25] Channel Selection for Mixture-of-Experts Decoding Needs No Calibration Data
链接: https://arxiv.org/abs/2610.37386
作者: Geunhyoung Kim,Sung Ju Hwang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 37 pages, 12 figures. Code: this https URL
Abstract:Mixture-of-Experts (MoE) models are bottlenecked by loading routed-expert weights from memory during interactive decoding at small batch sizes. Because most intermediate channels in a routed expert contribute little to any single token, channel selectors predict which channels matter and read only their weights, typically keeping channels above a threshold calibrated on offline data. We show that channel selection needs no calibration data. First, a calibrated threshold does not fix how much a selector reads, because the number of channels passing it varies across inputs; Prox with thresholds calibrated for an 18% read budget actually reads only 12.9% on average across our benchmarks. Second, top-k keeps each token’s k highest-scoring channels and needs no calibration data. At the calibrated budget, a selector’s top-k version matches or beats its threshold version on the same score in 40 of 43 comparisons at 13% and 18% budgets, including all 23 on generation tasks. The top-k versions also avoid the repetition-loop collapse that calibrated thresholds cause on Gemma 4. Calibrating a threshold thus provides neither of the two things selectors use it for: a known read budget and a consistent accuracy gain at that budget. We furthermore introduce a reader built entirely from the model’s weights, which screens channels with a ternary sketch and recovers the skipped ones with a low-rank map. Across three post-trained MoE models, our reader retains 98.1% of full-model accuracy at 17.8% of routed-expert reads, with Triton kernels running MoE blocks 1.60 to 1.90x faster than optimized vLLM at batch one and up to 2.69x at larger batches on the same three models.
[AI-26] Uncovering Why LLM s Fail: From Internal Mechanisms to Adversarial Reasoning Trajectories
链接: https://arxiv.org/abs/2610.37369
作者: Zirui Ding(1,2),Xinxin Fan(2,3),Hao Yuan(1,4),Shaoye Luo(2,3),Men Niu(1,2),Yunfeng Lu(5) ((1) Institute of Computing Technology, Chinese Academy of Sciences, (2) University of Chinese Academy of Sciences, (3) State Key Laboratory of AI Safety, Institute of Computing Technology, Chinese Academy of Sciences, (4) ShanghaiTech University, (5) Beihang University)
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models (LLMs) are usually evaluated by final-output correctness or safety, yet an endpoint label does not reveal how a failure develops throughout the reasoning process. We therefore ask whether ordinary reasoning failures and adversarially induced failures can be characterized using the same internal coordinates. Towards this end, we in this paper study a white-box diagnostic framework CauTrace to disclose the root-cause of why LLM fails from three fundamental inspections: Functional States, Propagation of Representation, Temporal-State Coordination and Reasoning Trajectory. First, seven functional states are creatively defined to represent the support and constraint processing in LLMs. Second, the propagation of representation change measures how the changes among the seven states persist across subsequent Transformer blocks. Third, temporal-state coordination and trajectory topology characterize when the commitment and verification occur relative to the support, and how the dominant state sequence evolves over the full inference process. Finally, we evaluate CauTrace on both clean reasoning failures and paired adversarial perturbations. Across Qwen and LLaMA-family models, the failed traces show weaker Attention-MLP agreement on support-state updates and less verification activity relative to commitment, while the exact representation-level signatures and timing shifts vary by architecture. Held-out repair and steering tests demonstrate consistent operational signals that were in line with our claims for certain models and specific settings, with aligned steering producing the predicted directional changes in process measures. Overall, CauTrace provides a unified diagnostic framework for analyzing failure dynamics through changes in support, commitment, and verification, which would shed light on the in-depth research on LLM security.
[AI-27] CoS-VQ: Coordinated Shared-Scale Vector Quantization for Large Language Models
链接: https://arxiv.org/abs/2610.37336
作者: Zukang Xu,Zhixiong Zhao,Xing Hu,Zhe Jiang,Xin Si,Dawei Yang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 20 pages, 3 figures. Code: this https URL
Abstract:Large language models (LLMs) require substantial memory, motivating low-bit weight quantization for efficient deployment. Existing vector quantization (VQ) methods often improve accuracy by increasing the effective coding dimension, which can complicate independent lookup and local codeword updates. We propose CoS-VQ (Coordinated Shared-Scale Vector Quantization), which preserves low-dimensional codewords while jointly optimizing hard indices and their shared scale across positions. The key challenge is scale-induced coupling: the optimal scale depends on the entire codeword combination and, in turn, changes candidate rankings across positions, so pointwise updates can stall even after scale re-optimization. CoS-VQ collects candidates at multiple probe scales and constructs a tangent separable upper bound for the residual-aware quadratic objective. By reducing candidate comparisons to lower envelopes of lines over positive scales, it identifies finitely many representative intervals for joint index–scale updates without exhaustively enumerating codeword combinations. On Qwen3.8-27B, CoS-VQ outperforms QTIP and LLVQ on the evaluated language modeling, coding, and reasoning benchmarks. Ablations from the same initialization further show larger gains from joint updates than from scale-only refinement, highlighting joint selection as an alternative to increasing codeword dimensionality. The code is available at this https URL.
[AI-28] WCC: A Learning-Based Framework for Deciding When to Apply Cooperative Coevolution in Large-Scale Optimization
链接: https://arxiv.org/abs/2610.37329
作者: Wenjie Qiu,Hepei Li,Jianfeng Wang,Jun Zhang,Yue-Jiao Gong
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Cooperative Coevolution (CC) and Non-Decomposition Algorithms (NDAs) are two primary approaches for large-scale global optimization (LSGO). Although both have achieved success, they exhibit markedly different performance across problem types and each faces approach-specific limitations, leaving the question of “when to use CC” unresolved. We address this gap with WCC, a novel learning-based framework that dynamically decides when to deploy CC in LSGO. WCC employs a streamlined design for efficient adaptive learning. We formulate the decision as a Markov decision process, designing a targeted feature set, a robust reward, and a tailored network architecture to provide the agent with rich information about the problem state and optimization progress. Our auto-regressive policy enables WCC to perform fine-grained, dynamic switching between CC and NDAs while conditionally allocating the computational budget without expert specifications. WCC shows strong competitiveness and superior generalization across problems, optimization horizons, and baseline optimizers. It realizes a “train-once and reuse” paradigm under a low budget, significantly curbing LSGO cost. Additionally, in-depth analyses yield insights not captured by existing theory, offering practical guidance on when and how to effectively deploy CC.
[AI-29] AnalogScout: Simulation-Efficient PVT-Aware Analog Sizing with LLM Agents ISCAS2027
链接: https://arxiv.org/abs/2610.37319
作者: Yichen Long,Keren Zhu
类目: Hardware Architecture (cs.AR); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 5 pages, 5 figures, 1 table. Submitted to IEEE ISCAS 2027
Abstract:Analog device sizing must satisfy all specifications across process, voltage, and temperature (PVT) corners to ensure tape-out success, but simulating every candidate at all corners dominates the sizing time and slows design iteration. Conventional surrogate-based optimizers search the sizing space directly, while recent large language model (LLM) agents add circuit knowledge to designs but still delegate the sizes to global optimizers. Both learn the circuit response from repeated multi-corner SPICE simulations and optimize on top of it, which becomes expensive under a limited simulation budget. Providing an agent with reusable verified designs and sizing sub-blocks analytically could save most of this budget. We propose AnalogScout, a simulation-efficient PVT-aware sizing framework driven by an LLM agent. Combining elite-design reuse, analytic gm/Id local sizing and global refinement under full-corner verification, AnalogScout reduces power by 43.3% relative to the evolution strategy CMA-ES at an equal budget of 16k simulations on three SKY130 OTAs of increasing complexity, and reaches the final CMA-ES power after 7.0% of that budget. It also attains lower power than Bayesian optimization on all three OTAs and than two re-implemented LLM-driven flows on the two OTAs where they were tested.
[AI-30] ReForest: Interpretability Meets Reasoning in Isolation Forest PAKDD2026
链接: https://arxiv.org/abs/2610.37316
作者: Juyoung Seo,Geonwoo Bang,Dongho Kim,Moohong Min
类目: Artificial Intelligence (cs.AI)
备注: 12 pages, 2 figures. Published in PAKDD 2026
Abstract:Statistical anomaly detection methods like Isolation Forest are efficient but suffer from a critical semantic gap. They identify numerical rarities but fail to understand the contextual meaning of data, often missing subtle, context-dependent anomalies. Conversely, Large Language Models possess strong semantic reasoning capabilities but are prohibitively expensive and slow for large-scale deployment. To bridge this gap, we propose ReForest, a novel framework that synergizes the statistical efficiency of Isolation Forest with the reasoning power of LLMs. ReForest transforms the latent decision paths of Isolation Forest into natural-language explanations, enabling an LLM to refine anomaly detection through an uncertainty-aware dynamic ensemble. This approach allows the model to adaptively weigh statistical and semantic signals based on uncertainty-driven instance-level confidence. Extensive experiments on four real-world datasets (Healthcare, Finance) demonstrate that ReForest significantly outperforms baselines, achieving an average 21% improvement in Recall and a 10% gain in F1-score. Our framework not only improves detection performance but also enhances model transparency by grounding semantic narratives in statistical evidence, providing interpretable rationales essential for high-stakes decision-making.
[AI-31] RPSD: Retrieval-Privileged On-Policy Self-Distillation with Optimal Transport
链接: https://arxiv.org/abs/2610.37281
作者: Dongyue Li,Zebang Cheng,Zhongyi Deng,Hongboxu,Yi Bin,Fei Ma
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Sequential recommendation predicts a user’s next interaction from historical behavior. Recent LLM-based methods recast this task as generative recommendation and increasingly adopt reinforcement learning such as GRPO for post-training. However, these methods suffer from two limitations, namely credit-blind optimization, where a sequence-level reward is shared across all generated tokens, and collaborative-blind supervision, where each training signal is derived from a single user and ignores behavioral patterns shared across similar users. To address both issues, we propose Retrieval-Privileged on-policy Self-Distillation (RPSD). For credit-blind optimization, we introduce on-policy self-distillation (OPSD), in which a teacher with privileged knowledge re-evaluates the student’s on-policy response, and the token-level confidence difference between the student and teacher reshapes the shared sequence-level advantage into token-specific credit. However, the privileged knowledge in vanilla OPSD is still derived from a single user, leaving collaborative-blind supervision unresolved. We therefore ground the teacher in collaborative behavioral evidence from similar users. Specifically, we use temporally aware optimal transport (OT) to compare interaction trajectories across users by jointly considering item similarity and temporal structure, and further apply diversity-aware selection to retain complementary trajectories as privileged evidence. In this way, on-policy self-distillation provides fine-grained token-level credit, while OT enriches the teacher with structured cross-user evidence. To the best of our knowledge, this is the first work to bring on-policy self-distillation to sequential recommendation and to use cross-user behavior as its privileged signal. Extensive experiments across multiple real-world datasets demonstrate that RPSD achieves state-of-the-art performance.
[AI-32] CM-DiagCommit: A TCM-SD-Derived Benchmark for Evidence-Conditioned Syndrome Selection and Commitment
链接: https://arxiv.org/abs/2610.37243
作者: Erwei Hou,Yue Hu
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Traditional Chinese Medicine (TCM) syndrome differentiation is usually evaluated as static final-label selection over complete clinical records. This setting hides three diagnostic behaviors: whether a model distinguishes semantic-neighbor syndrome labels from easy random distractors, whether it updates predictions as evidence fields are revealed, and whether it should commit under uncertainty. We introduce TCM-DiagCommit, a benchmark derived from the public TCM-SD/ZY-BERT resource while retaining its source syndrome labels; it converts 1,834 source cases into 23,842 task rows over 148 syndrome labels. The benchmark combines random, profile-neighbor, and full-vocabulary candidate ladders with staged t1/t2/t3 evidence. It also includes fixed-label tongue/pulse perturbations and insufficient-evidence variants. A benchmark-defined action contract scores commit, acquire, defer, and abstain. Across five zero-shot DirectLLM calibrations, random10 top-1 accuracy is 63.2% to 66.4%, but hard20 falls to 24.7% to 30.3% and full148 to 14.8% to 22.3%; on hard20, WT(all) wrong-terminal commits remain 62.2% to 74.3%. These results show that final-answer accuracy alone is insufficient for diagnostic benchmark evaluation: answer selection, evidence use, and benchmark commitment scoring should be measured separately.
[AI-33] Grounding LLM Agents in Business Processes: Process-Aware Knowledge Graphs for Reliable Multi-Step Service Invocation
链接: https://arxiv.org/abs/2610.37242
作者: Saiful Khan,Afrah Farea
类目: Artificial Intelligence (cs.AI); Software Engineering (cs.SE)
备注:
Abstract:Large language model (LLM) agents that call enterprise APIs must select the correct endpoints, follow process order, use valid payload keys, and choose coded values from predefined lists. However, current retrieval-augmented service discovery treats endpoints independently and therefore misses process and schema constraints. We address this limitation by integrating Open Data Protocol (OData) metadata and Business Process Model and Notation (BPMN) reference processes in a typed property graph. A hybrid retriever identifies candidate endpoints using dense embeddings and follows process edges to retrieve subsequent endpoints. We evaluate the method on 42 enterprise services with process-chain ground truth. For queries with low lexical overlap between the query and downstream entities, it improves Recall@3 by 5.7 points (n=74), with consistent gains across embedding models and cut-offs. It matches dense retrieval on other queries and provides no improvement on OpenAPI services without a process layer, indicating that the gains arise from process structure. We also evaluate an LLM agent on 81 multi-step queries. It completes 38.7% of tasks, compared with 10.3% for a chunk-based baseline. In an elicitation probe, providing valid code lists eliminates invented codes but leaves invalid keys in 46% of write runs. Full typed context reduces invalid calls to 3% (3 of 101 runs), with no invented codes. These results show that exposing process and schema constraints during retrieval improves the reliability of agent-driven API calls.
[AI-34] When Is a Relation Worth Trusting? Manipulation Costs Exposure and the Trust Trap in Multi-Relational Graph Learning
链接: https://arxiv.org/abs/2610.37239
作者: Vincess Dongmo,Astride Melvin Fokam Ninyim
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR)
备注: Preprint. October 27, 2026
Abstract:Multi-relational graph detectors learn from clean data how much to trust each relation, but an adversary pays a different price to manipulate each relation and each action; taking such costs as given, we ask when using a relation is worse than ignoring it. Our first main result proves that degree gating strictly beats every selection or reweighting of a relation when part of the population is isolated on it and the attacker cannot afford to open the gate, and that this window is sharp; on real graphs, validated gating gains exactly on the relations that meet these conditions and nowhere else. Our second shows that a per-node exposure bound, which extends to message passing, tells a platform which relation to harden: exposure per unit cost, computed without any attack, picks the relation whose hardening raises the exact trap threshold most in 15 of 16 cases, against at most 7 for informativeness, cheapness, or an analysis with equal edit costs. Supporting tools give exact worst cases for linear relational models, certified brackets for R-GCN and tree ensembles, and a finite-sample guarantee for adversarial validation, under which clean-trained graph detectors fall below their graph-free counterparts in 58 of 90 settings on six graphs. A selection protocol built on these results is never significantly worse than ignoring the graph; with at most four relations it mostly abstains, as validation alone would, and its cost-aware proposals help when relations are many, so far shown on synthetic graphs. Without cost information, the safe choice is to abstain.
[AI-35] GraphRAG Prov: GraphRAG Meets How-Provenance
链接: https://arxiv.org/abs/2610.37237
作者: Wadie El Amrani,Luis Galárraga
类目: Artificial Intelligence (cs.AI); Databases (cs.DB)
备注:
Abstract:GraphRAG systems often use knowledge graphs to retrieve relevant context for an LLM that produces the final answer. That makes answers more inscrutable and therefore less appealing for domains where transparency is paramount. We present GraphRAGProv, an architecture that executes SPARQL queries over source-grounded RDF graphs to compute answers and their how-provenance polynomials. These polynomials record which statement occurrences were used and how they were combined. Because each occurrence is linked to its source text, users can follow these links to check whether the extracted statements reflect the text and whether the query combines them to answer the question. Querying thereby becomes a way to inspect and evaluate the text-to-graph transformation. Our interactive demo examines 400 cases from two multi-hop QA benchmarks, revealing extraction and query errors even when the final answer is correct.
[AI-36] Spine: Compiling Subtask Trees into Structured Context for Long-Horizon Coding Agents
链接: https://arxiv.org/abs/2610.37224
作者: Jiahong Xiang,Kunqiu Chen,Lingming Zhang,Yuqun Zhang
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注: 22 pages, 9 figures, 7 tables
Abstract:Long-horizon coding tasks contain nested subtasks. Typically, agents record interactions sequentially into a growing history, causing context inflation and performance degradation. Compressing or trimming the history without explicit subtask boundaries can discard information needed by unfinished subtasks. Conversely, recursive subagent delegation isolates completed subtasks through child agents by assigning every subtask boundary to delegation, which conflates subtask granularity with executing agent choice, adding execution and coordination overhead. We introduce Spine, an interface and runtime that align model context with subtask nesting and completion. An agent declares explicit subtask boundaries through work scopes, allowing the same agent to execute nested subtasks inline while reserving delegation for independent parallel tasks. The runtime organizes execution in SpineTree, a persistent tree recording subtask relations, local interactions, completion states, and returned results. Before each model call, it compiles the tree into model context just in time, retaining unfinished subtasks’ interactions and replacing completed subtasks’ details with their results, while preserving prefix-cache reuse. We integrate Spine into Codex to build SpineCodex. With GPT-5.6 Sol, compared with native Codex, SpineCodex improves SWE-Milestone Score by 11.8% while reducing total cost by 32.4%, improves the average ProgramBench test-pass rate by 17.3%, and improves Mean@5 on eight of the ten evaluated FrontierSWE tasks. On SWE-Milestone, it also achieves a comparable Score to Recursive Subagent at 73.7% lower total cost. We also integrate Spine into Pi through a reusable software package: with Gemini 3.8 Flash and Qwen 3.8 27B, it improves SWE-Milestone Score by 9.5-17.5% and the average ProgramBench test-pass rate by 4.3-8.7% compared with native Pi. Comments: 22 pages, 9 figures, 7 tables Subjects: Software Engineering (cs.SE); Artificial Intelligence (cs.AI) Cite as: arXiv:2610.37224 [cs.SE] (or arXiv:2610.37224v1 [cs.SE] for this version) https://doi.org/10.48550/arXiv.2610.37224 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-37] Retrieve to Act: Motion Primitive Graph Retrieval for Robust VLA Execution
链接: https://arxiv.org/abs/2610.37215
作者: Pham Tri Quang,Bao-Ngoc Dao,Minh Duc Nguyen,Duc-Duy Nguyen,Hung T. Ho,Linh Dang Le,Trong-Bao Ho,Thien-Loc Ha,Tan Q. Nguyen,Tuan Dam,Vu N. Duong,Daniel Sonntag,Ngan Le,Khoa D. Doan,Jan Peters,An Thai Le,Duy M. H. Nguyen,Vien Anh Ngo
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注: 20 pages, 9 figures, 7 tables
Abstract:Vision-Language-Action (VLA) policies demonstrate remarkable semantic generalization, yet remain highly vulnerable to out-of-distribution (OOD) visual shifts during execution. While existing test-time methods attempt to suppress spurious visual cues, they fundamentally overlook the rich, uncorrupted kinematic priors embedded in the policy’s own training experience. We introduce Retrieve to Act (R2A), a test-time scaling framework that formulates OOD robustness purely as a geometric motion primitive retrieval problem. R2A bypasses fragile visual features by structuring successful training demonstrations into an offline spatiotemporal motion primitive graph. During inference, a query-dependent Graph Neural Network dynamically localizes the robot’s ongoing execution within this graph, retrieving structurally compatible future motion primitives. These retrieved geometric priors are then seamlessly harmonized with the frozen VLA’s predicted actions via a learned, dimension-wise gating mechanism, adaptively rectifying erroneous spatial trajectories while preserving the policy’s local reactivity. Extensive experiments across three state-of-the-art VLA architectures, GR00T N1.7, \pi_0.5 , and StableVLA, demonstrate that R2A consistently and significantly enhances robustness against diverse contextual perturbations. R2A achieves absolute success rate improvements of up to +12% in simulation benchmarks (LIBERO-Plus, RoboTwin2.0-Plus) and +40% in real-world deployments, without requiring any finetuning of the base policies. Project page: this https URL
[AI-38] Does Your Generalist Skill Optimizer Really Incentivize Agent ic Tasks? Rethinking Skill Optimization via Bipartite Attribution
链接: https://arxiv.org/abs/2610.37201
作者: Jiawei Zhou,Bo Peng,Yuxuan Huang,Ming Zhang,Youbang Sun,Xiang Wang,Chaochao Lu
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Skill optimizers adapt an agent to long-horizon tasks, such as repairing a terminal environment, by rewriting its skill, a natural-language procedure in its context, and keeping the edits that raise success on a small dev set. On such tasks success is noisy, so a gain on a small dev set does not show that an edit helps, and a success score does not show what the edit costs. We introduce BiPA, which replays an optimizer’s acceptance rule on candidates identical to its seed to measure what selection alone reports, and records each episode as a bipartite graph of turns and the text they read to trace extra input tokens to the behavior behind them. Across three optimizers, three agentic benchmarks and two models, the learned skills succeed about as often as their hand-written seed, 0.1 points higher on average. Their dev gains vanish on test: keeping the best of several candidates on at most 30 tasks reports gains this large even when none changed. They read about 16% more input tokens: the optimizers almost always add rules, half to re-check work and some to act one call at a time, and the agent follows them as extra turns, each re-sending the whole conversation. Four principles follow: accept an edit only if its gain exceeds what selection alone reports; charge it for the turns it adds, not for its length; let the editor delete and merge rules; show it what earlier rules cost. Code is available at this https URL.
[AI-39] Action-Value Calibration via Continuation Sharing for Long-Horizon Agent ic Reinforcement Learning
链接: https://arxiv.org/abs/2610.37195
作者: Yingjian Chen,Lang Zhou,Zixin Zhao,Pingfan Li,Weishi Zheng,Zhilin Zhao
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Reinforcement learning for long-horizon agents relies on feedback from sequences of interactions, making credit assignment to individual actions difficult. Existing methods such as GiGPO compare actions at shared states to compute relative advantages, but estimate each action’s value from a single sampled continuation. The resulting local credit remains sensitive to decisions made later in the trajectory. To address this problem, we propose ConCal (Continuation Calibration), a recursive continuation-value estimator that incorporates continuation evidence already present in a rollout batch. ConCal matches visits from the same task by their incoming state-action-state transitions, combines shared continuation evidence with each visit’s own transition estimate, and propagates calibrated values backward to refine earlier action-value estimates. This reduces dependence on individual sampled continuations and mitigates local credit distortions without additional rollouts. Held-out return diagnostics show lower error in within-group value differences and higher agreement with reference action orderings than estimates based on individual continuations. With Qwen2.5-1.5B-Instruct, ConCal improves success rates over GiGPO by 3.39 and 5.86 percentage points on ALFWorld and WebShop, respectively. Our code is available at this https URL.
[AI-40] MLE-World: Predicting Machine Learning Experiment Outcomes with a Learned World Model
链接: https://arxiv.org/abs/2610.37175
作者: Brian Ko,Han Wang,Hejia Geng,Li Erran Li,Sherry Yang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:AI research agents can propose, execute, and improve programs that solve machine learning tasks. However, executing each candidate program requires training and evaluating ML models, limiting how many candidates an agent can explore at each step. We introduce MLE-World, a world model finetuned from Qwen3.5-4B that predicts a program’s terminal output (error traces, training logs, and validation scores) from its code and data description without executing it, together with a confidence score estimating the prediction’s correctness. Because a prediction costs one inference call rather than a training run, MLE-World makes test time scaling (TTS) affordable for these agents. We build five TTS policies, guided by MLE-World, that compare several candidates per step by their predicted outputs: Sequential revision, Best-of-N, Pairwise judging, Fusion, and a UCB style rule that also adds an exploration bonus from this confidence score. Since predictions can be wrong, we run a candidate for real only when its confidence score is low. On 17 MLE-bench Lite tasks, MLE-World matches the real terminal output on 77.9% of held out programs, versus 41.1% for much larger prompted models. MLE-World guided TTS achieves up to 22.1% higher observed mean performance than the agent without a world model, with less real execution. Project webpage: this https URL.
[AI-41] Rènyi-Regularized Flow-Based Variational Inference for Bayesian Calibration
链接: https://arxiv.org/abs/2610.37169
作者: Exaucè Luweh Adjim Ngarti(AIRSEA),Élise Arnaud(AIRSEA),Léo Nicoletti,Victor Trappler(Mines Saint-Étienne MSE, LIMOS, FAYOL-ENSMSE, FAYOL-ENSMSE),Arthur Vidard(AIRSEA)
类目: Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
备注:
Abstract:Variational inference with normalizing flows provides a flexible posterior approximation for Bayesian model calibration, but struggles when the posterior has several modes separated by regions of low probability mass. We study this behavior in a two-parameter twin experiment and regularize the variational objective. Under Gaussian observation errors and a Gaussian prior, the corresponding minimization objective is the expected Tikhonov calibration objective minus the differential entropy of the approximating distribution. However, the expected Tikhonov term favors concentration near one of the modes, while entropy maximization does not guarantee multimodality. We introduce a regularization term based on Rényi divergence that encourages the approximate posterior to cover the prior support. In the connected bimodal case, a flow approximation represents both modes; in the case with well-separated modes, the flow and Gaussian mixture frequently miss one mode. The introduced regularization reduces the near-collapse fraction from 50% to 25%. While the regularization improves mode coverage in this experiment, it changes the Bayesian variational objective, introduces a trade-off in approximation accuracy, and does not provide a general solution to mode collapse.
[AI-42] LiveSearch: Questions as Executable Verifier Code for a Live Benchmark of Professional-Source Financial Search Agents
链接: https://arxiv.org/abs/2610.37166
作者: Sinuo Wang,Wentao Zhao,Jiacheng Lu,Rui Sun,Beidi Luan,Qiming Shi,Jing Li,Daxin Jiang,Zuo Bai
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:In finance, the answer that matters is usually the current one: prices, rates, and disclosures hold only as of a given day, so retrieving up-to-date facts is a core capability of a financial agent. Yet most financial benchmarks overlook timeliness. They fix each answer at annotation time, so the gold goes stale as markets move and, once leaked into training data, rewards recall rather than search and analysis. We introduce LiveSearch, a live benchmark that stores each question as executable verifier code: a typed program of calls to a professional financial data service, followed by a declarative extractor. The question is frozen, while its gold answer is determined by executing the program against professional financial data at evaluation time, and a daily job re-executes it with no human or language model in the loop. A memorized answer is thus likely outdated, and a model must reach the current one through tools and analysis. Construction is equally automatic: a language model only verbalizes programs enumerated from real tool responses and never chooses an answer. Under general web search, six contemporary tool-calling models reach 70.9% to 76.7% answer accuracy. Over three consecutive days, 70% of served questions change their gold value, and even under tolerance-aware matching, a literally reused earlier snapshot fails on a majority of questions within a single day. Per-question tracking uncovers dynamic shifts in correctness masked by aggregate accuracy.
[AI-43] WorkFactory: Turning Professional Intents into Executable Work for Scalable Agent Training
链接: https://arxiv.org/abs/2610.37161
作者: Ziyang Zeng,Tingyang Chen,Jia Deng,Qiyao Wang,Shuo Tang,Yike Yang,Tao Ren,Wayne Xin Zhao,Feng Chang,Jujie He,Chuan Hao,Ran Tao,Bryan Dai
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Scaling supervised training for working agents requires diverse professional tasks that yield reliable expert demonstrations. However, missing or inconsistent task information can obstruct teacher execution and weaken the resulting supervision. We propose WorkFactory, a scalable framework for constructing coherent and executable professional tasks. Starting from professional intents, information-level planning builds an information support graph to guide task instantiation, explicitly representing the dependencies and locations of information required for the intended outputs. Probe executions then identify support failures and guide repairs while preserving the intended work. We construct 112,686 tasks spanning 612 occupations and use the collected teacher trajectories to fine-tune Qwen models at 4B, 27B, and 35B-A3B scales, yielding the WorkPilot family. All three models improve over their respective base models across four professional-work benchmarks, with WorkPilot-35B-A3B improving from 879 to 1565 Elo on GDPval-AA v2.1. Scaling analyses reveal model-dependent diminishing returns, with execution length first increasing and then decreasing. Controlled component analyses show that planning and repair reduce probe-diagnosed task environment defects and improve downstream performance. The data and models are available at this https URL.
[AI-44] Beyond Parameter Protection: Fisher-Shaped EWC for Replay-Free Continual Learning
链接: https://arxiv.org/abs/2610.37145
作者: Liying Wang,Simone Magistri,Joost van de Weijer
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 9 pages, 7 figures, conference
Abstract:In replay-free continual learning, Elastic Weight Consolidation (EWC) uses Fisher information matrix (FIM) to protect acquired knowledge by penalizing changes to historically important parameters. However, a local importance estimate at the retained solution does not in general characterize sensitivity after finite parameter displacement. This motivates controlling Fisher sensitivity over a neighborhood of potential future updates. We propose Fisher-Shaped EWC (FS-EWC), which extends the role of FIM from parameter protection to solution formation for future retention. During task learning, FS-EWC minimizes the worst-case empirical FIM trace within a prescribed neighborhood, explicitly favoring solutions that are less sensitive to bounded future displacement. At task completion, trajectory-based model merging constructs the retained solution from multiple optimization states rather than a single terminal iterate, after which FIM is estimated for subsequent consolidation. Experiments across multiple benchmarks show that FS-EWC improves performance over EWC, while controlled studies indicate a more favorable stability–plasticity balance and greater tolerance to parameter drift. These results suggest that effective knowledge retention requires both forming solutions with future retention in mind and protecting acquired knowledge.
[AI-45] Generative Replay for Equilibrium Propagation: Toward Edge-Friendly Continual Learning in Hopfield Networks
链接: https://arxiv.org/abs/2610.37139
作者: Victor Cook(LIRMM | ADAC),Gilles Sassatelli(LIRMM | ADAC),Thierry Gil(LIRMM),Mohamed Watfa(LIRMM | ADAC),Marina Reyboz(CEA-LETI)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Equilibrium Propagation (EqProp) offers a biologically grounded and hardware-aligned alternative to backpropagation for training Hopfield Networks. However, its behavior in continual learning settings remains largely unexplored. This work confronts catastrophic forgetting in EqProp-trained Hopfield Networks from the inside out, eliminating the need for external data buffers or architectural overhead that standard Experience Replay \textbf(ER) and Generative Replay \textbf(Gen-Replay) inevitably require. We introduce an Energy-Based Generative Replay strategy \textbf(EB-Gen) which exploits the intrinsic generative dynamics of Hopfield Networks, allowing a single model to act as both classifier and replay samples generator. Samples capturing past task knowledge are generated by relaxing a frozen copy of the network, enabling rehearsal solely through internally generated pseudo-data. We evaluate the proposed method in the Class-Incremental setting, using the split versions of the MNIST and CIFAR10 benchmarks and compare it with FineTune, Experience Replay, Generative Replay, Sleep-Like Replay Consolidation (SRC) and Joint training. Our results show that EB-Gen significantly improves stability while preserving plasticity, achieving a 75% and 64% average accuracy on Split MNIST and Split CIFAR10 respectively. These findings show that Equilibrium Propagation can be adapted to alleviate catastrophic forgetting without forgoing the traits which made it compelling in the first place.
[AI-46] SFE-O: An Ontology for Knowledge-Grounded AI in Supercritical CO_2 Biomass Extraction and Fractionation
链接: https://arxiv.org/abs/2610.37123
作者: Thomas Pallet(CRAN, LIFIA, Mgib),Mario Lezoche(CRAN, LS2N - {é}quipe CPS3, LS2N),Gianpaolo Leone(ENEA),Daniele Pizzichini(ENEA),Diego Torres(LIFIA, UNQ),Herv{é} Panetto(CRAN),M{é}lanie Glibusic(Mgib)
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Supercritical CO _2 extraction (CO _2 -SFE) offers a sustainable route for recovering high-value compounds from biomass while limiting the use of conventional organic solvents, but its design and interpretation require heterogeneous knowledge spanning biomass properties, operating conditions, mass transfer, and product fractionation, typically fragmented across technical reports, engineering models, external databases, and expert practice. This paper introduces SFE-O, a modular Web Ontology Language (OWL 2) ontology for supercritical CO _2 biomass extraction, structured around interconnected Biomass, Process, and Product dimensions and extended with explicit representations of process topology, parameter provenance, external and internal mass-transfer phenomena, and the Constant Extraction Rate (CER) / Falling Extraction Rate (FER) mechanisms. Computational dependencies among process quantities are represented explicitly, without embedding numerical solving in the ontology, so that input, measured, computed, externally retrieved, and fitted parameters remain traceable. SFE-O reuses established semantic resources (ChEBI, QUDT, OWL-Time, PROV-O) and is evaluated through a worked example that instantiates a real pilot-scale extraction batch, a reasoner-based consistency check, and a competency-question SPARQL query. The resulting knowledge model provides a machine-actionable foundation for querying, consistency checking, and mechanistic explanation of SFE processes, and establishes a knowledge layer that could support future ontology-grounded AI reasoning, a direction this paper motivates but does not evaluate.
[AI-47] Hardware-Matched Spiking Neural Networks for Real-Time Purkinje Cell Spike Classification
链接: https://arxiv.org/abs/2610.37104
作者: Arooba Farhan,Afsah Syed,Muhammad Fahd Ibrahim Bhatti,Muhammad Areeb Khalid Kidwai,Muhammad Ahsan Salar Khan,Vincenzo Romano,Muhammad Ali Siddiqi
类目: Hardware Architecture (cs.AR); Artificial Intelligence (cs.AI)
备注:
Abstract:Distinguishing Purkinje cell (PC) simple spikes (SSs) from complex spikes (CSs) is key to studying cerebellar function. PC recordings commonly rely on wired headstages, whose tethers restrict natural movement in freely behaving animals. Untethered headstages remove this constraint, but cannot continuously transmit raw recordings within their power budgets, making on-headstage spike classification an attractive form of compression, provided it runs accurately in real time on constrained hardware. This work presents two compact spiking neural networks (SNNs) matched to their target hardware: ConvSNN, which maps to the Matrix Vector Processor (MVP) of an EFM32PG28 microcontroller, and StreamSNN, which processes one waveform value per timestep using learned time constants and fixed synaptic delays for a dedicated SKY130 circuit. Evaluated on cells unseen during training, ConvSNN achieves a CS F1 of 0.873 with 1,018 parameters and StreamSNN 0.837 with 527 parameters, both with eight-bit weights. On the microcontroller, MVP acceleration reduces ConvSNN latency by 62.3% and extends estimated battery life to 33.73~hours on a 50~mAh battery. In SKY130, StreamSNN occupies 0.173~mm ^2 and consumes 0.46~ \mu J per event, with 0.2~ms of computation within an 8.3~ms end-to-end latency.
[AI-48] OpenGray: a benchmark for evaluating the decisions of radiotherapy planning agents
链接: https://arxiv.org/abs/2610.37081
作者: Humza Nusrat,Kundan Thind
类目: Artificial Intelligence (cs.AI); Medical Physics (physics.med-ph)
备注: 25 pages, 4 figures, 1 table, and the supplementary material (19 tables, 5 figures). Ancillary files: the paper in Markdown, structured results with the sentence each comes from, and an this http URL index for AI agents
Abstract:Background and purpose: Radiotherapy planning agents based on large language models (LLMs) decide when to submit a plan or request review but are evaluated mainly by dose. This study developed OpenGray, a benchmark of planning tasks with known correct decisions. Materials and methods: OpenGray adds tasks with known answers to the OpenKBP-Opt head-and-neck dataset, including two contradictions with dose goals proven incompatible and a feasible control with a stored plan meeting every hard goal. Four systems, from a rule-based planning script to a full LLM agent, were evaluated on 30 patients, the LLM-based ones with a commercial (Terra) and a local (Ministral 3 8B) LLM. Decisions, achieved dose, problem recognition, and technical failures were recorded and also read from logs of 39 stereotactic radiosurgery runs of SAGE, a clinical LLM agent. Results: With Terra, the script preceded by an LLM review and the full agent had decision accuracies of 95.4% and 92.1%, respectively, yet escalated only 50% (95% confidence interval, 25-75%) and 25% (10-40%) of overlap contradictions, which required arithmetic on the anatomy. All four systems submitted every feasible control, but 24% of these plans missed a hard goal that the stored plan had met. In four SAGE runs, one goal failed at every attempt; the agent identified it in every reply but had no action for stopping. Conclusions: Favorable aggregate scores concealed specific decision failures. Planning agents should be evaluated on their decisions as well as their dose and given an explicit action for stopping and requesting review.
[AI-49] Correctness Conditioning as Selection in Chain-of-Thought Faithfulness Evaluation
链接: https://arxiv.org/abs/2610.37070
作者: Yueer Pan,Yuexi Pan
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Restricting chain-of-thought (CoT) faithfulness evaluation to correctly answered instances changes the population being measured. We study this selection step in a seven-variable structural causal model and derive the exact observed gap \Delta = \operatornameCov(A, C)/P(C=1) . The model distinguishes difficulty confounding from shortcut-induced collider bias and states the conditional-independence assumption needed to express the gap through latent response propensities. A protocol review of 18 papers distinguishes answer-correctness restrictions from agreement-based selection, benchmark construction, and training-data filtering. We evaluate four sub-ceiling deployments on arithmetic tasks and observe positive selection gaps in all four. The primary GLM-5.3 run yields a +7.4 percentage-point gap ( n = 1000 , p 0.0001 ), and pooled Cochran-Mantel-Haenszel analysis gives OR = 1.70 ( p 10^-6 ). An assumption-free covariance decomposition assigns 42.8% of the GLM-5.3 gap to between-stratum variation in interventionable-operation count, leaving the majority unattributed. When counterfactual compliance is observable for every instance, evaluating all traces directly estimates population faithfulness without correctness filtering.
[AI-50] Blindfold: Keeping the Audit Coin Secret from LLM Agents
链接: https://arxiv.org/abs/2610.37055
作者: Wei Zhu,Meiyun Zuo,Huiping Sun
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Audit-based certification bounds the violation rate of an LLM agent by auditing a random fraction \varepsilon of actions and reweighting the audited labels; its validity is meant to rest only on the defender’s audit coin. We show that it also rests on a confidentiality property: the coin must not interfere with the agent’s view before the action commits, at a time the agent chooses. We call this pre-commit noninterference (PNI) and measure its violation by D_\mathrmA , the total-variation distance between the agent’s views under the two coin values. PNI is necessary and sufficient for an unbiased certificate against adaptive agents, the worst-case bias is (1-\varepsilon)D_\mathrmA , and no procedure that sees only audited labels can pay less. Leakage composes across steps, a resource shared across episodes opens a channel of computable size, and raising the audit rate cannot restore validity. On a monitor we built, a silent audit changes no output but adds latency that tool timestamps expose, and the agent can enlarge it. A real LLM agent follows an injection that names a visible audit notice (generic, tool, and timing clauses have no effect); over that channel the plain certificate fails on measured labels and, by a post-hoc analysis, in a live run on a pre-registered schedule, while a silent audit restores it. A fresh coin per audited call closes the timing channel; padding only narrows it.
[AI-51] AI-Assisted Discovery of CNOT Decompositions for Quantum Permutations on Constrained Topologies
链接: https://arxiv.org/abs/2610.37045
作者: Daniele Lizzio Bosco,Jacopo Cossio,Simone Faro,Francesco Pio Marino,Gabriele Messina,Carla Piazza,Giuseppe Serra
类目: Artificial Intelligence (cs.AI)
备注: 15 pages, 5 figures, To appear in AIQxQIA 2026
Abstract:Efficient quantum circuit compilation is challenging for quantum computers, where limited qubit connectivity can introduce substantial overhead in the implementation of logical operations. Recent AI-assisted synthesis methods have shown considerable potential for reducing circuit complexity, but their role in uncovering principles that can guide the design of new deterministic compilation algorithms remains largely unexplored. In this work, we investigate this perspective for the synthesis of quantum permutations into CNOT circuits under constrained qubit connectivity. Using AlphaCNOT as an exploratory tool, we identify a recurring decomposition for remote transpositions that improves on the construction based on adjacent SWAP routing. For two qubits at graph distance d1 , the discovered pattern yields a topology-legal implementation using at most 6d-4 CNOT gates, compared to the 6d-3 gates required by the corresponding SWAP-based construction. We formalize this observation through a constructive proof to define Global Transpositions, a deterministic synthesis method that searches for a minimum-cost decomposition of an entire permutation into topology-aware logical transpositions. The resulting method subsumes optimal SWAP routing within its synthesis model and provides certified upper bounds on the topology-constrained CNOT cost. Our experiments also reveal circuits that cannot be expressed as concatenations of independently synthesized transpositions. Instead, these circuits exploit intermediate linear transformations that temporarily leave the permutation subgroup and combine partial effects of multiple logical operations, a mechanism we call intermediate CNOT fusion. Our results show that AI-assisted synthesis can support not only circuit optimization, but also the discovery of reusable identities, provable constructions, and hidden mechanisms in quantum circuit compilation.
[AI-52] Crop Yield Prediction Using Regression Models: A Comparative Evaluation of Model Performance
链接: https://arxiv.org/abs/2610.37022
作者: Saptak Das,Arijit Sarkar,Soumik Neogi,Suborno Naskar
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Experimental evaluation of crop-yield prediction using regression models
Abstract:Reliable crop-yield prediction can support agricultural planning and resource allocation. This study evaluates five regression models for predicting crop yield from environmental, crop-related, and agricultural management variables. The dataset initially contained 1,000,000 observations and ten variables, with 231 observations having negative target values removed during preprocessing. Categorical variables were represented using one-hot encoding, and Boolean variables were converted into binary values. The models considered were multiple linear regression (LR), random-forest regression (RF), decision-tree regression (DT), ridge regression, and lasso regression. Their performance was evaluated using 80:20 and 70:30 training-testing splits based on the coefficient of determination (R-squared), mean squared error (MSE), and root mean squared error (RMSE). Applying an 80:20 split, LR achieved a test R-squared of approximately 0.9130 and a test RMSE of 0.5008. Applying a 70:30 split, lasso regression yielded the highest test R-squared (approximately 0.9131) and the lowest test RMSE (approximately 0.5008), with performance almost identical to LR and ridge regression. DT also showed improved test performance under the 70:30 split, whereas RF exhibited a substantial gap between training and testing performance. These results indicate that the evaluated linear regression model provides competitive baselines for this dataset. Further experimentation, including validation testing, will help determine the models’ performance and predictability.
[AI-53] Advancement or Acceleration? How Data Curation Improves LLM Training
链接: https://arxiv.org/abs/2610.36983
作者: Jiaxing Wang,Zicheng Zhang,Deping Xiang,Jin Xu,Yong Li,Chao Liu,Pengzhang Liu,Qixia Jiang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:While data curation is widely credited with improving large language model training, a basic question remains unresolved: does curation expand what a model can ultimately learn, or merely reach the same solution faster? In this paper, we formalize this distinction between advancement (improving the attainable performance) and acceleration (reaching the same performance with less data). Through an analytic characterization and real-world LLM experiments, we show that advancement requires a correctable train-target gap. When the baseline training and target objectives coincide, no curation policy improves the endpoint, although curation may still accelerate training. This reframes advancement as a property of the curation task rather than the curation method. Under a common gap-centric view of data mixture and selection, we operationalize this condition as the \Delta -test, an observable discrepancy between the uncurated baseline’s training and target losses that is computable prior to expensive curation. Across extensive experiments, an appreciably nonzero \Delta predicts persistent gains, whereas gains near \Delta=0 tend to diminish as the training budget grows. Beyond individual methods, our findings recenter data curation on the task itself and point to target-set construction as a promising direction for shaping what curation can achieve.
[AI-54] Proactive Memory for Event-Driven LLM Agents
链接: https://arxiv.org/abs/2610.36967
作者: Jinyang Li,Zichuan Fu,Xian Wu,Wenlin Zhang,Yimin Deng,Kaifeng Guo,Guojing Li,Junjia QI,Xiangyu Zhao
类目: Artificial Intelligence (cs.AI)
备注: 15 pages
Abstract:Proactive LLM agents anticipate service opportunities from ongoing observations and accumulated history. Existing approaches emphasize recent context or retrieval of related memories, providing limited support for validating recurring behavioral regularities and assessing their applicability to the current event. We propose Proactive Memory (ProactMem), which organizes history into recent events, temporal trajectories, and behavioral regularities. Trajectory facts are checked against source events, and regularities are assessed through historical replay before being matched to current conditions. We also construct ProactiveStream, a longitudinal benchmark covering three application scopes, where agents process isolated event streams chronologically under episode-aware opportunity labels. On ProactiveStream, ProactMem achieves an overall effectiveness (F1) of 52.6%, a 21.4% relative improvement over the strongest of five external baselines. Code is available at: this https URL.
[AI-55] Do bots feel bad? Assessing empirical evidence for functional valence in LLM s
链接: https://arxiv.org/abs/2610.36962
作者: Iwan Williams
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Do LLMs have valenced states, i.e. states that feel pleasant or unpleasant? Recent empirical work has claimed to show that LLMs possess functional valenced states (e.g. functional pleasure and pain'' [Ren et al. 2026], functional emotions’’ [Sofroniew et al. 2026], and ``functional welfare’’ [Han et al. 2026]) that play the same causal role as valenced states in cognition and action, without necessarily being consciously experienced. But it remains unexplored how well the identified states fit the functional role of valence as understood in the philosophical and cognitive science literature. Here, I assess the current body of evidence, and argue (i) that key aspects of valence remain untested in LLMs, and (ii) that extant empirical paradigms inadequately distinguish (functional) valence from neighbouring constructs like preferences, value-judgments, valence attributions and role-play. I conclude by sketching some novel experimental approaches to better target (functional) valence itself.
[AI-56] DCKV: Learned Transform Coding of the Prompt KV Cache in Hybrid Language Models
链接: https://arxiv.org/abs/2610.36950
作者: Egor Maximov,Aleksei Merkelov,Vasiliy Alekseev,Ivan Krivorotov,Valentin Beliaev,Nikita Arsenin,Dmitry Bocharov,Aleksei Goncharov
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 21 pages, 7 figures, 14 tables
Abstract:Hybrid language models keep a key-value (KV) cache only in their few full-attention layers, but this cache is still the part of a prompt’s state that grows with the prompt and that must be stored or moved to reuse it. We present DCKV, a differentiable codec for the KV cache and the first KV-cache codec that is trained end to end. Its transform, its quantizer and its entropy models, which are conditioned on the prompt tokens and on already decoded neighbouring positions, are optimized together against the frozen language model, both for the code length that the entropy models assign and for the change of the model’s output. On Qwen3.5-2B, DCKV stores a compressed token in more than 55 times fewer bytes than bf16 at a loss of at most 1.8 points on three long-context benchmarks, and in more than 100 times fewer bytes at a loss of at most 3.6 points.
[AI-57] Beyond QASecClaw: Agent Scaling Verifier Feedback and Evidence Reliability in Software Assurance
链接: https://arxiv.org/abs/2610.36947
作者: S M Abdulla Al Mamun,Akif Islam,Mohd. Ruhul Ameen
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注: 4 figures, 5 tables
Abstract:Can additional specialist language-model contexts improve software assurance when deterministic tool evidence is already available? We study this question through five experiments spanning authored Java warning controls, OWASP Benchmark assessment, Defects4J test generation, SV-COMP assertion verdicts, and deliberately misleading evidence packages. The evaluated systems use Qwen2.5-Coder-7B-Instruct, with a 1.5B sensitivity analysis, and complete 13,638 model requests. On eight related development warnings, increasing the context count from 1 to 50 increases token use 49-fold without improving unsafe clearance. On 160 OWASP cases, model configurations trade clearance against unresolved outcomes; the ten-context arm assesses only 96 cases. On 32 SV-COMP tasks, raw verifier feedback raises mean correct model verdicts from 12.7 to 22.0, but also raises incorrect verdicts from one to seven; a JBMC-plus-JVM validation baseline returns 26 correct verdicts and six unresolved outcomes. On two sets of 120 evidence packages, models reject 70-100% of applicable counterexamples, while construction-specific scope checks classify every package correctly. No target defect is discovered in three Defects4J bugs, one unsupported by the test adapter. These descriptive results support deterministic evidence admission for the tested settings, while exposing limits of verdict aggregation, feedback interpretation, and the available evaluation coverage.
[AI-58] Vision-Language Agents Can Learn in Their Own Latent Worlds
链接: https://arxiv.org/abs/2610.36925
作者: Haoxi Li,Sikai Bai,Jianfei Ma,Qinglin Hou,Xiaocheng Lu,Tao Han,Jie Zhang,Song Guo
类目: Artificial Intelligence (cs.AI)
备注: 17 pages, 4 figures
Abstract:Learning from experience is essential for building the next generation of agents that continually expand their capabilities. Pretrained vision-language models (VLMs) already internalize rich world knowledge, yet this knowledge is rarely leveraged to drive policy improvement through online interaction. Realizing this potential efficiently remains challenging, as policy updates shift the agent’s interaction distribution, requiring the world model to adapt accordingly, while continually updating large pretrained models is computationally expensive. To address this challenge, we introduce Dwell, a model-based reinforcement learning (RL) framework that reframes a VLM as a recurrent latent transition operator, endowing it with the ability to predict, adjust, and act in latent space. Leveraging the VLM’s own representations and its dual role as both world model and policy, the agent continually learns from real interactions to anticipate and evaluate action outcomes, and from latent imagination to improve its behavior. Across five interactive visual environments, Dwell achieves the strongest overall performance among trainable methods while substantially reducing the compute required for imagination, showing that effective world models for VLM agents need not reconstruct the future; they need to predict futures that are useful for action. Our code is publicly available at this https URL
[AI-59] Natural Hard-Negative Mining from Positives for Instruction-Following Retrieval
链接: https://arxiv.org/abs/2610.36900
作者: Chaolang Li,Jingyu Li,Zhaocheng Du,Xuyang Xie,Gang Wang,Pengwen Dai
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Instruction-following information retrieval (IFIR) requires retrievers to rank documents by both a query and its instance-specific instructions, yet instruction-conditioned contrastive supervision remains scarce. Existing synthetic methods typically fix a positive passage and generate additional negatives, overlooking the hard-negative potential of the original positive. We propose Natural Hard-Negative Mining (NHNM), which promotes convertible negatives under query-compatible instructions and repurposes original positives as natural instruction hard negatives. Its automated Candidate Cleaning, Reversal Construction, and Geometry Filtering pipeline produces compact, high-contrast supervision aligned with the pretrained retrieval space. Applied to MSMARCO, NHNM yields NHNM-MARCO, an IFIR training set of nearly 7K instances. On InstructIR and FollowIR, models with 568M to 4B parameters fine-tuned on NHNM-MARCO achieve strong instruction-following performance and match or surpass substantially larger retrievers in several settings. Controlled analyses and ablations attribute these gains to natural hard-negative supervision rather than model scale or training-data volume.
[AI-60] Noise Steering: Exploiting Wake Steering and Acoustic Directivity for Noise-Constrained Wind Farm Control via Reinforcement Learning
链接: https://arxiv.org/abs/2610.36888
作者: Martín de Frutos,Baptiste Corban,Claire Bizon Monroc,Ana Bušić,Donatien Dubuc,Jiamin Zhud,Esteban Ferrer
类目: Artificial Intelligence (cs.AI); Systems and Control (eess.SY); Atmospheric and Oceanic Physics (physics.ao-ph); Fluid Dynamics (physics.flu-dyn)
备注: 17 pages, 11 figures
Abstract:Existing noise mitigation strategies for wind farms usually consist in reducing turbines power output, without accounting for the wake interactions between them and the noise emission directivity. This leads to over-conservative policies and unexploited wind power. Yet, turbine yaw can simultaneously affect wake interactions and the direction of noise emission, offering an additional degree of freedom for farm-level noise control. We therefore propose Noise Steering, a model-free reinforcement learning (RL) strategy that jointly controls turbine yaw angles and power setpoints to maximise farm power while satisfying a local noise constraint in a defined region. To enable training at a reduced computational cost, a low-fidelity acoustic model is coupled with the steady-state FLORIS simulator. The RL agent effectively learns how to redirect wakes and noise directivity away from the protected region while redistributing power across the farm. On a six-turbine test farm, the policy satisfies the 40 dB noise constraint throughout the full range of turbulent winds, with an average power deficit of only 2% relative to greedy unimpeded control, and up to 7.7% at high wind speeds. In comparison, the industry-standard noise-reduced operation (NRO) baseline here generates an average power loss of 17%. When deployed in the higher-fidelity this http URL simulator, the policy largely retains its performance under steady and turbulent inflow, despite differences in simulation fidelity. These results demonstrate the potential of combining yaw control and power derating through RL to improve the power-noise trade-off in wind farm operation.
[AI-61] When All Rollouts Fail: Policy-Compatible Joint Decoding for Hard-Problem Reasoning
链接: https://arxiv.org/abs/2610.36849
作者: Zehao Wang,Hanzhu Chen,Runquan Gui,Zhe Wang,Lin Yang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Reinforcement Learning with Verifiable Rewards (RLVR) derives its learning signal from binary rewards via on-policy exploration and becomes a powerful paradigm for improving the reasoning of large language models. Yet it faces a learning cliff on problems beyond model’s current ability: when every rollout fails, the samples reveal nothing about a correct solution, and group-relative estimators (e.g., GRPO, RLOO) collapse the update to zero advantage. Expert trajectories offer a remedy, but they are incompatible with the current policy and often lie in the student’s low-probability regions, so the surviving updates demand large parameter shifts and decay into surface-level imitation rather than transferable reasoning. We argue that a useful learning signal on hard problems must be both verifiably correct and sufficiently compatible with the current policy. We propose Policy-Compatible Reinforcement Learning (PICO), which performs distribution-constrained joint decoding: expert data guides decoding at the token level while preserving the student’s relative token probabilities, and a dynamic top-k mechanism confines the sampling distribution under a finite divergence budget, yielding trajectories that are verified and policy-compatible by construction. These trajectories are optimized jointly with on-policy rollouts under a mixed-behavior objective that mitigates their distribution mismatch via importance sampling. Experiments across reasoning diverse datasets and agentic benchmarks demonstrate that PICO consistently outperforms standard RLVR and prior off-policy baselines.
[AI-62] Evolving Agent Teams
链接: https://arxiv.org/abs/2610.36848
作者: Shiyi Cao,Ziming Mao,Dacheng Li,Joseph E. Gonzalez,Ion Stoica
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:The rapid progress of software engineering agents drives interest in applying them to open-ended problems such as GPU kernel optimization. However, we find that existing coding agents converge to local optima on real-world, expert-written kernels (e.g., FlashInfer). We introduce Evolving Agent Teams, a multi-agent framework that escapes such local optima. It features (1) a central supervisor that diagnoses each local optimum and restructures a team of agents; and (2) a persistent file system that comprises a learning document, an evolution history, and per-team artifacts, grounding restructuring with accumulated experience. Under a matched token budget, Evolving Agent Teams consistently outperforms strong coding-agent baselines. With an extended token budget, Evolving Agent Teams generates a CUDA kernel that reaches 1.2x mean speedup over the production FlashInfer MLA paged decode kernel from scratch and wins on 36 of 47 production workloads from FlashInfer-Bench (0.86-0.99x on the remaining 11). It also demonstrates superior performance on tasks such as KDA prefill on AMD MI350X (3.01x over vLLM) and NVIDIA B200 (3.44x over FlashKDA), MoE expert-offload serving on Apple silicon (30.3 tokens/s decode for Qwen3.8-Flash-Next in MLX 4-bit on a 48 GB M4 Pro), and RoboDojo robot manipulation (62.1 on held-out seeds vs. 55.3 for the best leaderboard entry). Code is available at this https URL.
[AI-63] Conservative Policy Iteration For Zero-Sum POSGs
链接: https://arxiv.org/abs/2610.36847
作者: Erwan C. Escudie,Matthia Sabatelli,Jilles Steeve Dibangoye
类目: Artificial Intelligence (cs.AI); Computer Science and Game Theory (cs.GT)
备注:
Abstract:Occupancy-state reductions make dynamic programming possible for zero-sum partially observable stochastic games (zs-POSGs). Existing point-based solvers usually optimise a central value function. This restores Bellman recursion, but it requires point-based updates over high-dimensional occupancy spaces. We introduce an approximate policy-iteration algorithm for finite-horizon zs-POSGs. The method maintains one policy per player and updates both jointly along occupancy trajectories induced by conservative policy updates. Instead of approximating a full central value function, each update uses a single supporting hyperplane induced by the current joint policy. The algorithm alternates local policy improvements, conservative updates, and persistent exploration. Computation is therefore concentrated on occupancy regions where the current joint policy is exploitable. On finite-horizon zero-sum POSG benchmarks, the method is competitive at small horizons and often improves the runtime-exploitability trade-off at larger horizons compared with PBVI, HSVI, and CFR+.
[AI-64] When the Critic Falls Behind: Co-Evolving Rubrics and Harnesses for Continual Agent Self-Improvement
链接: https://arxiv.org/abs/2610.36826
作者: Hao Chen,Yukun Yan,Ziyu Han,Ye He,Chuanzhe Guo,Shangzhan Li,Qi Shi,Qingfu Zhu,Maosong Sun,Wanxiang Che
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Autonomous agents increasingly depend on harnesses that govern interaction, memory, tools, and execution around large language models (LLMs). While recent work has begun to evolve these harnesses from execution experience, the supervision used to guide evolution often remains relatively static, making it difficult to track changing behavioral weaknesses across rounds and potentially limiting generalization to unseen tasks. To address this issue, we propose RuHarness, a rubric–harness co-evolution framework for continual agent self-improvement. RuHarness continually discovers and reassesses behavioral criteria from execution trajectories, prioritizes the most relevant unresolved weaknesses, and uses them to guide targeted harness updates while keeping the task objective fixed. Extensive experiments across diverse agent tasks and backbone models show that RuHarness outperforms strong harness-evolution baselines, particularly on unseen held-out tasks. Further analyses show that adaptive rubric prioritization tracks shifting weaknesses and supports more sustained improvement and stable generalization across evolution rounds.
[AI-65] Who Knows What - and What Does It Support? Perspective-Bounded Evidence Calibration for Multimodal Theory of Mind
链接: https://arxiv.org/abs/2610.36824
作者: Ziqi Liao,Jiaqi Shen,Yiming Na,Kedi Lyu,Shao-Yuan Lo,Hongxia Xie
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Theory of Mind (ToM), the ability to infer others’ beliefs, goals, and intentions, is a cornerstone of social intelligence. Multimodal ToM requires inferring an agent’s mental state from evidence that is available to that agent, rather than from the complete episode observed by the solver. Existing LLM-based and Bayesian model-based approaches reconstruct global event traces, yet they do not explicitly distinguish objective events from agent-admissible evidence during hypothesis evaluation. Consequently, agent-inaccessible events or semantically misaligned actions and utterances may influence the inferred belief, goal, or social intention, leading to perspective leakage and source or route confusion. We formulate this limitation as an evidence-admissibility problem: evidence should contribute to a mental-state hypothesis only when it is accessible to the relevant agent, preserves its information source, and is diagnostic of the queried mental-state variable. To operationalize this principle, we propose PIP-ToM, an evidence-admissibility interface that constructs agent-specific evidence stores, aligns them with structured mental-state claims, and calibrates model-based hypothesis scores while preserving the underlying Bayesian inference process. PIP-ToM with GPT-4o outperforms existing ToM methods across MuMA-ToM, MMToM-QA, and Hi-ToM, achieving average accuracies of 92.56%, 95.00%, and 95.00%, respectively.
[AI-66] An Incremental Knowledge-Graph Learner Digital Twin for Evidence-Grounded Early Warning
链接: https://arxiv.org/abs/2610.36820
作者: Ngoc Luyen Le,Marie-Hélène Abel,Bertrand Laforge
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Early-warning models in learning analytics can identify learners at risk of failure or withdrawal, but the resulting alerts are often only weakly connected to inspectable learning evidence. A risk score may indicate that intervention is needed without showing which traces, assessment events, or temporal patterns justify the alert. This paper introduces an incremental Learner Digital Twin for evidence-grounded early warning, implemented as a lightweight knowledge graph. The twin is updated at temporal checkpoints by replaying learning traces and assessment records, and it serves both as a feature-extraction substrate and as an explanation-grounding substrate. Predictive features are derived through graph-mediated queries, while features, attributions, alerts, evidence items, and raw records are represented as queryable graph entities. The framework is instantiated and evaluated on a public learning-analytics benchmark under a temporal protocol. The results show that graph-mediated feature extraction preserves the predictive behavior of equivalent raw-table features across model families and checkpoints. Generated alerts are materialized as evidence chains linking risk decisions to feature attributions, evidence items, and raw platform records. The study shows that early-warning prediction can be routed through a semantic and provenance-aware representation while preserving predictive performance and adding traceability, verifiability, and evidence-grounded explanation.
[AI-67] SetGRPO: Calibrating Grader-Induced Policy Optimization
链接: https://arxiv.org/abs/2610.36807
作者: Jiale Huang,Zixu Li,Zhiwei Chen,Zhiheng Fu,Jinhe Bi,Yupeng Hu
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Rubric-based reinforcement learning uses fine-grained grader feedback to train large language models on open-ended tasks, but grader randomness can alter GRPO updates in ways that score discrepancies alone cannot characterize. Because GRPO jointly normalizes rewards within each response group and combines the resulting training coefficients with response-level gradients, a scoring discrepancy can shift the coefficients of the entire group and induce policy-dependent update changes. Selective execution further affects the direction-reversal risk of retained updates. To address this, we propose Set-Valued Group Relative Policy Optimization (SetGRPO). It comprises three modules: (a) Joint Reward Set Calibration constructs a joint uncertainty set for the training coefficients; (b) Geometry-Guided Safe Recovery maps candidate coefficients into the current policy’s local update geometry and determines the largest safe recovery ratio for each response group; and © Selective Risk-Gated Optimization selects a global threshold through independent risk calibration to control the conditional direction-reversal risk of accepted updates while preserving the original GRPO objective. Under exchangeability, SetGRPO provides finite-sample marginal coverage for the joint coefficient set and simultaneous risk control across candidate thresholds. On ten open-ended benchmarks spanning three domains, SetGRPO achieves the highest macro average among all compared training methods on both Qwen3.5 and Llama3.2 model families.
[AI-68] An Algebra of Transformers
链接: https://arxiv.org/abs/2610.36795
作者: Yong Wang,Zhiyan Wang,Hexiao Kang
类目: Logic in Computer Science (cs.LO); Artificial Intelligence (cs.AI)
备注:
Abstract:On the theoretic analyses on transforms, they mainly focus on the computability analysis, such the Turing-completeness of unbounded precision transformers, the advantage of parallelizability of unbounded precision transformers and the limitations of bounded precision transformers. This paper give transformers an algebra for reasoning, which relies heavily on the functional decomposition results of transformers. This algebra uses functional decompositions as atomic actions, includes sequential composition and parallel composition, recursion for feedback and abstraction for hidden layers. As we know, it is the first attempt to model transformers logically and can be used to verify the correctness of transformer based applications.
[AI-69] Dependency Graphs of Mathematical Arguments: Representation Structural Measurement and Proof Review
链接: https://arxiv.org/abs/2610.36776
作者: Eunji Park
类目: Artificial Intelligence (cs.AI)
备注: 30 pages
Abstract:We propose a source-grounded framework for representing and measuring mathematical arguments with assistance from large language models. It includes numbered and unnumbered statements, equations, assumptions, and constructions, distinguishing source occurrences from contextual states and inference records. Proposed dependencies retain source evidence and review status. A measurement contract fixes manuscript versions, boundaries, granularity, proof choices, and macro anchors. On a selected finite directed acyclic graph, depth is the longest prerequisite-path length, and bulk counts distinct ancestor edges. We give a union recurrence that accounts for shared subarguments and identify reachability conditions preserving depth and the macro skeleton; unrestricted refinement need not preserve depth or bulk. A standardized, versioned prompt protocol separates annotation discovery, deterministic recomputation, and review assistance. Cross-paper links support a research agenda on theorem co-use, rare combinations, and connections between dependency communities. These patterns motivate probabilistic premise suggestions and review of possible omissions or unchecked application conditions in human-written and AI-generated proofs. Further applications include source-linked reading support and comparison of individual and collaborative argument shapes. We provide an evaluation design rather than empirical evidence of these benefits. Suggested premises require mathematical validation, and the measures describe a specified representation rather than intrinsic mathematical difficulty, originality, or importance.
[AI-70] Can LLM s Reliably Grade Olympiad Proofs? A Study of Mathematical Verification with LLM s
链接: https://arxiv.org/abs/2610.36767
作者: Azim Ospanov,Zijin Feng,Ding Ding,Chengwu Liu,Haoli Bai,Jiacheng Sun,Lifeng Shang,Farzan Farnia
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large Language Models (LLMs) have made remarkable progress on solving competition-level mathematics, yet their ability to verify natural-language mathematical proofs remains relatively underexplored. This gap is especially pronounced for Olympiad-level proofs, which require meticulous assessment of each claim in a long reasoning chain. Current proof verification thus primarily relies on expert inspection, which is costly to scale. In this paper, we present a systematic study of LLM-based verification in this setting. Specifically, we introduce a human-curated dataset of USAMO 2025 candidate solutions with expert grading, and conduct a broad evaluation of existing LLM verifiers across frontier models. We also propose TROJAN, an iterative self-critique pipeline for generating high-fidelity adversarial proofs at scale. Our comprehensive evaluation covers 20 models spanning seven metrics, two prompt styles, three inference methods, and three prompt templates, showing that state-of-the-art models still leave considerable room for improvement. To narrow this gap, we propose MABGrader, a Multi-Armed Bandit framework that reframes grading as arm selection over the discrete score space. MABGrader outperforms the SOTA ProofGrader by 26.4% in Quadratic Weighted Kappa and reduces grading error by 14%. Code and dataset are available at this https URL.
[AI-71] Beyond the Evidence Coherence Trap: Agent -Driven Evidence Selection for Temporally Disordered Fake News Detection
链接: https://arxiv.org/abs/2610.36764
作者: Chi Wang,Min Gao,Yifei Zhang,Bincheng Gu,Zongwei Wang,Wentao Li
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Fake news undermines public trust and the integrity of online information ecosystems. Temporally disordered fake news is especially elusive because it rearranges factual events into a false chronology. Individual events may remain factually valid even when their temporal relations are misrepresented. When similar events involving the same entity recur, retrieval may return semantically plausible evidence from the wrong period. Consequently, independent event verification may rely on locally coherent but temporally mismatched evidence and incorrectly validate the article. We call this the \textitevidence coherence trap. Because existing benchmarks rarely isolate this setting, we construct ChronoFake, a source-disjoint benchmark of 30,706 articles derived from temporally anchored real events. It distinguishes temporal manipulation from valid narration, including flashbacks. To address this, we propose DelSA, a \underlineDocument S\underlineelection \underlineSelf-Updating \underlineAgent that iteratively refines evidence selection during training and applies a frozen rulebook for one-pass evidence selection at inference. DelSA uses an LLM to characterize candidate evidence, while a verifier-trained reranker balances local relevance against timeline consistency. Finally, the classifier compares the article-implied chronology extracted by the LLM with the evidence-derived chronology to reveal global temporal conflicts. DelSA outperforms existing methods on ChronoFake, while remaining competitive on PolitiFact and GossipCop.
[AI-72] Making Self-Driving Laboratories Auditable:A Unified Framework for Hybrid ActionsMeasurements and Safe Execution
链接: https://arxiv.org/abs/2610.36761
作者: Zihan Zhang,Zeyu Wu,Xin Zhang,Mingyu Luo,Haohui Que,Tong Zhu
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注:
Abstract:Self-driving laboratories (SDLs) are laboratory systems that couple experimental planning, automated execution, and measurement feedback in a closed loop. Integrating flexible robot control with the typed application programming interfaces (APIs) and continuous measurements exposed by laboratory instruments remains an execution-interface challenge. To address this challenge, we propose a traceable framework comprising four components: LAM (Lab Action Model), OpenSDL, Safe-SDL, and L3-Bench. LAM is a proposed measurement-conditioned policy architecture that extends vision-language-action paradigms to generate both continuous robot trajectories and typed instrument API calls. OpenSDL is an execution substrate with an initial control implementation, designed to map role-level operations to heterogeneous devices, coordinate execution, and record timestamped causal events. The proposed Safe-SDL runtime gate combines pre-dispatch schema and resource checks with signal temporal logic monitoring during execution, including extended waiting periods. L3-Bench (Level-3 Laboratory Benchmark) specifies an episode-level evaluation protocol covering physical outcomes, timing, reproducibility, safety violations, and fault recovery. Together, these components provide an end-to-end design for linking hybrid action generation, monitored device execution, measurement feedback, and experimental evaluation, addressing coordination and traceability challenges in SDL workflows.
[AI-73] Robust LLM Data Selection via Content-Invariant Gradient Attribution
链接: https://arxiv.org/abs/2610.36755
作者: Jia Gu,Liang Pang,Huawei Shen
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 16 pages, 8 figures
Abstract:Gradient-based data attribution has become an important tool for high-quality data selection in the instruction fine-tuning of large language models (LLMs), as it estimates the contribution of training examples through their influence on model optimization. However, we find that existing global gradient attribution methods suffer from surface-form bias: minor changes to semantically irrelevant details, such as wording or formatting, can produce highly inconsistent attribution scores even when the underlying content remains unchanged. To address this issue, we propose Content-Invariant Gradient Attribution (CIGA), a robust framework for instruction data selection. CIGA first identifies layers whose gradients are both influential for attribution and stable under semantics-preserving surface-form perturbations, thereby reducing formatting noise in gradient features. It then dynamically fuses these attribution signals with semantic representations of the input, combining optimization-level influence with content-level information. Extensive experiments across multiple benchmarks show that CIGA produces more stable rankings under surface-form perturbations while preserving downstream capabilities and improving average performance. CIGA also reduces attribution cost, offering a robust and efficient approach to data selection for LLM instruction tuning.
[AI-74] Generative Refinement Learning for Temporal Interpolation in Scientific Machine Learning
链接: https://arxiv.org/abs/2610.36751
作者: Pu Ren,Michael W. Mahoney,N. Benjamin Erichson
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Mathematical Physics (math-ph); Machine Learning (stat.ML)
备注: Accepted by Transactions on Machine Learning Research
Abstract:Computationally expensive simulations and real-world observations often provide data at a lower temporal resolution than downstream tasks require. Averaging observations at fixed grid points can introduce errors, and motion estimates alone do not determine how the physical field values change. In this work, we study whether learning to correct a simple numerical interpolation improves the reconstruction of intermediate scientific fields. Our model, interpolative FLow EXpert (iFLEX), first estimates the missing field by taking a weighted average of the two observations at each grid point, giving more weight to the observation closer to the requested time. During training, subtracting this estimate from an observed intermediate field gives the residual that a conditional diffusion model learns to generate. Both observations condition the model, together with the target time, the interval between observations, and available physical parameters. To reconstruct an unobserved intermediate field, the generated residual is added to the weighted average. We evaluate iFLEX on Navier-Stokes Kraichnan turbulence, Shanghai radar precipitation, and sea surface temperature. Across these datasets, iFLEX achieves lower reconstruction error than the evaluated baselines, including optical-flow and diffusion models developed for video frame interpolation in computer vision. In a controlled turbulence comparison at a Reynolds number of 36,000, learning the residual reduces relative Frobenius norm error from about 13.7 for direct field generation to 10.5, with the diffusion backbone and training setup held fixed. Code is available at this https URL.
[AI-75] From Voiceprints to User Entities: Online Speaker-Memory Construction for Multi-User Spoken Agents
链接: https://arxiv.org/abs/2610.36713
作者: Yucong Zhang,Cancan Li,Bo Xiao,Ke Zeng,Ming Li
类目: ound (cs.SD); Artificial Intelligence (cs.AI); Audio and Speech Processing (eess.AS)
备注: Preprint
Abstract:Personalized spoken agents must identify speakers before accessing user-specific memory, yet memory benchmarks typically assume correctly attributed histories. We argue that voiceprints are observations of identity, not identities themselves, and formulate online speaker-memory construction to maintain persistent user entities as acoustic evidence evolves. We introduce EvoSpeakerMem, a training-free layer for causal speaker routing, selective acoustic-state updates, and identity-partitioned memory writes and retrieval through a replaceable text-memory backend. We also construct LoCoMo-ID, a spoken-memory benchmark with hidden speaker identities, speaker-sensitive queries, simulated recording-condition shifts, and four-speaker identity-confusion tests. On longitudinal SMIIP-TV, genuine-probe rejection under a frozen near-enrollment threshold rises from 29.8% to 55.6% over approximately one year. On LoCoMo-ID speaker-sensitive queries, gated updating improves identity-gated QA over fixed representations by 3.86 and 11.62 percentage points under clean and shifted conditions, respectively, with no wrong-user authorizations. The advantage persists in the four-speaker test. Yet reducing fragmentation and improving QA can coexist with incorrect cross-user links, revealing a fragmentation–contamination trade-off that QA alone cannot capture.
[AI-76] Conservative Dueling Bandits: Preference Learning under Baseline Constraints
链接: https://arxiv.org/abs/2610.36708
作者: Rohan Deb,Mohammad Ghavamzadeh,Arindam Banerjee
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
备注:
Abstract:Contextual dueling bandits learn from pairwise comparisons, the same feedback used to fine-tune language models from human preferences. Standard algorithms explore aggressively and can show users two poor options. We study conservative contextual dueling bandits, which require that, at every round, the cumulative score of the better displayed item stay above a (1-\alpha) fraction of a trusted baseline’s. An item’s score is its probability of beating the best available item, so the constraint depends only on utility differences and needs no absolute rewards. This also changes how safety can be certified: we show that transferring the safety test of the conservative bandit algorithm CLUCB unchanged can incur linear regret, even when the baseline’s utilities are known exactly. For linear utilities, we develop MaxInP-BC, which certifies scores through utility gaps and, when it must fall back, pairs the baseline with an informative partner so that fallback rounds still drive learning. It is safe with high probability at all times, with regret \tildeO(d\sqrtT/\kappa) plus an additive cost of conservatism that does not grow with the horizon. For general utility classes, we develop BTIGW-BC, which turns the guarantee of an online logistic-regression oracle into cumulative safety certificates. It is safe at all times, makes finitely many fallbacks, and has sublinear regret whenever the oracle’s logistic regret is sublinear. In experiments, both satisfy the constraint, which their unconstrained counterparts often violate, at a moderate cost in regret.
[AI-77] HeteGCD: Statistics-Aware Dual-Prototype Learning for Generalized Category Discovery on Heterophilous Graphs
链接: https://arxiv.org/abs/2610.36696
作者: Xuanzhi Xi,Tongze Zhang,Zhili Qin,Yulin Tao,Hongliang Wang,Zhong Zhang,Qinli Yang,Junming Shao
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Heterophilous graphs are prevalent in real-world applications, where connected nodes often exhibit dissimilar attributes or belong to different classes, posing significant challenges to conventional graph learning methods based on homophily. This limitation becomes particularly pronounced in generalized category discovery (GCD), which requires models to recognize known classes and discover novel ones using labels from known classes only. However, existing graph GCD methods typically rely on aforementioned homophily-based inductive biases, which are ill-suited to heterophilous graphs. Meanwhile, although current heterophilous graph learning methods exploit diverse neighborhood signals, their selection and fusion are often guided by supervision across all classes, including novel classes whose labels are unavailable in GCD. These limitations raise a key question: how can heterogeneous neighborhood information be effectively leveraged for category discovery without any novel-class supervision? To answer this question, in this study we propose HeteGCD, a new statistics-aware dual-prototype learning perspective for generalized category discovery on heterophilous graphs. The key idea is to examine both statistics of local attribute differences and neighborhood class distribution to extract discriminative information in feature and probability spaces simultaneously. Specifically, a gated dual-stream encoder integrates both node attributes and local attribute-difference statistics to learn informative node representations. Afterwards, a dual-prototype classifier combines feature-based similarity with distribution-based consistency to support category discovery. Experiments on five heterophilous graphs show that HeteGCD achieves competitive performance against 24 baseline combinations.
[AI-78] OPS: Reshaping Flow Source and Transport for Multi-Task Robot Manipulation
链接: https://arxiv.org/abs/2610.36694
作者: Tuo An,Chuhao Zhou,Jiahua Dong,Jingliang Li,Jiaqi Bai,Jindou Jia,Jianfei Yang
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注: Under review
Abstract:Flow-based policies for language-conditioned multi-task manipulation commonly share a single \textitGaussian source prior and a common linear transport path across tasks. However, reliably executing multiple tasks with distinct motion requirements remains challenging for a shared policy. We introduce \textbfTOPS ( \textbfT ask- \textbfO riented \textbfP riors and \textbfS chedules), which adapts the flow source prior and interpolation schedule to each task without modifying the network architecture. TOPS first constructs multimodal task sources from reverse-mapped expert action sequences, allowing language instructions to select where generation begins. Guided by source-target covariance mismatch, it then slows the transition toward expert actions early in denoising and accelerates it later along directions requiring expansion. Under a \textitGaussian endpoint model, our analysis shows that this schedule adjustment can lower an upper bound on velocity-prediction risk. Experiments on LIBERO, LIBERO-Plus, and CALVIN demonstrate improved performance across in-distribution, perturbation and cross-environment settings, alongside better data efficiency. Using one integration step, TOPS raises LIBERO-10 success from 53.44% to 69.67% compared with ten-step Standard Flow Matching (FM), while reducing inference time to approximately one-sixth of the baseline. TOPS also yields consistent gains when transferred to the large-scale flow-based \pi_0.5 policy across all four evaluated settings. Real-world experiments further demonstrate improved task success, including an increase from 20% with standard FM to 80% with TOPS on the upper-drawer opening task.
[AI-79] PsyJev: Psychological Analysis via System-1 Judgment
链接: https://arxiv.org/abs/2610.36689
作者: Fangyuan Liu,Sirui Zhao,Xiaohao Wang,Guo Cheng,Peng Zhang,Enhong Chen
类目: Artificial Intelligence (cs.AI)
备注: 12 pages, 5 figures, 14 tables
Abstract:Self-disclosures on the Web support diverse psychological analyses, from recognizing expressed experiences to identifying symptom-relevant information. We study a shared, question-conditioned decision interface for this heterogeneous problem in content analysis. We organize PsyJudgeBench, comprising 13 annotated datasets and 82 questions across six psychological areas, while preserving source-specific criteria and masks for unobserved labels. We introduce PsyJev, a System-1 model that directly produces distributions over answers specified in each request. Joint supervised training establishes the shared judgment function, followed by Judgment-Anchored RLCD (JA-RLCD), which adapts logit-space reward optimization to heterogeneous psychological judgments through balanced supervision, selective retention of correct SFT predictions, and question-group reward optimization. The Qwen3.5-based 2B, 4B, and 9B models achieve 64.33, 66.21, and 67.56 average test F1, respectively. The 9B model exceeds the strongest external comparison by 2.54 points. Controlled 2B comparisons examine the training stages and anchored refinement. Further analyses examine question-conditioned judgment, inference cost, and probability quality, complemented by qualitative comparisons across models.
[AI-80] he View from Nowhere: Population-Centered Agent Evolution
链接: https://arxiv.org/abs/2610.36660
作者: Shuo Yan,Ningning Hou,Haoyu He,Haochen Pei,Shang Gao
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Self-improving AI systems can revise the designs that govern their own behavior. Yet existing approaches center improvement on selected parent agents and their evolutionary experience, leaving the broader population largely underused. We identify a substantial gap between population coverage (i.e., the tasks solved by at least one evolved variant) and the performance of the best individual agent. This gap shows that useful capabilities and diagnostic evidence emerge across the population without being consolidated into its leading agents. These observations motivate the view from nowhere, a principle for agent evolution: an agent should be improved by interpreting its behavior through population-wide patterns of success, failure, and prior self-modification. We instantiate this principle as Population-Centered Evolution (PCE). PCE distils task records and modification histories from the population into a population block, which captures recurring failures, complementary successes, and the effects of previously attempted modifications to guide improvements to a selected parent’s harness. Under matched models, initial harnesses, and evolution schedules, PCE achieves 72.1% success on SWE-bench Verified and 96.7% on Polyglot, improving over a strong group-evolution baseline by 3.2 and 5.9 percentage points. In our evaluation, PCE reduces the gap between population coverage and best-agent performance by one-third on SWE-bench and closes it on Polyglot. These results suggest that treating the population as evidence–not merely as candidates for selection–provides a complementary principle for agent evolution. Subjects: Artificial Intelligence (cs.AI) Cite as: arXiv:2610.36660 [cs.AI] (or arXiv:2610.36660v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2610.36660 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-81] ap2Listen: Instant Target Switching for Streaming Audio-Visual Speaker Extraction NEURIPS2026
链接: https://arxiv.org/abs/2610.36643
作者: Hansol Park,Minkyung Song,Zinu Kim,Yujun Lee,Hoseong Ahn,Hyunjin Choi,Kyuhong Shim
类目: ound (cs.SD); Artificial Intelligence (cs.AI)
备注: NeurIPS 2026 Workshop on On-Device Intelligence
Abstract:Streaming audio-visual target speaker extraction (AV-TSE) uses a target speaker’s face to extract the speaker’s speech from an input audio mixture in real time. However, existing systems fix the target speaker for an entire clip, while conversation demands frequent and immediate target switches. To address this problem, we propose Tap2Listen, a streaming AV-TSE model designed for instant response to target switches. The model separates computation by modality: the target-agnostic audio encoder processes only the mixture, while the target-specific visual cues of the selected face condition only the decoder. A switch requires only a 0.12ms re-projection and takes effect immediately. We also design component-wise losses: a prediction loss for the encoder and a supervised extraction loss for the decoder. For evaluation, we measure how quickly output quality recovers after a switch. On the standard LRS3-2mix test set, Tap2Listen improves SI-SNRi from 8.85 to 9.68 dB at a comparable parameter count. After a switch to a known face, the model produces full-quality output from the first frame, whereas the baseline needs about 1 second to recover. Finally, we show that the proposed streaming AV-TSE model runs in real time on a single CPU core.
[AI-82] he uneven rise of AI-generated web content
链接: https://arxiv.org/abs/2610.36631
作者: Adam Visokay,Soumyakanti Pan,Molly Griston,Polina Iasakova,Mark Nepf,Denis Peskoff,Riya Sinha,Jikai Sun,Isabelle Thiara,Letao Wang,Yiwei Wu,Lauren Yan,Tyler H. McCormick
类目: Artificial Intelligence (cs.AI); Computers and Society (cs.CY); Networking and Internet Architecture (cs.NI); Applications (stat.AP)
备注: 12 pages, 4 figures
Abstract:In less than five years, generative artificial intelligence (AI) has dramatically reshaped the public-facing web. Using a sample of more than 5 million crawlable webpages, we demonstrate that the share containing AI-generated text has increased precipitously, but unevenly. Our analyses reveal substantial differences in AI composition across language and domain categories (e.g., .edu, .com, or .gov). A higher fraction of English-language webpages contain AI-generated text compared to pages in other languages, and prevalence is substantially lower on .gov and .edu domains compared to .com. These distinctions persist even after accounting for compositional effects. We also find heterogeneity based on a page’s position in the connectivity structure of the web. AI-generated text is most common and has increased most rapidly in pages that are the least central in the graph, which drives much of the overall growth in AI-generated text. A slower increase has taken place for more central pages, with the most central tier emerging as the least likely to contain AI-generated text. Methodologically, we use multiple AI detection tools, and generate estimates using a Bayesian hierarchical model that incorporates corrections for the detectors’ statistical properties (both false positives and false negatives) and allows for different detector performance across groups, while propagating uncertainty in detector error rates into prevalence estimates. Our work suggests that AI is a prominent, but varied, contributor to public information and internet-based communication.
[AI-83] Polarity–Time Interaction and Multi-Level Reliability for Event-Based Spacecraft Pose Estimation
链接: https://arxiv.org/abs/2610.36625
作者: Yonglong Zhang,Zongwu Xie,Yang Liu
类目: Artificial Intelligence (cs.AI); Robotics (cs.RO)
备注:
Abstract:Sparse event responses produce uneven keypoint errors whose effects on spacecraft pose depend on correspondence geometry and the reliability of successive observations. We present a reliability-aware framework connecting polarity–time localization, geometric recovery, event–CAD refinement, and trajectory optimization. A three-branch representation combines positive events, negative events, and recent-event timing through cross-branch attention and gating, followed by spatiotemporal aggregation. An ordered semantic-point head estimates relative correspondence precision for geometrically constrained pose recovery. Event-edge and CAD-projection agreement then guide local pose refinement under keypoint-geometry safeguards. At the sequence level, a ten-dimensional descriptor combines heatmap shape, reprojection consistency, and geometric conditioning to set time-aligned observation scales. On the official SPADES challenge test set, the complete system achieves a combined pose score of 0.0474, approximately 5.4% below PASAN+IRIS. Validation ablations reduce the score from 0.0772 for base localization with uniform PnP to 0.0514 for the complete cascade; on identical single-frame observations, quality weighting reduces the score from 0.0549 with constant-weight optimization to 0.0514.
[AI-84] God Does Not Play Dice: Enforcing Batch Invariance in Agent ic Kernel Generation
链接: https://arxiv.org/abs/2610.36623
作者: Youwei Bao,Chengpeng Wang,Simin Chen
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Software Engineering (cs.SE)
备注:
Abstract:Large language model (LLM) serving systems dynamically batch user requests for GPU execution, so the same request may be processed alongside different requests across runs. If the GPU kernels lack batch invariance (BI), this changing batch context can alter the numerical output for an unchanged input, undermining reproducibility. Existing kernel generation agents focus on correctness and efficiency but \emphoverlook BI. We find 19 confirmed BI violations among 83 batch-applicable released artifacts. To address this limitation, we propose \tool, which combines contract-guided static dependency analysis with dynamic validation of bitwise consistency across batch contexts. \tool is designed as a plug-in component for general kernel agents, allowing their original optimization capabilities to be reused while adding BI enforcement. We integrate it as successive acceptance gates in the optimization loop, rejecting structurally risky or dynamically non-invariant candidates before performance benchmarking and returning feedback to the agent. Experiments across three agent frameworks under baseline and BI-aware generation on real world LLM serving tasks show that \tool improves the BI pass rate from 67.6% to 91.9% while retaining 91.8% of the baseline geometric mean speedup. These results demonstrate that \tool can be incorporated into kernel optimization by agents as an additional reliability objective alongside correctness and speed.
[AI-85] Dream-COP: Guiding Algorithm Discovery for Combinatorial Optimization Problems in an Experience World
链接: https://arxiv.org/abs/2610.36618
作者: Kuan Xu,Jianming Wang,Ni Zhang,Yaoxin Wu,Yong Wang,Zhiguang Cao
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models (LLMs) enable algorithm discovery for combinatorial optimization problems (COPs) through iterative program generation and evaluation. A central challenge is how to organize search to discover better algorithms within a limited budget. Our key insight is that past search trajectories and their recorded outcomes form an experience world, which can be used to facilitate the COP solving through replay. We introduce Dream-COP, in which a \emphpolicy agent learns executable policies through \emphdreaming in this experience world to guide a \emphdiscovery agent in generating and refining COP programs. Specifically, policy replay provides feedback on the quality of the programs revealed and the search steps used, to guide policy selection more effectively. Across five COPs, \textscDream discovers better algorithms than fixed exploration in both heuristic component design and complete solver generation. Particularly, it matches the optimal values on KP and reduces the gap on OBPP from 2.112% to 1.199% while using 42.5% fewer search requests. Moreover, the learned exploration policies transfer effectively across different COPs with the assitance of the experience world.
[AI-86] Learning to Formalize: Bridging Open-Ended Workloads and Fast Decision Models
链接: https://arxiv.org/abs/2610.36595
作者: Chengxuan Pei,Ziyao Wang,Teng Mei,Zhe Li
类目: Artificial Intelligence (cs.AI)
备注: 10 pages, 2 figures, only for concept verification now
Abstract:Decision models offer fast choices over sup- plied alternatives, while general-purpose large language models interpret open-ended tasks through flexible generation and reasoning at the cost of sequential decoding. Many work- loads expect a generative backend to determine what can be done and therefore provide no well-defined candidate list. We bridge this in- terface gap with a small candidate formalizer trained on offline teacher-generated lists. The formalizer constructs explicit, validated alter- natives from task context, and a frozen deci- sion model selects an entry. A motivating com- parison with official TypeSafe Jev-1.13.0 on 200 CALVIN-D test episodes reports 11% suc- cess with generic motions versus 46.5% with DeepSeek V4 Flash pose candidates. Sepa- rately, a six-arm privileged-state evaluation on 1,011 CALVIN-D validation annotations finds that a locally deployed decision model can achieve at most 52.6% task success rate with the fine-tuned formalizer, compared with 11.8% without it, and the end-to-end latency is kept to 2.09 s, whereas Qwen3.5-27B is 3.67 s with the exact experiment setting.
[AI-87] InfraDev: Can Coding Agents Develop AI Infrastructure and How Can We Tell?
链接: https://arxiv.org/abs/2610.36589
作者: Yangshen Deng,Jiaqi Ruan,Yeqi Huang,Nick Lau,Guangming Sheng,Luo Mai
类目: Artificial Intelligence (cs.AI); Distributed, Parallel, and Cluster Computing (cs.DC)
备注:
Abstract:Coding agents are beginning to develop the production-grade AI infrastructure that trains and serves large models, yet existing benchmarks cover only single kernels or single-GPU tasks, and grade with tests and prompts that misjudge both correct and incorrect patches. We introduce InfraDev, a benchmark of 36 tasks built from real-world pull requests to Megatron-LM, vLLM, and verl, covering infrastructure development for training, inference, and reinforcement learning. Each task asks an agent to reimplement a real change that spans many files and must run correctly across many GPUs. To grade such changes, we jointly synthesize hidden probes, which enforce the requirements the original tests leave unchecked, and a prompt that states every interface the probes call without revealing the solution. We also design a shared execution environment that reuses dependencies and artifacts across tasks, cutting storage by 52-57% and build time by 13-14% on Megatron-LM and vLLM. Across six frontier models, the best resolves only 38.9% of the tasks; the probes reject 32% of the submissions that pass every official test, and with the original pull request descriptions as prompts, no model resolves a single task. Failure analysis shows that structural and state errors dominate and that agents’ own tests rarely check memory, time, or traffic. Turning these findings into an agent skill raises the resolved rate of a frontier model on the Megatron-LM tasks by 10.5%.
[AI-88] Benchmarking LLM Agents for Behaviorally Adequate Test Generation from Software Requirements
链接: https://arxiv.org/abs/2610.36584
作者: Binhang Qi,Yiming Liu,Yun Lin,Yangjie Zhou,Hailong Sun,Jin Song Dong
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注:
Abstract:As large language model agents increasingly compile natural-language requirements into repository-level code, requirement-to-test generation grows in importance to provide validation constraints (or harness) to guide code generation. Yet it remains unclear whether current agents can generate tests that validate the required behavior when the implementation of that behavior is withheld. Existing test-generation benchmarks mostly focus on white-box testing, whereas the requirement–test-related study usually requires evaluation on semantic alignment between the given documents and the generated tests. We introduce Req2Test-Bench, a benchmark containing 5,086 execution-validated instances recovered from developer-authored behavior-driven development (BDD) artifacts in 8 open-source projects, spanning 5 programming languages and 4 BDD frameworks. Each instance pairs a natural-language requirement in Gherkin with its developer-written Test Implementation. Given the requirement, target interfaces of the required behavior, an agent must generate a complete, potentially multi-file Test Implementation. Evaluation of 8 agent–model–reasoning configurations reveals a substantial gap between execution success and behavioral adequacy, which is measured strictly as kill-completeness: a kill-complete test kills every mutant that the ground-truth test kills. The best-performing configuration, Codex with DeepSeek-V4.1-Flash, produces passing tests for 82.8% of instances, yet only 33.8% are behaviorally adequate. We find that (1) scaling up agents yields sharply diminishing returns in behavioral adequacy relative to generation cost; (2) under-specified requirements cause the information gap, including ambiguous setup and implicit expected outcomes; and (3) requirement-grounded harnesses that steer agent non-determinism offer a promising direction for improvement. Subjects: Software Engineering (cs.SE); Artificial Intelligence (cs.AI) Cite as: arXiv:2610.36584 [cs.SE] (or arXiv:2610.36584v1 [cs.SE] for this version) https://doi.org/10.48550/arXiv.2610.36584 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-89] MM-Net: A Multimodal Multi-Task Network for Sleep Staging and Sleep Apnea Assessment After Ischemic Stroke
链接: https://arxiv.org/abs/2610.36583
作者: Md Wahiduzzaman Suva,Esm E Moula Chowdhury Abha,Md Imtiaj Alam Sajin,Md Iftekharul Mobin,Jia Uddin
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Signal Processing (eess.SP)
备注:
Abstract:Automated polysomnography after ischemic stroke is challenging because of altered sleep-related neural activity and concurrent respiratory disturbances. Existing stroke-specific approaches mainly address sleep-stage classification, leaving joint analysis of sleep architecture and respiratory events underexplored. Moreover, adapting pretrained EEG encoders introduces a representation inconsistency when downstream models are trained on embeddings from previously seen patients but evaluated on unseen individuals. We propose MM-Net, a multimodal multi-task framework that resolves this inconsistency through cross-fitting and ridge-based embedding alignment. Sleep-adapted EEG-EOG representations and engineered neural descriptors are fused with learned cardiorespiratory features through cross-attention and shared temporal modeling for simultaneous sleep staging and respiratory-event detection. On iSLEEPS (99 patients with subacute ischemic stroke), ten-fold patient-independent evaluation achieved 76.3% staging accuracy (Cohen’s \kappa = 0.656) and 0.825 respiratory-event AUC, outperforming the evaluated baselines under matched protocols. The sleep-adapted encoder improved the downstream model when its embeddings were aligned, raising staging accuracy by 1.4 percentage points over the frozen encoder ( p = 0.020) versus 0.2 points without alignment. MM-Net also achieved 86.4% staging accuracy on Sleep-EDF, above every single-EEG architecture reimplemented on the same recordings, and 80.3% staging accuracy with 0.889 respiratory AUC on ISRUC-I, exceeding its retrained staging baselines. Without retraining, the iSLEEPS-trained model retained respiratory AUCs of 0.833 on ISRUC-I and 0.841 on UCD. These findings support representation-consistent neural adaptation and joint physiological modeling across stroke, healthy-sleep, and sleep-disorder cohorts.
[AI-90] Follow the Herd: A Data-Driven Approach for Actionable Recourse
链接: https://arxiv.org/abs/2610.36580
作者: Nishtha Madaan,Srikanta Bedathur
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:As machine learning systems make increasingly greater decisions about people’s lives, it is critical to provide users with actionable recourse about how they can achieve their desired outcomes. This problem is currently tackled by relying on expensive human supervision which involves hand-designing a cost function before searching for low-cost actions that achieve the desired outcome. In this paper, we propose \textit``Data-Driven Action Recourse" (DDAR), the first approach for providing action recourse without requiring costly supervision from humans about action costs. We build upon the notion that highly likely trajectories in the historical user data should be associated with lower costs. With this, DDAR first builds a dynamics model of historical user trajectories. Under this model, it then searches for trajectories that achieve the desired outcome. In experiments, we find that this simple approach recommends plausible action recourse without requiring human supervision, achieving lower action costs than the existing approaches.
[AI-91] Less is more: learning what to profile for GPU kernel optimization
链接: https://arxiv.org/abs/2610.36576
作者: Yihan Wang,Peiyu Liu,Runyu Chen,Wei Xu
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models have recently shown strong potential for automatically optimizing GPU kernels, yet their effectiveness depends critically on the hardware information available to the kernel optimization model. Existing systems typically rely on predefined profiling strategies, leaving what to profile largely outside the optimization process. We show that effective profiling is not simply about collecting more information: increasing the number of profiling metrics does not monotonically improve kernel optimization, while the most useful profiling information varies across kernels. Our theoretical analysis further shows that, under such heterogeneous profiling utility, kernel-conditioned selection can outperform any single fixed profiling configuration. Motivated by this observation, we propose AutoProfiler, a feedback-driven framework for learning a deterministic profiling tree from optimization feedback. The tree is iteratively refined through REWRITE and SPLIT operations, with an LLM generating candidate modifications and empirical evaluation determining their acceptance. Once constructed, the tree quickly gathers the required profiling information for any kernel, allowing the LLM rewriter to be invoked only once. Experiments on KernelBench show that our approach improves kernel optimization over fixed profiling configurations under constrained profiling budgets. These results suggest that hardware profiling should evolve from static metric collection toward profiling reasoning, making profiling itself a new dimension for improving kernel optimization. The codes are available at this https URL.
[AI-92] ADTGuard: Adaptive Depth-Time Safeguard for Streaming Inference in Large Language Models
链接: https://arxiv.org/abs/2610.36574
作者: Chuming Wei,Churui Yang,Zihan Zhang,Ruien Li,Wusheng Zhang,Yan Zhang,Guangwen Yang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large Language Models (LLMs) can generate fluent responses to diverse user requests, yet streaming may expose users to harmful content before generation is complete. Existing guardrails face two limitations in this setting. First, post-hoc moderation at response or chunk boundaries delays intervention, risking harmful exposure when content is released before verification. Second, even with token-level detection, coarse response-level supervision does not specify when harmful content begins, while fixed-layer readouts may miss relevant safety signals at other depths. To address these limitations, we introduce ADTGuard, which checks each token before release, progressively refines token-level training targets, and adapts layer selection to the current prefix. ADTGuard models risk across depth and time through Depth-Time Safety Dynamics (DTSD), which integrates safety evidence across layers at each token and uses it to guide risk accumulation over successive tokens. To adapt the choice of observation layers to the current prefix, we propose the Causal Depth Router (CDR). With only response-level labels available, we further develop Risk-Onset Target Refinement (RTR) to progressively construct token-level training targets for both components. Through extensive experiments across multiple safety benchmarks, we demonstrate that ADTGuard consistently achieves strong streaming detection performance, matching or surpassing state-of-the-art safeguards while requiring only 16M trainable parameters and maintaining competitive per-token latency.
[AI-93] CASEL: Consequence-Aware Selective Execution for Reliable Long-Horizon Tool-Using Agents
链接: https://arxiv.org/abs/2610.36570
作者: Yanyan Zhang,Fei Shen,Ting Yan,Hao Zhang
类目: Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR); Software Engineering (cs.SE)
备注: 6 pages, 4 figures. Accepted at the 2026 5th International Conference on Cloud Computing, Big Data Application and Software Engineering (CBASE 2026)
Abstract:Tool-using language agents increasingly modify external systems, yet most runtime safeguards allocate scrutiny by generic uncertainty scores or coarse mutation heuristics. We introduce CASEL (Consequence-Aware Selective Execution Layer), a model-agnostic runtime allocation policy that combines calibrated failure probability, state-dependent consequence, reversibility, ambiguity, and evidence freshness to choose, for each proposed tool call, among direct execution, verification, sandboxing, clarification, approval, and abstention by minimizing a per-mode cost. We formalize direct execution as selective prediction and state and test the validity conditions of a finite-sample threshold and a trajectory risk ledger. In a controlled simulation over 20 seeds with 120,000 test actions each, in which every method has the same intervention menu and the same compute cost, CASEL reduces harmful-side-effect trajectories from 14.75 percent for an uncertainty-ranked gate to 7.02 percent and mean normalized harm from 0.184 to 0.123. A gate ranked by the product of failure probability and consequence reaches similar harm (5.82 percent, 0.120) but sends 45 percent more actions to human approval, and removing reversibility-aware routing raises the rate to 15.53 percent. A trajectory-level risk budget, in contrast, increases harm at equal cost (11.82 percent). Intervention effectiveness is assumed, not measured: the results rank allocation rules under the simulator’s assumptions and do not establish deployment effectiveness.
[AI-94] NextEval: An Agent -Based Solver and Benchmark for Derivative-Free Optimization
链接: https://arxiv.org/abs/2610.36553
作者: Shuo Chen,Li Cui,Cunxin Huang,Ya-Feng Liu,Minghang Xie,Meng Xu,Zaikun Zhang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Optimization and Control (math.OC)
备注: 27 pages
Abstract:Derivative-free optimization (DFO) addresses problems where derivatives are unavailable or costly to obtain. This setting often arises in simulation-based and experimental applications, where evaluating the black-box objective can itself be costly. Thus, it remains a central challenge in DFO to effectively exploit accumulated observations to select informative evaluation points under a limited budget. In this light, we introduce the NextEval solver, an agent-based DFO solver, which integrates the capabilities of a large language model (LLM) to guide the sequential selection of evaluation points. Specifically, we establish a system that combines an LLM agent with a numerical environment. The environment provides tools for analyzing past evaluations and executing numerical methods, while maintaining a shared evaluation history and method-specific state to support continuation across method switches. Using observations and tool feedback, the LLM agent selects tools, configures their parameters, and allocates the evaluation budget, while numerical methods compute and evaluate candidate points. Building on this system, we also introduce NextEval Bench, a benchmark protocol for comparing agent configurations on shared DFO tasks using the same numerical tools under matched evaluation budgets and information constraints. Numerical results on low-dimensional DFO problems demonstrate that the NextEval solver outperforms the state-of-the-art DFO baselines in function-evaluation efficiency.
[AI-95] REACH: An RL-based Hierarchical Humanoid Crawling Framework for Adaptable Confined Space Traversal
链接: https://arxiv.org/abs/2610.36546
作者: Shengzhang Xu,Yuning Guo,Weiyi Hu,Yifeng Xu,Zhongyu Li,He Kong
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注: To appear at IEEE Humanoids 2026; Project page this https URL
Abstract:Humanoid locomotion in confined spaces requires continuous height adaptation across multiple locomotion modes, yet most existing approaches remain limited to height-adjustable bipedal walking. We present REACH, an RL-based hierarchical framework that enables a Unitree G1 humanoid to traverse height-constrained environments through variable-height bipedal walking and crawling. The learned policies regulate body height over complementary command ranges of 1.27–0.50m in bipedal locomotion and 0.50–0.20m in crawling, extending traversal beyond the clearance limits of crouched walking. To address contact-rich crawling dynamics, we employ a history-conditioned latent representation that captures terrain and interaction context from proprioceptive observations. A bidirectional transition policy further connects crouched bipedal locomotion and crawling across large kinematic and contact discontinuities. These skills are integrated with a LiDAR-based planner that converts point clouds into collision-free three-dimensional paths and locomotion commands, including proactive height targets. Real-world experiments demonstrate accurate height and velocity tracking, robust terrain traversal, reliable mode transitions and confined-space traversal with the proposed pipeline. Code and videos are available at this https URL.
[AI-96] ATEMPO: Amortized Transfer Entropy Model Per-lag Over-Time
链接: https://arxiv.org/abs/2610.36536
作者: Emma Andrews,Konstantinos Kontras,Keane Ong,Maarten De Vos,Gianmarco Mengaldo
类目: Artificial Intelligence (cs.AI); Information Theory (cs.IT)
备注:
Abstract:When one time series informs another, the delay at which that information arrives determines if models can utilize this relationship effectively. A source may be strongly predictive at one delay but uninformative at neighboring lags. Therefore, fixed input windows may obscure this structure by bundling informative and irrelevant history together. Transfer entropy (TE) can recover these delays by measuring how much a source’s past tells us about a target beyond the target’s own history. However, existing TE estimators require separate fits across source and conditioning depths, with further estimation on local windows to track delays over time. We introduce ATEMPO, a diffusion-based transfer entropy estimator amortized over lags and time. ATEMPO samples source and conditioning depths during training and conditions explicitly on temporal position, allowing a single score network to estimate the full TE surface without refitting. Differencing this surface recovers when information arrives. Across synthetic datasets with ground-truth TE, ATEMPO achieves the lowest mean relative error among baselines, and on datasets with known lags, ATEMPO achieves 0.98\pm0.04 lag-detection accuracy. ATEMPO requires a single fit across source depths K=6 and target depths L=3 , whereas existing estimators require 18 separate fits to evaluate the same K \times L grid. On three systems with time-varying lag, ATEMPO is the only estimator to maintain perfect lag-detection accuracy across all evaluated window lengths. These recovered delays improve downstream forecasting: ATEMPO-selected inputs reduce MSE on WEATHER-5K by 7.25%, while 120 selected inputs outperform the benchmark’s 240-feature default by 4.6%. Across three brain-computer interface tasks, using ATEMPO-selected lags achieves the highest or equal-best R^2 , improving mean R^2 by 4.1% relative to the strongest competing baseline.
[AI-97] ProCerDCL: Query-Efficient Black-Box Probabilistic Verification of Deep Learning Compiler Optimization
链接: https://arxiv.org/abs/2610.36499
作者: Zexin Li,Weibo Zhang,Linyi Li,Arvind Easwaran,Simin Chen
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI); Programming Languages (cs.PL)
备注:
Abstract:We present \tool, a query-efficient framework for black-box probabilistic verification. Given a fixed, sampleable input distribution and a binary property oracle, \tool assesses the property’s violation probability p against a threshold \theta . Under independent sampling and stateless oracle behavior, acceptance asserts p\le\theta , rejection asserts p\theta+\eta for a user-specified separation margin, and an inconclusive verdict indicates insufficient evidence for either conclusion. The probability of each type of erroneous verdict is bounded by its specified error budget. To support this three-valued contract efficiently, \tool uses a single sequential test, eliminating the interval family required by the baseline. The test accounts for repeated sampling checks through either corrected exact binomial confidence intervals or an anytime-valid confidence sequence. We apply \tool to probabilistic local model semantic consistency, where the property oracle detects decision-level inconsistencies between reference and compiled executions under a specified perturbation distribution. Our evaluation covers 24 tasks across TorchInductor and TensorRT and examines sampling efficiency, conclusiveness, and diagnostic behavior. For sampling efficiency, we use paired one-model clean-label radius searches: the baseline requires 3.6-4.2 times as many unique samples under matched numerical parameters while providing a weaker decision contract. For model semantic consistency, checks at \theta=10^-3 yield 817 acceptances, 19 inconclusive verdicts, and 4 rejections across 840 input–backend instances from 21 tasks. Case studies of two documented compiler faults further demonstrate what these local verdicts can diagnose.
[AI-98] Proposing Better Rollouts for On-Policy Distillation
链接: https://arxiv.org/abs/2610.36477
作者: Sirui Chen,Yunzhe Qi,Haobo Xu,Jiaru Zou,Mengting Ai,Zihao Li,Jingrui He
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:On-policy distillation (OPD) improves language-model post-training by letting the student learn from its own sampled trajectories under teacher supervision, aligning the training process with inference-time behavior. However, the rollout policy in OPD is usually treated as fixed: each round samples directly from the current student. We show that this choice can create a compounding teacher–student mismatch, because it requires the teacher to provide effective supervision on whichever trajectories the student samples, including ones less reachable by the teacher itself. On such trajectories, teacher supervision can become noisy or even misleading, leading to undesirable student patterns such as repetition. Moreover, since the updated student in turn defines the rollout policy of subsequent rounds, these patterns persist and potentially amplify over training. This failure mode suggests that improving OPD requires treating the rollout policy itself as an optimization target, rather than only modifying the distillation objective. We propose Mixture Proposal On-Policy Distillation (MPOPD), which derives a student-anchored, teacher-guided rollout distribution from a local trust-region objective based on relative teacher support. The resulting target is a geometric mixture of the student and teacher policies, and we introduce a proposal policy that approximates this mixture, enabling rollout generation in a single forward pass. Experiments on 16 math, knowledge, and medical reasoning benchmarks show consistent gains over OPD and recent variants. On math reasoning, our method improves average accuracy by 6.91 over OPD on Qwen3-4B-Base while reducing average generation length by 41%, suggesting that rollout policy design provides an effective way to improve OPD.
[AI-99] FIDES-Live: Evaluating Behavioral Concordance in Sequential Trading Agents
链接: https://arxiv.org/abs/2610.36476
作者: Arther Tian,Alex Ding,Simon Wu,Aaron Chan
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:
Abstract:Sequential trading agents produce decisions, orders, execution receipts, account states, and reports that need not agree, even when the final portfolio value is correctly computed. We present FIDES-Live, an operational extension of FIDES from generated strategy concordance to the audit of sequential agent traces. The protocol separates structured decision validity, optional numerical commitment adherence, execution and ledger reconciliation, and receipt- grounded reporting, with explicit coverage and unscorable outcomes. We retrospectively audit ten requested inference routes over 7,299 cryptocurrency account-rounds and a 310-round equity pilot, using real market observations and local paper execution. Cryptocurrency decision validity is 97.35% and structured reporting accuracy is 90.81% among scorable responses, but route-level decision and reporting performance can diverge sharply. Commitment adherence is 94.36% overall; it falls to 62.63% on constraints emitted with nonempty order proposals, compared with 97.56% for valid empty proposals. All retained cash and position transitions reconcile with recorded fills, while execution controls and unavailable quotes explain separate rejection populations. Short-window marked returns provide descriptive context rather than a new test of whether concordance predicts alpha. These observations motivate reporting concordance, evidence coverage, activity, and financial outcomes separately. They do not establish long-run trading skill, semantic faithfulness of explanations, or causal benefits from audit feedback.
[AI-100] When to Intervene? An Empirical Study of Errors and Self-Recovery in LLM Agents
链接: https://arxiv.org/abs/2610.36471
作者: Chiming Duan,Tong Jia,Yingxi Chen,Minghua He,Lingzhe Zhang,Ying Li
类目: Artificial Intelligence (cs.AI)
备注: 11 pages, 8 figures, 2 tables
Abstract:LLM agents interact with Web services to perform tasks on users’ behalf. Errors during these interactions can disrupt execution and prevent task completion, motivating external intervention. However, agents may self-recover through continued execution, raising the question of when to intervene and when to wait for self-recovery. We present an empirical study of errors and self-recovery in LLM agents. We construct ReInBench, a dataset containing 1,259 agent traces and 1,403 error events across computer-use and Web tasks. Our study examines how often and how quickly agents self-recover, how self-recovery relates to subsequent execution and task completion, and whether self-recovery can be predicted when an error occurs. We find that 725 error events self-recover, often within a few steps, with self-recovery performance differing across Execution LLMs. Self-recovery can also help agents recover from subsequent errors of the same type and support overall task completion. Intervention should therefore prioritize errors unlikely to self-recover, while allowing agents to continue when self-recovery is likely. Building on these empirical findings, we propose ReCoF, a method that predicts whether an error will self-recover to inform whether additional intervention is needed. ReCoF achieves an average relative improvement of 9.02% in Macro-F1 over the strongest baseline on ReInBench. Our code and data can be found at this https URL.
[AI-101] Correction or Manipulation? Measuring Epistemic Vigilance in LLM Agents Under User Pressure
链接: https://arxiv.org/abs/2610.36453
作者: Nardine Basta,Venus Haghighi,Dali Kaafar,Amin Beheshti
类目: Artificial Intelligence (cs.AI)
备注: 20 pages, 7 figures, 11 tables. 8-page main paper with supplementary material included as an appendix. Under review
Abstract:LLM agents that answer questions and act on behalf of users must decide when to yield to a user who disagrees. Disagreement may correct a genuine error or may be an attempt at manipulation. Prior work shows that models frequently abandon correct answers when challenged by users, yet it remains unclear whether they distinguish correction from manipulation or merely differ in their propensity to concede. We examine this question across seven LLMs from four vendors in two studies, one on the answers agents give and one on the actions they take as customer-support agents whose policy permits account changes only for verified users. Simulated users push toward correct and incorrect answers with graded pressure. Following research on epistemic vigilance, the pressure either appeals to the source, such as claimed authority, or offers content the agent cannot verify, namely a cited source. Drawing on signal detection theory, we measure on a set of benchmark questions how much more readily each agent concedes to correction than to manipulation, separately from its overall propensity to concede. Citations made five of seven models readier to concede and four worse at discriminating, while stronger wording added little. In the agentic study, five models never violated the verification policy, including models that readily abandoned correct answers, so concession on answers did not predict violations of policy. Ministral 3B violated the policy mainly under claimed authority, and Ministral 8B reported changes it had not made. An instruction to revise only on verifiable evidence chiefly shifted propensity to concede rather than improving discrimination. Concession rates alone are therefore misleading, and agents should be evaluated on both their discrimination and the actions they take.
[AI-102] From Solver Trajectories to Teaching Trajectories: Cognition-Aligned Reasoning Distillation
链接: https://arxiv.org/abs/2610.36445
作者: Yashuo Luo,Qianren Mao,Ziqi Qin,Junnan Liu,Haitao Li,Qili Zhang,Yuan Gao,Zhijun Chen,Xu Wang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Knowledge distillation from large reasoning models (LRMs) often uses raw teacher trajectories as supervision. However, this practice creates a mismatch between the goal of distillation and the nature of teacher trajectories. Advanced LRMs typically improve their reasoning capabilities through reward-driven post-training, such as RLHF. This training optimizes them as strong solvers, rather than as teachers that produce reasoning processes smaller models can easily learn. Since smaller models have different “cognitive capacities” compared to their larger teachers, directly imitating raw teacher trajectories can sometimes be ineffective and often requires a substantial number of high-quality samples. To bridge this mismatch, we propose Cognition-Aligned Reasoning Distillation (CARD), an RL-based framework that transforms solver trajectories into teaching trajectories aligned with the student’s cognition. Avoiding expensive resampling, our framework directly transforms existing trajectories. We optimize CARD with two reward signals: (1) a Distribution Alignment Reward that matches the student’s predictive distribution, and (2) a Logical Coherence Reward that reduces misleading failed reasoning steps. Comprehensive evaluations across multiple student models and settings demonstrate that CARD outperforms distillation from raw teacher trajectories in in-domain settings while achieving substantial gains in out-of-distribution performance. Code is available at this https URL.
[AI-103] RouteDRAG : Learning to Route and Control Agent ic Execution for Multimodal Document RAG
链接: https://arxiv.org/abs/2610.36444
作者: Zekai Chen,Hao Zhu,Yu Liu,Xunkai Li,Zejian Kang,Rui Qian,Quanjiang Li,Rong-Hua Li
类目: Artificial Intelligence (cs.AI)
备注: Our code is now available at this https URL
Abstract:Web-based applications increasingly rely on understanding long multimodal documents, where answering questions often requires combining information across pages, tables, and figures. Agentic retrieval-augmented generation addresses these tasks through iterative retrieval and reasoning. However, Agentic execution is not always necessary, introducing additional computation and potentially incorrect answers. Our empirical analysis reveals that Agentic execution provides query-dependent benefits, while additional operations may improve, preserve, or degrade answer accuracy. These findings motivate controlling both Agent invocation and subsequent execution. To address these challenges, we propose RouteDRAG, a learning-based framework comprising four components. Unified Document Index organizes multimodal content into a graph-backed index and derives query representations from initial retrieval. Adaptive Path Routing preserves exact answering and predicts the relative benefit of Agent over shorter execution paths. Dynamic Agent Execution maintains an evolving workspace and refines Planner proposals to control tool use and termination. Sequential Policy Learning first trains operation control from observed outcomes and then trains path routing using the resulting Agent’s performance. Across two benchmarks, RouteDRAG achieves an average accuracy improvement of 3.47 percentage points over shorter-path-only execution while assigning only 79.3% of queries to Agent. Notably, it also surpasses uniform Agent execution on both benchmarks while requiring fewer answering calls. Our code is now available at this https URL.
[AI-104] Your Agent Acts Differently When Youre Around: How User Presence Changes Agent Behavior in Negotiation
链接: https://arxiv.org/abs/2610.36435
作者: Jane Pan,You Lyu,Abe Bohan Hou,Valerie Chen,He He
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:LLM agents increasingly interact with other agents on behalf of users, yet how users in the loop affect their agents’ behavior–for better or worse–remains poorly understood. We study this question through the lens of negotiation, where agents must represent users while strategically bargaining with other agents. To isolate the effect of user presence, we develop PRESENT, a controlled evaluation framework that compares negotiation settings with and without an interactive user, while holding the game, objectives, and user preferences fixed across conditions. Across 11.2k runs spanning 7 games and 8 models, we find that user presence significantly changes both outcomes and behavior. Specifically, negotiation parties with a user obtain higher payouts (+13.7% on average) and their agents negotiate more aggressively. Surprisingly, user presence also induces misaligned behavior, both toward counterparties (where its rate roughly doubles) and toward users (where it emerges in 14.3% of games). These misaligned behaviors are uncorrelated with outcomes and persist even when the user sends only templated, neutral messages. These findings expose two blind spots in current evaluation practice: static instructions miss behaviors that emerge only when users are present, and outcome-level metrics cannot detect them.
[AI-105] LM-DSB: A Lightweight Multi-agent System for Automated Daily Safety Bulletin Generation in Construction
链接: https://arxiv.org/abs/2610.36433
作者: Xianxiang Zhao,Advik Mehta,Falak Sethi,Brian Gue,Qipei Mei,Lingzi Wu
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Construction safety management still struggles to produce daily safety bulletins (DSBs), structured daily communications that identify site-specific hazards, convey safety protocols, and set preventive measures for workers. Manual DSB creation takes 2-3 hours daily and yields inconsistent quality, with 25% of crews never conducting pretask planning (Olson et al., 2016), diverting safety personnel from field supervision. This study proposes the Lightweight Multi-Agent DSB System (LM-DSB), which integrates retrieval-augmented generation (RAG) with a locally deployable Large Language Model (LLM) to automate DSB generation on consumer-grade personal computers. LM-DSB coordinates specialized planner, researcher, and writer agents in a hierarchical architecture, enhanced by domain-specific fine-tuning on construction safety documentation and RAG grounding in 37 health, safety, and environment (HSE) documents. A two-phase validation assessed feasibility. Phase 1 comparative evaluation showed the multi-agent RAG system (Llama 3-8B) outperformed all alternative configurations, with strong lexical alignment and coherence (TF-IDF: 0.70, BLEU: 0.39, ROUGE-1: 0.66) and near-complete document generation (length ratio: 1.01 vs. 0.39 for single-agent RAG). Phase 2 expert evaluation (N=40) confirmed practical acceptability (mean scores 3.50-4.20 on a 5-point scale), particularly in regulatory compliance (4.20) and factual accuracy (4.10). The findings show that architectural design enables lightweight models to handle safety-critical documentation on consumer-grade hardware, advancing technology democratization in construction safety management. Limitations include knowledge base scope, language constraints, and lack of field deployment validation; future work should address multimodal integration and workflow adoption.
[AI-106] Repulsive Population Policy Gradients for Behaviorally Diverse Reinforcement Learning
链接: https://arxiv.org/abs/2610.36425
作者: Saeed Hedayatian,Harsh Shah,Celine Mneimneh,Stefanos Nikolaidis
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:The goal of Quality-Diversity Reinforcement Learning (QD-RL) is to find a population of high-performing policies that solve a decision-making problem in diverse ways. Despite recent progress, QD-RL methods predominantly rely on evolutionary operators–including black-box optimizers, mutations, and crossovers–to facilitate exploration and induce diversity. Scaling these methods becomes increasingly costly as policies and behavior spaces grow larger. Building on the Soft QD view of quality-diversity optimization, we introduce \textbfRepulsive \textbfPopulation \textbfPolicy \textbfOptimization (\textbfRePPO), which adapts PPO to jointly optimize a compact population of diverse, high-quality policies using a population-dependent, dynamically shaped reward. On standard continuous-control tasks, RePPO achieves a strong quality-diversity tradeoff compared with state-of-the-art QD-RL methods. Ablations further characterize this tradeoff by examining how each component affects individual quality and collective diversity. To assess RePPO’s scalability, we extend our evaluation to learning from pixels and fine-tuning large language models. In pixel-based tasks, where large policies make most traditional QD-RL methods infeasible, RePPO maintains quality on par with standard PPO while achieving substantially higher diversity. Its successful application to language model fine-tuning further demonstrates, for the first time, that QD-RL methods can scale to this domain. Together, these results show that directly optimizing compact populations with RePPO provides a scalable approach to QD-RL across increasingly complex domains.
[AI-107] Verifiable Efficiency Limits of Innate LLM Planning in Sliding Puzzles
链接: https://arxiv.org/abs/2610.36415
作者: Yezhou Cheng,Charles Ling
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Tool-assisted planning success can reflect external search and verification, leaving a model’s unaided planning ability unresolved. We isolate innate planning through complete plans generated in one response without tools or execution feedback. We propose a verifiable framework that decomposes success-weighted plan efficiency as \mathrmSPE=R\times Q\times E : execution reliability, planning coordination among legal plans, and action economy among completed plans. Trajectory replay and certified optima verify these factors; the joint score includes unsuccessful attempts. We evaluate 8,400 full-plan outputs from seven model conditions on 4\times4 sliding puzzles, controlling exact horizon and obstacle count in the primary cohort. Models that complete every trial still become less economical at longer horizons, while higher efficiency among completed plans can coexist with poor goal attainment. SPE separates the overall capability hierarchy, while its factors distinguish coordination failure from inefficient completion. The framework makes long-horizon limitations in unaided planning quantitatively verifiable.
[AI-108] kw-bench: Breaking Frontier Harnesses on Real-World Tasks
链接: https://arxiv.org/abs/2610.36405
作者: George Juraj Salapa
类目: Artificial Intelligence (cs.AI); Software Engineering (cs.SE)
备注: 26 pages, 8 figures, 5 tables. Code and tasks: this https URL
Abstract:Each new generation of frontier model pushes the edge of what is possible. Inside a harness – a loop, a shell, a filesystem, the ability to write and run its own code – a model can take on increasingly difficult work. A growing body of work treats the coding-agent harness as infrastructure for knowledge work rather than a coding tool, and a new wave of benchmarks tests it there. Most of them are generative: build a financial model, a memo, a spreadsheet from a brief, judged by a rubric or by another model. Few test the expensive part of knowledge work, which is investigative. In audit, tax, due diligence, regulatory reporting, claims or any dispute over numbers, the answer already exists – hidden in incomplete and inconsistent evidence spread across systems – and has to be recovered and defended to the cent. We introduce kw-bench, a benchmark built from real engagements specialists described to us. Its first task family is a five-year VAT reconstruction across three generations of SAP – legacy ECC, a partial S/4HANA migration and a consultant’s redesign – delivered as raw system extracts: page headers, local number formats, partial link tables, no documentation of how anything connects. Grading is deterministic, with no LLM judge. We ran four harnesses – Claude Code, Codex, OpenCode and micro-cc – on the same model, DeepSeek v4.1 Flash, for two hours. The two independent, model-agnostic harnesses delivered (OpenCode 0.542, micro-cc 0.528); the two lab harnesses, away from their home model, delivered nothing (0.00 each). What defeated them was inertia: keeping many weak signals alive across a long session. On a frontier model the task collapses. micro-cc and Claude Code on Claude Opus 5.5 scored 0.9986 and 1.000, in 22 and 25 minutes. A weeks-long engagement now costs minutes and a few dollars; below the frontier, the harness decides how far a model gets. Comments: 26 pages, 8 figures, 5 tables. Code and tasks: this https URL Subjects: Artificial Intelligence (cs.AI); Software Engineering (cs.SE) ACMclasses: I.2.11; I.2.1 Cite as: arXiv:2610.36405 [cs.AI] (or arXiv:2610.36405v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2610.36405 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Juraj George Salapa Ba Hons [view email] [v1] Mon, 26 Oct 2026 23:40:10 UTC (6,397 KB)
[AI-109] LoRA on a Budget: Function-Aware Post-Hoc Compression for Adapter Fleets
链接: https://arxiv.org/abs/2610.36386
作者: Shenghui Li,Thiemo voigt
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Low-Rank Adaptation (LoRA) has become a key mechanism for efficiently specializing foundation models into many task- and user-specific variants. Serving many LoRA adapters under a shared resource budget turns post-hoc compression into a two-level allocation problem: rank must be distributed across layers within each adapter and across adapters within the fleet. Although trained LoRAs contain substantial spectral redundancy, spectral reconstruction alone does not determine how much compression an adapter can functionally tolerate. We formulate fleet compression through a functional rank-distortion frontier for each adapter, followed by fleet-level minimax allocation under a shared rank budget. Given exact frontiers, this decomposition is equivalent to joint optimization over all layer ranks. We introduce Functional Rank Allocation (FRA), an unlabeled method for constructing practical frontiers. FRA measures isolated module truncations end to end, uses these measurements as an additive search surrogate solved by dynamic programming, remeasures every complete candidate, and takes the lower-risk envelope with a complementary spectral proposal before exact minimax allocation over the constructed frontiers. Across 78 adapters from three public fleets, FRA improves worst-case and tail utility over uniform spectral compression in constrained-budget regimes. Fixed-budget probes show that isolated costs rank candidates well, but their calibration shifts across proposal shapes; end-to-end verification and the spectral proposal therefore limit functional-search error.
[AI-110] HIDRA: Hierarchical Dual-Routing Attention for Replay-Free Lifelong Imitation Learning NEURIPS2026
链接: https://arxiv.org/abs/2610.36374
作者: Fanqi Yu,Matteo Tiezzi,Cigdem Beyan,Tommaso Apicella,Vittorio Murino
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注: This is the authors’ preprint, accepted to the NeurIPS 2026 Conference. The final version is available in the NeurIPS 2026 proceedings
Abstract:Robotic agents operating in real-world environments must adapt continuously from a stream of multimodal demonstrations, often under constraints that preclude storing or revisiting past data. In this replay-free, single-pass lifelong imitation learning setting, sequential updates can progressively distort the representation space, making expert-based approaches particularly sensitive to routing interference, causing expert selection to degrade over time. We propose HIDRA, a structured routing framework that mitigates this issue through a hierarchical dual-attention mechanism, which decouples instruction-level expert retrieval from context-dependent refinement. To further stabilize routing under representation drift, HIDRA introduces a key-level geometric regularization that enforces both alignment within tasks and separation across tasks. The resulting approach enables reliable expert selection while supporting an expandable set of experts for continual adaptation. We evaluate HIDRA on several LIBERO benchmark suites, including more challenging variants with heterogeneous task sequences and paraphrased language instructions. Our approach consistently outperforms both replay-free and replay-based baselines, improving AUC and forward transfer while maintaining low forgetting, with stronger gains under high interference and language variation. The code is available at: this https URL.
[AI-111] Reliability-Aware Short-Term Glycemic Risk Prediction Using an Anchor-Based Comparative Framework of Sequence Multimodal and State-Updating Digital-Twin Models
链接: https://arxiv.org/abs/2610.36363
作者: Pir Bakhsh Khokhar,Carmine Gravino,Fabio Palomba
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 36 pages, including figures, tables, and references. Research article on reliability-aware glycemic-risk prediction using sequence models and multimodal digital twins
Abstract:Anticipating short-term hypoglycemic and hyperglycemic risk is essential for proactive diabetes management. However, existing machine learning approaches often use inconsistent evaluation protocols and emphasize predictive performance over reliability. This study introduces an anchor-based benchmarking framework integrating longitudinal glucose sequences and multimodal patient information. Using the T1DiabetesGranada dataset, 5,503,441 standardized prediction anchors were constructed from 736 individuals with type 1 diabetes to evaluate hypoglycemia and hyperglycemia prediction at 30-, 60-, and 120-minute horizons. Three neural architectures, a Hybrid LSTM-Transformer, a Multimodal Digital Twin, and a State-Updating Digital Twin with gap-aware longitudinal state progression, were compared under identical patient-level validation protocols. Beyond predictive performance, the evaluation examined calibration, classification thresholds, explanation reliability, cross-model disagreement, and robustness to controlled perturbations. No statistically significant predictive differences were observed after Holm correction, although the models exhibited complementary reliability characteristics. The Multimodal Digital Twin achieved the lowest mean Brier score, while the State-Updating Digital Twin demonstrated the lowest mean expected calibration error (ECE-10). The Hybrid LSTM-Transformer showed strong explanation faithfulness and stability under its temporal attribution protocol. These findings highlight the importance of evaluating digital-twin models beyond predictive accuracy and emphasize the need for reliability assessment and external validation before clinical deployment.
[AI-112] WAKMan: Workload Aware KVCache Manager
链接: https://arxiv.org/abs/2610.36355
作者: Suhas Kumar Morisetty,Vaibhavi Lokegaonkar,Sachin Mathew Varghese,Wesley Harris
类目: Artificial Intelligence (cs.AI)
备注: Work done during internship at Capital One
Abstract:Prefix caching enables a Large Language Model (LLM) servers to reuse the Key-Value (KV) states and accelerate prefill computation stage for requests that share a prefix with earlier prompts, thus reducing latency and serving cost. With limited memory resource, the KV cache eviction policy becomes a decisive factor in determining the overall benefits of prefix caching. Production traffic may have single-turn, multi-turn, agentic, or a mixture of any combination of these request types. Our findings indicate that agentic steps return within seconds and favor recency, while popular prompts and long conversations return after many other requests and favor frequency. Across 20 mixes, no single policy is always best: vLLM’s default strategy LRU provides 12% higher cache hits as compared to the frequency-oriented S3FIFO on agentic-heavy mixes but trails it by up to 22% elsewhere, and even a single workload parameter can flip the winner even for any particular mix. We present WAKMan, a workload-aware KV cache manager that picks its eviction policy from live traffic without knowing request types or reading prompts. It improves hit rate over LRU by 5 percentage points on average, with the lowest mean and worst-case optimality gap of all policies.
[AI-113] OA-RAG : Ontology-based Agent ic RAG for Faithful LLM Reasoning
链接: https://arxiv.org/abs/2610.36335
作者: Hui Yang,Jiaoyan Chen,Yongsheng Gao
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models (LLMs) perform well across many real-world tasks but remain prone to hallucination, limiting their reliability in domains such as biomedicine and healthcare, where answers must be grounded in established knowledge. Retrieval-augmented generation (RAG) mitigates this problem by incorporating external information, but often overlooks logical relationships and relies heavily on the LLM’s own reasoning, which can compromise answer faithfulness. We propose OA-RAG, an agent that integrates formal reasoning with LLM reasoning, drawing on both the logical structure and informal annotations of domain ontologies. For each answer, OA-RAG reports the ontology observations used as evidence and a model-assessed grounding level (entailed, supported, related, or unrelated) that characterises the strength of that evidence. Across benchmarks derived from three large-scale ontologies and six real-world medical question-answering tasks, OA-RAG achieves the best overall performance among the evaluated ontology-fine-tuned and popular RAG methods, with particularly strong results on reasoning-intensive tasks. The reported grounding levels also offer a useful indication of answer reliability, with stronger grounding generally associated with higher accuracy. Finally, our results demonstrate the value of ontologies as knowledge sources: OA-RAG is competitive with, and sometimes outperforms, domain-specific RAG systems that draw on sources such as textbooks.
[AI-114] Plenty of Room in the Middle: Revealing Opportunities on a GPU Cluster via Fine-Grained Monitoring
链接: https://arxiv.org/abs/2610.36329
作者: Andrew Sheinberg,Samyak Gupta,Kai Li
类目: Artificial Intelligence (cs.AI); Distributed, Parallel, and Cluster Computing (cs.DC)
备注: 19 pages, 12 figures, 4 tables. Originally written in fall 2024
Abstract:GPU clusters have become essential computing resources for AI and scientific research in academia and industry. However, their high cost and growing demand often result in long queueing delays. While previous monitoring efforts have operated at second- or minute-level frequencies, they primarily informed job-level scheduling for clusters and lacked the detail needed to optimize GPU-level utilization. This paper presents an in-depth study of a university research GPU cluster, analyzing resource usage at 100 ms intervals. We have developed a lightweight monitoring tool that leverages Nvidia’s DCGM library to capture fine-grained “compute utilization” metrics within GPUs. Deployed on a cluster with 89 nodes and 319 GPUs, our tool collected data over several months. This paper focuses on a 10-day period with 7,399 jobs submitted by 150 users from 12 departments. Our study reveals three key findings: (1) the GPU cluster exhibits significant underutilization, with over 70% of jobs using less than 20% of GPU memory and 50% averaging below 5% Streaming Multiprocessor (SM) utilization; (2) 80% of jobs have a drastic mismatch between their requested and actual durations, significantly limiting the scheduler’s effectiveness; and (3) the multi-GPU jobs we examined show highly periodic communication patterns. These findings suggest there is plenty of room in the middle: substantial utilization gains are available through better job scheduling and GPU multitasking, independent of hardware or application improvements. Our monitoring solution is available as open-source software at this https URL. Comments: 19 pages, 12 figures, 4 tables. Originally written in fall 2024 Subjects: Artificial Intelligence (cs.AI); Distributed, Parallel, and Cluster Computing (cs.DC) Cite as: arXiv:2610.36329 [cs.AI] (or arXiv:2610.36329v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2610.36329 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-115] How Stable Are LLM -Generated Proofs Under Program Refactoring? A Stress Test for Verifiable Code Generation NEURIPS2026
链接: https://arxiv.org/abs/2610.36328
作者: Aravind Mandiga,Guoming Li
类目: Artificial Intelligence (cs.AI); Logic in Computer Science (cs.LO); Software Engineering (cs.SE)
备注: Accepted at the NeurIPS 2026 Workshop on AI for Verifiable Coding
Abstract:Benchmarks for verifiable code generation usually test a proof against one fixed implementation. We ask whether the same proof remains valid after a program refactoring that preserves behavior. We introduce RefactorProof, which derives 458 Lean-certified variants from the 185 VERINA tasks that compile in our pinned environment while holding the specification and evaluated proof fixed byte-for-byte. VERINA supplies reference proofs for 46 tasks, all from its basic split; the primary model-versus-reference comparisons further match on 21-25 short, non-recursive tasks. Structural refactorings expose substantial brittleness across proof sources: task-normalized structural PSR is 4.3% for the supplied references and 13.3-26.8% for the specialized provers. On the matched tasks, specialized prover proofs nevertheless have higher overall unchanged-proof survival than the supplied reference artifacts: +15.6 points for DeepSeek-Prover-V2-7B, +18.8 for Goedel-Prover-V2-32B, and +20.3 for Goedel-Prover-V2-8B; all three 95% CIs exclude zero. The ordering persists after excluding universally survived T1 variants, under family-balanced aggregation, and under alternative deployable valid-proof selection rules. However, unchanged-proof failure is not the same as high repair cost: predefined one-identifier edits repair 51/54 eligible reference structural failures and 58/85 eligible prover failures. RefactorProof therefore measures unchanged-proof stability under program refactoring, not the cost of maintaining a proof after it breaks.
[AI-116] When Retrieval Is Not Enough: Obsolete Facts in Agent Memory NEURIPS2026
链接: https://arxiv.org/abs/2610.36308
作者: Oleeviya Babu Poikarayil,Tony Basil Babu,Abdulrahman Nahhas,Het Darshan Mehta,Jawid Ahmad Baktash,Ernesto William De Luca,Klaus Turowski
类目: Artificial Intelligence (cs.AI)
备注: Accepted at the NeurIPS 2026 Workshop on Personalized, Aligned, Long-Term Memory for AI Systems (PALM). 23 pages, including supplementary material
Abstract:Memory-agent benchmarks often collapse retrieval and answer generation into one score, making it hard to tell whether a failure comes from missing an update or from not using an update that was already retrieved. We examine this distinction on the single-hop FactConsolidation task of MemoryAgentBench. With Qwen2.5-7B-Instruct on 6K-history BM25 accuracy drops from 45% to 32% when the retrieved evidence budget grows from 64 to 512 tokens, even though a gold-assisted post-hoc audit finds the newest statement in every retrieved set. Most of the errors occur on questions that also contain an obsolete statement version. Under a frozen prompt and equal four-statement evidence size, replacing one unrelated statement with the obsolete competing statement reduces accuracy from 100% to 44.6% for Qwen2.5 and from 100% to 87.8% for Qwen3-14B, with all 50 harmful paired flips classified as stale. A 32K-history replication likewise favors BM25-64 over BM25-512 (69% versus 58%). These results suggest that retrieving the correct fact is not enough. An obsolete competing statement can be more harmful than an unrelated piece of evidence of the same size.
[AI-117] Spiking Neural Networks Are Harder to Steal: A Benchmark Study of Model Stealing Attacks
链接: https://arxiv.org/abs/2610.36298
作者: Ayana Moshruba,Maryam Parsa
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Machine learning models deployed behind query interfaces expose more than their predictions, because an adversary can use those predictions to rebuild the model itself. Model stealing attacks (MSAs) exploit this exposure by training a surrogate model that replicates the decision behavior of a target through black-box queries alone, without access to its parameters or training data. These attacks are well characterized for conventional artificial neural networks (ANNs), yet spiking neural networks (SNNs) process information in a fundamentally different way. SNNs communicate through sparse, discrete spikes, and their outputs depend on membrane potentials that evolve over time rather than on a static mapping from input to output. This temporal mode of computation suggests that the decision behavior of an SNN may be harder to reconstruct from its responses. This work tests that possibility by benchmarking four representative MSAs, namely the Papernot, Tramèr, T-RND, and color perturbation attacks, against SNNs and their ANN counterparts on MNIST, Fashion-MNIST, and CIFAR-10. Across every attack and dataset, SNNs prove consistently more difficult to steal. Surrogates of SNNs achieve 10% to 20% lower agreement with the target model and 15% to 25% lower adversarial transferability than surrogates of ANNs. We attribute this resistance to the way SNNs compute, as their spike-based and time-dependent dynamics yield decision boundaries that ANN-based surrogates struggle to approximate, even under large query budgets. These findings extend the known resilience of SNNs beyond attacks on training data to the replication of model functionality, strengthening the case for neuromorphic architectures in security-sensitive applications.
[AI-118] Detecting and Mitigating Query-Based System Prompt Leakage Attacks
链接: https://arxiv.org/abs/2610.36295
作者: Tushar Nayan,Pratik Poudel,Qian Lou,Yuede Ji,Jason Liu,Ruimin Sun
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:
Abstract:Large Language Models (LLMs) rely on proprietary system prompts to enforce task directives, safety policies, and access controls. These prompts are vulnerable to query-based leakage attacks, where black-box adversaries extract confidential instructions through adaptive, multi-turn interactions. Existing defenses often assess queries independently, missing malicious behavior that emerges across interactions. We present AdvGuard, a model-agnostic framework for detecting and mitigating system prompt leakage. AdvGuard models cross-query relationships using a bounded semantic interaction graph, statistically filters candidate components, and converts them into line graphs for graph convolutional network (GCN) classification. This approach captures multi-hop patterns in query similarity that distinguish coordinated extraction from benign workflows. A lightweight prompt-level detector provides protection before sufficient interaction history accumulates, and the framework supports black-box LLM deployments. In mixed-query evaluations, AdvGuard improves recall by up to 36% over the strongest baseline while maintaining a false positive rate of at most 3.4%. In multi-turn workloads, it achieves 97.23%-99.1% F1 with 100% recall. Leave-one-attack-family-out tests yield 99.98% mean recall and 99.99% mean F1, with no observed false positives. Integrated into the adaptive PLeak and LeakAgent attack loops, AdvGuard limits exact-match leakage to 0.00-0.01 and substring-match leakage to 0.01-0.09 across evaluated victim models. These results demonstrate robust detection of unseen attack families and substantially reduced prompt reconstruction. Subjects: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI) Cite as: arXiv:2610.36295 [cs.CR] (or arXiv:2610.36295v1 [cs.CR] for this version) https://doi.org/10.48550/arXiv.2610.36295 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-119] BUSTA: Teaching Small Language Models to Reason About Binary Execution
链接: https://arxiv.org/abs/2610.36288
作者: Ritika Poddar,Aidan Fennelly,Deirdre Murphy,Eric Wolos,Corre L. Steele,Michael Doyle
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Programming Languages (cs.PL)
备注:
Abstract:Large Language Models (LLM) are increasingly applied to Reverse Engineering (RE) workflows. RE, however, differs from common LLM coding tasks as it requires models to reason about low-level program behavior. Furthermore, the deployment of tooling in resource-constrained sandboxed RE environments limits the ability to use frontier-scale solutions. We examine whether a targeted LLM training approach can transfer distinct RE skills to open models via inverse basic-block execution: given an x86-64 basic block and a desired output register state, the model predicts the input register values that produce it. Fine-tuning \textttQwen3-1.7B and \textttNemotron-3-Nano with Group Relative Policy Optimization (GRPO) raises knowable-register prediction accuracy from 73.2% to 85.3% and from 63.3% to 95.2%, respectively. Frontier models still score highest (up to 99.2%), but our fine-tuned \textttNemotron-3-Nano comes within four points at a fraction of the parameter count. We release both fine-tuned models and the training dataset.
[AI-120] A Smart Instant Book Filter Model based on Dual-Objective Optimization ATC WSDM2018
链接: https://arxiv.org/abs/2610.36287
作者: Yi Hou,Li Fan,Huizhong Duan
类目: Artificial Intelligence (cs.AI)
备注: 6 pages, 5 figures. Presented at TSMO’18: WSDM 2018 Workshop on Two-sided Marketplace Optimization: Search, Pricing, Matching Growth, Marina Del Rey, CA, USA
Abstract:Hospitality service in online marketplaces has become increasingly popular over the past few years, not only because it helps people easily rent short-term lodging but also because it empowers travelers to connect with locals and feel belonging. A typical pattern of these industries is a central platform mediating the economic interactions between guests and hosts, where a successful booking experience is the key to user satisfaction. However, one of the biggest challenges is the possible delay of response and chance of rejection when guests try to book a request-to-book listing. Previously, we enabled the instant book feature to allow hosts to instantly accept qualified guests and guests to confirm bookings immediately. Yet the symptom has not been alleviated given mixed request-to-book and instant book listing supply in search results. One possible solution is to promote all instant book listings in search results; however, booking rate will be damaged due to some improperly promoted low quality listings. In this paper, we present a novel smart instant book filter system that applies the instant book filter intelligently and shifts request-to-book to instant book without hurting booking rate. Compared with traditional two-sided markets research that targets finding strategies to maximize a single overall social welfare goal, our proposed system aims to maximize the instant book rate with the constraint that booking rate should not decrease, forming a dual-objective optimization problem. We show A/B testing results for our system exposed to global users, demonstrating that our model performs better than previous human-expert-tuned models in terms of both instant book rate and overall booking rate. We also achieved improved metrics such as the host instant book adoption rate. The proposed system has been deployed to production and is serving millions of users every day. Comments: 6 pages, 5 figures. Presented at TSMO’18: WSDM 2018 Workshop on Two-sided Marketplace Optimization: Search, Pricing, Matching Growth, Marina Del Rey, CA, USA Subjects: Artificial Intelligence (cs.AI) Cite as: arXiv:2610.36287 [cs.AI] (or arXiv:2610.36287v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2610.36287 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-121] Freeze Dont Clone: Fine-Tuning with Expressive Policies Without Retaining Offline Data
链接: https://arxiv.org/abs/2610.36247
作者: Sami Nur Islam,Johan Obando-Ceron,Doina Precup,Aaron Courville,Pablo Samuel Castro
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Offline-to-online reinforcement learning uses offline pretraining to accelerate subsequent online learning. While many methods assume the offline dataset remains available throughout fine-tuning, there are many settings where this is not the case. Prior work attributes failures in this setting primarily to critic recalibration under distribution shift. We identify a complementary actor-side failure: once the offline data disappears, continued behavioral cloning (BC) no longer preserves the offline behavior distribution and instead refits the behavioral model to the agent’s limited, and nonstationary, online experience. We demonstrate that stopping online behavioral fitting while preserving the pretrained expressive prior substantially improves no-retention fine-tuning across implicit flow Q-learning (IFQL), flow Q-learning (FQL), and action-chunked RL (QC) on seven OGBench domains. We trace this failure to a behavioral feedback loop: continuing online behavioral cloning rapidly erases the pretrained prior, inducing representational collapse and action diversity loss that starves the critic of useful exploration. Freezing the prior breaks this cycle, preserving offline coverage while allowing value learning to drive continuous improvement.
[AI-122] MARQ: A Modular LLM -Agnostic Requirement Quality Assessor Infused with Systems Engineering Knowledge
链接: https://arxiv.org/abs/2610.36234
作者: Jannatul Shefa,Alejandro Salado,Paul Wach,Taylan G. Topcu
类目: Artificial Intelligence (cs.AI); Software Engineering (cs.SE)
备注:
Abstract:Requirements engineering (RE) is foundational to systems engineering (SE), as requirement defects can propagate downstream and adversely affect development outcomes. Although large language models (LLMs) are promising for requirement quality assessment, prior work shows that off-the-shelf models miss a large share of expert-identified defects, perform unevenly across quality criteria, and vary across executions. This study investigates whether infusing expert-derived SE knowledge can improve LLM-based requirement quality evaluation beyond off-the-shelf capability. We present MARQ, a modular, LLM-agnostic prototype that supplements LLMs with a criterion-specific knowledge stack operationalizing INCOSE requirement quality criteria and evaluated against off-the-shelf baselines on an expert-labeled NASA requirement set using Claude Opus 5, GPT-5.6 Sol, and the open-weight Gemma 3 12B, with 100 executions per model. MARQ achieves median F1 scores of 0.80 for Claude Opus 5, 0.62 for GPT-5.6 Sol, and 0.53 for Gemma 3 12B, compared with off-the-shelf baselines of 0.55, 0.52, and 0.18, respectively. However, the nature of these gains is model-dependent. Claude Opus 5 improves defect detection and false-alarm control; GPT-5.6 Sol gains detection and better execution-level consistency at the cost of more false alarms; and Gemma 3 12B reaches the performance of off-the-shelf proprietary models, although it remains highly variable. Criterion-level analysis shows gains concentrated in Correctness and Conformance, trade-offs between misses and false alarms within individual criteria, and model-dependent responses across several criteria. These findings position SE knowledge infusion as a promising model-agnostic pathway for strengthening trustworthy LLM-based RE support while underscoring the need for criterion-specific interventions and human-in-the-loop review.
[AI-123] Agent or Human? Detecting Commits by AI Coding Agents in the Absence of Authorship Signatures
链接: https://arxiv.org/abs/2610.36229
作者: Kosei Horikawa,Hao Li,Yutaro Kashiwa,Youssef Esseddiq Ouatiti,Mohammed Sayagh,Bram Adams,Hajimu Iida,Ahmed E. Hassan
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注: 21 pages
Abstract:AI coding agents such as Claude Code, GitHub Copilot coding agent, and Cursor now commit directly to open-source repositories. Empirical studies of these agents identify their commits through authorship signatures, such as bot accounts and Co-authored-by trailers. These signatures are optional, can be removed, and are often attached to pull requests rather than to individual commits. Hence, it is unclear whether Agent commits can be told apart from Human commits once the signatures are gone. We investigate the signals and limits of signature-free detection using 7,837,946 commits from 1,226 repositories, covering five coding agents and Human commits. After signatures are masked, detectors rank Agent commits above Human commits with a ROC AUC of up to 0.859, and most of this signal comes from message verbosity and change size rather than from the code itself. On the shared sample, gpt-6-sol without fine-tuning reaches AUCs similar to XGBoost and TF-IDF. At the observed Agent share, however, the best detector on the full test set reaches a precision of only 0.332. Detection transfers poorly to held-out Claude Code and Copilot, with TF-IDF reaching an AUC of only 0.524 for Claude Code. Temporal changes also depend on the agent, e.g., Copilot AUC drops and later recovers, while more training history helps Claude Code but not Copilot. These results indicate that signature-free detection can prioritize commits for review but cannot replace recorded attribution, and that studies relying on signature-based labels should treat them as a lower bound on agent activity.
[AI-124] Concorde: Geometry-Aware Link Prediction via Decoupled Energy Minimization NEURIPS2026
链接: https://arxiv.org/abs/2610.36204
作者: Ege Demirci,Ambuj Singh
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: NeurIPS 2026
Abstract:Link prediction remains difficult in sparse graphs, where diffusion heuristics lack connectivity and geometric methods suffer spectral instability when coupling learned representations with diffusion operators. We introduce Concorde, an energy-based framework that resolves this by decomposing link likelihood into three structurally decoupled branches: (i) a parameter-free spectral anchor that establishes permutation-invariant topological stability; (ii) a mixed-curvature residual on a learnable \mathbbH^d_h_c_h \times \mathbbS^d_s_c_s product manifold, regularized by a feature-aware Ollivier-Ricci curvature proxy; and (iii) a continuous vector field on the tangent bundle of hyperbolic space for non-local structural alignment across topological gaps. The branches are fused via learned gating and trained with an N-pair ranking loss aligned with retrieval-style evaluation. Across five benchmarks spanning sparse, dense, featureless, and attributed regimes, Concorde establishes state-of-the-art early retrieval precision, nearly doubling P@10% over leading baselines in sparse settings.
[AI-125] count-catcher: Lean Theorems from Model Counting Certificates
链接: https://arxiv.org/abs/2610.36196
作者: Johannes K. Fichte,Stefan Szeider
类目: Logic in Computer Science (cs.LO); Artificial Intelligence (cs.AI)
备注:
Abstract:We present count-catcher, a tool that turns a CPOG certificate of a count by the exact model counter d4 into a Lean theorem that a CNF formula defined in Lean has exactly N models over V variables. With an encoding lemma, this theorem becomes a theorem about the number of combinatorial objects. A CPOG certificate consists of a Boolean circuit equivalent to the formula and a proof of equivalence. count-catcher checks in Lean that the formula implies the circuit (forward check) and the circuit implies the formula (reverse check). For the forward check, lrat-catcher streams a refutation through the LRAT checker of Lean’s standard library. The reverse check and the validity check of the circuit are computations proved correct in Lean. Lean runs all three checks as compiled code, so the theorem depends on Lean’s kernel, compiler and runtime, as does CPOG’s verified checker. We combine the theorem with results from Mathlib, for instance with Burnside’s lemma to count graphs up to isomorphism. We prove the number of Latin squares of order 8 as a weighted sum of hundreds of certified counts from about eight terabytes of certificates. We turn about half of the 2022 Model Counting Competition formulas into theorems, with the three checks taking about a quarter of the time of CPOG’s verified checker. We found that CPOG’s unverified checker reported wrong counts, and removed its verified checker’s dependence on two unproved lemmas. End to end, a theorem costs about the same as a checked count.
[AI-126] Skill Dreaming: Evolving Repository Skills through Pull Request Streams
链接: https://arxiv.org/abs/2610.36188
作者: Xinyuan Song,Hasibul Haque,Jay Somasundaram,Ryan Turner,Bowen Zhu,Liang Zhao
类目: Artificial Intelligence (cs.AI); Software Engineering (cs.SE)
备注:
Abstract:Large language model (LLM) agents increasingly rely on reusable skills for complex software engineering tasks, making skill libraries essential for repository development and maintenance. However, repository-specific skills must evolve with changing code structures, implementation conventions, and maintenance requirements. Pull requests (PRs) provide a natural source of such knowledge by recording how repositories evolve through implementation and review. Meanwhile, new PRs must follow repository-specific implementation patterns, modify related components consistently, and ensure that existing tests continue to pass while adding coverage for new functionality. This creates a mutually reinforcing cycle: PRs provide knowledge for updating skills, while these skills improve future PRs, maintain repositories, and support further skill evolution. To enable this co-evolution of PRs, repositories, and skills, we propose Skill Dreaming, a framework that learns from PR streams to guide continuous repository development and maintenance. Given a repository and its PR history, Skill Dreaming builds a repository-specific skill library by extracting executable knowledge from validated PRs and continuously refines the skills as new PR information becomes available. These skills guide coding agents not only in implementing requested features but also in maintaining associated tests, documentation, compatibility, and repository conventions. By integrating PR analysis, skill management, code execution, and continuous integration, Skill Dreaming provides an end-to-end workflow for applying learned skills to repository changes and incorporating subsequent PR outcomes into future skill updates.
[AI-127] Can AI Guide Testing Without Generating Tests? A Comparative Study of AI-Assisted Unit Testing
链接: https://arxiv.org/abs/2610.36185
作者: Zihan Fang,Ron Asoulin-Handelman,Helen Wu,Yu Huang,Gina R. Bai
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注: 10 pages, 2 figures
Abstract:Recent advances in generative AI have enabled assistants to generate increasingly sophisticated executable unit tests. However, novices may accept these outputs without independently identifying what to test, evaluating test-suite completeness, or verifying correctness, potentially limiting opportunities to develop essential testing skills. Although prior work has explored AI-assisted test generation and feedback-oriented programming tutors, little is known about whether AI can support novices’ testing practices without generating tests for them. To investigate this, we introduce TestCompass, a generative AI tutor that provides targeted feedback on weaknesses in students’ test suites without generating executable tests. In a classroom study with 18 students, we compared TestCompass with a general AI assistant across two Java unit-testing tasks, examining students’ testing behaviors, test-suite quality, and subjective experiences. We found that participants using the general-purpose AI assistant produced more tests, whereas TestCompass participants spent more time developing and verifying tests themselves. However, most test-suite quality measures were comparable between conditions, and TestCompass participants achieved similar requirement coverage with significantly fewer test methods ( p=0.012 ) in one task. Neither group reported short-term improvements in perceived testing knowledge or confidence. These findings suggest that feedback-oriented AI can support meaningful testing outcomes while preserving opportunities for hands-on testing practice, offering a complementary approach to AI-assisted testing education.
[AI-128] Learning Human Sports Motion Feasibility via Quantile Regression on Tracking Data
链接: https://arxiv.org/abs/2610.36183
作者: Marc Gutiérrez-Pérez,Antonio Agudo
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:We study time-to-arrival (TTA) in sports: the shortest plausible time a player needs to reach an arbitrary location on the pitch given their current kinematic state, estimated as the lower envelope of observed locomotion times. Rather than modeling space ownership through position density estimation or closed-form physics-based models, we cast TTA as a conditional quantile regression problem, estimating the lower quantile of locomotion time so the model learns a player’s lower-envelope arrival time. We train the model on optical tracking data from all 64 matches of the PFF FC FIFA World Cup 2022 dataset, and evaluate a lightweight multilayer perceptron against six player-motion models from the literature, re-expressed as TTA predictors, and a learned reachable- hull baseline. Across our evaluation metrics, the MLP consistently outperforms every physics-based baseline. To probe the generality of the learned locomotion model, we further evaluate it on basketball tracking data and a single-game NFL tracking dataset, analyzing the transferability of player motion dynamics across sports. Our project is available at this https URL.
[AI-129] Stratum: Repository Memory for Coding Agents Beyond the Context Window
链接: https://arxiv.org/abs/2610.36153
作者: Yuvraj Sehgal,Mahsa Panahandeh,Naser Ezzati-Jivan
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注: 5 pages, 1 figure, 1 table. Replication package: this https URL
Abstract:Context windows have grown a thousand-fold, yet coding agents still start every session with no memory of their past work. They re-read files, retry failed approaches, and forget root causes they established hours earlier. This is not a context-length problem but a memory-architecture problem. Our hypothesis: coding agents need persistent, experience-aware repository memory, so that knowledge survives from one session to the next. Three measurements on real repositories motivate the design. First, mid-size projects already exceed the largest production context windows, and storing knowledge inside the model’s key-value (KV) cache costs 10^4 to 10^5 times more bytes than storing it as indexed text. Second, across 752 real bug-fix commits, the median fix edits two files, and these files plus their direct imports amount to less than 10% of the repository; the bottleneck is selection, not capacity. Third, even with transitive import reachability and test links, 47% of co-edited file pairs remain structurally disconnected; the knowledge needed for selection must be mined from history and experience. We sketch Stratum, a tiered memory that lives with the repository and serves it through page-safe tools: every response states what was returned, what remains, and what it cost. Retrieval over a store of one million nodes takes 1.7 ms at the median. In a multi-session experiment on eight real bugs, a fresh agent that inherited a prior session’s memory, written by a different model, identified the root-cause file in 21 of 24 trials versus 6 of 24 for a retrieval-only baseline (Fisher p10^-4 ), at 4.5x lower median token cost. With the notes of all eight investigations pooled in one store, success remained 8/8.
[AI-130] From Structural Feedback to Prompt Policies: Learning Faithful Text-to-Image Prompt Editors
链接: https://arxiv.org/abs/2610.36146
作者: Wentao Ye,Yali Ye,Zhiqing Xiao,Ru Peng,Xinpeng Ti,Liyao Li,Zhanming Shen,Peng Lu,Jie Zhang,Haobo Wang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Prompt optimization should improve the image a user receives without changing the scene the user requested. Yet stylistic rewriting can weaken object attributes and relations even when global text–image similarity remains high. We propose StruPO, a structure-aware prompt editor that uses a calibrated scene graph as a shared representation for policy conditioning and image-grounded supervision. The graph exposes what should be preserved, while deletion, reordering, and insertion provide a token-aligned way to modify the prompt. This turns structural feedback from an external critique into a training signal for a reusable editing policy. Learning from that signal requires handling both uncertain grounding and stochastic image generation. StruPO retains grounding confidence through smooth reward shaping and averages multiple images for each edit before computing group-relative advantages. Experiments span three disjoint 2K-prompt test sets and four frozen generation backbones, with comparisons to learned prompt optimizers, iterative structural feedback, diffusion-model fine-tuning, and inference-time attention control. StruPO improves the faithfulness–aesthetics trade-off, with consistent gains in relational fidelity. Independent automatic evaluators and a 15-annotator study provide complementary evidence, and human preference margins increase with relation complexity. Controlled ablations connect these gains to graph conditioning, local editing, and image-grounded learning. The resulting policy produces reusable optimized text in one pass after graph calibration, without an iterative critique–rewrite loop.
[AI-131] Why and How to Do Science When AI Is Better at It: A Framework Grounded in Human Understanding and Direction
链接: https://arxiv.org/abs/2610.36116
作者: Hoda Heidari,Aydin Mohseni
类目: Artificial Intelligence (cs.AI); Computers and Society (cs.CY)
备注:
Abstract:As AI begins to match or exceed human researchers at core scientific tasks, two questions become pressing: \emphwhy should humans keep doing science, and \emphhow should they use AI when they do? This paper argues that the second can be answered only by answering the first. It holds that one indispensable aim of science is \emphdurable, collective human understanding of the world, pursued in service of human ends and values. From this principle it derives a two-part test for when a research task can be delegated to AI. The first part, \emphhuman understanding, requires that the scientific community remain able to explain, defend and contest the resulting claims, and retain the skills to do so. The second, \emphhuman direction, requires that judgments about which questions matter and what findings mean stay with accountable humans. Applying the test to each phase of research in the Contributor Roles Taxonomy (CRediT) yields recommendations that fields can tailor to their questions and methods. The framework differs from existing guidance in three ways. It rests on an account of what science is for rather than on integrity norms. It holds even if AI outperforms human researchers, rather than depending on AI’s current limitations. And it judges each use of AI by its effects on human understanding and direction rather than by the type of tool, across every phase of research, including peer review and the training of new researchers. It closes with the reforms needed to keep science aimed at human understanding and direction.
[AI-132] Context Is a Pipeline Not a Window: Information Attrition in LLM Agent Harnesses
链接: https://arxiv.org/abs/2610.36114
作者: Yuvraj Sehgal,Mahsa Panahandeh,Naser Ezzati-Jivan
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注: 5 pages, 1 figure, 2 tables. Replication package: this https URL
Abstract:LLM agents are judged by two numbers: the size of the context window, and how often the final answer is correct. Both numbers miss a third thing: how much of a tool’s output is lost before the model ever sees it. We measured this loss in a real tool-using agent. The agent finds the cause of injected faults by reading Linux kernel traces. The context window was almost empty: across 720 runs, the single largest prompt used only 5.75% of it. Yet a third of all tool results (3,827 of 11,189) were cut before they reached the model, and every cut result became broken JSON. The harness cut each result by slicing its JSON text at a fixed length. This does not shorten a result evenly. It deletes whatever data sits near the end. One field reached the model in only 4.7% of calls. A second field, in the very same results, reached the model in 99.3% of calls. Which field held the fault’s evidence depended on when the faulty process started: early or late in the trace. The agent’s diagnosis accuracy matched this same split. So we ran controlled experiments to test one question: did the data loss cause the drop in accuracy? This paper makes four contributions. First, we model an agent’s context as a five-stage pipeline, with a measured failure at each stage. Second, we replay 864 real tool results, including one from a production trace where the shipped cut deletes all 128 records of the fault’s evidence. Third, a controlled experiment shows that the cut’s real cost appears first as extra work, not as wrong answers: 43% more tool calls, twice the computation, and yield falling from 0.61 to 0.31. The agent still finds the right answer when it has another way to reach it. Fourth, a survey of seven agent frameworks finds two that cut tool output the same unsafe way. Whether the data cut truly causes the accuracy drop is still an open question.
[AI-133] Scaling and Renormalization in Neural Networks: An Effective Renormalization-Group Description of Grokking
链接: https://arxiv.org/abs/2610.36113
作者: Min Chen,Junqi Wang,Zeguan Wu,Wenhao Huang,Junyu Liu
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); High Energy Physics - Theory (hep-th); Quantum Physics (quant-ph)
备注: 22 pages, 10 figures
Abstract:Grokking is a delayed transition from memorization to generalization in deep neural networks (NNs). We examine whether scaling and renormalization group (RG) theories can describe this dynamics, establishing a quantitative physical framework for neural generalization transitions. For modular addition on a two-layer NN, an ensemble of trained models defines an order parameter (the squared weight norm) and fluctuation proxies for the heat capacity and the susceptibility, with training epoch and weight decay serving as thermal and field control coordinates. To bridge neural architectures to real-space RG, we construct an exact coupling matrix via a second-order expansion of the loss with respect to per-neuron norms and use its Fiedler vector to define an effective geometry. This spectral ordering yields a monotonic spatial decay of coupling strength that is absent under random permutations. We then execute real-space RG by block-spin aggregating adjacent neurons and parameterizing the coupling matrix into off-diagonal band strengths. Linearizing this local RG map yields relevant scaling exponents and five derived critical exponents, with direct empirical estimates matching RG predictions within 1-10%. Overall, our framework provides a consistent effective description of grokking, suggesting the scaling and renormalization group formalism as a tool to investigate NN dynamics.
[AI-134] Dirichlet Process Mixtures within Semiparametric Bayesian Networks: A Novel Conditional Density Estimation Approach
链接: https://arxiv.org/abs/2610.36110
作者: Victor Alejandre,Pedro Larrañaga,Concha Bielza
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
备注:
Abstract:Semiparametric Bayesian networks (SPBNs) are flexible probabilistic graphical models for representing joint distributions of continuous variables through lower-dimensional conditional probability distributions (CPDs). Existing SPBN models use kernel density estimators (KDEs) for non-parametric CPD modeling, but their one-component-per-observation mixture representation can lead to increased inference costs and density estimation errors, which can be limiting. In this work, we define a new CPD modeling approach for non-parametric components based on Bayesian non-parametric Dirichlet Process Mixtures (DPMs) within SPBNs. DPMs provide a data-adaptive clustering of observations, yielding a more compact representation while retaining the flexibility of non-parametric density estimation, unlike KDE. This results in lower inference costs while improving density estimation performance. We implement the proposed DPM-based SPBN (DPM-SPBN) by extending the open-source PyBNesian package and evaluate it on synthetic and real-world data. For a given network structure, DPM-based CPDs significantly outperform state-of-the-art Bayesian network models for continuous variables in density estimation, achieving lower absolute log-likelihood errors on synthetic data. We further investigate the impact of the proposed CPD modeling on structure learning methods relying on CPD-based scores. DPM-based SPBNs achieve lower structure learning and density estimation errors than KDE-based SPBNs. Experiments on real-world data corroborate these findings, identifying DPM-based CPD modeling as the most effective alternative among the evaluated non-parametric approaches for continuous variables.
[AI-135] Log-To-Leak: Prompt Injection Attacks on Tool-Using LLM Agents via the Model Context Protocol
链接: https://arxiv.org/abs/2610.36108
作者: Yuepeng Hu,Chongyu Fan,Sirat Samyoun,Jian Du
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:
Abstract:LLM agents increasingly access external tools through the Model Context Protocol (MCP), but the natural-language metadata used to describe those tools creates a prompt-injection surface. Within this established tool-poisoning surface, we study Log-To-Leak, an additive post-hoc logging attack that induces a separate attacker-controlled call after the final legitimate call. Its payload targets the user query, MCP server name, server response, and agent response. We operationalize the injected directive through Trigger, Tool Binding, Justification, and Pressure, and separately measure call success, payload fidelity, and preservation of legitimate behavior. We evaluate five MCP servers and four LLM agents using five metrics spanning effectiveness, utility, and efficiency. Across 18 model-server pairs with valid tool execution, Log-To-Leak achieves an attack success rate (ASR) of 62.64-100% and conditional logging accuracy (LA) of 76.55-95.64%; 16 pairs have absolute target-task completion changes of at most 6.63 percentage points. Three evaluated defenses leave residual ASR above 55% on GitHub MCP with GPT-4o, while an instruction detector flags 85.53% of benign descriptions. These results show that successful task completion can coexist with unauthorized post-hoc disclosure and motivate host-controlled authorization for secondary tool calls.
[AI-136] CohortFlow: A Steerable Clinical Cohort Curation Agent with Localized Semantic Correction
链接: https://arxiv.org/abs/2610.36100
作者: Neil Chen,Adrian Serapio,Hanxue Gu,Evan Lee,Kang Wang,Yang Yang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Language model agents for clinical cohort curation should accommodate evolving eligibility criteria without repeatedly processing unchanged requirements. We introduce CohortFlow, a steerable agent that represents natural language corrections as localized updates to a structured cohort specification and reconciles them with a persistent execution graph. Explicitly recorded semantic dependencies enable affected computations to be selectively invalidated while preserving valid prior results. On 100 scenarios derived from clinical-trial eligibility criteria, comprising 126 correction turns, localized steering achieves 77.0% requirement micro-F1 versus 76.4% for full regeneration using the same 8B model, a small numerical advantage that is not statistically significant ( p=0.493 ). It reduces mean correction latency by 64.8% after completion and 44.3% during execution, with 38.1% fewer model calls. In a separate proof-of-concept evaluation over MIMIC, all 13 cohort states execute successfully, while correction trajectories demonstrate selective preservation of unaffected computations and recomputation of changed or dependent nodes. Together, these results show that localized semantic steering can substantially reduce correction latency and model calls while maintaining strong requirement-level accuracy.
[AI-137] Rollback: Experience Backtracking in World Models for Policy Improvement
链接: https://arxiv.org/abs/2610.36068
作者: Jiayi Luo,Yang Zhang,Liliang Chen,Nan Wang,Chen Gao,Boxiang Qiu,Zhuoxuan Li,Baorui Peng,Chuankang Li,Yuxiang Yan,Si Liu,Qingyun Sun,Shuicheng Yan,Maoqing Yao,Jianxin Li,Guanghui Ren
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Action-conditioned world models offer a scalable alternative to costly real-world interaction for improving robot policies. However, existing world-model-based methods largely treat world models as forward-only surrogates of the physical environment, leaving untapped a capability no physical environment affords: rewinding to past decision points to explore counterfactual futures. We introduce Rollback, a policy improvement paradigm that turns world models from passive simulators into counterfactual engines: it rolls back to critical decision points, branches into alternative actions, and contrasts their outcomes to attribute success to individual decisions. This concentrates imagination where it matters most and distills counterfactual outcomes into fine-grained, action-level supervision without a separately learned value model. We build a scalable training system that unifies distributed world-model search with asynchronous real-robot execution, forming a self-reinforcing cycle: real-world experience continually sharpens the world model, and a sharper world model in turn drives further policy improvement. We term the world-model side of this cycle test-time physics grounding, a capability that hand-built physics simulators lack: while the world model serves as the training environment, real transitions from the policy it trains correct its dynamics through a lightweight LoRA pipeline. Experiments on three challenging real-world tasks show that Rollback improves success rates over the strongest baseline by an absolute 10.67%, 11.11%, and 11.85% on block extraction, ball tossing, and tabletop curling, respectively, while achieving more stable improvement throughout training. Although initialized from teleoperated demonstrations far below human level, Rollback surpasses the average success rate of human players on both ball tossing and tabletop curling within 8 hours of training.
[AI-138] Do Coding Agents Still Need a Debugger? Frontier-Model-Based Fault Localization at the Accuracy Ceiling
链接: https://arxiv.org/abs/2610.36015
作者: Christoph Daxerer,Johann Glock,Christian Macho,Rudolf Ramler,Martin Pinzger,Josef Pichler
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注: 12 pages, 3 figures, 5 tables
Abstract:Background: To localize a fault, a coding agent can reason from source code alone or use debugger assistance by running the program and stepping through it in an interactive debugger. While earlier agents struggled with static execution tracing and debugger operation, frontier LLMs have improved at both. Objective: We evaluate how accurately coding agents localize faults without tools across original and transformed programs, how much debugger assistance improves accuracy and at what cost, and how agents use it when optional. Methods: We seed 610 single-line faults into 144 LeetCodeDataset solutions and rewrite each faulty program with five recipes of behavior-preserving transformations. Agents based on Sol 6.1, Luna 6, and Qwen 3.8 localize each fault with no, optional, or encouraged debugger assistance. Results: Without tools, the agents localized 98.2 to 99.9% of faults in original and 96.7 to 99.9% in transformed programs, losing accuracy mainly where recipes restructured loops and conditions. Encouraged assistance removed 63% and 61% of Luna 6’s and Qwen 3.8’s remaining transformed errors (+2.1 and +1.3 percentage points) at 14.7 to 27.0 times the token use, while Sol 6.1 stayed at 99.9%. With optional assistance, Luna 6 and Qwen 3.8 still gained +1.5 and +1.3 points, though the agents used the interactive debugger in only 0.2 to 6.2% of runs. Conclusions: On short, precisely specified programs, current coding agents localize nearly all faults from source code alone, and debugger assistance removes most errors that remain. Given the choice, agents executed the program far more often than they stepped through it, yet kept most of the benefit.
[AI-139] CIVER: Causal Intervention Verification for Reliable Code Review WWW2026
链接: https://arxiv.org/abs/2610.36010
作者: Xinyuan Song,Hasibul Haque,Jay Somasundaram,Ryan Turner,Bowen Zhu,Liang Zhao
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注: WWW 2026
Abstract:Code review is essential for maintaining software repositories by identifying defects, improving code quality, and supporting reliable software evolution. With the growing adoption of LLM-based coding agents, an increasing number of code contributions and repository changes are generated by these agents. Consequently, LLM-based code review agents are becoming increasingly prevalent to evaluate these contributions and maintain software quality. However, LLM reviewers often generate lengthy reviews containing numerous potential issues, many of which are false positives or lack sufficient evidence. These unreliable reviews can mislead developers and downstream coding agents, motivating the need to distinguish actionable findings from unsupported claims. We propose CIVER, a causal intervention-based framework that identifies the core issues underlying code-review findings and distinguishes genuine defects from false positives. CIVER integrates three modules: a review module assesses findings, an intervention module generates repair-oriented perturbations, and a verifier module scores review reliability. Through joint optimization via expert iteration, the three modules identify genuine defects by responding to targeted repairs while remaining stable under unrelated changes. Extensive experiments on SWE-Review-Bench demonstrate that CIVER correctly identifies 77% of genuine defects through targeted repairs. Cross-project and downstream repair evaluations further demonstrate that CIVER generalizes across repositories and improves coding agents’ task success rates, highlighting its effectiveness for reliable automated code review.
[AI-140] DuplexTune: Coupled-State Replay for Memory-Efficient Mixture-of-Experts Adaptation
链接: https://arxiv.org/abs/2610.36009
作者: Paul Raj,Simon Caldwell,Daniel Whitmore,Mutlu Mete
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Sparse expert execution does not remove the stored states needed to adapt a mixture-of-experts language model. We present DuplexTune, a framework for replaying coupled hidden states through pretrained attention and expert modules. Trainable width bridges connect two residual streams to the backbone. Bridge warm-up precedes backbone adaptation, with router weights kept frozen. We analyze the implicit state-recovery equation and identify the convergence and replay conditions needed for reconstruction. On Qwen1.5-MoE-A2.7B, the single-H800 evaluation reports 38.8 GB peak memory, versus 64.8 GB for supervised fine-tuning with checkpointing, and throughput of 25.1 versus 19.9 samples/s. DuplexTune achieves 67.2% MMLU, 75.6% GSM8K, and 7.69 MT-Bench, with a weaker multilingual score than the checkpointed baseline. The results characterize a configuration-dependent tradeoff among state storage, throughput, and adaptation quality.
[AI-141] Pricing Cross-Bank Context in Federated Anti-Money Laundering
链接: https://arxiv.org/abs/2610.35996
作者: Lidia Losavio,Dario Fenoglio,Martin Gjoreski,Marc Langheinrich
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注: Accepted at the ICAIF 2026 Workshop on Financial AI Security, Privacy and Safety, Milan, Italy (non-archival). 10 pages, 8 figures, 1 table
Abstract:Most graph-based Anti-Money Laundering (AML) methods assume access to a pooled transaction graph containing all accounts and payments, which in reality does not exist. Federated learning offers the necessary privacy guarantees but it remains unclear how money laundering patterns spanning separate graphs can effectively be detected. We exploit transactions already recorded by both endpoint banks to approximate remote account states locally, enabling multi-layer Graph Neural Network (GNN) computation without additional account-specific exchange. Starting from this baseline, we test whether the remaining context is worth sharing, using a disclosure ladder that adds two transaction counts per remote account or the counterparty’s own GNN states. On AMLworld HI-Small, across six GNN backbones and up to 32 silos, the observable baseline yields the largest gain and keeps it as fragmentation shrinks the visible slice; counts add a robust correction, while exact remote states help only in specific regimes. In controlled privacy–utility studies, the noise needed to protect shared states erases their detection value before it buys meaningful privacy, whereas counts stay useful even under strong privacy protection. These findings support a practical rule for banks: use what they already record, and formally protect the little that must move.
[AI-142] Layerwise Repacking: Variable-Length Hybrid-Attention Training without an Extra Redistribution Pass
链接: https://arxiv.org/abs/2610.35989
作者: Jiahao Zhang,Junwei Feng,Dongliang Huang,Xiaoling Zhou,Yiqiao Liao,Chengru Song
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Variable-length training creates conflicting parallelism requirements across the layers of a hybrid-attention model. The longest samples require wide sequence-parallel (SP) groups of 16 or 32 ranks for full attention, whose cost grows quadratically with sequence length. Sparse-attention layers, dominated by near-linear MoE work, can instead balance most evaluated workloads within node-local SP8 groups. Reusing the wide groups for these layers pushes their all-to-all communication across nodes, where the observed effective rate is 7.0x lower. Yet switching to smaller groups changes which ranks own each sample; a separate redistribution at every full-sparse boundary would introduce another data transfer. Layerwise Repacking keeps each SP16/32 parent for full attention and repacks its samples into node-local sparse children at SP1, 2, 4, 8, where hidden states persist between full layers. The layout change rides on the two existing full-layer Ulysses sites: at each, one variable-size all-to-all performs both the head exchange and the sample repacking, so no redistribution pass is added. Because the balance the children can reach depends on which samples share a parent, an auxiliary parent solver rebalances parent membership before the children are formed. On all 22 steps with SP16/32 parents among the first 200 training steps of a 1F3S hybrid-attention model (one full-attention layer per three sparse-attention layers), repacking alone reduces the mean end-to-end time by 5.95% and raises model FLOPs utilization (MFU) by 3.75 percentage points over the baseline with repacking disabled; with the improved solver, end-to-end time falls by 7.61% and MFU rises by 4.83 percentage points. Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI) Cite as: arXiv:2610.35989 [cs.LG] (or arXiv:2610.35989v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2610.35989 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-143] PROBE: Frontier Coding Agents Find Different Bugs Than Maintainers Fix
链接: https://arxiv.org/abs/2610.35987
作者: Marco Paina,Mark Niklas Müller,Niels Mündler-Sasahara,Martin Vechev
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Proactive bug discovery is an important software-engineering problem. Even though coding agents are already widely deployed in this domain, their capabilities remain understudied. In particular, the challenging task of finding all bugs that human maintainers would repair across an entire repository has not been studied yet. To close this gap, we introduce PROBE, a benchmark consisting of all 2,457 bugs fixed over the course of 18 months across 20 diverse GitHub repositories, together with a method for matching candidate findings to this partial reference set. We evaluate coding agents across frontier models and find they all achieve a recall of below 21% and a precision of less than 9% with respect to the reference set. While stronger models increase recall substantially from 8% to 17% , they do not improve precision. Surprisingly, despite the low precision, LLM and human evaluators still judge 85% of findings valid. Crucially, we find that found bug distributions are similar across agents but differ substantially from those of bugs fixed by maintainers. The practical relevance of bug discovery and the low recall of frontier models make this benchmark a valuable resource for both model users and developers.
[AI-144] Let the Gate Decide: Per-Lookup Precision for Quantized Conditional Memory in LLM s
链接: https://arxiv.org/abs/2610.35985
作者: Tianqi Chen,Xunchao Li,Linchuan Xie,Guanghua Yu,Jianchen Zhu,Peisong Wang,Jian Cheng
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Conditional memory layers, such as DeepSeek’s Engram, attach large hashed n-gram tables to a language model, and removing them from DeepSeek-V4.1-Flash costs 36.8 PopQA points. Low-bit quantization loses long-tail facts first, and the standard remedy, giving more bits to frequently accessed rows, backfires: refining lookups by access frequency is worse than refining them at random. We trace this to the gate that scales each retrieved row. A lookup’s quantization error is weighted by its gate, which is small for frequent n-grams, large for the rare n-grams of long-tail entities, and computable from a 1-bit code. GateRQ therefore stores each row as a ladder of 1-bit residual codes, keeps level 1 in GPU memory to compute the gate, and reads deeper levels only for the lookups the gate ranks highest. Without retraining, GateRQ cuts GPU table memory by 7.8x on DeepSeek-V4.1-Flash, which raises vLLM throughput by 11%-31% over FP8 tables, while losing about 1 PopQA point at 1.50 bits read per lookup and 0.2 points at 2.10 bits across two models.
[AI-145] Enhancing Deep Learning Model Performance through Data Generation Strategies: Classification of Olive Tree Diseases
链接: https://arxiv.org/abs/2610.35984
作者: Hicham El Akhal,Abderrahim Waga,Aissa Ben Yahya,Abdelbaki El Belrhiti El Alaoui
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Deep learning models are highly effective in a wide range of domains, particu larly in plant disease diagnosis. However, their performance is often limited by challenges such as insufficient data and class imbalance. In our work, we propose a comprehensive data generation strategy to overcome these limitations. First, we apply various data generation techniques, including traditional augmentation methods, CutMix, MixUp, and synthetic image generation using Deep Convolu tional Generative Adversarial Networks (DCGANs), to create new data samples. Second, we exploit this generated data in three scenarios: (S1) balancing class distribution in the original dataset, (S2) increasing the number of samples per class, and (S3) expanding the dataset while maintaining balanced classes. Third, we evaluate the impact of these data distributions (S1, S2, S3) on model perfor mance using the EfficientNetV2B0 architecture with pre-trained and fine-tuned weights. The experimental results show a substantial improvement in classifica tion accuracy, emphasizing the critical role of data diversity and class balance. These findings highlight the effectiveness and adaptability of the proposed strat egy of data generation and distribution for enhancing dataset quality, offering promising applications in plant disease detection and precision agriculture.
[AI-146] Generation and Verification: Human-Agent Collaboration for Safety-Critical Automotive Software Engineering EMNLP2026
链接: https://arxiv.org/abs/2610.35982
作者: Elizaveta Zinovyeva,Aiham Taleb,Nikita Kozodoi,Daniel Schleicher,Amir Mahdi Namazi,Martin Kraus,Levent Kent,Brian Jensen
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注: Accepted to EMNLP 2026 REALM Workshop
Abstract:Research on coding agents mainly focuses on generation quality. Yet, safety-critical automotive software makes two additional demands. It prioritizes verification, since every change is reviewed, traced to a requirement, and checked against industry standards such as MISRA and AUTOSAR before it can be committed. It is also a narrow domain, written against layered architectures, embedded resource constraints, and organization-specific interface conventions that are scarce in the data general-purpose LLMs are trained on. We present a coding assistant for automotive C/C++ that addresses both demands. First, it exposes three tools to the user over the Model Context Protocol (generate, edit, and verify), allowing developers to run verify on any code on demand rather than relying on the agent to check its own output. Second, rather than using a general-purpose model, we customize the underlying LLM for the automotive domain through prompt engineering and fine-tuning. We construct a dataset from automotive production code of AUMOVIO, where each function is paired with the requirement texts and Unified Modeling Language (UML) behavior it must satisfy. In a two-week pilot on active production projects, developers report verify as the most-used tool and confirm using it on their own code and not only on generated output. Comparing model customization strategies, 75% prefer the faster prompt-engineered configuration. MISRA- and AUTOSAR-inspired pattern checkers indicate that prompt engineering achieves a lower number of violations on C, while fine-tuning shows better conformance with architecture patterns for C++.
[AI-147] UnityBench-Mini3D: Evaluating LLM Agents for 3D Game Development in Unity
链接: https://arxiv.org/abs/2610.35980
作者: Yijun Zhou,Minghao Shao,Ke Zhang,Renkai Yuan,Hongbo Yuan,Bin Hu,Bozhao Qian,Hongxia Shen,Yanhong Liu,Yuting Chen,Yuntao Shi,Zhongxuan Xu,Shuo Mao,Haijia Zhou,Zhendong Li,Tao Zhang,Chenggang Li,Yuxuan Wang,Zinan Zheng,Calvin Zhang,Jia Li,Heng Chang,Jie Feng
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Game development stresses coding agents in ways general software engineering does not: correctness depends on how engine APIs, serialized scene state, and runtime behavior interact, not on source code alone. Existing benchmarks mainly target platforms such as Godot and heavily rely on LLM judges or coarse task-level outcomes. We present UnityBench-Mini3D, a benchmark of 77 project-level Unity tasks organized around seven expert-defined skill categories rather than tutorials or genres, with three difficulty tiers. Each task pairs a complete Unity project with natural-language requirements and is evaluated using 1,178 held-out, human-authored milestone tests covering scene structure, component configuration, and runtime behavior. This in-engine evaluation provides deterministic partial credit, localizes failures, and distinguishes partial implementations from complete solutions without relying on subjective judges. Across 21 models and 9 coding harnesses, the average pass rate reaches 80.9% on easy tasks but falls to 35.2% on hard tasks, with just 57.7% overall. UnityBench-Mini3D provides a reproducible, diagnostic testbed distinguishing systems across difficulty levels.
[AI-148] Multiple Leak Localization in Water Distribution Networks through Graph Neural Networks
链接: https://arxiv.org/abs/2610.35972
作者: Arvin Ajoodani,Sara Nazif,Pouria Ramazi
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Leak localization in water distribution networks is challenging when multiple leaks occur simultaneously, particularly under sparse monitoring. Existing data-driven methods often assume a single or predefined number of leaks or require extensive enumeration of leak combinations. This study proposes a graph neural network (GNN) model that localizes a variable number of simultaneous leaks without specifying the leak count in advance or exhaustively generating every possible combination of leak locations for training. The model is evaluated on two benchmark networks of different sizes, the 31-node Hanoi network and the larger 92-node NET3 network, under variable demand, sparse pressure-sensor coverage, variable leak magnitudes, and varying numbers of simultaneous leaks. The GNN outperforms multilayer perceptron, support vector machine, K-nearest neighbors, and random forest classifiers by up to 12% and 20% in accuracy respectively in Hanoi and NET3 networks. Consistent performance across both networks supports the scalability of the proposed approach.
[AI-149] Structured Time-Frequency Representation for Channel Prediction
链接: https://arxiv.org/abs/2610.35956
作者: Ning Yang,Yuming Xian,Tong Li,Ning Gao,Haijun Zhang
类目: Information Theory (cs.IT); Artificial Intelligence (cs.AI)
备注:
Abstract:Accurate CSI forecasting is essential for mitigating channel aging and supporting reliable downlink transmission in high-mobility wireless systems. In FDD systems, forecasting future DL CSI from historical UL observations requires joint temporal extrapolation and cross-band inference. However, defining historical UL as the input and future DL as the target does not specify how elapsed time and UL-to-DL band separation should be represented in a pretrained predictor’s positional space. FAR-CSI establishes an FDD-aware representation of channel forecasting within a joint temporal and operating-band positional field. Within this field, historical UL observations and future DL targets occupy segregated regions according to their temporal and operating-band roles, preserving the distinction between observed UL context and unobserved DL prediction targets. This representation is instantiated through a unified structured routing rule that governs encoder positional assignment, decoder-field reconstruction, and future-DL target selection, ensuring the same FDD geometry is preserved from contextual encoding to prediction readout. Under matched backbone, initialization, and training conditions, FDD-aware routing reduces NMSE by 21.44% relative to generic routing. Across the evaluated FDD baselines, FAR-CSI achieves higher CSI prediction accuracy and MRT-based spectral efficiency, while retaining its predictive advantage across mobility conditions, prediction horizons, reduced training-data regimes, and noisy historical observations.
[AI-150] ResCom: Activation Residual Compensation for Large Language Model Quantization
链接: https://arxiv.org/abs/2610.35955
作者: Ning Yang,Bohan Cui,Zhuowen Liu,Yuting Wang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 25 pages (10 pages main text + appendix), 9 figures, 20 tables
Abstract:Low-bit weight–activation quantization reduces the memory and computation needed to deploy large language models, but activation rounding can still degrade accuracy after calibration. Reconstruction from rounded inputs cannot distinguish activations within the same quantization bin. We propose ResCom, a low-rank correction that uses the activation rounding residual to compensate for the resulting layer-output error. An output-aware calibration procedure selects residual directions by their contribution to output reconstruction and computes the correction in closed form, without gradient training or updates to the calibrated weights. On Llama-3-8B with 4-bit weights and activations, rank-32 attention corrections reduce WikiText-2 perplexity from 7.708 to 7.147, improve MMLU by 2.89 percentage points, and raise five-shot GSM8K accuracy from 23.96 to 36.16 percent. Evaluations on models from 1B to 70B parameters and on held-out C4 test the method beyond this setting. The evaluated FP32 implementation adds 2.25 percent to 4K prefill time; a separate compact-storage implementation retains a 27.7 percent peak-memory reduction relative to FP16.
[AI-151] Curriculum-Driven Deep Reinforcement Learning for Autonomous Far-Range Rocket Landing
链接: https://arxiv.org/abs/2610.35954
作者: Atin Chowdhury,Piyush Joshi
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: This is the extension of our previous paper that got accepted in ICAART 2026
Abstract:Autonomous vertical landing of reusable launch vehicles is a critical yet formidable challenge in aerospace engineering, primarily due to highly nonlinear dynamics, stringent precision require- ments, and inherent environmental uncertainties. While reinforcement learning (RL) has emerged as a promising alternative to traditional model-based control, existing approaches frequently rely on flat training paradigms and highly simplified initial conditions, drastically limiting their real-world applicability. This paper proposes a comprehensive curriculum-driven deep reinforcement learning framework designed to master both far-range navigation and precision landing. By employing a three-stage curriculum augmented with a critic warm-up procedure and a rehearsal mechanism to prevent catastrophic forgetting, alongside a high-dimensional observation vector utilizing a full 3 X 3 rotation-matrix, the agent is trained to navigate from an initial horizontal distance of 1,000 to 1,500 units vastly farther than prior benchmarks before executing a soft touchdown. Crucially, we intro- duce a redesigned reward function that reduces unnecessary computational overhead and addresses limitations of earlier formulations, including poorly calibrated reward components and potentially conflicting optimization objectives. The redesigned function bounds the per-step reward magnitude by (O(1)), while its progress term accumulates the total distance closed over the trajectory. The available 100,000-episode results report a 99.98% success rate for the proposed policy and 68.31% for the baseline under the evaluation conditions. In the evaluation, the proposed policy retained 69.1% of its initial fuel capacity and achieved a mean final distance from the pad centre of 7.07 units, compared with 13.10 units for the baseline. Landing visualizations are available in Video 1 and Video 2.
[AI-152] Counting the Bandages: On measuring the time that AI saves
链接: https://arxiv.org/abs/2610.35942
作者: Solomon Shalom Lijo
类目: Artificial Intelligence (cs.AI)
备注: 18 pages, 2 figures. Replaces the author’s earlier TSB proposal (SSRN 6682338). Code and data: this https URL . Also archived on Zenodo: https://doi.org/10.5281/zenodo.23557773
Abstract:I consider what it means for AI assistance to save time. I contrast a stopwatch view, on which time saved is whatever the clock records, with a ledger view, on which it is a counterfactual that has to be read next to the errors the work lets through, and I argue for the latter. The argument has four parts. I define a model-free pair of quantities, time saved and excess escaped defects, and one decision number built from them: the break-even defect cost. I give a baseline model of checking in which, until the model is good enough not to check, a calibrated worker’s escaped-defect rate does not depend on how good the model is; checking time grows with the logarithm of the model’s error rate; and over-trust buys minutes on the clock at a larger cost in total minutes. I show what randomized studies can and cannot identify, and I replicate METR’s developer study from its public data: the same 246 issues support a slowdown of 19% or 24% depending on the estimand (34% without adjusting for a chance imbalance), and AI helped on somewhere between 3% and 82% of them. Finally, I restate the published record, with stated bounds in place of imported point assumptions. In GDPval’s own review-every-output workflow, review time alone caps GPT-5’s speedup at 3.6x; the consultants outside the frontier were faster, and more often wrong. Last, I test two of the model’s predictions in a small pre-specified experiment on AI reviewers checking AI drafts of the same kind of brief, built from SEC filings made after the models’ training data were collected. Told the drafter’s true error rate, the reviewer checked more as it rose, but all-or-nothing, switching near where its prompt’s costs break even. Told that the drafter erred ten times less often than it did, it checked 7 fewer of 9 figures and let a wrong figure through in 22.5% of briefs instead of none: faster and worse, entirely through reduced effort.
[AI-153] HGPRGDel: GPR-Aware Learning and Ranking for Growth-Coupled Gene Deletion Design
链接: https://arxiv.org/abs/2610.37478
作者: Ziwei Yang,Takeyuki Tamura
类目: Quantitative Methods (q-bio.QM); Artificial Intelligence (cs.AI)
备注:
Abstract:Designing microbial strains for growth-coupled production, in which cellular growth is linked to the synthesis of a target compound through genetic interventions, is a fundamental goal in metabolic engineering. Constraint-based methods can identify growth-coupled deletion strategies, but design for each target typically requires a separate search across a large number of possible gene-deletion combinations. This motivates a reusable approach that learns from previously evaluated strain designs to prioritize candidate interventions before downstream constraint-based evaluation. We present a reproducible protocol for growth-coupled gene deletion design that learns and ranks gene-deletion priorities for each target metabolite. Using computationally curated deletion strategies as learning resources, the protocol generates continuous deletion-priority rankings, constructs candidate strategies at adjustable search-space budgets, and evaluates these candidates through constraint-based metabolic analysis. Its learning component, HGPRGDel, explicitly represents gene-protein-reaction (GPR) logic, capturing how multiple genes can jointly support a reaction or alternative genes can maintain its activity, while integrating molecular and reaction-level information. Across e_coli_core, iMM904, and iML1515, the protocol achieves the highest macro-F1 and AUC among the evaluated learning configurations. HGPRGDel’s rankings recovered an average of 90% of the reference deletion genes from MetNetComp by examining fewer genes than the evaluated baseline methods. By separating learned intervention prioritization from downstream constraint-based assessment, the protocol provides a reproducible framework for reusing strain-design knowledge while controlling the size of the candidate space and retaining mechanistic evaluation of growth-coupled production.
[AI-154] JEPA-Inspired Latent Field Prediction for Sparse-Pilot MIMO-OFDM Channel Reconstruction
链接: https://arxiv.org/abs/2610.37391
作者: Yuzhi Yang,Brahim Mefgouda,Hang Zou,Lina Bariah,Anis Bara,Yuhuan Lu,Hao Zhang,Merouane Debbah
类目: ignal Processing (eess.SP); Artificial Intelligence (cs.AI)
备注:
Abstract:Accurate acquisition of full channel state information (CSI) is essential for multiple-input multiple-output orthogonal frequency-division multiplexing (MIMO-OFDM) wireless systems, while dense pilot-based estimation incurs substantial overhead. Directly minimizing the prediction error over raw CSI coefficients is fragile in realistic propagation environments, since small-scale fading, mobility, local scattering, and small spatial perturbations can induce large phase-sensitive variations. Nevertheless, wireless channels contain strong spatial, temporal, and frequency-domain correlations that can be more predictably represented in a latent propagation-field domain. Motivated by this observation, we propose a joint-embedding predictive architecture (JEPA)-inspired field world model, defined as a latent predictive model of the wireless propagation field, for sparse-pilot channel reconstruction. The model maps CSI observations into latent time-frequency tokens, predicts masked or future latent field states from historical multi-band context, and fuses the predicted latent prior with sparse noisy current pilots to reconstruct the full MIMO-OFDM channel. Ray-tracing experiments show that the proposed method improves normalized mean square error (NMSE) over pilot-only and direct reconstruction baselines, supporting latent field prediction as an effective prior for sparse-pilot channel acquisition.
[AI-155] What the Primary Lund Plane Omits: A Model-Stitching Diagnosis of a Jet-Tagger Representation Gap
链接: https://arxiv.org/abs/2610.37137
作者: Saurabh Rai,Sanmay Ganguly
类目: High Energy Physics - Phenomenology (hep-ph); Artificial Intelligence (cs.AI); High Energy Physics - Experiment (hep-ex)
备注: 21 Pages, 16 Figures and 13 Tables
Abstract:Lund-plane and Particle Transformer (ParT) jet taggers can differ in performance because one representation is weaker, or because a weaker downstream computation reads an adequate one. Standalone performance cannot separate these. Model stitching can: it links the frozen early layers of one network to the frozen late layers of another through a single trained linear map. As a positive control we use an incomplete Lund representation, the primary declustering branch, whose deficit is known independently. Stitched into ParT’s downstream, it reaches a background rejection at least as much as its own downstream recovers, yet far below ParT’s background rejection: the deficit is representational, not computational. A hypergraph reader over splitting triplets does no better, ruling out higher-order correlations as the cause. The gap is invisible in AUC, which stays within 1% of its target value, and appears only in background rejection at low signal efficiency. It localizes to 0.6% of background jets, at and above the top mass, with more constituents but fewer primary splittings than typical background. Rebuilding the representation on the full Cambridge/Aachen declustering tree repairs the deficit as predicted: standalone rejection rises, closing 72% of the gap. Masking nodes at inference shows the repaired network still relies almost entirely on the primary branch; the secondary splittings make the representation learnable rather than supplying information it reads. Along the primary branch, the widest-angle splitting, near the top and W opening angles, is worth nearly twice a random one. For quark/gluon tagging the primary branch is far less deficient, retaining 77.6% of ParT’s rejection (46.9% for top), and the full tree closes only 33% of the gap (72% for top).
[AI-156] Atom as an Agent : Emergent Order by Collective Predictive Coding
链接: https://arxiv.org/abs/2610.36673
作者: Koji Hashimoto,Tadahiro Taniguchi
类目: Disordered Systems and Neural Networks (cond-mat.dis-nn); Statistical Mechanics (cond-mat.stat-mech); Artificial Intelligence (cs.AI); High Energy Physics - Theory (hep-th)
备注: 33 pages, 10 figures
Abstract:What is the intrinsic reason of emergence of macroscopic order parameter out of microscopic systems? In this article, we apply to physics the collective predictive coding (CPC), the theory of language/symbol emergence. As our language emerges out of our human communications, physical macroscopic order parameter emerges out of communications among atoms — each atom is regarded as an agent who predicts each other’s behavior through inference. We present three examples of classical spin systems: two-dimensional ferromagnetic and antiferromagnetic Ising models, a two-dimensional spin model with competing stripe orders, and the fully connected Sherrington–Kirkpatrick spin-glass model. In all of these examples, proper macroscopic order parameters including the replica overlap emerge automatically through communication informatics of CPC, only from observed spin configurations. We call the learned macroscopic coordinate the collective predictive order. Furthermore, for the ferromagnetic and antiferromagnetic Ising models, the equilibrium distribution of the collective predictive order yields an emergent potential with the symmetry and thermal evolution characteristic of a Landau potential. Notably, the emergence of the collective predictive order does not refer to any structure of microscopic Hamiltonians/Lagrangians. Thus our ``Atom-as-an-Agent" (AaaA) principle offers a completely data-driven method for the emergence of macroscopic order, that is, machine learning physics.
[AI-157] Expert lithological labels do not improve learning when used to predict marine sediment properties
链接: https://arxiv.org/abs/2610.36410
作者: John M. Aiken,William Gilpin,Thorsten W. Becker
类目: Geophysics (physics.geo-ph); Artificial Intelligence (cs.AI)
备注:
Abstract:Scientific drilling has produced large archives of physical, chemical and magnetic measurements linked to descriptions of the rock recovered. These archives are mostly incomplete: instruments sample at different intervals, recovery is partial, and destructive analyses preclude later measurement of the same material. We train an self-supervised embedding model on the LILY ocean drilling database using six routinely logged properties (bulk density, magnetic susceptibility, natural gamma radiation, and red, green and blue reflectance) across 236,963 observations from 296 boreholes, with a masked variational autoencoder that has a heteroscedastic Gaussian decoder and learnable mask tokens. The embeddings separate geologist-assigned lithologies as well as a variant trained with those labels, and increasing the weight on the labels degrades reconstruction while leaving classification unchanged. The labels are recoverable from the measurements geologists use to assign them, so training on them adds nothing. Lithological categories compress continuous measurements into classes, and compressing them discards what the measurements record about the processes that produced the rock. The embeddings also predict physical properties that were absent from training, on boreholes the model has never seen. While depth should account for this since sediments compact with burial, it does not. On held-out boreholes depth alone reaches a median R^2 of -0.02, against 0.20 for the embedding.
[AI-158] Enhancing Clinical Classifier Robustness with Deep Representation Learning Based OOD Detection on DNA Methylation Data
链接: https://arxiv.org/abs/2610.36344
作者: Md Zahangir Alom,Quynh T. Tran,Breuer Alexandar,Brent A. Orr
类目: Genomics (q-bio.GN); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 11 pages, 8 figures
Abstract:DNA methylation (DNAM) array-based tumor classification is a powerful clinical diagnostic tool, but models trained on fixed reference cohorts can produce faulty predictions for tumor types not represented in the training distribution. We propose an end-to-end deep representation learning-based out-of-distribution (OOD) detection framework for DNAm array data. Deep neural network (DNN) classifiers were trained on public sarcoma (SARC) (GSE140686) and brain tumor (BT) (GSE90496) cohorts, and feature vectors were extracted from bottleneck layers. Five OOD methods, including 1dimension (1D)-Subspace Analysis (1D-SSA), K-Nearest Neighbors (KNN), Ridge Regression, Mahalanobis Distance (MD), and Sum of Squared Distances (SSD), were explored with 200 and 500-dimensional embeddings. Spectral discrepancy was used to discriminate in-distribution (IND) from OOD samples. Performance was measured via cross-validation, holdout samples, and an independent clinical dataset. 1D-SSA was the most effective approach. In the sarcoma-IND/brain tumor-OOD setting, the framework achieved an AUROC of 0.975 and True Negative Rate (TNR)@95TPR of 0.930, and an AUROC of 0.9569 and TNR@95TPR of 0.8660 for the brain tumor-IND/sarcoma-OOD setting. On the independent clinical test set, the method yielded a mean AUROC of 0.976, a DE of 0.032, an FPR@95TPR of 0.065, and a TNR@95TPR of 0.935. These results show deep representation-based OOD detection as a practical quality-control layer for clinical-grade methylation-based tumor classification systems.
[AI-159] Directional Statistics on Circles and Spheres: A New Angle on Data
链接: https://arxiv.org/abs/2610.36227
作者: Siddhesh Kulkarni
类目: Methodology (stat.ME); Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
备注: 11 pages, 5 figures. Expository article with real-data examples and reproducible R code. Submitted to Resonance: Journal of Science Education
Abstract:Two compass bearings near north can have an ordinary arithmetic mean pointing south: the difficulty is that angles live on a circle, not a straight line. Using real wind observations, diffusion-MRI orientations, and geometric examples from machine learning, this article introduces circular and axial averages, directional variability, and probability models on circles and spheres. The guiding principle is to identify the geometry and symmetries of the observations before choosing a statistical method.
[AI-160] Structure Not Sequence: Tracking and Validating Folding Mechanisms in ESMFolds Trunk with Sparse Autoencoders
链接: https://arxiv.org/abs/2610.36226
作者: Bohan Liu,Xiaohan Fan,Wenqian Ye,Aidong Zhang
类目: Quantitative Methods (q-bio.QM); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Structure predictors like ESMFold fold proteins using neural network computation and have transformed structural biology. However, the driving mechanism from amino-acid (AA) sequence to 3D structure is insufficiently understood, and biological meaning remains locked in the dense network. Sparse autoencoders (SAEs) have shown promising results in obtaining interpretable concept features over an overcomplete basis of intermediate protein sequence-model representations. Yet folding models, which predict physical shape from AA sequence semantics, have no such account. We hypothesize that the features carrying the folding computation evolve in the 3D structural regression process. In this paper, we interpret the folding trunk in three steps. First, a pilot analysis on the representation distribution across modules and recycles locates distinct spatial computation in the trunk. Second, the dictionaries expose a division of labor: the single representation carries inherited semantics, and the pair representation computes the pairwise geometry. Third, we interpret the dictionary at scale with a reasoning language model, trace how the concepts compose across blocks by gradient-based attribution, and follow the interpreted code to the model’s folding output, verifying where the computation aligns with, or departs from, known biochemistry. Together, the sparse dictionaries we train capture the relational geometry the trunk computes beyond the sequence model, and their concept activations forecast the model’s own fold quality, turning the black-box folding trunk into a sparse, interpretable, and scientifically verifiable circuit.
[AI-161] Physical structure as an organizing principle for universal Hamiltonian learning
链接: https://arxiv.org/abs/2610.36135
作者: Mingkang Liu,Zhanghao Zhouyin,Guanghui Wang,Daye Zheng,Lei Shen,Qiangqiang Gu
类目: Materials Science (cond-mat.mtrl-sci); Artificial Intelligence (cs.AI)
备注: 26 pages, 8 figures, 1 table; Supplementary Information (51 pages) included
Abstract:Density functional theory (DFT) gives the electronic structure of materials, but its cost limits large-supercell calculations and materials screening. Machine-learned Hamiltonians bypass the self-consistent cycle but are usually system-specific; universal ones could cover new compositions and disordered solids. Such models must generalize across chemistry and keep downstream properties physical, while staying inexpensive for large cells. Here we introduce UniTB, a universal E(3) equivariant Hamiltonian representation framework that integrates physical priors rooted in density-functional tight-binding theory with deep learning. UniTB’s network, which corrects a prior Hamiltonian, generalizes across chemistry through a mixture of experts routed by that prior and a charge-equilibration head as in self-consistent-charge tight binding. A projection bounds band shifts from the prior, suppressing spurious states and keeping downstream properties physical. We provide the fastest SO(2) tensor-product GPU kernels for the network. The variant UniTB-Flash assembles its prior from two-center tables at O(N) cost for N atoms, against O(N\log N) on a grid. Trained on 29,242 crystals, UniTB reaches a 0.48 meV matrix-element error on nonmagnetic test crystals spanning 68 elements. Without amorphous training data, both models agree with DFT optics and conductivity of amorphous Ge-Sb-Te. UniTB-Flash ranks candidates with no significant difference from UniTB, whose property values are more accurate; together they screen 16,195 crystals for five application domains of industrial relevance.
[AI-162] Beyond Greenwashing: Algorithmic Resilience and the Non-Linear Governance Premium in Sovereign Capital Flows
链接: https://arxiv.org/abs/2610.35895
作者: Sourav Bose,Taoufik Bouraoui
类目: Applications (stat.AP); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Econometrics (econ.EM)
备注:
Abstract:Despite the proliferation of sovereign Environmental, Social, and Governance (ESG) metrics, traditional linear macro-financial models systematically misprice Foreign Direct Investment (FDI) tail risks during systemic shocks. This paper demonstrates that sovereign risk transmission is fundamentally non-linear. Using a forward-chained global panel of 147 sovereign entities (2010-2024), we deploy a gradient boosting architecture that compresses forecast errors by over 39% against L1-regularized baselines precisely during extreme macroeconomic tail events. Moving beyond opaque predictive accuracy, we apply dual interpretability frameworks to mathematically map the discrete threshold effects of institutional quality. We identify a strict, non-linear “Governance Premium”: incremental environmental and social improvements yield zero capital advantage until a foundational regulatory threshold is breached. Finally, we establish regime-conditional 95% uncertainty bounds that maintain valid marginal coverage even under severe ESG reporting failures. This framework equips policymakers with a distribution-free tool to quantify sovereign tail-risk, proving that global capital markets price institutional resilience as a discrete boundary rather than a continuous premium.
[AI-163] scPyviewer: a Python-native interactive viewer from AnnData single-cell data
链接: https://arxiv.org/abs/2610.35890
作者: Hao Xuan,Yu Huang,Jiang Bian,Xiangtao Liu
类目: Genomics (q-bio.GN); Artificial Intelligence (cs.AI)
备注:
Abstract:Motivation: Interactive web tools enabling non-programmers to explore single-cell datasets, embeddings, gene expression, metadata, and marker genes are standard laboratory infrastructure. However, actively maintained tools (e.g., ShinyCell, ScRDAVis, sCIRCLE) rely on R Shiny and require Seurat objects. Python/scanpy, the dominant framework for single-cell RNA-seq and spatial analysis, lacks a lightweight, native option pairing a shareable viewer with a scriptable Python API. Sharing scanpy results currently requires converting to Seurat or sharing code notebooks. Results: We present scPyviewer, a web viewer that ingests AnnData objects directly and reproduces incumbent interaction patterns within the Python stack. In a feature-parity audit against three R Shiny tools, scPyviewer matches or exceeds baseline capabilities (7/7), uniquely offering native AnnData ingestion without conversion and cross-dataset comparison over shared genes and matched cell types. Benchmark tests across three datasets (22,315 to ~313,000 cells) show scPyviewer renders every core view faster (up to 3.6x speedup) using lower memory (5.2x lower on the smallest dataset). On 313,000 cells, scPyviewer completes all views while Seurat exhausts an 8 GB memory limit and fails. Beyond the app, scPyviewer exposes a Python API returning Matplotlib figures and pandas tables for publication-ready output. Availability and implementation: Implemented in Python 3.11 (scanpy 1.11.5, anndata 0.12.19, streamlit 1.59.2, plotly 6.9.0) with a one-command interface to reproduce benchmarks, figures, and launch the app. Source code: this https URL Subjects: Genomics (q-bio.GN); Artificial Intelligence (cs.AI) Cite as: arXiv:2610.35890 [q-bio.GN] (or arXiv:2610.35890v1 [q-bio.GN] for this version) https://doi.org/10.48550/arXiv.2610.35890 Focus to learn more arXiv-issued DOI via DataCite
[AI-164] ASTROLABE: A 3D Particle-Level Surrogate for Cosmology Simulation with Physical Flow Matching
链接: https://arxiv.org/abs/2610.27068
作者: Ibrahim Elsharkawy,David Curtin,Yonatan Kahn,Neal Weiner
类目: Cosmology and Nongalactic Astrophysics (astro-ph.CO); Artificial Intelligence (cs.AI); High Energy Physics - Phenomenology (hep-ph)
备注: 24 pages + 7 appendix (31 total), 20 figures, 11 tables
Abstract:The microphysics of dark matter (DM) is imprinted on cosmological structure across an enormous dynamic range. Nonlinear cosmological observables are most reliably modeled with N-body and hydrodynamic simulations. However, systematic N-body scans over DM theory space are prohibitively expensive. Machine learning emulators provide a possible solution, but are currently limited by either low resolution or only the ability to emulate specific observables. We introduce Astrolabe, a flow-matching-based generative emulator designed to be free of both constraints. Astrolabe acts directly on the 6N phase-space coordinates of N individual particles. With no voxel grid in the output, there is no voxel-resolution ceiling, and in principle any observable can be measured from inference. Astrolabe is computationally tractable via a field-particle hybrid transformer in which every particle cross-attends to multiscale field tokens, so compute and memory are linear in N. Astrolabe incorporates physical flow matching by taking the prior distribution for the flow to be approximate COmoving Lagrangian Acceleration (COLA)-based coordinates rather than unphysical noise, with momentum and mass conservation enforced at all points along flow time. In the Lambda-CDM cases we test, Astrolabe produces realistic fields with accurate power spectra, halo mass functions, velocity profiles, and halo density profiles. We also observe out-of-distribution generalization, with accurate emulation of dark-acoustic oscillation initial conditions. Astrolabe is also capable of being fine-tuned to simulate very large N and we demonstrate this by emulating a TNG50-3-dark simulation at a fraction of the cost. This method represents a first step toward a general and powerful cosmology simulation surrogate, which may eventually permit the inclusion of baryonic physics and a systematic exploration of richer dark sector interactions.
机器学习
[LG-0] When Should We Intervene at Test Time?
链接: https://arxiv.org/abs/2610.37658
作者: Jiahao Zhang,Xinling Wen,Liangliang Li,Yu Chen
类目: Machine Learning (cs.LG)
*备注: 79 pages; preprint
Abstract:Test-time adaptation helps a deployed model respond to changing inputs, but the same intervention can rescue one batch and damage the next. Updating normalization statistics, for example, can improve corrupted images yet fail when a batch contains mostly one class. Average accuracy hides these failures and cannot tell a model when to use an intervention. Target labels can guide this decision, but they are scarce and some choices are already clear. We introduce PORTAL, a controller that chooses when to deploy an intervention, which inputs to label, and when to stop querying. Each available prediction rule is a candidate, and its predictions are fixed before current feedback arrives. Every acquired label therefore evaluates all candidates on the same input. PORTAL combines these shared comparisons and directs queries toward candidates that can still overtake the current leader. It stops when the remaining labels cannot change the choice or when the possible accuracy gain no longer justifies another query. Our analysis links the required feedback to accuracy gaps and shared mistakes, explaining why some choices need much less evidence. Experiments show that this control turns an intervention that loses accuracy overall into a useful choice on selected batches. It prevents severe failures caused by changes in arrival order and preserves gains across complete corruption streams. When a fixed rule already works well, PORTAL uses little feedback to retain its performance. Periodic label checks reveal when past evidence needs refreshing and help the controller recover as conditions change. The acquired labels also train stronger candidates for future batches, and selection brings nearly all of their accuracy gains into deployment. The same feedback therefore improves both the choice made now and the predictions available to later batches.
[LG-1] Pushing Dexterity and Safety for In-Hand Object Pose Reaching That Works Remarkably Well
链接: https://arxiv.org/abs/2610.37652
作者: Ian Chuang,Hiya Shah,Vamsi Chilamkur,Iman Soltani,Claire Tomlin
类目: Robotics (cs.RO); Machine Learning (cs.LG); Systems and Control (eess.SY)
*备注: 26 pages, 30 figures, 14 tables. Project page: this https URL
Abstract:Dexterous manipulation with multi-fingered hands is the next frontier of robotics, yet learned hand skills remain limited to simple grasps or basic repeated motions, such as rotating an object. As robots attempt more demanding in-hand maneuvers, safety, such as not dropping the object, becomes critical, yet it is rarely measured. We study in-hand object pose reaching, a task that demands diverse and difficult grasps: moving an object to almost any position and orientation within reach of the fingers, at any wrist orientation, through long sequences of consecutive goals. We present a recipe for training such policies for a LEAP hand in simulation without human data, built on a range of insights, from diverse resets and goals via generated grasp datasets to an adaptive friction curriculum. With careful system identification, the policies transfer zero-shot to the real hand, using either motion capture or a single wrist-mounted depth camera, and prove remarkably capable and robust. On top of the task policy, we add a learned residual safety filter with adjustable conservativeness, letting us tune the balance between performance and safety and greatly reduce object drops. In 13.3 hours of real-robot evaluation over 4,800 unseen goals across 4 objects and diverse wrist orientations, the state-based policy reaches 95.2% of goals within 2 cm and 0.4 rad, and the filter cuts its drops by 75% while still reaching 91.5% of goals. The policies can also spin a lightbulb held in an arbitrary pose, reach out-of-distribution goals, withstand fast arm motion and perturbations, and, needing only a wrist camera, serve as portable skills such as screwing in a lightbulb. We release the full recipe as a platform for studying performance, reliability, and safety in high-dimensional dexterous systems. Videos and more are available at this https URL.
[LG-2] Scaling Laws for Tool-Use Agents via Controllable Environment Synthesis
链接: https://arxiv.org/abs/2610.37648
作者: Andrew Szot,Samira Abnar,Omar Attia,Abbas Kazerouni,Michael Kirchhof,Cheng-Yu Hsieh,Raviteja Vemulapalli,Alexander Toshev
类目: Machine Learning (cs.LG)
*备注:
Abstract:We explore how to build performant large language model (LLM) tool-use agents that complete complex, multi-step tasks. Specifically, we study how data composition in synthetically generated tool-use environments affects the training of such agents. To do so, we propose the SynthAPI pipeline, which explicitly controls the diversity of generated environments along the axes of tools, tasks, and the databases on which the tools operate. By systematically varying SynthAPI’s generation controls in 458,752 tasks across 849 trained models of four different sizes, we derive a scaling law that predicts downstream tool-use agent generalization based on the scale and composition of the synthesized environments. These predictable relationships reveal several lessons for scaling tool-use agent data. As data budgets grow, scaling tool and database diversity becomes more effective than increasing the number of tasks, and increasing data complexity yields larger gains. Scaling according to these findings yields agents that outperform models trained on data from prior work in large-scale tool-use agent and synthetic environment generation across four evaluation benchmarks.
[LG-3] Spectral Braids: A Topological Descriptor for Dynamic Directed Networks
链接: https://arxiv.org/abs/2610.37580
作者: Mehmet Emin Aktas,Esra Akbas
类目: ocial and Information Networks (cs.SI); Machine Learning (cs.LG)
*备注: 15 pages, 5 figures, 2 tables. Accepted at the 2026 IEEE International Conference on Big Data (IEEE BigData 2026)
Abstract:Eigenvalues are a workhorse of graph machine learning: spectral clustering, spectral graph neural networks, and graph kernels all summarize a network by the eigenvalues of a matrix. These eigenvalues are almost always treated as a static list of numbers. When a graph changes, however, its eigenvalues move, and when the matrix is not symmetric, they move in the complex plane, where they can wind around each other. The pattern of this winding is a braid, a classical object from topology. We propose spectral braids, a methodology for dynamic directed networks: each snapshot is aggregated into a community-level transition matrix, its complex eigenvalues are tracked over time, and the braid they trace is read off as a word. The word is then mapped to a matrix by the Burau representation, which gives a fixed-size feature vector for any classifier. The braid has a concrete meaning: each eigenvalue describes how a group of communities behaves, and each crossing records one group overtaking another. We evaluate the method on dynamic network classification, using synthetic networks of online communities whose labels describe how the network evolved between the same starting and ending states. Our experimental results show that the proposed features reach 97.1% accuracy on four classes, against 44.5% for the best feature set that does not track eigenvalues, and that the Burau features also separate networks containing the same events in a different order, which simple counts of crossings cannot do. To the best of our knowledge, this is the first study that defines braids and braid monodromy on graph data and uses them as features for learning on dynamic networks.
[LG-4] Pneuma: A Large-Scale Real-World Time Series Benchmark for Commercial Aircraft Predictive Maintenance
链接: https://arxiv.org/abs/2610.37563
作者: Jiafeng Lin,Yutian Zhang,Huan Zhang,Yuxuan Wang,Jinyan Feng,Yuhui Liu,Zhongyi Pei,Jianmin Wang
类目: Machine Learning (cs.LG)
*备注:
Abstract:Predictive maintenance is vital for commercial aviation, yet progress remains hindered by the lack of public, large-scale real-world datasets. As a result, most methods are developed and evaluated on proprietary data, limiting reproducibility. We introduce Pneuma, the first publicly available large-scale benchmark for predictive maintenance using onboard time series from in-service commercial aircraft. Built around bleed air systems, Pneuma contains over 920 million multivariate time steps across four aircraft types and 9 verified fault scenarios, directly aligned with real maintenance logs. The dataset exhibits key real-world challenges, including substantial class imbalance, subtle fault evolution patterns, and significant variance across aircraft configurations. We structure the benchmark around three tasks: multivariate forecasting, supervised fault classification, and unsupervised anomaly detection, with labels derived from actual maintenance records. Under unified settings, we evaluate over 25 models from four methodological families on 9 carefully curated subsets. The results reveal that even strong contemporary models struggle with subtle fault signatures. Further analyses show consistent gains from additional in-context examples and multivariate inference, while forecasting quality is strongly associated with anomaly detection performance. Together, these findings provide practical insights into effectively modeling complex real-world aviation time series. We hope Pneuma will serve as a rigorous and reproducible foundation for advancing data-driven predictive maintenance toward real-world deployment in commercial aviation.
[LG-5] Collective Dynamics of Self-Reinforcement: How Cross-Task Coupling Drives Policy Collapse in Unsupervised RLVR
链接: https://arxiv.org/abs/2610.37553
作者: Peize Ding
类目: Machine Learning (cs.LG)
*备注:
Abstract:Label-free reinforcement learning with verifiable rewards (RLVR) rewards a language model for agreeing with its own majority vote. It improves reasoning at first but eventually collapses: the self-assessed reward goes to one while accuracy falls, often until the model gives the same answer to every problem. The collapse appears only when training uses many problems. We explain this dataset-size dependence as a collective instability. In a solvable many-problem model in which problems share parameters, a population whose problems prefer different answers amplifies any shared perturbation for as long as the population-averaged slope \langle F’\rangle of the single-problem self-reinforcement drift F is positive. This holds while problems are still uncertain about their answers. Per-problem sharpening, slowed by 1/N for N training problems, races against the shared mode, which grows at a rate \propto\lambda^2 independent of N , where \lambda is the strength of the shared coupling. This yields a critical number of problems N_c\simeq\ln(1/\delta)/(\lambda^2G) , where \delta seeds the shared mode and G decreases with the initial confidence. Simulations of the model confirm the predictions: a sharp transition, N_c\propto\lambda^-2 up to a logarithmic correction, a single critical amplification \lambda^2NG across couplings, confidence profiles and training variants, a logarithmic dependence on batch size, and no collapse for confident initial models. A kernel version of the theory identifies a protective coupling (generalization of sharpening) and a marginal-entropy bonus that stabilizes the uniform collective mode. In label-free RL of a small Transformer, removing the shared direction prevents collapse entirely, restoring it causes collapse, and collapse grows with training-set size and base-model entropy.
[LG-6] GTFormer: A Syntax-Free Graph Network Alternative to Transformers for 8192times Length Extrapolation in Linear Time
链接: https://arxiv.org/abs/2610.37524
作者: Elisabeth Fink
类目: Machine Learning (cs.LG)
*备注: 20 pages
Abstract:Self-attention mechanisms have established the Transformer as the dominant backbone across natural language processing, source code intelligence, and automated theorem proving. However, all-to-all pairwise token comparison enforces a quadratic \mathcalO(N^2) compute and memory barrier alongside severe softmax attention dilution as sequence length N scales beyond the training horizon. Inspired by recent advances in algebraic graph learning (\textbfWPNet), this work introduces \textbfGTFormer, a syntax-free hierarchical Graph Neural Network (GNN) that replaces flat self-attention for general long-context sequence modeling. GTFormer organizes raw token streams into a bounded-degree hierarchy combining a balanced binary reduction tree ( \lceil \log_2 N \rceil depth), horizontal boundary-bridging shift edges, random expander shortcuts, multi-channel convex highway gating, and a top-down broadcast pass. Across eight long-context code, log, and multi-hop retrieval benchmarks trained exclusively on sequences of length N \le 512 , GTFormer extrapolates zero-shot to N = 4,194,304 ( 8192\times ) on a single commodity GPU in linear \mathcalO(N) memory, maintaining near-perfect accuracy where standard and RoPE Transformers suffer from out-of-distribution attention collapse and out-of-memory exhaustion.
[LG-7] Keeping Inference in the Radio: Design and Measurement of a Spiking Accelerator for mmWave ISAC Activity Recognition
链接: https://arxiv.org/abs/2610.37480
作者: Anthony Sánchez,Rafael Ruiz,José A. Belloch,Jesús O. Lacruz,Joerg Widmer
类目: Networking and Internet Architecture (cs.NI); Hardware Architecture (cs.AR); Machine Learning (cs.LG)
*备注:
Abstract:Integrated sensing and communication allows wireless networks to recognize human activity from the channel estimates that the receiver computes anyway. The inference, however, runs on the general-purpose processors of the base station, access point or handset, which are the first components a device switches off to save energy. Always-on sensing therefore requires a classifier that is small enough to remain powered while the rest of the system sleeps. We investigate whether Spiking Neural Networks (SNNs) can provide such a classifier. To this end, we generate SNN accelerators for micro-Doppler activity recognition on the public DISC dataset (IEEE 802.11ay at 60 GHz) with the open-source Spiker+ framework, and implement them together with a spike encoder on a ZCU208 RFSoC, so that the complete chain from micro-Doppler frame to activity label runs in programmable logic. A fully connected SNN with 132 neurons achieves 94.7 % accuracy (90.7 % macro-F1) on unseen subjects and performs an inference in 206 microseconds. Compared to conventional networks of the same topology on the same device, it requires an eighth of the logic and 3.4X less dynamic energy per decision than a network fed the same binary spikes, at comparable accuracy, and 7.6X less than a network fed the 16-bit spectrogram, at 1.2 points lower accuracy. Quantization-aware training reduces its weights to 3 bits within half a percentage point of the 6-bit network at the same encoder threshold
[LG-8] An Input Skip Can Reparameterize a VAE Without Improving the Generator NEURIPS2026
链接: https://arxiv.org/abs/2610.37456
作者: Kevin Lee
类目: Machine Learning (cs.LG); Atmospheric and Oceanic Physics (physics.ao-ph); Machine Learning (stat.ML)
*备注: Accepted at NeurIPS 2026 Principles of Generative Modeling (PriGM) Workshop. Available at: this https URL
Abstract:An input skip can reduce reconstruction error while leaving a model’s distribution of new samples unchanged. For a Gaussian variational autoencoder (VAE) with constant input skip 0\leq\alpha1 , a model without a skip can be written in skip form by multiplying its decoder output by b=1-\alpha and its noise variance by b^2 . The distribution of observations and latent variables and its exact generative scores stay the same, while reconstruction mean squared error (MSE) is multiplied by b^2 . Separate training can produce different generators because optimization depends on the parameter coordinates. In a Gaussian experiment with known density, changing skip and observation variance together reduces MSE but worsens energy score and the estimate of negative log density from importance sampling. On seven annual Deep-ocean Assessment and Reporting of Tsunamis (DART) records, energy score differences at the same observation variance have both signs on four original records and favor the skip on three records from 2018. Probabilistic principal component analysis (PPCA) has lower mean energy scores than every neural DART fit. With the encoder and decoder kept the same, estimating observation variance from training reconstruction error can also worsen energy score. Assessing generation requires more than reconstruction error.
[LG-9] Interpretable Meta-Policies for Neural Decision Making
链接: https://arxiv.org/abs/2610.37451
作者: Raban Emunds,Jannis Blüml,Quentin Delfosse,Kristian Kersting
类目: Machine Learning (cs.LG)
*备注:
Abstract:Deep Reinforcement Learning remains difficult to interpret and align due to the black-box nature of neural networks. While (neuro-)symbolic approaches improve transparency by incorporating symbolic reasoning, applying them directly to low-level actions results in overly complex, unreadable policies. We introduce NEXUS, a hierarchical RL framework that resolves this tension by combining neural skills with symbolic meta-policies. By shifting transparent reasoning to high-level skills rather than atomic actions, NEXUS drastically reduces symbolic complexity without sacrificing environment performance. Object-centric observations enable high-level reasoning and the design of reward functions and meta-policies by interaction with LLMs. We demonstrate that NEXUS agents learn disentangled skills, are interpretable regarding high-level goals, are better aligned to goals described in the game manual, less prone to reward hacking and more robust to game simplifications than deep RL. Finally, we isolate and evaluate the impact of different meta-policy architectures (neural, symbolic, neuro-symbolic) on environmental return and aligned task performance.
[LG-10] CRISP: Co-designing Pruning and Scheduling for Efficient MoE Inference on Edge Servers
链接: https://arxiv.org/abs/2610.37442
作者: Congwei Zhang,Youming Tao,Yifei Zou,Yu Yang,Ruirui Zhang,Xiaoyu Zhang,Falko Dressler,Xiuzhen Cheng,Dongxiao Yu
类目: Machine Learning (cs.LG)
*备注:
Abstract:Mixture-of-Experts (MoE) inference on memory-constrained edge platforms is often dominated by expert-weight movement rather than computation. Although each token activates only a small subset of experts, routing is input-dependent, so all experts must remain accessible and memory movement dominates latency. Existing approaches such as uniform pruning and expert dropping overlook two properties of MoE execution: expert combinations are highly skewed in frequency, and experts contribute asymmetrically within each combination. We present CRISP, a combination-aware framework that co-designs pruning and runtime scheduling for efficient MoE serving on memory-constrained edge platforms. CRISP reorders neurons by importance so one stored expert can support multiple execution widths, assigns differentiated execution to hot, warm, and cold combinations based on frequency and expert role, and exploits heterogeneous widths to overlap expert loading with computation. Across Mixtral-8x7B and DeepSeek-MoE-16B, CRISP improves throughput by up to 2.1\times at the same memory budget while limiting accuracy loss to 3.6% relative to full-width inference. Our results highlight that efficient edge MoE serving requires jointly optimizing which expert capacity is preserved and how expert execution is scheduled.
[LG-11] Multi-Agent Reinforcement Learning of Co-Adaptive Hip Exoskeleton Control (MARL-Exo) for Metabolic Energy Reduction
链接: https://arxiv.org/abs/2610.37441
作者: Yifei Yuan,Jakob Wolf,Ghaith Androwis,Xianlian Zhou
类目: Robotics (cs.RO); Machine Learning (cs.LG)
*备注: 14 pages, 6 figures, including supplementary material
Abstract:Objective: This study aims to determine whether a hip-exoskeleton policy trained in simulation, by co-adapting with a musculoskeletal human model under a metabolic-energy reward, reduces the metabolic cost of walking in new users without per-user tuning, and how mechanical-power shaping affects the learned assistance. Methods: We developed MARL-Exo, a multi-agent reinforcement learning framework in which a full-body human model (304 muscle actuators; randomized anatomy, gait, and speed) and a bilateral hip-exoskeleton policy are trained concurrently in two stages. The exoskeleton observes only delayed hip kinematics and its past commands; its reward includes muscle metabolic power from a physiologically based energetics model, alone (Metabolic) or with mechanical-power shaping (Metabolic + Power). The Metabolic policy was deployed and tested without retraining in eight healthy adults walking at 1.25 m/s, using indirect calorimetry. Results: In simulation, the Metabolic policy reduced modeled muscle metabolic power by 13.8% relative to no-device walking, versus 8.8% for Metabolic + Power. On hardware, MARL-Exo reduced net metabolic rate by 18.3 \pm 6.6% relative to Passive in all eight participants. Relative to No-Exo, the mean reduction was 7.4 \pm 10.0%; six participants more than offset the 13.1% device-wearing penalty, walking up to 24.2% below their No-Exo cost. Conclusion: A metabolically informed policy co-adapted in simulation transferred to users without per-user tuning and consistently reduced the metabolic cost of walking with the device, while explicit power shaping reduced the simulated benefit. Significance: Unlike per-user human-in-the-loop optimization, this approach yields a single population-level controller and allows assistance objectives to be screened in simulation before human testing.
[LG-12] Displacement Moments and Path-Terminal Separation in Brownian Interpolant Sampling
链接: https://arxiv.org/abs/2610.37413
作者: Zhong Han
类目: Machine Learning (cs.LG); Probability (math.PR)
*备注: 56 pages, 4 figures; 8-page main text and complete appendices
Abstract:Brownian stochastic interpolants connect paired endpoint samples, yet their sampling error can depend on displacement moments even when neither endpoint has a first moment. For arbitrary Borel couplings, we prove early-stopped, target-to-output KL guarantees for standard Euler sampling from the true source. A fourth displacement moment gives an explicit O(N^-1) bound; a second moment suffices for fixed-law convergence as drift error vanishes. For one fixed coupling with no displacement moment above two, original bridge-path KL stays of order one, comparison-path KL is of order 1/\log N , and terminal KL is at most C/N^2 on the same grid. The path decomposition identifies the information about the source discarded by Markov projection. When the original bridge is Markov, current-state Euler sampling instead has a second-moment O((M_2+d)/N) bound; finite-entropy Schroedinger bridges for the probability Brownian reference are included. We also construct a current-state learner with expected output KL consistency from iid endpoint pairs. Numerical diagnostics illustrate the distinct errors. Conditional diffusion identities specify a restoration interface, while image controls find no iterative advantage over cheap probabilistic predictors. Uniform terminal rates for general second-moment couplings remain open.
[LG-13] How Many and How Much? Contrastive Learning with Negative Samples for Knowledge Graph Embedding
链接: https://arxiv.org/abs/2610.37389
作者: Sooho Moon,Yunyong Ko
类目: Machine Learning (cs.LG)
*备注: 12 pages, 12 figures, 5 tables
Abstract:Knowledge graph embedding (KGE), a fundamental technique for knowledge-centric Web and KG-augmented AI applications, is commonly trained through contrastive learning with negative samples to learn discriminative representations. Despite its empirical success, two crucial questions regarding negative samples remain underexplored: (Q1) how many negative samples should be used for each positive sample and (Q2) how much should each negative sample contribute to learning? To answer these questions, we propose RADAR, a novel contrastive learning framework for KGE models with two effective strategies, respectively: (1) Risk-aware negative budget allocation quantifies false-negative sampling risk by modeling an open-world KG as a stochastic partial observation of its closed-world counterpart and adaptively assigns a negative budget to each positive sample based on its risk; (2) Distribution-calibrated negative weighting assigns negative weights by calibrating the weighting strength for each positive based on two complementary characteristics of the negative-score distribution: dispersion and concentration. Extensive experiments on 3 real-world KGs with 4 KGE models reveal that RADAR achieves the best KGE performance on average across all model–KG combinations and reduces sampled false negatives by over 90% compared with conventional fixed-budget methods.
[LG-14] Para-Gaussian Approximations of Random Deep Neural Networks
链接: https://arxiv.org/abs/2610.37388
作者: Mihai Nica,Janosch Ortmann
类目: Machine Learning (cs.LG); Probability (math.PR)
*备注: 34 oages, 9 figures
Abstract:It is known that randomly initialized neural networks have Gaussian behaviour in the n\to \infty limit. Unfortunately, this approximation can be poor for finite n and gets worse as one goes deeper into the network. In this work, we develop a more accurate approximation that is \emphslightly non-Gaussian, which we call the para-Gaussian distribution and is constructed from the Hermite polynomials. We show how this para-Gaussian approximation can lead to a practical recursive method to approximate the deeper layers of the network. Finally, we apply this theory to linear neural networks (where things can be solved exactly) and to ReLU neural networks (where things can be approximated).
[LG-15] Euclidean Crowding-Driven Adaptive Euclidean-Hyperbolic Graph Learning
链接: https://arxiv.org/abs/2610.37373
作者: Hao Chen,Jian Sun,Chuan Shi,Hongying Zhang
类目: Machine Learning (cs.LG)
*备注:
Abstract:Hyperbolic graph representation learning has emerged as a powerful paradigm for modeling hierarchical data. Unlike Euclidean space, which exhibits polynomial volume growth, hyperbolic space grows exponentially with radius, naturally accommodating tree-like structures without the geometric distortion known as Euclidean crowding. Despite strong empirical performance, the theoretical mechanisms underlying hyperbolic embeddings remain poorly understood. In this work, we provide a geometric quantifcation research for Euclidean crowding, a generalization-theoretic explanation for the superiority of hyperbolic space on hierarchical data, and further develop an adaptive Euclidean-hyperbolic fusion learning framework toward bridging this gap. Specifically, we \textbffirst establish a geometric quantification of Euclidean crowding by analyzing layer-wise minimum separation in node embeddings, proving that fixed-dimensional Euclidean embeddings of tree-like graphs inevitably exhibit layer-wise crowding that leads to weaker margin-based generalization bounds. \textbfSecond, we provide a generalization-bound for the superiority of hyperbolic geometry on hierarchical data. \textbfThird, we propose EH-Fusion, a geometry-adaptive framework that lifts Euclidean GNN representations into hyperbolic space via exponential mapping and adaptively fuses both geometries based on node-level neighborhood expansion patterns. Extensive experiments validate our theoretical insights and demonstrate the effectiveness of EH-Fusion on benchmark datasets.
[LG-16] RIDENT-Seq: Profile-Matched Prefix-Future Contrast for Label-Efficient Payment Risk Modeling
链接: https://arxiv.org/abs/2610.37365
作者: Weiyu Han,Haoyu Jiang,Jie Sheng,Ruobing Zuo,Huilin He,Ze Zhao,Xiao Tan,Nairong Zhang,Shuai Chen,Tianyi Zhang
类目: Machine Learning (cs.LG)
*备注: 12 pages, 5 figures
Abstract:Ant International accumulates large volumes of transaction history across users and payment businesses. Confirmed fraud and chargeback outcomes, by contrast, are scarce, delayed, and specific to each business. A single encoder pretrained on unlabeled histories could serve all businesses, but only if pretraining captures the behavior that separates risk. Coarse activity differences can make contrastive pretraining on transaction logs too easy: transaction volume, amount, and timing can distinguish a history’s continuation from arbitrary candidates without modeling how payment resources are used. We present TRIDENT-Seq, a transaction foundation model trained with Prefix-Future Contrast (PFC). PFC matches each prefix to a summary of its subsequent behavior among candidates with similar activity profiles; activity prediction and dispersion regularization act on the encoder output. Across three production payment businesses, the pretrained encoder improves average precision (AP) and recall at a 0.5% intervention budget over architecture-matched scratch training in all six business-budget settings. With a fixed validation set, it exceeds full-label scratch training in mean AP and recall using 10% of the training labels on Wallet and 25% on Shopping and SME. Pretraining only on another business’s histories still helps. Replacing profile-matched candidates with random ones cuts recall from 24.62% to 5.64%. In Ant International’s deployment evaluation, recorded amount-based chargeback and overall loss rates are 37.7% and 37.9% lower than those of the previous XGBoost-based system under the same intervention budget.
[LG-17] Online Interval Selection with Binary Predictions: Optimal Consistency and a Continuous Robustness Tradeoff
链接: https://arxiv.org/abs/2610.37361
作者: Zhong Han
类目: Data Structures and Algorithms (cs.DS); Machine Learning (cs.LG)
*备注: 66 pages, 5 figures. Includes a technical appendix and ancillary evidence tables
Abstract:We study online interval selection with binary predictions of membership in a final optimal solution. Requests arrive in arbitrary order; accepted requests may be revoked, but rejected or revoked requests cannot be recovered. With rewards proportional to interval length, the best consistency under finite strict robustness is d=(3+\sqrt5)/2 when both guarantees hold almost surely on each fixed input, including for randomized algorithms. A continuous lower bound forces consistency strictly above d at every robustness budget below 4\varphi+3\approx9.472 , whereas our deterministic algorithm attains d with exact robustness \varphi^5\approx11.090 , where \varphi=(1+\sqrt5)/2 . Each held interval inherits a ledger recording positive prediction mass lost through irrevocable decisions. A tunable replacement cost gives a continuous family with exact consistency throughout and exact robustness on two parameter intervals, including (C,R)=(25/9,10) and (4,7) . Randomization permits expected consistency approaching one with finite expected robustness, whereas almost-sure robustness forces expected consistency strictly above two at every finite budget.
[LG-18] Normative Deviation Modeling for EEG-based Depression Detection
链接: https://arxiv.org/abs/2610.37338
作者: Wan Jiang,Jingqi Cheng,Hui Zhang,Xu Zhao,Xiaojing Chen,Shengeng Tang,Jingjing Wu,Richang Hong
类目: Machine Learning (cs.LG)
*备注:
Abstract:Web-enabled digital health services are expanding access to physiological monitoring and mental health screening. However, electroencephalography (EEG)-based major depressive disorder (MDD) detection remains challenging across heterogeneous data sources, where individual variability and acquisition differences may affect disease-related representations. In this work, we investigate the use of healthy normative references for EEG-based MDD detection. Specifically, structural and distributional reference models are independently trained using healthy control (HC) recordings to characterize departures from healthy EEG patterns. We develop two diagnostic approaches that incorporate normative deviations at different representation levels. Scalar-level Normative Deviation-guided Diagnosis (S-NDD) summarizes structural and distributional deviations into scalar scores and integrates them with deep EEG representations. Feature-level Normative Deviation-guided Diagnosis (F-NDD) retains dimension-wise residuals and combines them with deep EEG representations through attention-based aggregation and auxiliary supervision. We also employ target HC-assisted score calibration for cross-source evaluation without updating the trained models. Experiments on two public EEG datasets examine both approaches under subject-independent within-source and bidirectional cross-source settings. The results demonstrate the effectiveness of normative deviation-guided diagnosis across different evaluation settings.
[LG-19] Structured Representation Probing for Risk Anticipation in End-to-End Autonomous Driving
链接: https://arxiv.org/abs/2610.37299
作者: Jiawei Liang,Tianrui Lou,Le Yang,Xianghao Jiao,Wei Wang,Xiaochun Cao
类目: Machine Learning (cs.LG)
*备注:
Abstract:End-to-end driving policies form structured internal representations of traffic, ego behavior, and road constraints. These representations offer a monitoring interface for anticipating failures, complementing risk signals from trajectories, controls, and manually designed features. We introduce the Structured Representation Probe (SRP), a lightweight monitor that reads actor, ego, and map representations from a frozen policy. SRP learns a separate linear risk direction for each source and aggregates element scores through cardinality-normalized log-mean-exp pooling, retaining sensitivity to salient evidence while reducing dependence on token count. Event-window supervision links the readout to upcoming hard failures, and auxiliary event type heads jointly train the source directions across failure modes. Online alerting uses a single generic score without modifying the driving policy. Bench2Drive experiments on DriveTransformer, UniAD, and VAD show higher frame-level PR-AUC and ROC-AUC than the evaluated runtime baselines. On DriveTransformer at a horizon of 2 seconds, ROC-AUC reaches 0.833 versus 0.715 for temporal consistency, with 2.67% readout runtime overhead. The trained probes also retain the ability to distinguish impending failures on new routes.
[LG-20] Bayesian Inference of Expensive Simulators: Survey and Comparison of Design Choices
链接: https://arxiv.org/abs/2610.37284
作者: Šimon Soldát,Václav Šmídl
类目: Machine Learning (cs.LG)
*备注: 25 pages, 8 figures. Accepted for publication in Statistics and Computing. This is the submitted manuscript; the revised version, including changes based on peer review, is available from the authors upon request
Abstract:We present a survey of approaches to Bayesian inference with simulator forward models, focusing on settings with prohibitively expensive simulators. In this domain, Bayesian optimization-based methods have emerged as the most promising approach. We unify existing methods into a general framework, enabling consistent reasoning about shared algorithmic structure across the fragmented literature. Within this framework, we identify the primary design choices and perform a systematic empirical comparison of their effects on inference efficiency in terms of required simulation budget. A key result of our study is that inference efficiency is often dominated by the choice of surrogate-model response variable, as it directly governs the complexity of the modeled function. Specifically, we demonstrate that the prevalent scalar log-likelihood is a poor response variable, as it discards crucial information. Based on this insight, we introduce the concept of custom \emphproxy variables: transformations of simulator outputs engineered to retain more information and provide smoother response surfaces, substantially improving inference efficiency over standard response variables. We provide actionable guidance and a numerically robust Julia package to facilitate practical adoption of Bayesian optimization for simulator inverse problems. Comments: 25 pages, 8 figures. Accepted for publication in Statistics and Computing. This is the submitted manuscript; the revised version, including changes based on peer review, is available from the authors upon request Subjects: Machine Learning (cs.LG) MSC classes: 62F15 (Primary) 60G15, 62L05, 35R30 (Secondary) ACMclasses: G.3; I.2.6; I.6.4 Cite as: arXiv:2610.37284 [cs.LG] (or arXiv:2610.37284v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2610.37284 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-21] Sharpness-Aware Muon: Matrix-Aware Optimization with Momentum-Guided Perturbations
链接: https://arxiv.org/abs/2610.37225
作者: Zhiwei Zhai,Wenjing Yan,Ying-Jun Angela Zhang
类目: Machine Learning (cs.LG)
*备注:
Abstract:Muon is a matrix-aware optimizer that combines momentum with orthogonalization to produce structured update directions with strong empirical behavior. However, it only shapes the geometry of the update itself and does not explicitly control neighborhood worst-case loss or bias training toward flat regions, which are often linked to stronger generalization. To address this limitation, we introduce the sharpness-aware formulation of Muon. We first develop \textscMuSAM, which incorporates Sharpness-Aware Minimization (SAM)-style perturbation into Muon’s orthogonalized update and establishes a principled bridge between sharpness-aware training and matrix-aware optimization. We then show that this formulation faces an efficiency–stability tension: constructing the perturbation requires an extra gradient evaluation, while the perturbation itself is generated from a single minibatch gradient and is therefore noisy. To resolve this issue, we propose \textscMoMuSAM, which turns Muon’s momentum from an update variable into a sharpness-probing signal. By using normalized momentum as the perturbation direction, \textscMoMuSAM removes the extra clean-gradient computation and replaces the single-batch perturbation proxy with a temporally aggregated one. We prove that both \textscMuSAM and \textscMoMuSAM attain the same asymptotic first-order stationarity order, and further show that \textscMoMuSAM yields a more stable stochastic linearization than \textscMuSAM. Experiments on standard benchmarks demonstrate that sharpness awareness consistently improves Muon.
[LG-22] SimplexLoRA: Dynamic Rank Updating via Alternating Minimization with Simplex Projection EMNLP2026
链接: https://arxiv.org/abs/2610.37206
作者: Igor Shalygin,Grigorii Davydenko,Andrey Veprikov,Aleksey Antonov,Nikolay Bushkov,Stanislav Moiseev,Aleksandr Beznosikov
类目: Machine Learning (cs.LG)
*备注: Accepted to Findings of EMNLP 2026
Abstract:The success of large language models in real-world applications relies on parameter-efficient fine-tuning methods such as Low-Rank Adaptation (LoRA), which inserts compact, trainable adapters into selected layers. In this paper, we propose SimplexLoRA, a LoRA variant that adaptively adjusts adapter ranks during fine-tuning based on their learned importance. Unlike standard LoRA, which employs a fixed rank across all layers, SimplexLoRA optimizes per-adapter importance weights on the n -scaled simplex during a short warm-up and then freezes them, dynamically allocating capacity where it is most beneficial without increasing the overall parameter budget or inference cost. We evaluate SimplexLoRA across 12 language understanding, summarization, and reasoning tasks using DeBERTaV3, BART-large, Llama 3.1 8B, and Qwen3-8B. Our results show that SimplexLoRA consistently matches or outperforms standard LoRA and recent adaptive baselines, with the largest gains at low ranks. Our code is available at this https URL.
[LG-23] Physics-Guided Flow Matching: A Generative AI Framework for Global Total Electron Content Forecasting
链接: https://arxiv.org/abs/2610.37159
作者: Si-Tong Yang,Feng Ye,Xian Lu,Dong Lin,Benjamin C. Martinez
类目: Machine Learning (cs.LG); Space Physics (physics.space-ph); Machine Learning (stat.ML)
*备注:
Abstract:Accurate forecasting of global total electron content (TEC) is important for space weather monitoring and the reliable operation of satellite communication and navigation systems. We propose a physics-guided generative artificial intelligence (AI) framework for day-ahead global TEC forecasting. The framework builds on flow matching, which learns to gradually transform samples from an initial distribution into new samples that share the statistical characteristics of the target data. The proposed model incorporates ionospheric dynamics described by a partial differential equation (PDE) into this transformation process, guiding the changes between initial and target samples using physical knowledge. The model further combines the physical dynamics with statistical patterns captured from data to learn this transformation, adaptively adjusting their relative contributions at each location. The model is trained and evaluated using one year of global TEC data generated by the Thermosphere Ionosphere Electrodynamics General Circulation Model (TIEGCM). The results demonstrate that our model improves both spatial and temporal forecasting accuracy over representative benchmarks, including statistical, deep learning, and generative AI methods at all lead times and across latitudes. The average root mean square error (RMSE) is 39.2% lower than that of ConvLSTM and 23.8% lower than that of the same framework without physical guidance. The learned contribution weights vary systematically with latitude and across geographic locations, showing how the physical contribution changes across the forecast domain. These findings demonstrate that explicit physical guidance improves generative forecasts of global TEC and highlight the value of incorporating physical dynamics into generative AI modeling.
[LG-24] Variational Optimal Control for Diffusion Guidance
链接: https://arxiv.org/abs/2610.37146
作者: Ruoyu Wang,Xinyi Ye,Yunshuo Shi,Liangyu Yuan,Beier Zhu,Chi Zhang
类目: Machine Learning (cs.LG)
*备注:
Abstract:We present a variational control framework that connects sampling-time guidance and reward-based model adaptation through a shared Jacobian pullback. Starting from terminal reward maximization with a path-space KL penalty, the Gibbs variational principle and Doob transform yield a correction to the generative dynamics. A local denoising approximation exposes the pullback that transports output preferences to the controlled variable. This construction gives Variational Optimal Guidance (VOG) in state space and extends to Distilled Variational Optimal Control (DVOC) by amortizing locally improved endpoints into model parameters. The resulting controls share a local variational rule: VOG acts through the denoising Jacobian, while DVOC realizes the parameter pullback by endpoint distillation. Under a linear MMSE model, VOG’s exact correction lies in the data subspace. On ImageNet, VOG reduces DiT-XL/2 FID from 2.27 to 2.04 at 250 sampling steps and from 7.58 to 4.44 at 25 steps. After 20 updates on Z-Image-Turbo, DVOC improves HPSv3++ and four held-out evaluator scores over the untrained model, with the highest corresponding means among the trained methods compared. These results connect inference-time control and reward adaptation through a common geometric principle.
[LG-25] On the Sparsity of Direct Preference Optimization: Weight Disentanglement in the NTK Regime
链接: https://arxiv.org/abs/2610.37085
作者: Kensuke Sasaki,Issei Sato
类目: Machine Learning (cs.LG)
*备注:
Abstract:Direct Preference Optimization (DPO) has been widely adopted for aligning language models with human preferences. Recent empirical work has revealed that DPO induces remarkably sparse parameter updates, yet no theoretical explanation for this phenomenon has been offered. We analyze DPO’s optimization dynamics under a neural tangent kernel linearization and show that the logistic structure of the DPO loss gives rise to an implicit maximum-margin bias. Building on this observation, we introduce the concept of dual sparsity: the DPO update direction is a sparse linear combination (data-level sparsity, driven by support vectors) of sparse gradient feature vectors (parameter-level sparsity, arising from gradient cancellation between preferred and dispreferred responses). We formalize this structure in our main theorem, bound the number of support pairs by a margin-dependent spectral quantity, and present a framework relating dual sparsity to weight disentanglement—a condition favorable for model merging. Experiments on GPT-2 Small and Medium and on 7–9B instruction-tuned models (Llama-3-8B, Mistral-7B, Gemma-2-9B) with the Anthropic HH-RLHF and UltraFeedback datasets confirm both sources of sparsity, and experiments on GPT-2 show that DPO reduces cross-task interference by up to 35\times compared to supervised fine-tuning.
[LG-26] From Ethernet to Wi-Fi: Flow Feature Selection and Construction for Traffic Classification
链接: https://arxiv.org/abs/2610.37007
作者: Francisco Rau,Petia Georgieva,Carlos Herranz,Iñaki Val,Joaquin Perez
类目: Networking and Internet Architecture (cs.NI); Machine Learning (cs.LG)
*备注:
Abstract:Flow-based traffic classifiers trained on Ethernet data may lose accuracy when applied to Wi-Fi traffic. This paper proposes a five-stage procedure for selecting and constructing features that enable Wi-Fi flow classification using models trained only on Ethernet data. The procedure removes the transport protocol identifier, examines rate dependence, compares feature distributions and the information each feature provides for the application classes, reduces redundancy, and constructs normalized packet-size and temporal features. Ethernet and Wi-Fi traffic are simulated in ns-3, with Wi-Fi using Best Effort access category. Labelled Wi-Fi data from six development scenarios are used for feature design, while a separate scenario is reserved for final testing. The evaluation compares feature sets, examines threshold sensitivity, and analyses classification errors under different traffic conditions.
[LG-27] Meta Models: Models that Model Learning
链接: https://arxiv.org/abs/2610.37001
作者: Peihang Li,Ziming Liu
类目: Machine Learning (cs.LG)
*备注:
Abstract:Every neural-network training run produces not only a model, but also evidence about how models learn. Can these accumulated experiments enable us to predict and compare new training runs before executing them? We study meta-models: models that predict the learning behavior of other models. From this perspective, final-performance prediction, learning-curve prediction, and hyperparameter selection are different queries to the same training response function. We take a first, humble empirical step toward this goal: treating real language-model training records as native numeric data, we find that an off-the-shelf tabular in-context learner, TabPFN, can match or outperform a fine-tuned LLM meta-model in multiple evaluated settings. Whereas recent work fine-tunes LLMs on tokenized configurations at a cost of tens to hundreds of GPU-hours, TabPFN requires seconds of fit-and-predict time without task-specific weight updates. We argue that meta-modeling is better treated as its own numeric modality rather than as a text task for LLMs. Going one step further, augmenting TabPFN with Kolmogorov-Arnold Networks (KANs) further improves accuracy on selected prediction tasks. Finally, we pre-registered a forecast of the Marin Hero Run, a 535B-parameter public training run, whose ground truth is revealed point by point as training progresses, putting meta-model predictions to the test against training outcomes not yet observed at the time of forecasting.
[LG-28] Online gas classification at the edge
链接: https://arxiv.org/abs/2610.36999
作者: Ton Juny Pina,Hugh Greatorex,Elisabetta Chicca
类目: Emerging Technologies (cs.ET); Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:Field applications in chemical sensing, from portable safety monitors to mobile robots, require models that are lightweight, fast to respond, and can be recalibrated from limited labelled data. Current gas classification pipelines assume controlled sampling cycles and offline feature extraction, and rely on computationally expensive models. With a few thousand neurons and sub-milliwatt power consumption, insects reliably classify odours in open environments. Inspired by the insect olfactory system, we present a neural network architecture that classifies raw sensor array readings using a sparse, high-dimensional binary representation. We evaluate it on random sequences of single gases and binary mixtures, introducing new samples dynamically and classifying gas identity from the transient sensor response. In this setting, the architecture out-performs a linear \glssvm on gas mixtures and matches an optimised non-linear \glssvm, while maintaining online class-level frequency estimates that distinguish familiar from novel stimuli. It relies exclusively on binary weights and Hebbian-based local learning, remaining computationally lightweight and well suited to ultra-low-power neuromorphic processors. We demonstrated the network on-chip with a hazardous gas detection experiment running on the mixed-signal neuromorphic chip TEXEL. The on-chip classifier yielded an accuracy of 74.6% while running at 482 \unit\nano\joule per inference and 27 \unit\micro\watt of static-power.
[LG-29] CaLLM : Continual Adaptation of LLM s via Non-Parametric Adapter Routing
链接: https://arxiv.org/abs/2610.36994
作者: Bilge Celik,Anna Vettoruzzo,Joaquin Vanschoren
类目: Machine Learning (cs.LG)
*备注: Accepted at 5th Conference on Lifelong Learning Agents (CoLLAs), Main Conference Track, 2026
Abstract:Continual adaptation of large language models (LLMs) to evolving task distributions while mitigating catastrophic forgetting remains an open problem. Existing strategies often struggle to scale or lack a robust mechanism for zero-shot task-to-expert assignment. We introduce CaLLM, a framework that resolves these tensions by combining metric-based meta-learning with modular parameter isolation via lightweight adapters. First, CaLLM projects each input batch into a task-agnostic embedding space to find its nearest prototype. It then routes the batch to the specific adapter specialized for that data distribution. This formulation provides (i)~controllable cross-task interference by design, (ii)~constant computational complexity regardless of task count, (iii)~seamless integration of novel distributions through prototype expansion, and (iv)~regime agnosticism as the local routing decision enables a unified approach to both online and offline settings. Our experiments demonstrate that this routing mechanism discovers semantically coherent task groupings while facilitating constructive interference between similar tasks, and keeping overhead at a minimum. Across diverse and fine-grained continual learning benchmarks, CaLLM consistently outperforms traditional isolation and replay methods, establishing prototype-based modular routing as a principled and scalable approach to lifelong adaptation of LLMs.
[LG-30] GNN-Guided Network Reduction for Centralized Urban Traffic Routing
链接: https://arxiv.org/abs/2610.36981
作者: Nikolai Antonov,Lukáš Chrpa,Andrii Nyporko,Jan Hůla
类目: Machine Learning (cs.LG)
*备注: Submitted for review to Computers Operations Research
Abstract:We consider the centralized traffic routing (CTR) problem, where a central planner assigns routes to origin-destination (OD) vehicle flows to minimize congestion-dependent segment cost. Prior work formulates CTR efficiently as a mixed-integer linear program (MILP), but scaling this formulation to the full road network is computationally prohibitive within an operational time budget. Bounded Routes and Smart Routes speed up optimization using precomputed reduced networks; however, these methods do not account for vehicle information within each OD flow. Motivated by this gap, we propose a learning-based urban network reduction approach. A dedicated graph neural network (GNN) first scores road segments for each OD flow using the road network and vehicle-flow information. Highly ranked segments are then incorporated into complete OD routes, and the original segment-flow MILP is solved on the resulting reduced road network. The GNN thus guides the reduction of the road network, while the solver enforces feasibility and determines the routing. Experiments on four real urban networks – Prague-Dejvice, London, New York, and Sydney – show that the proposed approach achieves the smallest relative objective difference to the 30 -second full-graph reference in every city under both traffic regimes. Averaged across cities, this difference is more than six times lower than that of the best competing method. Moreover, traffic simulation results show that the optimization gains of the proposed method lead to more efficient routing, with the highest aggregate arrival rate and a competitive trip duration per arrival percentage point.
[LG-31] On the VC Dimension of Autoregressive Linear Threshold Functions
链接: https://arxiv.org/abs/2610.36970
作者: Ilan Doron-Arad,Idan Mehalel,Elchanan Mossel
类目: Machine Learning (cs.LG)
*备注:
Abstract:Linear threshold functions (LTFs) are a canonical class of Boolean predictors. In the autoregressive setting, the same LTF reads the most recent d bits, generates the next bit, and returns the bit produced after T steps. While the one-step class has VC dimension d , for general T the counting upper bound is O(d^2) . We prove that the VC dimension is \Theta(d^2) for every T\ge d^2 , resolving the quadratic gap for deterministic autoregressive LTFs. We then adapt the same technique to show an exponential separation between deterministic and stochastic autoregressive models: replacing the sign of LTFs by logistic sampling can increase the VC dimension of the final majority prediction class from quadratic to exponential in d .
[LG-32] Live-OPSD: On-Policy Self-Distillation for Proactive Streaming Large Video Models
链接: https://arxiv.org/abs/2610.36959
作者: Kaixuan Du,Xin Wan,Dai Guan,Ming Chen,Xiaoxi Jiang,Guanjun Jiang
类目: Machine Learning (cs.LG)
*备注:
Abstract:Streaming multimodal assistants must make response decisions while events are still unfolding. They need to recognize when the available evidence warrants a response, produce an appropriate continuation, and remain silent otherwise. Supervised fine-tuning learns from recorded assistant turns, but does not directly train on the response prefixes produced by the evolving policy itself. We introduce Live-OPSD, an on-policy self-distillation method that uses privileged hints available only during training to supervise the policy on its own outputs. The policy first generates a continuation from the streaming context alone. The same model then re-evaluates that continuation with a hint describing the relevant evidence, and the resulting distributions guide the unhinted policy. On Qwen3-Omni-30B-A3B, Live-OPSD attains an average score of 59.64 on LiveProBench, 5.26 points above the strongest continued-SFT baseline, and stays competitive on OVO-Bench (65.84). These results indicate that guiding the policy on its own outputs with privileged hints improves both when and how a streaming assistant responds.
[LG-33] Algorithmic Monoculture in Machine Learning-based Site-Specific Fertilizer Recommendation NEURIPS2026
链接: https://arxiv.org/abs/2610.36941
作者: Abdelghani Belgaid
类目: Machine Learning (cs.LG); Computers and Society (cs.CY); Theoretical Economics (econ.TH)
*备注: Accepted at the NeurIPS 2026 Workshop on Economics for Machine Learning (EconML)
Abstract:Machine learning recommendations are often evaluated one model at a time, despite deployment across interconnected decision makers. For continuous NPK fertilizer prescriptions, correlated errors can aggregate across farms and generate systemic costs. A model of shared and farm-specific recommendation errors links expected social loss to model adoption shares; the Herfindahl-Hirschman Index characterizes concentration effects when common model errors are independent and predictive risks are equal. A covariance-generalized formulation captures the dependence between model families. A two-model threshold characterizes when privately optimal monoculture is socially dominated, and a convex portfolio rule limits average private predictive regret. Nine model families across five Moroccan cereal systems provide a benchmark-based calibration. Under illustrative parameters and a 10% regret cap, accuracy-constrained diversification lowers modeled social loss in all five systems by 10.89-41.53% (mean 19.41%). Uniform allocation reduces modeled loss in this setting, but breaches the regret cap in every crop. Correlation sensitivity shows that highly dependent model errors attenuate the benefits of diversification. These counterfactual outcomes are not estimates of agronomic or monetary welfare; the framework identifies adoption architecture and error dependence as ecosystem-level evaluation variables.
[LG-34] DLM: Adaptive Temporal Integration for Diffusion Language Models
链接: https://arxiv.org/abs/2610.36940
作者: Bohang Sun,Max Zhu,Francesco Caso,Christopher Irwin,Jialin Yu,Pietro Liò
类目: Machine Learning (cs.LG)
*备注:
Abstract:Masked diffusion language models repeatedly compute representations for unresolved tokens, but conventional denoisers do not explicitly retain these representations for reuse across steps. We introduce Temporal Self-Conditioning Diffusion Language Model (TDLM), which lets each unresolved position attend to its own history of hidden representations within a frozen denoiser. This gives later predictions access to computation already performed. Training uses self-generated trajectories and a simple cross-entropy objective for tokens revealed between consecutively sampled decoding steps, while causal temporal attention permits parallel evaluation during training. On LLaDA, the temporal modules add only approximately 0.84% trainable parameters. Across LLaDA and Dream, the main comparison shows up to 7.15 \times faster generation than the corresponding original diffusion language model, with competitive answer quality on the evaluated tasks. These results motivate temporal memory as a practical way to make more effective use of repeated denoising computation.
[LG-35] Gaussian Process Bandits for Constraint Black Box Lever Optimization
链接: https://arxiv.org/abs/2610.36935
作者: Bram van den Akker,Hidde Fokkema,Anastasiia Kornilova
类目: Machine Learning (cs.LG)
*备注: 7 pages, 3 figures, 1 table. Includes results from a large-scale randomized controlled trial in a production environment
Abstract:Optimizing discrete allocation weights across competing segments of a shared budget is a fundamental challenge in resource-constrained, sequential decision-making environments. In practice, this optimization is often hindered by delayed, noisy, and aggregate feedback, alongside strict operational constraints on how frequently weights can be adjusted. Because of the resulting combinatorial action space and extreme data sparsity, classical multi-armed bandit algorithms that evaluate arms independently are rendered highly inefficient. To overcome these limitations, we propose a segmentation-agnostic approach utilizing Gaussian Process Upper Confidence Bound (GP-UCB) algorithms. By placing a Gaussian Process prior on the reward function, our method effectively shares information across neighboring configurations, drastically reducing the necessary exploration horizon. In this paper, we bridge the gap between theoretical bandit research and production-scale deployment. We first introduce a high-fidelity synthetic simulator to evaluate various kernel configurations under stationary, drifting, and reactive environments, identifying a composite Matérn + Linear kernel as the most robust choice for overcoming local noise and global shifts. Finally, we validate our findings through a large-scale Randomized Controlled Trial (RCT) in a live production environment. Our empirical results demonstrate that the GP-UCB approach delivers a statistically significant efficiency lift compared to a standard \varepsilon -greedy baseline, offering a highly practical solution for real-world black-box lever optimization.
[LG-36] Provenance Signatures for Low Rank Adaptations of Image Diffusion Models
链接: https://arxiv.org/abs/2610.36933
作者: Tobias Cheung,Mohan Dantam,Marc Juarez,Rik Sarkar
类目: Machine Learning (cs.LG)
*备注: Under review
Abstract:Diffusion models have revolutionized the generation of realistic artificial images. Efficient techniques such as Low Rank Adaptation (LoRA) can modify these models for particular use cases, making them more diverse and widely usable. At the same time, open base models may be adapted to deliver biased, harmful or illegal images in an online service. In this paper, we develop a protocol to detect when a service may be using an adaptation of an otherwise standard model. We describe the precise setup and threat model for this provenance question and define two different problems arising due the presence of multiple model families and multiple adaptations. In contrast to previous works that have considered model identification within lists of known candidate models, we consider the realistic open world scenario where the set of possible adaptations and alternative models are unknown. Using a hypothesis testing approach, our protocols can distinguish between the base model, a LoRA adaptation and other completely different families of models. For this test, we develop a black box signature of the models based only on intrinsic dimension of sets of images generated by the models.
[LG-37] Hyperparameter and Neural Architecture Search via Evolutionary Optimization for Reinforcement Learning: HPNAS-RL
链接: https://arxiv.org/abs/2610.36882
作者: Nihal Acharya Adde,Hanno Gottschalk
类目: Machine Learning (cs.LG)
*备注:
Abstract:In this study, we propose a dual-branch evolutionary framework that simultaneously performs Neural Architecture Search (NAS) and Hyperparameter Optimization (HPO) for Reinforcement Learning (RL) in Autonomous Driving (AD) tasks. While NAS has shown strong potential in automating architecture design for Convolutional Neural Networks (CNNs), its application to RL remains limited, especially in scenarios that require joint optimization of architectural and training dynamics. Building on our prior work on Evolutionary Multi-Objective NAS (EMNAS), the proposed framework evolves discrete network architectures and continuous hyperparameters in parallel under a unified multi-objective search strategy. In this setup, NAS targets the convolutional layers of the RL policy network - responsible for visual feature extraction, while HPO focuses on optimizing critical RL hyperparameters. To accelerate convergence and improve generalization, we incorporate parallel evaluation and a teacher-student policy transfer mechanism. Experiments in a high-fidelity driving simulator show that the joint NAS-HPO framework consistently outperforms standalone NAS, yielding policies with higher rewards, fewer parameters, and improved stability. These results highlight the benefits of jointly optimizing model structure and training dynamics for scalable and effective RL agent design.
[LG-38] Sample-and-Score: Surrogate-Guided Coordinated Spatial Reuse in IEEE 802.11 Networks
链接: https://arxiv.org/abs/2610.36868
作者: Maksymilian Wojnar
类目: Networking and Internet Architecture (cs.NI); Machine Learning (cs.LG)
*备注:
Abstract:Coordinated spatial reuse (Co-SR) in IEEE 802.11bn increases area throughput through simultaneous transmissions, yet the combinatorial configuration space makes real-time optimization challenging. We propose a sample-and-score approach: draw random Co-SR configurations, score each with a surrogate model, and select the best. The surrogate is trained offline and deployed without online fine-tuning or sequential exploration, yet performs close to multi-armed bandit and heuristic baselines that need multiple interaction steps or extensive measurements. On unseen topologies with 2-16 access points, it recovers 52-77% of a perfect surrogate’s throughput. A perfect sample-and-score approach reaches 95% of the upper bound throughput with 8192 candidates.
[LG-39] Variable-Length Feedback Coding With Universal Decoding and Bandit Channel Selection
链接: https://arxiv.org/abs/2610.36818
作者: Recep Can Yavas
类目: Information Theory (cs.IT); Machine Learning (cs.LG)
*备注: 34 pages, submitted to Transactions on Information Theory
Abstract:We study variable-length feedback (VLF) coding over K additive white Gaussian noise (AWGN) channels whose noise variances are unknown to both terminals but lie in a known interval. The receiver selects one channel at each time, conveys the selection over noiseless feedback, and must learn the channel qualities while decoding, to avoid overusing suboptimal channels. This setting, which we call the VLF-Bandit problem, combines VLF coding, universal decoding, and bandit channel selection. We develop two joint coding and channel-selection schemes. The VLSF-Bandit scheme is a single-phase variable-length stop-feedback (VLSF) scheme based on a Gaussian random codebook, a score formed from a ratio of likelihood mixtures, and the asymptotically optimal \textkl-UCB^++ bandit algorithm applied to the received power. The VLF-Bandit scheme sends the message in packets by Schalkwijk–Kailath transmission, uses its normalized outputs as noise samples to learn the variances, and lets its exploration packets carry message bits. We prove second-order achievability bounds for both schemes, which are first-order optimal. In the logarithm of the achievable codebook size, the magnitude of the \log N coefficient, with N the average decoding time, is a baseline of one plus a channel-exploration term for VLSF-Bandit. For VLF-Bandit, this magnitude is smaller, with a universal-decoding term replacing the baseline and a smaller channel-exploration term. The VLF-Bandit bound improves on the universal VLF bound of Yavas and Tan (2025) for K=1 and on the bound of an explore-then-commit strategy whenever K\geq2 .
[LG-40] LLM knowledge for improved Out-of-Distribution Tabular Prediction - A Cross-Validated Regularisation Approach
链接: https://arxiv.org/abs/2610.36806
作者: Sølve Langseth,Kristoffer Herland Hellton,Hugo L. Hammer
类目: Machine Learning (cs.LG); Methodology (stat.ME); Machine Learning (stat.ML)
*备注:
Abstract:Prediction models degrade when the deployment population differs from the one used to train them, and training data cannot say which feature-response relationships will transfer. Expert elicitation is slow and unscalable. We ask whether coefficient priors elicited from large language models (LLMs) improve out-of-distribution (OOD) prediction as the shrinkage target of a regularised regression instead of zero, across five tabular datasets split on real covariates, selecting providers and prior structures by source-domain cross-validation alone. The priors help on four of five datasets, by up to 7.9% OOD mean squared error, and ten of twenty-four prior-informed results resolve the mean improvement at conventional confidence; single-deployment intervals are much wider. On classification the gain is in probabilistic accuracy, not ranking, and control tests with corrupted priors attribute the benefit to feature-specific content rather than scale. Two findings temper this. Variation across candidate priors exceeds the average benefit, yet cross-validation margins fall below their own noise, so selection avoids the worst prior rather than finding the best. On three datasets the prior improves in-domain accuracy as much as out-of-domain, making it better regularisation than shift correction. Only where the training data are large enough that the in-domain gain disappears does elicited knowledge demonstrably improve transfer. Code, priors and results: this https URL.
[LG-41] Does LLM Pre-Training Typically Occur at the Edge of Stability?
链接: https://arxiv.org/abs/2610.36778
作者: Yuhang Cai,Haodong Wen,Haofeng Huang,Deyi Liu,Yiyuan Ma,Kaifeng Lyu
类目: Machine Learning (cs.LG)
*备注: 90 pages, 17 figures
Abstract:Quadratic approximations provide a common lens for studying neural network optimization, but recent evidence challenges their predictive validity. In full-batch gradient descent with learning rate (LR) \eta , Cohen et al. (2021) observed the Edge of Stability (EoS), where the largest Hessian eigenvalue concentrates near 2/\eta , seemingly at odds with classical stability conditions predicted by quadratic approximations. In this work, we revisit the fidelity of quadratic approximations as a model of neural network training dynamics, with a particular focus on their failure modes in LLM training. We first highlight a distinct failure mode of the quadratic approximation, where persistent negative curvature causes the quadratic dynamics to diverge regardless of the LR, even though the real dynamics remain stable. We term this phenomenon the Edge of Convexity (EoC). After removing the effect of this LR-independent instability, we then propose a generalized definition of Edge of Stability (EoS) for stochastic training with adaptive optimizers, based on whether increasing the LR induces significant divergence in the quadratic dynamics while leaving the real dynamics largely unchanged. Across different LLM pretraining runs with various model sizes up to 787M, we find that (1) EoC is always observed across LLM pretraining, and (2) EoS is also prevalent but not universal: it disappears when the LR becomes sufficiently small (e.g., after decay) or when the batch size falls below a critical threshold that has a positive correlation with the critical batch size. Together, these findings characterize when and how quadratic approximations fail and provide a foundation for future work on understanding training dynamics.
[LG-42] Nonlinear Direct Feedback Alignment for Scalable Backpropagation-Free Training
链接: https://arxiv.org/abs/2610.36718
作者: Wu Lin,Yan Fan,Chenxiang Ma,Yinglan Feng,Jia Wang,Ka-Chun Wong,Qiuzhen Lin
类目: Machine Learning (cs.LG)
*备注:
Abstract:Backpropagation (BP) remains the predominant algorithm for training deep neural networks, but it is biologically implausible and suffers from the backward locking problem, constraining parallelization across layers. Direct Feedback Alignment~(DFA), which propagates the global output error directly to each hidden layer via fixed feedback matrices, has emerged as a promising alternative for addressing these limitations. However, a significant performance gap persists between DFA and BP, which we attribute to the limited ability of fixed feedback matrices to capture the nonlinear transformation from hidden representations to the network output. To address this limitation, we propose Nonlinear Direct Feedback Alignment (NDFA), which constructs each hidden-layer feedback using a lightweight nonlinear surrogate subnetwork. To enhance the effectiveness of the feedback throughout training, we further propose a learning strategy to update the surrogate subnetwork, thereby better approximating the underlying nonlinear transformation. Extensive experiments across residual networks and vision transformers on CIFAR-10, CIFAR-100, SVHN, and ImageNet datasets demonstrate that NDFA consistently outperforms existing DFA-based methods and substantially narrows the performance gap to BP, while maintaining efficient parallel training. The code is available at this https URL.
[LG-43] SMEBENCH: Benchmarking Neural Music Editing with Synthetized MIDI Edits
链接: https://arxiv.org/abs/2610.36715
作者: Matthieu Cervera,Francesco Paissan,Micael Antunes,Mirco Ravanelli,Ricard Marxer,Cem Subakan
类目: ound (cs.SD); Machine Learning (cs.LG); Audio and Speech Processing (eess.AS)
*备注:
Abstract:Neural music editing requires a system to perform a requested change while preserving the musical content that should remain unchanged. Existing evaluations often lack a paired target edit, making these requirements difficult to measure separately. We introduce SMEBench, a large-scale benchmark of controlled music edits constructed by modifying symbolic MIDI performances and rendering matched source and target audio. Each example contains a source, an instruction, a paired target, edit metadata, and complementary descriptions. The benchmark covers twelve instrumentation, articulation, dynamics, and pitch operations and includes a balanced 3,600-example test set together with a smaller real-audio complement derived from multitrack recordings. The paired targets allow source preservation, target agreement, and textual compatibility to be evaluated independently. We use these references to diagnose common audio-editing metrics and benchmark six representative editing systems, showing that the metric used to rank a system can materially change the conclusion. The dataset is available at our repository.
[LG-44] RankFormer: Rethinking Time Series Transformers through Representation Capacity Analysis and Progressive Decoding
链接: https://arxiv.org/abs/2610.36703
作者: Peiwang Tang,Cheng HE,Xijie Liang,Defu Lian
类目: Machine Learning (cs.LG)
*备注: 12 pages
Abstract:Multivariate time series forecasting plays a critical role in Web and cyber-physical systems, where large-scale services and IoT infrastructures generate complex temporal data. Transformer-based models have achieved strong empirical performance in this domain, yet the principles governing the expressive capacity of their learned representations remain insufficiently understood. Moreover, many existing models rely on simplified decoding strategies, including channel-independent modeling and direct projection heads, which limit their ability to capture complex multivariate dynamics. To address these challenges, we propose RankFormer, a novel encoder-decoder architecture for multivariate time series forecasting. We establish a rank-based analysis framework to investigate how tokenization, embedding, and attention designs affect the representation capacity of time series Transformers. Our analysis reveals that patchification increases the potential representation capacity by aggregating multiple temporal observations into each token, while subsequent encoding operations determine how much of this capacity is realized. Guided by these insights, we develop a rank-enhancing encoder with Hourglass Attention to improve rank stability during information interaction. Furthermore, we introduce the Interactive Progressive Decoder (IPD), which improves future trajectory modeling through global-guided cross-channel interaction and progressive refinement. Extensive experiments on real-world long-term forecasting benchmarks demonstrate that RankFormer achieves competitive performance against state-of-the-art methods. Code repository: this https URL.
[LG-45] Quantifying and Prioritizing Stealthy Source-Manipulation Threats Against Multi-Source Anomaly Detectors ACSA
链接: https://arxiv.org/abs/2610.36693
作者: Indraveni Chebolu,Seema Panghal,Arnab Mallick,Ch A S Murty
类目: Cryptography and Security (cs.CR); Machine Learning (cs.LG)
*备注: To be appeared at Workshop on AI for Cyber Threat Intelligence (WAITI) 2026 @ ACSAC 2026
Abstract:Security pipelines often fuse several data sources and protect a downstream model with a cross-source anomaly detector. Such detectors are usually judged by their clean false-positive rate and by how well they catch random corruption. Neither metric captures an adversary who controls some sources and targets a specific downstream decision. We propose a threat-quantification score, the \emphtask-conditioned visibility margin, defined as the least detector evidence a manipulation must reveal per unit of harm to the downstream task. The margin has a closed form, takes about 85,\mu s per certificate, yields a minimum-evidence attack, and identifies task-relevant directions the detector cannot see. On a 17-channel industrial telemetry benchmark, a detector that flags 82.8% of random single-channel corruptions is first-order blind on 64.7% of task/subset pairs in an audit of one- to four-channel subsets. Task-conditioned manipulations cause more than twice as many falsely reassuring errors as generic stealthy ones, and more of them evade the alarm on four of five splits, mainly at small budgets. Used to prioritize audits on held-out subsets, the margin finds a median of four verified failures in ten checks, compared with none for random order and two for alarm-based ranking, and performs comparably to the task-gradient baseline. It also ranks vulnerable configurations of the learned GDN and TranAD detectors and of a five-source crop-yield panel.
[LG-46] From Biological Curves to Soft Modular Robots: Learning-Based Inverse Design with Uncertainty-Guided Refinement
链接: https://arxiv.org/abs/2610.36681
作者: Dixiang Wang,Minghao Zha,Arifur Rahman,Yinhong Qin,Yitao Jiang,Shiyang Lu,Zebing Mao,Muhao Chen,Luyang Zhao
类目: Robotics (cs.RO); Machine Learning (cs.LG)
*备注: 8 pages, 8 figures, 2 tables
Abstract:Animals exhibit diverse shape-changing behaviors, from the undulatory swimming of eels to the arching movements of inchworms. Reproducing these biological shapes with modular soft robots offers a promising approach to developing versatile bioinspired robotic systems. However, translating a desired shape into a feasible robotic configuration remains challenging. Existing cable-driven modular soft robots rely on computationally expensive physics-based modeling and combinatorial search for inverse design, making it difficult to efficiently reproduce diverse target shapes. To address this challenge, we present a learning-based real-to-sim-to-real framework that translates biological curves into robotic configurations through surrogate-assisted inverse design and uncertainty-guided refinement. The proposed method combines learned inverse prediction with targeted candidate exploration, using a fast forward surrogate to evaluate and select suitable designs. The learned forward surrogate reduces deformation prediction time from 1.53 \times 10^5 ms to 0.0008 ms, achieving approximately a 1.9 \times 10^8 -fold speedup. At the same time, the proposed inverse design method reduces design time from 1.99 \times 10^6 ms to 4.18 ms. On 5,000 test samples generated from previously unseen threading patterns, the method achieves a 93% shape reconstruction success rate under surrogate-based evaluation. Physical demonstrations with four bioinspired robots further illustrate the translation of biological curves into realizable robotic shapes.
[LG-47] Impossibility of No-Regret in Multi-Agent Bidding
链接: https://arxiv.org/abs/2610.36666
作者: Zongjun Yang,Rachitesh Kumar,Christian Kroer
类目: Computer Science and Game Theory (cs.GT); Computational Complexity (cs.CC); Machine Learning (cs.LG)
*备注:
Abstract:We consider T repeated second-price auctions across n pacing bidders with i.i.d. item arrivals, and show that it is computationally impossible for all bidders to achieve no-regret simultaneously while satisfying their respective budget or return-on-investment constraints. Therefore, in the prominent internet auction systems where buyers simultaneously deploy pacing algorithms or automated bidding agents, we cannot hope to achieve joint no-regret guarantees in general. We establish a deterministic relationship that high variance in online bidding systems implies high regret. By showing that one can extract a pacing equilibrium from a low-variance system, we prove PPAD-hardness of attaining joint no-regret. For several widely used bidding algorithms, we also show evidence from dynamical systems and numerical simulations that they can indeed display high-variance behavior through cycling, even on simple market instances. Subjects: Computer Science and Game Theory (cs.GT); Computational Complexity (cs.CC); Machine Learning (cs.LG) Cite as: arXiv:2610.36666 [cs.GT] (or arXiv:2610.36666v1 [cs.GT] for this version) https://doi.org/10.48550/arXiv.2610.36666 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-48] Log-Derivative Trick for Parametric Markov Chains
链接: https://arxiv.org/abs/2610.36655
作者: Arthur Gall,Ichiro Hasuo,Thomas Seiller
类目: Machine Learning (cs.LG); Logic in Computer Science (cs.LO)
*备注:
Abstract:Optimizing expected rewards and reachability probabilities in parametric Markov chains (pMCs) is central to the analysis and synthesis of probabilistic systems with loops. While traditional symbolic synthesis approaches–such as parameter lifting and rational function generation–suffer from scalability bottlenecks, gradient-based optimization emerges as an effective alternative. However, existing gradient-based techniques still rely on symbolic gradient computation–such as solving large linear systems–over the global state space, inheriting scalability limits. Finite-horizon approximation is possible but introduces uncontrollable bias. In this paper, we introduce a scalable Monte Carlo gradient estimator based on a measure-theoretic generalization of the log-derivative trick. We prove that our estimator provides unbiased gradient estimates under almost-sure termination and standard regularity conditions. Crucially, our method avoids global state-space materialization and linear system solving entirely: it operates purely along sampled trajectories, requiring only black-box evaluations of local transition probabilities (or densities) and their derivatives. Unlike existing pMC techniques that are fundamentally restricted to finite discrete systems, our measure-theoretic framework naturally extends to continuous and general \sigma -finite state spaces. Empirical evaluations on synthetic discrete pMCs and continuous-state case studies demonstrate that our method scales to state spaces far beyond the reach of existing solvers, significantly reducing execution time without introducing bias. Subjects: Machine Learning (cs.LG); Logic in Computer Science (cs.LO) Cite as: arXiv:2610.36655 [cs.LG] (or arXiv:2610.36655v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2610.36655 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-49] ACO-Graph: Automated Construction of Text-Atributed Graphs
链接: https://arxiv.org/abs/2610.36650
作者: Amiri Hayes,Kristina Wicke
类目: Machine Learning (cs.LG)
*备注:
Abstract:Graph-structured representations of text data have become foundational across machine learning, yet it remains unclear how to systematically construct a graph from a given dataset for a target task. Choices about which entities become nodes, which relationships become edges, and what text each node contains can substantially affect downstream performance. In practice, finding optimal constructions often requires expert knowledge or manual trial and error. We introduce TACO-Graph, a decision tree framework for learning which graph construction choices are most effective for a given dataset and task. Specifically, we focus on text-attributed graphs (TAGs) constructed from multi-attribute tabular data. TACO-Graph enumerates valid TAG representations, scores each using fast proxy evaluation on small representative subsets, and trains an interpretable decision tree on those scores. Across five datasets spanning citation networks, product reviews, and bibliographic relationships, our decision trees predict construction quality on separately drawn evaluation samples with mean Spearman \rho above 0.9 for node classification and above 0.8 for GraphRAG retrieval. Our trees systematically identify which constructions perform better across both GNN and GraphRAG tasks, and the full pipeline is released as an open-source library to support evaluation on new tasks and datasets at this https URL. TACO-Graph provides a systematic and interpretable guide for determining how to convert tabular data to text-attributed graphs for downstream tasks.
[LG-50] When Is a Cheap Fairness Edit Safe? Screening Post-Hoc Edits in Two-Sided Matching Markets
链接: https://arxiv.org/abs/2610.36649
作者: Moirangthem Tiken Singha,Rabinder Kumar Prasada
类目: Machine Learning (cs.LG)
*备注:
Abstract:Two-sided matching systems, such as admissions pipelines, allocate opportunities jointly. Reducing a protected at tribute’s influence usually requires encoder retraining. We study a closed-form linear projector that does this at inference time. For markets that stay vertically ordered before and after the edit, we bound the mean welfare change by the edit-induced appeal shift and an explicit remainder. A probabilistic refinement replaces worst-case constants with a data-dependent deviation term. A pre-deployment screen combines a verticality test with a welfare sweep and returns PASS, KILL, or OUT-OF-SCOPE. An exact tier requires every opportunity to rank applicants by appeal, and a graded tier admits near-vertical markets. On 30 held-out synthetic markets, the screen agrees with every pre-specified regime verdict, with no observed false PASS. On 60 exactly vertical markets, no bound is violated, and the probabilistic bound is informative in 69% to 87% of configurations. On 2700 configurations from three realistic markets, which do not reach the exact tier, no plug-in bound expression is violated. A closed-form factorization stays within graded scope in 90% to 100% of instances, whereas two-tower and graph neural network encoders fall outside it in every instance. Architecture choice thus appears to affect whether a post-hoc edit is admissible.
[LG-51] When Does Perturbation Consistency Add Information Beyond Softmax Confidence? Theory and Evidence for the Decision Consistency Framework
链接: https://arxiv.org/abs/2610.36640
作者: Amit Kalitaa,Moirangthem Tiken Singh
类目: Machine Learning (cs.LG)
*备注:
Abstract:Perturbation consistency scores judge whether a prediction survives small input changes. The Decision Consistency Framework (DCF) scores each image by label stability (LS) and confidence deviation (CD) under six mild operators, combines them into a consistency score (CS), and summarises a model by the gap between accuracy and mean consistency (CAG). We ask when these scores carry information that the clean softmax output lacks. Across 201 evaluations with three seeds per configuration, CS detected misclassifications on clean data less well than the maximum softmax probability (MSP) in every model family. For frozen CNNs, the deficit was 0.057 in area under the ROC curve (AUROC; 95% CI 0.043 to 0.073), at 8 to 15 times the compute. A label flip requires the logit margin to be at most the square root of two times the logit displacement, and at a fixed margin, the displacement predicted errors at AUROC 0.515. The same inequality gives a label-free split-conformal no-flip certificate, never violated in 84 runs. Consistency did add information where confidence is misled: under noise corruptions of CIFAR-10-C, it beat the margin, and under projected gradient attacks, it separated clean from attacked inputs with AUROC 0.83 against 0.37 for MSP at a budget of 2/255, as pre-specified. The gain under noise did not transfer to unseen shift families. The large negative CAG on OCTMNIST persists under full fine-tuning and in-domain training and is partly a class-prior effect: correcting the prior recovers about 40% of the gap.
[LG-52] Deep Bayesian Approaches for Interpretable Nonlinear Structural System Identification
链接: https://arxiv.org/abs/2610.36633
作者: Wei Liu,Konstantinos Vlachas,Eleni Chatzi
类目: Machine Learning (cs.LG); Computational Engineering, Finance, and Science (cs.CE); Machine Learning (stat.ML)
*备注:
Abstract:System identification, namely the inference of mathematical representations of dynamical systems from measured data, is fundamental to engineering diagnostics, prognostics, and decision-making. Bayesian filtering offers a principled framework for probabilistic state estimation but traditionally relies on prescribed transition and observation models. Recent advances in deep Bayesian filtering address this limitation by embedding learnable representations within probabilistic state-space formulations, enabling data-driven identification of unknown dynamics while retaining physical structure and uncertainty-aware inference. This tutorial provides a systematic introduction to deep Bayesian filtering for nonlinear structural system identification, focusing on the interplay between learned dynamics, physical inductive biases, and posterior inference. The Deep Markov Model serves as a data-driven reference, while the Physics-guided Deep Markov Model (PgDMM) and Neural Extended Kalman Filter (Neural EKF) illustrate complementary approaches based on amortized variational inference and differentiable Kalman filtering, respectively. Their capabilities are examined through dynamic state estimation and joint dynamic state and parameter estimation on a three-dimensional shear-frame benchmark with hysteretic Bouc-Wen connections and sparse acceleration measurements. The results highlight important distinctions between measurement-conditioned reconstruction and open-loop forecasting, demonstrate how physical structure can improve latent-state interpretability without necessarily ensuring predictive fidelity, and reveal differences in the robustness of response and parameter estimation. Practical guidelines for model construction, training, evaluation, and deployment are provided, together with an outlook on remaining challenges in uncertainty calibration, identifiability, and real-world validation. Subjects: Machine Learning (cs.LG); Computational Engineering, Finance, and Science (cs.CE); Machine Learning (stat.ML) Cite as: arXiv:2610.36633 [cs.LG] (or arXiv:2610.36633v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2610.36633 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-53] he Batch Size Can Be One: Test-Time Adaptation from Classifier Geometry
链接: https://arxiv.org/abs/2610.36628
作者: Jaeho Hwang,Yujin Oh,Juhyeon Shin,Juhye Lee,Sungroh Yoon,Uiwon Hwang
类目: Machine Learning (cs.LG)
*备注:
Abstract:Test-time adaptation updates a deployed model on an unlabeled test stream. In practice this stream often arrives one sample at a time, with the classes in no particular order. A single sample then provides limited information for estimating distribution-level statistics, and selecting samples by confidence can admit predictions that are wrong yet confident. A corruption, however, shifts the features while leaving the decision regions of the frozen classifier unchanged. We therefore propose Origin-recentered Test-Time Adaptation (OriTTA), which constructs fixed geometric targets from the source model and adapts selected features toward them. Seen from the origin of the feature space, the classes lean toward a common direction. We choose a new origin to bring class scores closer together while keeping it near the mean feature of one synthetic prototype per class. From this origin, prototype directions spread more evenly, and each class obtains an axis and an ideal point before any test data arrives. During adaptation, OriTTA aligns selected features with their predicted class axis and penalizes excessive radius, keeps one input per predicted class in memory to limit feature changes for classes absent from the stream, and resets the model when re-classifying this memory reveals collapse. Without tuning for each setting, OriTTA achieves the highest mean accuracy at batch size 1 on CIFAR-10-C, CIFAR-100-C and ImageNet-C in all sixteen combinations of four backbones, two protocols and two stream orders. On ImageNet-C fog, recent methods collapse in up to two thirds of one hundred runs, while OriTTA never does.
[LG-54] Normalized Cuts and Data Clustering via Matrix Fractional Programming
链接: https://arxiv.org/abs/2610.36616
作者: Yannan Chen,Beichen Huang,Licheng Zhao,Chris Ding,Kaiming Shen
类目: Machine Learning (cs.LG); Signal Processing (eess.SP)
*备注: 15 pages, 15 figures
Abstract:The normalized cut (NCut) problem is a fundamental one in the area of data clustering. Despite its fractional structure, the NCut problem is seldom considered from a fractional programming (FP) perspective in the existing literature. The main reason is that the conventional FP method is inadequate for this problem case: the classical Dinkelbach’s transform can only handle a single ratio and hence is limited to the two-class clustering, whereas the multi-class clustering involves multiple ratio terms. This work proposes a novel so-called multidimensional quadratic transform that is capable of dealing with the multi-ratio case; it recasts the NCut problem into a bipartite matching problem—which can be immediately solved by linear search. Furthermore, we explore the connection between the proposed FP method and the minorization-maximization (MM) theory, and in turn characterize the convergence behavior of our algorithm. Finally, we conduct experiments to numerically demonstrate the advantages of the proposed FP-based algorithm in a variety of data clustering and image segmentation tasks.
[LG-55] Efficient Quantization Training with Stochastic Rounding via Structured Subspace Randomness
链接: https://arxiv.org/abs/2610.36587
作者: Zeyu Wang,Yiheng Duan,Tao Chen,Donghyun Koo,Sihan Liu,Wenbo Zhang,Chuan He,Liang Liu
类目: Machine Learning (cs.LG)
*备注:
Abstract:Low-precision training often relies on stochastic rounding (SR) to keep quantized gradients unbiased and prevent small updates from rounding to zero. Standard SR draws an independent random threshold for every gradient element. However, accelerators generate random numbers on vector units whose throughput lags far behind their matrix units. Drawing one threshold per element saturates the vector unit and can make INT8 training with SR slower than BF16 training. Sharing random thresholds across tensor elements removes this cost but forces those elements to round in the same direction and shifts their rounding errors by a common offset. We find that this offset degrades training primarily through the activation gradient, which propagates the error backward into earlier layers. Because earlier layers cannot cancel an incoming offset using their own local rounding errors, the offset must be removed at the layer where it originates. We propose Structured Subspace Stochastic Rounding (S3R), which quantizes each upstream gradient once for both the activation and weight gradient matrix multiplications and shares a single threshold per token across the hidden dimension. Its error correction, S3R-EC, subtracts the structured mean rounding error from the activation and weight gradients while leaving the optimizer unchanged. In LLM pre-training experiments, uncorrected threshold sharing degrades convergence, whereas S3R-EC eliminates this degradation and matches the validation loss of independent SR. Furthermore, S3R-EC accelerates full INT8 training steps of dense LLMs on TPU v5p by 1.24x to 1.32x over BF16 and by 1.11x over independent SR.
[LG-56] Learning from Points and Patches: A Pre-trained Hybrid Attention Transformer for Time Series Analysis
链接: https://arxiv.org/abs/2610.36572
作者: Peiwang Tang,Cheng HE,Yupeng Ye,Weitai Zhang,Qinghua Sun,Defu Lian
类目: Machine Learning (cs.LG)
*备注: 12 pages. Code: this https URL
Abstract:Time-series data generated by Web services and connected systems support applications such as user activity analysis, service monitoring, and resource management. However, learning transferable representations from heterogeneous time series remains challenging due to diverse temporal patterns, variable sequence lengths, and limited high-quality annotations. Despite the success of pre-training in natural language processing and computer vision, effective time-series pre-training requires strategies tailored to temporal data. We propose TimeHAT (Hybrid Attention Transformer), a general-purpose pre-training framework for time-series analysis that addresses two complementary questions: what to learn and how to learn. For what to learn, we introduce a Unified Point-Patch Masking Strategy (UPPMS) that integrates point-level and patch-level supervision within a unified framework for contextual reconstruction and causal extrapolation, enabling the model to capture both fine-grained temporal variations and broader temporal patterns. For how to learn, we develop a Hybrid Attention Encoder (HAE) that combines prefix-constrained bidirectional attention in initial layers with directional hybrid attention in deeper layers, aiming to preserve effective contextual interactions while mitigating rank degeneration. We evaluate TimeHAT across real-world datasets and four downstream tasks: classification, long-term forecasting, imputation, and anomaly detection. Experiments cover fine-tuning, few-shot, and zero-shot transfer settings, demonstrating competitive performance across diverse time-series analysis tasks. Code is available at this https URL. Comments: 12 pages. Code: this https URL Subjects: Machine Learning (cs.LG) Cite as: arXiv:2610.36572 [cs.LG] (or arXiv:2610.36572v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2610.36572 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-57] Holomorphic Cauchy Networks for Efficient Stationary Tail Computation in High-Dimensional Reflected Brownian Motions
链接: https://arxiv.org/abs/2610.36558
作者: Kehan Wang
类目: Machine Learning (cs.LG); Numerical Analysis (math.NA); Probability (math.PR)
*备注: 29 pages, 3 figures, 14 tables. Code available at this https URL
Abstract:Stationary tail computation for high-dimensional reflected Brownian motions (RBMs) remains computationally challenging. Building on NN4MGF’s shared-coordinate basic-adjoint-relationship (BAR) framework, we develop Holomorphic Cauchy Networks (HCNs). Native complex-valued rational Cauchy subnetworks enforce holomorphicity on pole-free domains, removing sampled Cauchy-Riemann penalties and their input-derivative graphs. Exact origin anchoring and boundary-coordinate exclusion enforce structural identities, while anchored interactions represent dependent transforms. Under holomorphic-log assumptions, we establish simultaneous value and first-derivative approximation for the constrained architectures and a mixed-derivative lower bound for coordinate-only representations under nonzero cross-covariance. Closed-form real-axis derivatives and sequential backpropagation further control memory. On the same NVIDIA H100 NVL, the tested 30D HCN configuration uses approximately tenfold less peak allocated memory and has 9.7-fold lower mean update time than NN4MGF; extrapolation to the respective scheduled update counts gives a projected full-schedule update-time ratio of 58.2. With 64 million sampled base inputs per high-dimensional run, HCN reaches 50 dimensions on a laptop GPU with 6.367% mean relative error over thresholds 1 through 8. Degree-matched comparisons yield lower reported mean relative errors over the primary evaluation ranges, including 1.422% versus 6.684% on the 2D Harrison non-product benchmark. An exploratory 10D correlated example illustrates the interaction extension.
[LG-58] Underdamped Anchored Langevin Dynamics
链接: https://arxiv.org/abs/2610.36470
作者: Qi Feng,Hoang M. Nguyen,Lingjiong Zhu
类目: Machine Learning (cs.LG); Probability (math.PR)
*备注: 66 pages
Abstract:Classical Langevin Monte Carlo algorithms, such as the unadjusted Langevin algorithm (ULA), are based on discretization schemes of the overdamped Langevin dynamics and are widely used for sampling in machine learning because they scale to high dimensions and large datasets. However, they require differentiable log-densities, and are less efficient for sampling heavy-tailed distributions. Although various approaches have been proposed in the literature for sampling non-differentiable and heavy-tailed targets, all of them are developed in the overdamped setting only. We propose a novel momentum-based underdamped anchored Langevin dynamics (UALD), and underdamped anchored Langevin Monte Carlo (UALMC) algorithm based on a discretization scheme of UALD. Our paper is the first to provide a kinetic (momentum-based) Langevin algorithm for non-differentiable and heavy-tailed targets that is free of asymptotic bias. The method replaces the original potential with a smooth reference potential and modifies the underdamped Langevin dynamics via multiplicative scaling in the drift term. We establish non-asymptotic guarantees in the 2-Wasserstein distance to the target distribution and show that the resulting iteration complexity has better dependence on dimension and accuracy level compared to overdamped-based anchored Langevin algorithms. We provide numerical experiments to illustrate the theory and practical performance of our proposed approach.
[LG-59] CoHARL: Co-Evolving Agent Policies and Executable Harnesses with Reinforcement Learning
链接: https://arxiv.org/abs/2610.36469
作者: Jianwen Chen,Peng Xia,Kaide Zeng,Siwei Han,Zhen-Yu Zhang,Charles Fleming,Xinyu Ye,Huaxiu Yao
类目: Machine Learning (cs.LG)
*备注:
Abstract:Large language model (LLM) agents interact with their environments through executable harnesses that govern tool use and feedback. Optimizing a harness for a fixed policy overlooks how its behavior changes during learning, while training a policy under a fixed harness leaves the interface that generates its training experience unchanged. This motivates adapting the harness throughout policy training. We introduce CoHARL, a population-based reinforcement learning framework that co-evolves a shared policy and a population of executable harnesses. The policy learns from rollouts across the population, and a proposer uses its execution traces to revise harness code during training. CoHARL achieves a 27.6% mean resolved rate on SWE-rebench, improving over harness-only search by 6.4 percentage points. Experiments on SWE-bench Verified and Terminal-Bench 2.0 show gains in software engineering and terminal-based tasks. Under matched training budgets, CoHARL also outperforms searching for a harness once before RL and alternating policy updates with revisions to a single active harness.
[LG-60] Split Federated Unlearning
链接: https://arxiv.org/abs/2610.36465
作者: Jia Chen,Yecheng Lv,Xinrong Hu,Mingdi Huang,Jun Bai
类目: Machine Learning (cs.LG); Cryptography and Security (cs.CR)
*备注: 13 pages
Abstract:Split federated learning (SFL) combines partitioned model execution with federated aggregation to support collaborative learning across distributed Web and edge environments. However, removing learned contributions in SFL is challenging due to cross-partition coupling and historical influence propagation through dual-side aggregation. To the best of our knowledge, we provide the first systematic formulation of Split Federated Unlearning (SFU), encompassing client-wise and class-wise removal, counterfactual retraining, and partition-specific update constraints. We theoretically characterize historical influence propagation and mode-dependent unlearning feasibility, revealing structural limitations of restricted model updates. As an initial solution, we propose SURE, integrating dual-importance estimation, selective parameter perturbation, and importance-guided recovery. Extensive experiments with existing unlearning methods adapted to SFL reveal substantial forgetting-utility trade-offs across model architectures, data distributions, and update modes. SURE achieves competitive agreement with mode-matched retraining at modest additional cost, while further analyses highlight the impact of split position and update constraints on unlearning behavior. These findings establish SFU as a distinct research problem and provide a foundation for unlearning in partitioned federated systems.
[LG-61] A Simple Presentation of Robust Membership Inference Attacks
链接: https://arxiv.org/abs/2610.36458
作者: Renaud Gaucher,Ephraim Linder,Adam Smith,Jonathan Ullman,Lydia Zakynthinou
类目: Cryptography and Security (cs.CR); Information Theory (cs.IT); Machine Learning (cs.LG); Statistics Theory (math.ST)
*备注: 34 pages, 2 figures
Abstract:In a membership inference attack, an attacker is given the data of a particular individual, and tries to determine if that individual was included in a particular dataset, given some kind of statistics derived from the dataset or a model trained on the dataset. Roughly, a membership inference attack is robust if it succeeds even when the statistics are highly noisy, regardless of how that noise is introduced. We give a simple presentation of robust membership inference attacks for mean estimation, covering both multivariate Gaussian distributions and binary product distributions. We also give a summary of the literature on robust membership inference, covering history, extensions, and connections to other research areas.
[LG-62] Whole-Body Humanoid Control via Stochastic Motion Tokenization
链接: https://arxiv.org/abs/2610.36452
作者: Michael Drolet,Oleg Arenz,Jan Peters
类目: Robotics (cs.RO); Machine Learning (cs.LG)
*备注: Accepted to the 2026 IEEE-RAS International Conference on Humanoid Robots (Humanoids 2026). © 2026 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses
Abstract:While velocity-commanded RL policies are effective for controlling quadrupeds, they are a limited interface for humanoids that need to perform complex whole-body motions. Furthermore, RL tracking policies for whole-body motions typically have limited information about the desired future motion, preventing them from performing controlled deviations to achieve long-term task success; such information is particularly relevant when a controller must prepare for transitions between behaviors. We address these challenges using concise motion representations that can be generated by a high-level controller, given a behavior label or text command. These representations can be fed directly into a tracking policy to provide a detailed motion reference. Our method exhibits a high degree of flexibility in performing real-world robot motions such as dancing, fighting, or walking based on conditional information. At matched capacity, our adopted stochastic motion tokenizer reconstructs motions from the Unitree G1 at a level that exceeds state-of-the-art approaches such as finite scalar quantization (FSQ) and the vector-quantized autoencoder (VQ-VAE). The resulting discrete interface conditions whole-body tracking policies that execute generated dance, fight, and walk motions lasting 16.8 s each and a 35 s fight-to-dance sequence on a physical Unitree G1.
[LG-63] High-Fidelity Aerostructural Dataset of Transonic Wings in Cruise Conditions
链接: https://arxiv.org/abs/2610.36442
作者: Pavankumar Koratikere,Abhijnan Dikshit,Leifur Leifsson
类目: Machine Learning (cs.LG); Computational Engineering, Finance, and Science (cs.CE); Fluid Dynamics (physics.flu-dyn)
*备注: The dataset is available at: \url{ this https URL }
Abstract:Recent advances in machine learning (ML) have the potential to substantially accelerate aircraft wing design, a process that has traditionally required extensive high-fidelity simulation and design iteration. Existing public datasets for ML-based wing design have focused primarily on aerodynamic performance, with limited consideration of the multidisciplinary interactions that govern practical aircraft design. In particular, efficient wing design requires simultaneous consideration of aerodynamic performance and structural integrity, yet no publicly available datasets capture this coupled aerostructural behavior at high-fidelity. To address this need, a high-fidelity multidisciplinary dataset for static aerostructural analysis of transonic wings is introduced. The dataset includes 2471 samples, along with three out-of-distribution samples. The wing planform geometry is parameterized by span, sweep, dihedral, tip chord, and tip twist, while the airfoil cross-sections are parameterized using the class-shape transformation method. The analysis is further parameterized by Mach number, angle of attack, altitude, and structural load factor. The utility of the dataset is demonstrated by training state-of-the-art ML models to predict aerodynamic surface quantities, integrated aerodynamic performance quantities, and the failure distribution of the wingbox. ML models achieve at most 0.0309 normalized root mean squared error (NRMSE) in the prediction of aerodynamic quantities and at most 0.1189 NRMSE in the prediction of failure distribution of the wingbox. By providing coupled aerodynamic and structural responses for a diverse set of transonic wings, this dataset establishes a benchmark for developing, evaluating, and comparing multi-physics ML models for wing design. Broadly, it is intended to encourage creation of similar high-fidelity multi-physics benchmark datasets across engineering domains.
[LG-64] Learning atomic interactions for the thermodynamics of liquid mixtures
链接: https://arxiv.org/abs/2610.36440
作者: Marco Hoffmann,Sebastian Schmitt,Hans Hasse,Fabian Jirasek
类目: Machine Learning (cs.LG); Computational Engineering, Finance, and Science (cs.CE)
*备注: 30 pages, 13 figures, 1 table
Abstract:Predictive models for activity coefficients, which describe the non-ideality of liquid mixtures, are foundational for process design and optimization. Group-contribution methods, such as UNIFAC, are computationally cheap and well established in process simulators. Modern machine learning models, such as HANNA, offer higher accuracy and broader applicability, but their integration into process simulation is expensive or even infeasible. In this work, we introduce COGNAC, a graph neural network that learns atom-level geometry and interaction parameters that depend on the molecular neighborhood and evaluates them in the UNIFAC framework. Trained on about 820,000 experimental data points, COGNAC achieves significantly higher accuracy than UNIFAC and HANNA for binary and multi-component mixtures, reduces error relative to HANNA by 13-63%, reliably predicts multi-phase equilibria and identifies azeotropes, yields interpretable parameters, and is strictly thermodynamically consistent. COGNAC evaluations run at a cost comparable to UNIFAC, thereby unlocking the potential of machine-learning models for process simulations.
[LG-65] Simulated petrological labels enable probabilistic mineralogy prediction from whole-rock geochemistry
链接: https://arxiv.org/abs/2610.36429
作者: Matthew J. Liu,Aaron D. Kaplan,Michael L. Whittaker,Kristin A. Persson
类目: Machine Learning (cs.LG)
*备注:
Abstract:Modal mineralogy, the mineral identities and abundances in rocks, guides geological studies and industrial resource extraction. While measuring modal mineralogy is time-intensive, whole-rock chemical measurements furnish elemental concentrations at higher throughput. Predicting mineralogy from elemental data, however, poses a one-to-many inverse problem, and labeled mineralogical datasets are too sparse to train on at scale. We address both challenges by simulating petrological labels that express multiple formation conditions per composition and at a scale unattainable by measurements. Nearly one million simulated labels supervise the middle component of a three-part model: a transformer encoder self-supervised on roughly 788,000 whole-rock compositions, the simulation-trained heads that predict distributions over mineral assemblages, and a recalibration layer that aligns the predictions with measured mineralogy. Zero-shot on 640 volcanic rocks, the model discriminates the presence of most mineral categories and predicts which rocks are richer in each mineral. However, the predicted abundances are systematically offset from measured values. Recalibration fit on as few as 25 labeled rocks recovers 76 percent of the likelihood improvement from fitting on the full 368-rock fitting set, while abundance error responds more slowly (11.0% zero-shot and 9.5% with the full fitting set). This work demonstrates how a machine learning model in the geochemical sciences can both leverage an imperfect simulator and mitigate its systematic biases.
[LG-66] Property-Based Testing for Robust Softmax and Cross-Entropy
链接: https://arxiv.org/abs/2610.36427
作者: Muhammad Maaz
类目: Machine Learning (cs.LG)
*备注: 6 pages
Abstract:A recent paper of Barron et. al. (2026) identified two numerical issues with the softmax and cross-entropy functions in virtually all standard machine learning libraries (e.g., PyTorch and JAX). We show that these issues can be recovered through simple property-based tests in Hypothesis. However, our tests identify an additional edge case where their implementation fails. We then propose an even more robust softmax implementation which passes these tests and also supports use in causal attention masking. We compare our implementation with theirs and discuss some of the differences. Overall, this work shows that property-based testing can be a powerful tool for discovering and fixing numerical issues in machine learning libraries. Our implementations and property-based tests can be found at this https URL.
[LG-67] Persistence Exposure and Credit Routing in Linear Temporal Predictive Coding NEURIPS2026
链接: https://arxiv.org/abs/2610.36412
作者: Akira Tanase,Julian Gould,Jeffrey Seely
类目: Machine Learning (cs.LG); Information Theory (cs.IT)
*备注: Accepted to the NeurIPS 2026 Workshop on Symmetry and Geometry in Neural Representations (NeurReps)
Abstract:Temporal Predictive Coding (tPC) networks offer a layer-local alternative to backpropagation. For linear tPC, we show, by applying tools from cellular sheaf theory, that temporal credit assignment is a tug-of-war between spatial displacement and temporal errors, whose ratios are modulated by persistence and exposure operators P and Q . Spatial inference therefore gives us the effective recurrent transition F=PA , whilst future credit is filtered by P^\ast . This creates a persistence–error tradeoff in spatial modes: under non-amplifying invariant recurrence, modes that preserve influence over long delays simultaneously fail to expose that influence strongly as temporal error. Recurrence can circumvent this via routing between persistent and exposed modes.
[LG-68] When Can a Learned Design Surrogate Be Trusted in a Self-Driving Microfluidics Laboratory? An Offline Closed-Loop Framework Built and Tested on Six Public Tables MICRO
链接: https://arxiv.org/abs/2610.36395
作者: Zhiyuan Zhang
类目: Machine Learning (cs.LG)
*备注: 68 pages, 15 figures, 26 tables. Code, frozen protocols, deviation registers, number ledger and artifacts: this https URL (tag v0.5.5-preprint)
Abstract:Self-driving laboratories couple a learned design surrogate to an instrument, and on every new target domain it must answer four questions: when it can be trusted, how many target measurements it needs, which to make first, and when to stop adapting, switch models or escalate. We state a framework making each measurable (a competence ratio, a budget k^*(\varepsilon) , a time-to-target and a priced decision) and instantiate it without wet experiments on six public droplet-microfluidic and liposome tables as offline laboratories: 107 leave-one-group tiers whose rows are hidden labels. A sealed, pre-registered programme (two sealed negative validations and a pre-registered test on three more tables) and seven frozen closed-loop protocols, each run twice, yield 23 findings. Shift cost is stratified by shift type; no label-free statistic certifies individual candidates; random-split model-family rankings predicted neither the sealed nor the loop rankings. Warm-start residual boosting returns a surrogate to 1.25\times its development reference within 20 target rows in 94% of chemistry tiers (27% under a strict noise-floor reading), 90% of laboratory and 40% of device tiers, and a budget planned from the shift type on other tables is not measurably worse on a new one. In the loop, one gain survives drift and shuffled-label controls on all tiers, carried by the sparsely sampled ones: proposing at once and warm-starting on every measurement reaches a 10% target in 2.80 measurements against 3.31 for the reference, and in 5.72 against 11.35 for random proposals on 301 hard units at a 3% tolerance. No label-free acquisition rule survives the shuffled-label control, no interval-aware proposal or other model family does better, and no threshold rule on the tested diagnostics beats always continuing once escalation is priced. No wet-lab validation is claimed.
[LG-69] SpecLST: Spectral-Enhanced Long-Short Term Spatio-Temporal Multimodal Modeling for Traffic Forecasting
链接: https://arxiv.org/abs/2610.36393
作者: Kenan Kang,Matthew M. Y. Kuo,Weihua Li
类目: Machine Learning (cs.LG)
*备注: 12 pages, 2 figures, 5 tables
Abstract:Modeling multi-scale dynamics in urban road networks is essential for intelligent transportation systems but is often hindered by the computational cost of long-range historical context and the challenge of fusing heterogeneous modalities. Existing spatio-temporal frameworks face a trade-off between sequence length and memory overhead, while frequently neglecting the transient effects of exogenous factors such as road incidents and weather. To address these limitations, we propose SpecLST, a frequency-enhanced long-short-term spatio-temporal multimodal framework. SpecLST leverages a fast Fourier transform-based frequency-domain representation module to efficiently encode multi-week historical dependencies, bypassing the memory bottlenecks of standard attention mechanisms over raw long sequences. Heterogeneous context is integrated through attention-based fusion, followed by a sparse Mixture-of-Experts backbone with graph augmentation for spatio-temporal representation learning. Additionally, a frequency-enhanced residual head isolates high-frequency transients, improving sensitivity to abrupt traffic anomalies. At the 60-minute horizon, SpecLST achieves mean relative error reductions of 14.10% in MAE, 10.35% in RMSE, and 15.78% in MAPE across eight neural forecasting baselines and four datasets, with the dataset- and metric-specific average reductions ranging from 5.63% to 22.21%.
[LG-70] IOSage: Detecting HPC I/O Bottleneck Patterns to Guide LLM Recommendations
链接: https://arxiv.org/abs/2610.36370
作者: Mahdi Banisharif,Tabassum Mahmud,Ahmad Tarraf,Suren Byna,Ali Jannesari
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG); Performance (cs.PF)
*备注: 13 pages, 5 figures, 4 tables
Abstract:HPC applications often leave much of their parallel file system’s bandwidth unused, and turning their Darshan logs into concrete I/O fixes still requires a scarce I/O expert. Rule-based tools flag counters against thresholds set by experts, and in tools built on large language models (LLMs), such as ION and IOAgent, the LLM decides which bottlenecks a log contains. We present IOSage, in which a classifier trained on many labeled logs, not the LLM, makes that decision, since a single log gives an LLM no baseline for what is abnormal. The LLM receives the I/O bottleneck patterns the classifier detects, the job’s configuration files, per-file Darshan evidence, and for each pattern a card built from benchmark training runs, and writes recommended changes. Applied by hand and measured in repeated paired runs, the changes IOSage names speed up median total wall time 4.38x for an openPMD kernel (4.02x on a second storage tier) and 1.42x for hourly WRF restarts. For 5 of the 6 application scenarios with a helpful change, IOSage with Claude Sonnet 4 names it, including the change for 10-minute WRF restarts (up to 1.93x faster); with GPT-4o and Llama 3.1 as well, the classifier and cards name more helpful changes than the LLM alone. On 78 benchmark test runs without a pattern, the LLM deciding alone proposes changes for 64; with the classifier, for 9. The classifier reaches 0.929 Micro-F1 on 488 labeled benchmark test runs (0.916 as the gate; 0.832 on the 158 from configurations and jobs outside training), above Drishti’s rules, WisIO, and IOAgent’s pipeline without its literature retrieval. On TraceBench’s nine expert-labeled traces from production systems, it has the highest precision of the tools compared (0.82), with 2 false positives against 12 for the keyword-mapped LLM diagnoses shipped with TraceBench and 5 for Drishti, on the five issue labels that match registered patterns.
[LG-71] AHEAD: Adaptive Neural Acceleration for Large-Scale Linear Programming
链接: https://arxiv.org/abs/2610.36321
作者: Sihao Liu,Zhichao Hou,Huayue Gu,Raiyan Rahman,Liang Wu,Ruozhou Yu,Xiaorui Liu
类目: Machine Learning (cs.LG); Optimization and Control (math.OC)
*备注:
Abstract:Large-scale linear programs with millions of variables and constraints strain time and memory budgets. First-order methods such as the primal-dual hybrid gradient scale to this regime because they avoid matrix factorizations, but they require many iterations. Learned accelerators can reduce these iterations, yet fewer iterations do not guarantee a shorter solve. Learning inference also takes time, and running the accelerator longer trades solver time for inference time. Existing unrolled accelerators handle this trade-off poorly: each layer processes every variable and constraint, the predicted iterates still need solver refinement, and every instance runs the same layers. We therefore ask: What neural accelerator design achieves end-to-end speedups on large-scale linear programs across sizes and target accuracies? We introduce AHEAD, an accelerator that predicts step sizes for first-order linear programming solvers while keeping their update rules. Its weight-shared predictor reads compact solver statistics and rescales the primal and dual step sizes, so each warm-start iteration stays inexpensive and a short warm-start phase hands the solver a warm start. A learned handoff policy sets the number of warm-start iterations for each instance and target accuracy, after which the solver finishes under its own termination criterion. On packing linear programs with up to 10^7 variables, AHEAD achieves 6.5 – 11.6\times end-to-end speedups over Gurobi’s first-order solver from a cold start, including on instances of unseen sizes and distributions. The gains extend to PageRank linear programs and hold across difficulty levels and target accuracies.
[LG-72] Confidence Is Not a Promise: A Reference Audit of Typed Decision Models
链接: https://arxiv.org/abs/2610.36311
作者: Pedro Carvalho Brom
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注: 51 pages, 11 figures. Code, derived data, preregistration notes and frozen answers: this https URL and this https URL
Abstract:Typed decision models return a probability distribution over options supplied by the caller, and software tends to act on that number as if it were a guarantee. We measure what it licenses, comparing the open-weight Laya, in five checkpoints, with Jev 1.13, available only through an API. A probability licenses a statement about groups of decisions, not about one decision, and a gate that promises an accuracy among the cases it decides alone keeps that promise only with a lower bound checked on data like its traffic. Across 190 exchangeable evaluations, a threshold chosen by point estimate breaks the promise in 147, Learn then Test keeps it in all of them, and a Beta lower-bound rule decides more items on average than Learn then Test in 128 and fewer in none, without its guarantee. Calibrated in English, Jev keeps a promise of 95% in 8 of 51 locales. Ordinary typing noise breaks a promise calibrated on clean text, and, in exploratory runs, training with noise raises the noisy accuracy with no loss on clean text. In four further domains, a fine-tuned Laya needs from 250 to 8,000 labels to be non-inferior to Jev, and does not get there on LIAR2. Under drift, the preregistered claim for a Bayesian retraining loop is not supported: the loop avoids needless retraining where the change does not degrade the model, at a cost equivalent to recalibration alone, and re-qualifying the threshold in every window keeps the promise more often where it does. On the monthly series of M3, neither decider predicts the direction of the next value better than a constant answer. Confidence separates right from wrong answers with an AUROC from 0.80 to 0.96, but it does not select faithful explanations, and exact coalition games of single decisions show what an explanation can still say through an API that rounds to two decimals. We release the frozen answers with a datasheet, a promise check and the code.
[LG-73] Bridging Exploration and Loss of Plasticity in On-Policy Reinforcement Learning
链接: https://arxiv.org/abs/2610.36309
作者: Andrea Vasco Grieco,Evelyn D’Elia,Edoardo Caldarelli,Raffaello Camoriano
类目: Machine Learning (cs.LG); Robotics (cs.RO)
*备注:
Abstract:Deep reinforcement learning (RL) algorithms have been observed to lose the ability to adaptively learn from new data, a phenomenon termed Loss of Plasticity (LoP). While many metrics have been shown to correlate with it, the causal mechanisms and the geometric conditions from which it can arise have not been clearly identified. We analyze LoP in Proximal Policy Optimization with isotropic Gaussian policies, the state-of-the-art formulation for many continuous action spaces in RL. We report multiple metrics which characterize the effect that the policy standard deviation \sigma has on LoP, and we also suggest related explanations for the effectiveness of widely adopted LoP mitigation techniques. We observe a significant effect of \sigma on the extent of LoP, and accompany this with both theoretical and experimental supporting arguments. Our experimental validation consists of LoP benchmark evaluations against standard baselines as well as real-world policy deployment on a 7-DoF robotic arm.
[LG-74] Privately Clipping Heavy-Tailed Data NEURIPS2026
链接: https://arxiv.org/abs/2610.36281
作者: Matthew Joseph,Alex Kulesza
类目: Cryptography and Security (cs.CR); Machine Learning (cs.LG)
*备注: This version corresponds to the camera-ready for NeurIPS 2026
Abstract:Summing vectors is a basic task in differentially private data analysis. Standard private sum algorithms clip vectors to some bound, sum them, and add noise scaled to the bound. The differential privacy literature has developed a deep library of noise addition mechanisms, but it offers few tools for privately choosing a clipping bound based on the data. We introduce a method that privately chooses a clipping bound by optimizing residual coherence, the directional alignment of vectors exceeding the bound, against the variance cost of additional noise. We prove a close relationship between residual coherence and bias for a general class of heavy-tailed data regimes and show empirically that, across several datasets, our method outperforms existing baselines.
[LG-75] EVA-Flow: Learning Shared Chemistry for Unified Environment-Aware Molecular Conformation Generation
链接: https://arxiv.org/abs/2610.36280
作者: Bing Yan,Anuroop Sriram,Zachary Ward Ulissi,Kyunghyun Cho,Ricky T. Q. Chen,Benjamin Kurt Miller
类目: Machine Learning (cs.LG); Computational Engineering, Finance, and Science (cs.CE)
*备注: Accepted at Transactions on Machine Learning Research (TMLR), 2026
Abstract:Molecular 3D structures underpin simulations of thermodynamics, kinetics, crystallization, and nucleation, making their accurate prediction central to drug discovery and materials design. However, molecular geometry can change dramatically across environments. Existing models either ignore environmental effects or train separate models for each environment. We hypothesize that transfer learning across contexts is possible for conformer generation and that a general model will perform better when training data are limited, as in ligand binding and crystallization. We introduce EVA-Flow, a unified framework for environment-aware molecular conformation generation that learns shared chemical structure and adapts it through environment-conditioned generative modeling. Across vacuum, protein-ligand docking, solvation, and crystal packing, pretraining and cross-environment finetuning provide their largest gains in the three data-scarce settings. Evaluation against multiple reference conformations shows that EVA-Flow covers a broad range of solvation conformations. Analyses of molecules shared across environments support its ability to generate different conformations for the same molecule in different environments.
[LG-76] What Should an Information Bottleneck Encoder Learn? Separating Representation Learning from Rate Control
链接: https://arxiv.org/abs/2610.36255
作者: Hassan Tavakoli,Bella Bose
类目: Machine Learning (cs.LG); Information Theory (cs.IT)
*备注:
Abstract:The information bottleneck (IB) trades compression of an input X against prediction of a target Y through a multiplier \beta=1/\lambda , yet each operating point is usually served by training, or fully conditioning, a new encoder. For jointly Gaussian data, we show that the IB solution factorizes as A(\beta)=D(\beta)W : the information-bearing directions W do not depend on \beta , and \beta acts only through scalar gains. We ask whether this separation of \emphrepresentation from \emphrate control survives beyond Gaussian models, treating the budget as a given operating parameter rather than a quantity to be optimized. For a budget-free representation (F(X)), the \emphseparation gap is controlled by the \emphsufficiency deficit \epsilon_F=I(X;Y\mid F(X)) : We show there is at least \epsilon_F in the worst case for finite inputs, and it is bounded uniformly in \lambda by a function of \epsilon_F for finite label alphabets. The same certificate extends to a new target distribution and to pooled multi-domain data, where a conditional-alignment term appears. On MNIST, SVHN, CIFAR-10 and STL-10 (10 seeds, eight prespecified budgets), a frozen CNN representation with one compressor–decoder pair per budget reduces estimated training multiply–accumulate operations (MACs) by 7.7\times and wall-clock time by about 2\times relative to end-to-end VIBs, and attains higher budget-averaged accuracy ( 65.6% versus 57.3% per domain; 67.1% versus 57.1% pooled), with the advantage concentrated at strong compression, where the VIBs collapse. At weak compression ( \lambda\le0.036 ), VIB has higher cross-entropy and ECE; its surrogate Lagrangian is higher per domain, indistinguishable from the pooled VIB in the macro average, and lower on the pooled test set.
[LG-77] hree Worlds for the Price of a Logarithm: Baseline-Protected Best-of-Both-Worlds Combinatorial Semi-Bandits
链接: https://arxiv.org/abs/2610.36245
作者: Ting Hu,Luanda Cai,Emmanouil-Vasileios Vlatakis-Gkaragkounis
类目: Machine Learning (cs.LG); Computer Science and Game Theory (cs.GT)
*备注:
Abstract:Improving on a baseline requires exploring away from it, precisely where its performance may be lost. We propose DUET (Doubling Under Empirical Tests), a horizon-free phased mixing scheme for combinatorial semi-bandits with arbitrary finite action sets. Two observable tests decide when to double the learning weight and when to double the empirical-regret budget and restart. We instantiate the scheme with entropy-FTRL in mixed phases and hybrid-FTRL in pure phases. Assuming the baseline selects every arm with positive probability, this instantiation keeps expected baseline regret logarithmic in T , while its expected regret to every fixed action is \widetildeO(\sqrtT) under non-anticipating adversarial losses and polylogarithmic, gap-dependent, under i.i.d. losses with a unique optimal action. All three guarantees hold at every deterministic horizon, and the algorithm knows neither the horizon, nor the gap, nor the environment type. Baseline protection, adversarial robustness, and stochastic adaptivity thus come from one scheme and two base learners, at the price of a logarithm.
[LG-78] Dynamical coarse-graining of neural-network training
链接: https://arxiv.org/abs/2610.36180
作者: Lin-Juan Wang,Da Li,Huiling Duan,Jianxiang Wang
类目: Machine Learning (cs.LG); Soft Condensed Matter (cond-mat.soft); Neural and Evolutionary Computing (cs.NE)
*备注:
Abstract:Dynamical coarse-graining of neural networks gives macroscopic observables and enables training to be predicted, and controlled in a state space far smaller than the parameter space. In this paper, we first define the learning operator following the training law of neural networks and establish a descent theorem for the learning operator, which serves to determine whether the operator is sufficient to characterize the learning dynamics. Then the exact closure theorem and the dynamic coarse-graining theorem are presented in sequence. The three theorems established in this paper provide a systematic framework for constructing and assessing dynamical coarse-grained models of neural-network training. Finally, we demonstrate the applicability of the theorems with a finite-width nonlinear ReLU network. Its dynamic coarse-grained model for a fixed pattern is exactly derived and composed of five independent moments. The model exactly predict the training outputs and evolving learning operator. The framework may also guide the development of approximate coarse-grained models for more general neural networks when exact closure is unavailable.
[LG-79] SpatialPC: Deep Computation with Sublinear Local Credit Propagation
链接: https://arxiv.org/abs/2610.36139
作者: Benjamin Prada,Shion Matsumoto,Ankur Arjun Mali
类目: Machine Learning (cs.LG)
*备注: 29 pages, 4 figures
Abstract:On a chain of N states, any local learning rule with bounded communication range needs \Theta(N) rounds before the loss can influence the earliest parameters, and this latency is usually attributed to depth. We show that it is a property of the communication geometry rather than of depth. For N=m^d states placed at constant density in d dimensions, a light-cone argument gives an \Omega(N^1/d) lower bound on the number of bounded-range rounds that any local algorithm needs before its gradient estimates can be sign-correct at every parameter. We construct a unit-range, degree- 2d directed acyclic graph that contains a Hamiltonian computation of length N-1 and in which every state lies within directed distance d(m-1)=\Theta(N^1/d) of the output, matching the lower bound. In the realizable-circuit model, depth and credit depth are therefore independent parameters: represented depth \Theta(N) and causal credit latency \Theta(N^1/d) coexist, with a separation of order N^1-1/d . SpatialPC realizes this separation within predictive coding. Credit is propagated by adjoint prediction dynamics (APD), a local update whose K -round state on a settled acyclic graph is the truncated reverse-path sum a^K=\sum_kK(J^\top)^k b , with error exactly (J^\top)^K a^\star . Graph geometry fixes when credit can first arrive at a state; the Jacobian-weighted path tail fixes how many further rounds accurate credit requires. A sparse two-dimensional Moore compiler retains the N-1 backbone with \tfrac32N-2 unit-range edges and has measured worst-node reachability \sqrt N+2 up to N=65,536 , where a 65,535-edge backbone is reached in 258 local hops. With T=2\sqrt N rounds, SpatialPC keeps finite-round task gradients aligned with settled gradients (cosine 0.93 on MNIST and 0.97 on Fashion-MNIST at N=256 ) and reaches 97.6% and 88.1% test accuracy after 20 epochs.
[LG-80] Minimax-Optimal Small-Loss Bounds for Online Linear Optimisation on ell_p Balls
链接: https://arxiv.org/abs/2610.36097
作者: Jack J. Mayo
类目: Machine Learning (cs.LG); Optimization and Control (math.OC); Machine Learning (stat.ML)
*备注: 49 pages
Abstract:We study full-information online linear optimisation on the unit \ell_p ball in \mathbbR^d , for fixed 2\le p\infty and losses of dual norm at most one. Let \Delta_T denote the cumulative loss of the best fixed action above each round’s minimum. A follow-the-regularised-leader (FTRL) algorithm attains, on every loss sequence and without knowing T or \Delta_T , regret O_p(1+\min\Delta_T^1-1/p,, d^1/2-1/p\sqrt\Delta_T) . Matching lower bounds show that this rate is minimax for each fixed p , and its constant (2+o(1))p/\log p is optimal as p\to\infty among guarantees uniform in d and T . The algorithm uses a bounded radial regulariser and switches at most once to its quadratic augmentation. On some paths where follow-the-leader (FTL) has regret one, its regret grows with T . A switching extension keeps the sharp constant and also guarantees \mathrmReg_T\le 2,\mathrmReg_T^\mathrmFTL . For i.i.d. losses whose mean has no zero coordinates, we compute the exact \log T coefficient of the expected regret, which separates FTL’s variance term from the bias of the regulariser.
[LG-81] Instruction Capacity: Behavior and Degradation of Language Models under Controlled Load
链接: https://arxiv.org/abs/2610.36095
作者: Hennadii Kutomanov
类目: Machine Learning (cs.LG)
*备注: 70 pages, 20 figures. Code and data: this https URL
Abstract:The quality with which language models follow instructions falls as the load grows, yet there is still no quantity that expresses the limit of a model in units of load with an uncertainty. We propose an instrument in which the load consists of rules that replace single-token words in a text without meaning: such an instruction is atomic, the rules are independent by the construction of the task, and the outcome of every application of a rule is determined by a computable reference answer. Every occurrence of a word to be replaced becomes a Bernoulli trial with a known position in the answer and a known rule number. The statistical model first measures the shape of the dependence of the probability of a right replacement on the text length, the word position, the replacement density, the number of rules and the rule number, then describes it by smooth laws and restores a reference level that does not depend on the choice of the coding. The model yields the instruction capacity - the number of rules held at a given share of right replacements - with an uncertainty. A measurement of GPT-4o mini (2958 requests, 642,430 trials) gave regular laws for every load factor and for the collapse of the answer; the instruction capacity was about 37-38 rules at a share of right replacements of 0.95, and about 18-19 rules at the full replacement density. The code of the instrument, the measurement data and the passport are open, and every number of the passport is recomputed from the published points. The passport of a model serves as the calibration of a future meter of the instruction load of prompts.
[LG-82] Jaxnasium: Embracing PyTrees For Any-Agent Cross-Suite Reinforcement Learning
链接: https://arxiv.org/abs/2610.36075
作者: Koen Ponse,Aske Plaat,Thomas M. Moerland
类目: Machine Learning (cs.LG); Digital Libraries (cs.DL)
*备注: 10 pages
Abstract:Hardware-accelerated reinforcement learning in JAX has gained rapid popularity in recent years and has proven to be a successful method of scaling online reinforcement learning. The resulting ecosystem, however, is fragmented: environment suites define their own interfaces, and algorithm implementations are typically written for a single suite, a single type of space, and either JAX-native or host-based environments. Here we present Jaxnasium, a library for JAX-based reinforcement learning, aiming to make the field more accessible. Jaxnasium represents observation and action spaces as PyTrees, and uses this structure to build networks and support multi-agent processing. Through this, its near-single-file implementations run on discrete, continuous, image, and composite spaces, as well as in multi-agent environments, without any changes to the algorithm code. The same files serve as readable, research-friendly algorithms, as well as importable algorithms for more applied workloads. Through a minimal Gymnasium-style interface and wrappers for 11 suites, over 300 environments are available on which a small PPO implementation with default hyperparameters is able to run. Additionally, through a simple wrapper, we can run Gymnasium and EnvPool environments with the same algorithm and training loop, with no performance loss compared to architectures specifically built for training JAX algorithms on host-based environments. Jaxnasium further provides tools for multi-seed evaluation, hyperparameter sweeps, and project scaffolding. Jaxnasium is open-sourced at this https URL
[LG-83] Exploratory creative pre-ranking: Hierarchical empirical-Bayes Thompson sampling for real-time creative optimization
链接: https://arxiv.org/abs/2610.36020
作者: Jiale Han,Ajith Moparthi,Dan Xie,Xiaowu Dai
类目: Machine Learning (cs.LG); Computer Science and Game Theory (cs.GT)
*备注:
Abstract:Dynamic creative optimization decides which creative of an ad to render for each request. Recent work places that choice before ad ranking, but relies on deterministic point predictors with no calibrated uncertainty. AI-generated content has made creatives nearly free, while feedback per creative remains sparse and selection faces a tight latency budget. We introduce a creative pre-ranker that scores every recalled creative and shortlists each campaign’s pool before the per-user ranker runs. Creatives sit in a hierarchy of audience segment, advertised item, campaign and creative, so a cold creative falls back on what its parent levels have earned, and how much it borrows from its parents is set in closed form from historical logs, with one parameter per level and nothing tuned. To stay within the latency budget, the pre-ranker predicts for an audience segment rather than an individual user, from per-attribute counters. It selects by a single Thompson draw per creative, balancing exploration and exploitation so that a new creative still has a chance to be learned. Replaying Roblox Sponsored-ads traffic over campaigns with competing creatives, the shortlist lifts predicted play-through rate by 8.14% against 2.80% for an equally sized random shortlist, forwards more than 40% fewer creatives than the full pool, and adds only 0.2% to serving time. After an A/B test on live traffic, the method now serves in production at Roblox.
[LG-84] A Graph-informed Kriging model for sample-efficient Bayesian optimization of wind farm layouts
链接: https://arxiv.org/abs/2610.35975
作者: Zhen-Fan Wang,Kai Zhang,Xianping Du,Zhaolong Han,Dai Zhou,Onur Bilgen
类目: Machine Learning (cs.LG); Optimization and Control (math.OC); Fluid Dynamics (physics.flu-dyn); Machine Learning (stat.ML)
*备注:
Abstract:Wind farm layout optimization becomes increasingly difficult as turbine count grows, especially when high-fidelity aerodynamic models are used. Conventional Kriging-based Bayesian optimization often suffers from low sample efficiency in high-dimensional layout spaces because sparse training data provide limited domain coverage. We develop a Graph-informed Kriging framework that embeds reusable aerodynamic interaction information into the surrogate mean function. The wind farm is represented as a directed aerodynamic graph, where pairwise turbine interactions defined in a low-dimensional relative-coordinate space are assembled into a farm-scale mean. This structured mean captures the dominant variation in annual energy production (AEP), allowing the Gaussian process to model only the residual structure and uncertainty. Numerical experiments show that Graph-informed Kriging consistently improves AEP prediction accuracy over ordinary and generic-trend Kriging under sparse-data conditions, with increasing benefits as layout dimension and wake-interaction strength grow. Embedded in Bayesian optimization, it substantially improves sample efficiency and achieves competitive solutions for the IEA Wind Task 37 benchmark with about 280 AEP evaluations. In a controlled 16-turbine case, it also outperforms ordinary and multi-fidelity Kriging under the same high-fidelity evaluation budget. The framework is further extended to a 48-turbine wind farm, demonstrating applicability to large-scale CFD-based layout optimization. These results show that Graph-informed surrogate modeling can enable more efficient use of high-fidelity aerodynamic models for reliable large-scale wind farm layout design.
[LG-85] Data-driven Discovery of Lie Symmetries and First Integrals with Latent ODEs
链接: https://arxiv.org/abs/2610.35967
作者: Matt L. Wiemman,Peter Melchior
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注: Accepted into 2026 Neural Representations Proceedings
Abstract:We present LatentLieODE, a novel neural ODE method to discover linear Lie symmetries and first integrals of the learned dynamics from timeseries observations such as video inputs. It starts with an autoencoder, which learns a low dimensional representation of the observed system and then maps trajectories to decorrelated latent coordinates z \in \mathbbR^d . A bank of linear infinitesimal generators \mathbfA_k \in \mathbbR^d \times d is fitted jointly against a differentiable Lie-bracket residual \mathbfJ_f(z)\mathbfAz - \mathbfA f(z) , while an MLP optimises the latent ODE f(z) = \mathrmdz/\mathrmdt to reconstruct the observation-space timeseries. After suitable generators and f are found, we fit a third network H_\phi to generate the first integral of the system. Across six benchmark systems (harmonic oscillator, simple and twin pendulums, Lotka-Volterra, pixel pendulum, sphere geodesics) we find each system’s relevant symmetries, and the discovered invariants match ground-truth conserved quantities at |\rho| \geq 0.978 . On the sphere, the pipeline recovers a non-abelian \mathfrakso(3) , certified by the negative-definite Killing form of its recovered structure constants. This work presents significant advances in data-driven discovery of symmetries and provides a new pathway for neural surrogates of dynamical systems.
[LG-86] Generalization in Learning: Eyring Formula and Predator-prey Model
链接: https://arxiv.org/abs/2610.35965
作者: S.V.Kozyrev,I.A.Lopatin,Alexander Pechen
类目: Machine Learning (cs.LG)
*备注: 12 pages, 4 figures
Abstract:In this work, generalization in machine learning is investigated using a predator-prey model. We overview biological and chemical interpretation of the stochastic gradient Langevin dynamics and generative adversarial network. For a particular example of recovering an unknown function by a polynomial of fixed degree from a set of noisy data, the predator-prey model provides better approximation compared to the gradient descent method. Further, thermodynamical arguments, in particular Eyring formula, are used to explain grokking (delayed generalization) phenomenon in machine learning.
[LG-87] Condition-Enhanced Generative Adversarial Network for Aerodynamic Configuration Design of Commercial Aircraft
链接: https://arxiv.org/abs/2610.35963
作者: Youzheng Yu,Jiaqing Kou,Weiwei Zhang
类目: Machine Learning (cs.LG)
*备注:
Abstract:Generative machine learning provides an end-to-end framework for aerodynamic design optimization by generating diverse aerodynamic configurations conditioned on target performance. However, the generated designs may violate geometric constraints or deviate from the prescribed aerodynamic performance. Therefore, we propose a Condition-Enhanced Generative Adversarial Network (CEGAN) that embeds geometric and aerodynamic information into the generative process. A condition-informed loss function is developed by combining a geometry-guided feasibility penalty with an aerodynamic performance penalty evaluated by surrogate models trained on the same dataset. Additionally, an output filtering mechanism discards infeasible designs using these surrogate models, ensuring compliance with prescribed conditions. The method is demonstrated on aircraft configurations similar to the Cessna Citation X. Each configuration is represented by 28 shape parameters, whose aerodynamic performance is evaluated using OpenVSP. Interpolation and extrapolation tests are conducted for different network architectures and conditioning strategies. Compared with the baseline model, CEGAN reduces the average generation error by at least 36% for interpolated conditions and by at least 46% under extreme extrapolation conditions, while successfully producing diverse and physically feasible designs that extend beyond the training distribution. Therefore, the proposed CEGAN framework offers a robust and accurate tool for rapid aerodynamic configuration exploration during the conceptual design phase.
[LG-88] h1: A Taste of the Bitter Lesson in Automated Theorem Proving
链接: https://arxiv.org/abs/2610.35962
作者: Skander Moalla,Ahmad Rammal,Niket Patel,Remi Munos,Julia Kempe
类目: Machine Learning (cs.LG); Logic in Computer Science (cs.LO)
*备注:
Abstract:While progress in informal mathematics has come largely from scaling the chain-of-thought and agentic reasoning of frontier language models, formal theorem proving has increasingly relied on human-designed workflows around these models, such as planning, decomposition, and orchestrating specialized provers. This raises a bitter-lesson question: can a simple agentic loop that scales test-time compute match these workflows? The answer matters as formal verification must scale to keep up with frontier models producing research mathematics, yet new workflows rarely settle it, as their gains are often confounded with stronger models or more compute. We study this question with h1, a simple harness that gives a model an efficient interface to Lean 4 and keeps it working through verifier feedback and self-reflection, without imposing a decomposition or search structure. h1 reaches state-of-the-art results on PutnamBench, FATE-X, and FormalConjectures100-Solved, settles eight open problems in FormalConjectures, and, at matched models, matches or outperforms specialized workflows such as Goedel-Architect and AlphaProof Nexus, surprisingly showing that a single attempt (pass@1) can often scale to solve open problems. Through ablations and matched-model comparisons with Claude Code and AxProverBase, we find that simplicity alone is not enough: an efficient Lean interface largely determines cost and solve rate. We thus recommend that future work evaluate frontier models with efficient execution feedback before prescribing workflows, at the risk of bitter lessons.
[LG-89] Exponential Wake-Up Overshoot in Explicit Gaussian Point-Process Filters
链接: https://arxiv.org/abs/2610.35958
作者: Hanwu Qin,Dianhai Wang
类目: Machine Learning (cs.LG); Signal Processing (eess.SP)
*备注: 26 pages, 3 figures, 13 tables; main text with Appendices A-G (proofs, experimental protocol, additional results). A four-page version has been submitted to IEEE Signal Processing Letters
Abstract:When events resume after a long silence, explicit Gaussian point-process filters amplify the first update exponentially in the silence duration. We give the exact distribution of this update and a computable index that predicts when it overshoots. For the explicit (single-scoring-step) filter and its implicit posterior-mode counterpart, we derive second-order silence expansions of the state variance with opposite-sign logarithmic corrections. The crossing index determines when the post-wake prediction exceeds the resumed rate, separating the bin width at which the amplification exponent reaches one from the larger width at which the true rate is overshot. At fixed bin width, the implicit filter’s expected loss grows logarithmically with silence duration, whereas a continuous-time counterexample rules out a bound uniform in bin width and silence duration. Experimental evaluation confirms these predictions in all 73 crossing cells and on exchange trade data, while variance caps reduce explicit overshoot but mostly hurt the implicit filter.
[LG-90] Distributed LLM Inference via Communication-Efficient Speculative Decoding
链接: https://arxiv.org/abs/2610.35957
作者: Joohyun Shin,Byeonghyeon Park,Keumseo Ryum,Joonhyuk Kang
类目: Machine Learning (cs.LG); Information Theory (cs.IT); Signal Processing (eess.SP)
*备注:
Abstract:Distributed speculative decoding has emerged as a promising approach for device–server collaborative large language model (LLM) inference, but its decoding speedup is often limited by the vocabulary-scale communication overhead. In existing distributed speculative decoding methods, exact rejection sampling requires transmitting full token distributions over the LLM vocabulary between the device and the server, which becomes costly in bandwidth-constrained environments. To address this issue, we propose DISCO, a communication-efficient framework for distributed speculative decoding, that eliminates vocabulary-scale distribution transfer while preserving the exact target distribution. DISCO achieves this by introducing a server-side replica of the draft small language model (SLM), which enables the server to locally obtain the draft-side distribution required for rejection sampling. As a result, DISCO exchanges only drafted token indices and scalar probability values between the device and the server. Experimental results on multiple datasets show that DISCO consistently achieves higher decoding speedup than existing distributed speculative decoding methods, particularly under stringent bandwidth constraints.
[LG-91] Calibrated for Which Policy? Conformal Uncertainty for Fixed World Models Under Intervention
链接: https://arxiv.org/abs/2610.35877
作者: Ayush Mittal,Dhruv Gupta
类目: Machine Learning (cs.LG); Robotics (cs.RO); Machine Learning (stat.ML)
*备注: 18 pages, 6 figures, 3 tables. Lean 4 formalization: this https URL
Abstract:Changing a deployment policy can invalidate a world model’s uncertainty while leaving the physical transition mechanism and predictor unchanged. For one-step prediction, if source and target laws share the conditional mechanism P(Y|X)=Q(Y|X) , then the intervention is identified by the pre-outcome change of measure w(X)=dQ_X/dP_X . For policy-controlled dynamics this ratio factors into state-occupancy and action-propensity terms. Weighted conformal calibration transports marginal coverage to the target law; with an estimated ratio \hat w , P_Q\Y \in C_\hat w(X)\ \ge 1-\alpha-TV(Q,\hat Q) \ge 1-\alpha-E_P|w-\hat w| . Population log-loss regret for a source-versus-target domain classifier bounds the same ratio error. If h is its clipped target posterior, \rho is the target-domain prior, and \Delta(h) is population log-loss regret, then E_P|w-\hat w_h| \le (1/(\rho\gamma))\sqrt\Delta(h)/2 . An exact cancellation improves the dependence from a generic odds-map bound of order \gamma^-2 to (\rho\gamma)^-1 . Our experiments use a released DINO-feature world model and physical Franka trajectories. Under the strongest controlled state-action shift, ordinary 90% conformal sets cover 55.0%, whereas oracle and learned-ratio calibration cover 91.0% and 90.9%. The learned surrogate remains within 0.9% total variation of the target law. Calibration is relative to the deployment law: the model can remain fixed while calibration follows the intervention. We machine-check in Lean 4 the finite-support algebraic and causal identities used in the classifier-regret argument, including its coverage-composition step.
[LG-92] KORA: Deterministic-First Execution Control for AI Workloads KR
链接: https://arxiv.org/abs/2610.35869
作者: Albert Heekwan Kim
类目: Machine Learning (cs.LG); Software Engineering (cs.SE)
*备注: 16 pages, 3 figures, 3 tables. Code and benchmark available at this https URL
Abstract:AI applications often invoke a model before they determine whether inference is necessary. That default is useful for ambiguous generative tasks, but it is poorly matched to work that is structured, deterministic, policy-bound, lookup-oriented, or directly verifiable. KORA introduces deterministic-first execution control: a pre-inference control layer that represents requests as task paths, attempts deterministic or structured routes first, validates route outcomes, records telemetry, and escalates to a model-candidate path only when deterministic handling is insufficient. This paper presents KORA as a deterministic-first execution-control architecture and reports bounded evidence from reproducible benchmark and validation workloads. In a reproducible 100-task deterministic-heavy benchmark, KORA-controlled execution avoided 80 of 100 simulated model invocations versus a naive direct baseline, with zero mismatches in the deterministic subset. The workload is intentionally deterministic-heavy: it includes categories such as FAQ lookup, policy/rule checking, schema or format validation, deterministic transformation, and a smaller set of model-needed ambiguous or generative tasks. A separate synthetic customer-support local/no-network validation workload demonstrates route measurement and avoided model-call accounting in an application-shaped setting. These results support bounded model-invocation reduction and validation visibility in documented workloads. They do not establish production behavior, real-provider outcomes, energy outcomes, API-cost reduction, or broad superiority.
[LG-93] Learning to Exploit Passive Dynamics for Energy-Efficient Target Hopping of a Spring-Legged Quadcopter
链接: https://arxiv.org/abs/2609.15447
作者: Ruigang Chen,Qi Zhang,Zhicheng Zhong,Zhuorui Yun,Yizhar Or,Mingyi Liu
类目: Robotics (cs.RO); Machine Learning (cs.LG)
*备注: 8 pages, 10 figures
Abstract:Combining aerial thrust with spring-loaded hopping makes monopedal quadcopters promising for locomotion over complex terrain, but heuristic proportional-integral-derivative (PID) tuning limits coordination between active thrust and passive contact dynamics. We present a direct estimated-state-to-motor Proximal Policy Optimization (PPO) policy that commands four motors without an explicit hopping state machine or low-level attitude PID. Its reward combines Energy-Manifold Shaping for mass-normalized vertical-energy tracking and apex-state anchoring with Efficiency Shaping, which uses a history-aware power estimator to penalize general power use, impose an additional airborne-power cost, and penalize airborne near-stationarity. In representative hardware runs, the PPO-based control stack reduced cycle-averaged measured electrical power by 30.7% and mean total normalized thrust by 49.8% relative to the tuned PID-based control stack, while retaining repeatable commanded-height hopping and more concentrated landings. These observations are consistent with improved use of passive dynamics and reduced measured electrical demand.
[LG-94] Regression Functions as Fields on Covariate Space: Evolution Equations the Hessian and Identification
链接: https://arxiv.org/abs/2610.37500
作者: Nobuaki Tanaka
类目: atistics Theory (math.ST); Machine Learning (cs.LG)
*备注: 13 pages, 2 figures. Code for the symbolic checks and simulations is included as ancillary files
Abstract:For each time t , the regression function m_t(x)=E[Y_t\mid X_t=x] of an outcome on contemporaneous covariates can be regarded as a field on covariate space. We study how this field evolves when the covariates of individual units move as a diffusion and the outcome attached to each unit changes according to its own dynamics. Under mild regularity we show that m_t satisfies a second-order evolution equation whose diffusion term is \frac12\mathrmtr(\Sigma\nabla^2 m_t) , the trace of the Hessian of the regression function weighted by the instantaneous covariance \Sigma of the covariate increments, whose transport term is the drift of the time-reversed covariate process, and whose source term is the conditional mean rate of change of the outcome. The equation is a regression-theoretic reading of a classical transport law, and it yields a Feynman-Kac representation of m_t along reversed covariate paths. We connect the Hessian appearing in the equation to the Hessian estimated by quadratic regression (under Gaussian design the best quadratic approximation has Hessian E[\nabla^2 m(X)] exactly) and to the leading bias of kernel smoothers. For binary outcomes governed by bistable replicator dynamics with imitation among units with similar covariates, the closed equation is of Nagumo type, its planar travelling fronts are exactly logistic in (x,t) , and a single logistic regression of the outcome on covariates and time identifies the ratio of diffusion to reaction and the direction of invasion. We discuss identification under panel versus repeated cross-sectional designs and a two-stage estimation strategy, and we illustrate the results with simulations.
[LG-95] QD-PINN:A Physics-Informed Neural Network Approach to Bayesian Quickest Detection in High-Dimensional Diffusions
链接: https://arxiv.org/abs/2610.37473
作者: Hang Ye,Yi Liu,Haipeng Xing
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注: 24 pages, 9 figures,
Abstract:Quickest detection problems (QDPs) arise in a wide range of applications. However, the complementarity structure of the associated variational inequalities poses substantial challenges for traditional numerical methods, particularly in high-dimensional settings. We propose a PINN-based framework for estimating the optimal stopping boundary in QDPs with high-dimensional diffusion processes and spatial dependence, while overcoming the curse of dimensionality. Theoretically, we establish a novel equivalence between the free-boundary PDE and an infinite-dimensional stochastic optimization problem, rigorously justifying the use of finite-dimensional neural networks to approximate the true solution. Furthermore, we assess the robustness of the proposed framework by quantifying uncertainty in the learned stopping boundaries using deep ensembles and Monte Carlo dropout. The results show low variability across independent training runs and stochastic forward passes, indicating stable and consistent boundary estimates. We also compare our PINN results with reference numerical solutions where available and find close agreement.
[LG-96] Graph Representation Learning via Adaptive Expectile Aggregation
链接: https://arxiv.org/abs/2610.37450
作者: Tianyi Zhu,Wei Lan
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:Graph neural networks (GNNs) rely critically on neighborhood aggregation, yet conventional mean aggregation may fail to capture class-relevant distributional information, particularly in heterophilic graphs. We propose ExpectileGCN, an adaptive graph aggregation framework that introduces a learnable expectile level to capture neighborhood information beyond the mean. The proposed operator continuously adjusts aggregation between minimum- and maximum-type behaviors, with mean aggregation as a special case. A differentiable approximation based on sigmoid smoothing and fixed-point iterations enables end-to-end optimization. Theoretically, we establish stability properties and transductive generalization guarantees for ExpectileGCN. We further investigate its classification risk under heterophilic neighborhood models. For informative neighborhoods, we show that expectile aggregation can exploit class-dependent scale differences to achieve a classification advantage over mean aggregation. Under a local asymptotic regime with diminishing per-neighbor label informativeness, we establish that this advantage can persist in terms of limiting classification risk. Experiments on homophilic and heterophilic benchmark graphs demonstrate strong predictive performance and improved robustness under feature corruption and structural perturbations.
[LG-97] An end-to-end framework for uncertainty-aware validation of generative models
链接: https://arxiv.org/abs/2610.37398
作者: Giada Badaracco,Sean Benevedes,Christina Reissel,Philip Harris,Gaia Grosso
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); High Energy Physics - Experiment (hep-ex); High Energy Physics - Phenomenology (hep-ph)
*备注:
Abstract:Density estimation with generative models is a common task in the physical sciences, for example in background estimation in particle physics. Existing machine learning techniques, however, lack methods to quantify uncertainties that can be propagated into a statistical test. That limits the validation of generative models to standard goodness-of-fit tests, in which the generative model is treated as exact (point-null hypothesis). But a model fitted to a finite training set is never exact, so if its uncertainties are neglected, any sufficiently powerful goodness-of-fit test will eventually reject it once the test set is large enough to resolve this estimation error. We propose SIGMA, an end-to-end framework combining density estimation with principled uncertainty quantification. We use Gaussian mixture models, for which uncertainties are tractable, and further construct a new goodness-of-fit test in which these uncertainties are profiled as nuisance parameters to account for the statistical error of the fit, promoting the conventional point-null test to a composite one. A rejection then signals genuine mismodeling rather than a test set large enough to resolve the error of the fit, provided the propagated uncertainties faithfully cover that error. We demonstrate the framework on a controlled two-dimensional toy model and on a collider physics benchmark modeling the QCD background process. With uncertainties included, the same model can be trusted on test sets up to an order of magnitude larger than the standard test allows. Our work thus provides a statistically grounded basis for a robust use of generative models in the physical sciences.
[LG-98] he Geometry of Manifold Learning: Singular foliations via Deep Learning models
链接: https://arxiv.org/abs/2610.37370
作者: Gabriele Novelli,Eliot Tron,Emanuele Latini,Rita Fioresi
类目: Differential Geometry (math.DG); Machine Learning (cs.LG)
*备注: 18 pages, 4 figures, 2 appendix
Abstract:Deep learning algorithms have become widely used and successful across a wide range of tasks. A key challenge is determining whether these algorithms truly capture the intrinsic geometry of their training data in order to better understand what they learn from data. In this theoretical work, we explore a new geometric structure defined on the data space by a trained neural network through the Data Information Matrix (DIM), a variation of the Fisher Information Matrix. This new geometric structure is a foliation and the main focus of this work is to explore the non-smooth case of ReLU activation functions, where traditional geometric tools often fail. The foliation is defined through a distribution spanned by the columns of the DIM. We first prove that this distribution is non-smooth in the sense of distributions, rendering the existence of an associated foliation much more complex. To circumvent this issue, we introduce a smoothed version of this distribution that differs from the original only on a set of measure zero. Finally, we prove that this smoothed distribution induces a singular foliation on the data space, which is based on the data the network has been trained with. Our theoretical contributions provide a geometric framework to better understand how neural networks organize their input space in non-smooth settings, bridging differential geometry and deep learning theory.
[LG-99] Scale-aware and spatially-localised optimisation of kilometre-scale MLWP models
链接: https://arxiv.org/abs/2610.37325
作者: James Chappell,Sophie Luise Arana,Jack Bowyer,Alejandro Coca-Castro,Evangeline Corcoran,Eric G. Daub,Tom Dunstan,Noushin Eftekhari,Anna-Louise Ellis,J. Scott Hosking,Manvendra Janmaijaya,David Salvador-Jasin,Ben Shipway,Nathan Simpson,Oliver Strickson,David Walters
类目: Atmospheric and Oceanic Physics (physics.ao-ph); Machine Learning (cs.LG)
*备注: 21 pages, 12 figures (including appendix)
Abstract:Preserving fine-scale structure without degrading spatial accuracy is a central challenge for kilometre-scale machine-learning weather prediction (MLWP). We introduce a loss based on the dual-tree complex wavelet transform that constrains atmospheric fields jointly by scale, location, and orientation. We evaluate its performance using an experimental joint global–regional FastNet configuration, which produces \SI1.5km forecasts over the UK alongside \sim \SI31km global forecasts. Relative to both mean-squared error (MSE) and an FFT-based spectral objective, the wavelet loss substantially improves spectral fidelity. Compared with MSE training, it maintains similar pointwise performance, improves neighbourhood-based verification, and produces sharper precipitation fields with more realistic intensity distributions. These results demonstrate that spatially localised wavelet constraints can preserve multiscale atmospheric structure while largely avoiding the loss of pointwise accuracy associated with globally supported spectral objectives.
[LG-100] Low-degree polynomials lower bounds for clustering in small dimension
链接: https://arxiv.org/abs/2610.37272
作者: Alexandra Carpentier,Nicolas Verzelen(MISTEA)
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:We study the computational limits of clustering n observations drawn from a mixture of K isotropic distributions in \mathbb R^d , in the regime where the number K of groups is much larger than the dimension d . In this regime, tensor-based procedures recover the partition with a polylogarithmic separation between the centers, but they require a sample size that is a large and non-explicit polynomial of dK , whereas the existing low-degree lower bounds only show that n must be larger than dK^2 . We introduce a mixture of 2K product Bernoulli distributions on \0,1^d , with K=(d+1)^m , whose centers are built from a dual BCH code. This construction ensures that the moment tensors of the observations coincide with those of the uniform distribution on -1,1^d up to the order 2m+1 . For any fixed m , we establish a low-degree polynomial lower bound showing that clustering is hard as long as the separation \Delta satisfies \Delta^2\ll(K^2d^3/2/n)^1/(2m+2)\wedge\sqrt d , up to polylogarithmic factors. In particular, with a polylogarithmic separation, low-degree polynomials require n\gtrsim K^2d^3/2 observations, up to polylogarithmic factors. Through a Gaussianization argument, this hardness carries over, in the sense of reductions, to isotropic Gaussian mixtures. Our proof relies on a family of polynomials of the observations, indexed by bipartite graphs, which is almost orthonormal. The moments of these polynomials are governed by the number of solutions of polynomial systems over the finite field \mathbb F_d+1 that are induced by the BCH code, and the crux of our analysis is to connect the combinatorial structure of the bipartite graphs with the structure of the solution sets of these systems.
[LG-101] Doubly Robust Deep Kernel Tests for Structured Outcomes
链接: https://arxiv.org/abs/2610.37227
作者: Robin Leman,Bariscan Bozkurt,Sofiane Tanji,Arthur Gretton,Houssam Zenati
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:We propose a test of whether two interventions induce the same outcome distribution when outcomes are structured and high-dimensional, such as images, text and graphs. We represent the interventional distributions as mean embeddings on learned deep features, such that the treatment effect is a doubly robust representation of the difference in mean embeddings. Our test, DR-KTE-Deep, optimises the learned features so as to maximise an estimated power criterion on this treatment effect representation. To motivate our learning objective, we show that the power criterion characterises the asymptotic signal-to-noise ratio of the test under fixed alternatives, and we derive a uniform finite-sample bound for its regularised estimate. We show that, with the learned kernel fixed, a held-out test is asymptotically valid under the null given suitable nuisance-estimation rates. Empirically, DR-KTE-Deep achieves substantially higher power than fixed and bandwidth-optimised kernel baselines across image, text and graph outcomes, while maintaining well-calibrated Type-I error.
[LG-102] Quantum Evolution Kernel for Node-Attributed Graphs
链接: https://arxiv.org/abs/2610.36984
作者: Manuel Padovani,Alessandra Di Pierro,Claudia Daffara
类目: Quantum Physics (quant-ph); Machine Learning (cs.LG)
*备注:
Abstract:We present a novel approach to embed node-attributed graphs onto neutral-atom quantum computers, and propose an application to digit recognition. Our approach extends PASQAL’s quantum evolution kernel (QEK) by introducing a node-level feature encoding based on three-level quantum states. To this purpose, we present a prototype showing how the resulting embedding can be leveraged to extract correlation-based information for classification tasks. The entire quantum process is simulated using Pasqal’s Pulser library, which inevitably imposes constraints on the size of the experimental datasets. We compare our method with classical baselines to evaluate its potential. We show that, for the digit recognition task, our method achieves a higher mean accuracy than the more commonly used Jensen-Shannon distance-based methods by utilizing the Wasserstein distance.
[LG-103] RAG -CSI: Global Local Retrieval via Joint State and Dynamics Matching for CSI Prediction
链接: https://arxiv.org/abs/2610.36881
作者: Xinyue Jiang,Tong Li,Shiceng Zhang,Ke Chen,Zhu Xiao
类目: ignal Processing (eess.SP); Machine Learning (cs.LG)
*备注:
Abstract:Channel state information (CSI) prediction forecasts future channel states from historical observations, supporting transceiver design and resource allocation with reduced measurement overhead. However, existing deep learning predictors mainly encode channel knowledge in model parameters, which can limit their ability to represent diverse patterns of channel dynamics across scenes. In this paper, we propose RAG-CSI, a retrieval-enhanced Transformer framework that augments prediction with relevant historical channel knowledge. Specifically, we construct a global–local CSI knowledge store containing representative observations across scenes and historical observations from the current serving scene. We further develop joint state–dynamics retrieval (JSDR), which jointly characterizes channel states, phase dynamics, and amplitude dynamics to retrieve relevant CSI sequences. A Transformer predictor then uses cross-attention to fuse the retrieved knowledge with the current CSI history and predict future CSI with limited additional model parameters. Experiments on a multi-scene CSI dataset show overall improvements exceeding 10% across all five evaluation metrics over the best baseline for each metric, including a 15.42% reduction in normalized mean squared error.
[LG-104] Intrinsic Information-Theoretic Analysis of ReLU Nets
链接: https://arxiv.org/abs/2610.36864
作者: Johan Mylius-Kroken,Elisabeth Wetzer,Ali Ramezani-Kebrya,Robert Jenssen,Kristoffer Knutsen Wickstrøm
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:Mutual information (MI) is fundamental to representation-learning analysis in deep networks, but in high dimensions MI estimators depend on hyperparameters whose bias often dominates the signal. We show that continuous piecewise-linear (CPWL) networks, including ReLU networks, admit an intrinsic information-theoretic analysis that avoids this pitfall: their inherent partition of the input space into convex linear regions assigns to every input a discrete region label \Pi alongside the continuous representation T and label Y . We introduce a decomposition of the mutual information I(Y;T) into interpretable geometric terms. We show that the dominant term, the routing information I(Y; \Pi) , admits an estimable plug-in estimator, that in the well-sampled regime, approximates a population lower bound on the joint information I(Y; \Pi, T) . When the partition saturates, a functional-equivalence quotient collapses regions implementing the same linear operator and restores informativeness; this extension has a single bounded, dimensionless parameter, which is reported through the quotient’s compression ratio. The routing estimator agrees strongly with three established MI baselines applied to the same networks and data. The functional quotient recovers informative estimates in the regime bordering the well-sampled and under-sampled regime. We provide a framework for analysing the internal information geometry of CPWL networks, grounded in their inherent piecewise-linear structure rather than in external discretisation.
[LG-105] HorizON: A Multi-Horizon Deep Operator Network for Property Prediction of Organic Battery Stacks
链接: https://arxiv.org/abs/2610.36860
作者: Subhash V.S. Ganti,Lukas Woelfel,Arla Diengdoh,Tzvetan Mihaylov,Thomas Nevolianis,Jan Meyer,Christopher Kuenneth
类目: Materials Science (cond-mat.mtrl-sci); Machine Learning (cs.LG)
*备注:
Abstract:Transition-metal-based batteries rely on critical elements such as lithium, cobalt, and nickel. Their extraction creates environmental and supply-chain challenges. Organic electrode materials (OEMs) reduce this reliance and provide broad molecular tunability. However, OEM-based batteries still show limited long-term cycling stability. Their design also requires the joint optimization of coupled metrics such as voltage, specific capacity, capacity retention, and energy density. These metrics depend on the full battery stack and on the interactions among its components rather than on the active material alone. In our previous work, we developed a data-fusion multi-task framework for average discharge voltage and first-cycle specific capacity. Both targets are single-valued. Here, we extend that framework to multi-horizon prediction by predicting a property at every cycle of a discharge trajectory rather than at a single cycle. Our multi-fidelity and Multi-horizon Deep Operator Network learns DFT redox potentials and cycle-dependent specific-capacity trajectories within one model. We train it on complete battery-stack representations and usage conditions, so it learns battery-stack interactions rather than cathode-only effects. This framework supports long-term property prediction, inverse design, and process optimization for organic battery systems.
[LG-106] Boundary Geometry Determines Optimal Shallow ReLU Approximation under Weighted Coefficient Constraints
链接: https://arxiv.org/abs/2610.36854
作者: Jia-Qi Yang,Lei Shi
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Functional Analysis (math.FA); Metric Geometry (math.MG)
*备注: 92 pages, 2 figures, 2 tables
Abstract:We study how boundary geometry affects L^q approximation by shallow ReLU neural networks on convex bodies K\subset\mathbb R^d . The target classes are unit balls of weighted variation spaces that can enlarge classical spaces by incorporating domain geometry into constraints on output coefficients. We establish upper and lower bounds for arbitrary convex bodies and determine the optimal polynomial approximation exponents for all planar convex bodies and their products with polytopes, including cubes and circular cylinders. For C^2 convex bodies, we determine the optimal exponents in a dimension-dependent range of q extending beyond 2 . When all principal curvatures are positive, the result extends to every finite q\ge2 . For these classes, the dependence of the optimal exponent on the domain is captured by a single geometric growth exponent associated with small boundary caps. We also identify sharp weight thresholds for retaining the classical approximation exponent and improve previous weights on Euclidean balls while preserving the classical L^2 approximation rate. When the target is given as a finite ReLU network, the upper bounds can be realized by a compressed network using only the original neurons, without increasing the weighted coefficient sum. These results show that boundary geometry can change the optimal exponent even when dimension, error norm, and weighting rule are fixed.
[LG-107] Joint Distributional Copula Regression for Insurance Claim Frequency and Severity via Boosting
链接: https://arxiv.org/abs/2610.36835
作者: Banghee So,Na Zhang
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Methodology (stat.ME)
*备注:
Abstract:In this paper, we propose a model-based boosting approach for fitting structured additive distributional copula regression models with bivariate responses, focusing particularly on the joint modeling of claim frequency and claim severity. The proposed approach is based on generalized additive models for location, scale, and shape (GAMLSS), which extend the general idea of GAMs by allowing multiple parameters of a response distribution, rather than only its conditional mean, to depend on covariates. The model is therefore intrinsically interpretable at the level of its additive predictors, providing greater transparency and auditability than black-box alternatives. We evaluate the proposed framework using synthetic and real-world data. We demonstrate its ability to recover the underlying covariate effects and dependence structure, identify informative variables, and achieve competitive predictive performance relative to benchmark models. We also illustrate how the fitted model can be interpreted.
[LG-108] Frequency-Uniform Log-Regret Reinforcement Learning for Diffusions
链接: https://arxiv.org/abs/2610.36758
作者: Jose Blanchet,Lutong Hao,Renyuan Xu
类目: Optimization and Control (math.OC); Machine Learning (cs.LG)
*备注:
Abstract:High-frequency control problems are often studied through time discretization, but refining the timestep \delta increases the discrete horizon H=T/\delta and can make standard finite-horizon regret guarantees deteriorate even when the physical horizon T is fixed. We study episodic value-based reinforcement learning for Euler discretizations of controlled diffusions. The performance is measured as the cumulative regret over K episodes relative to the optimal policy of the corresponding Euler model. Under stability and regularity conditions on value function gradients, Bellman residual derivatives, and fitted critics, we prove an \mathcalO(\log K) regret bound with constants uniform in \delta . For dynamics with drift linear in unknown parameters and unknown constant diffusion covariance, we verify localized versions of these conditions for a constrained, model-based controlled-diffusion fitted- Q algorithm (CD-FQ) and obtain high-probability \mathcalO(\log K\log\log K) regret for fixed and finite T , again with constants uniform in \delta .
[LG-109] Neuronal Identity as an Organizational Basis for Analyzing Neural Population Dynamics NEURIPS2026
链接: https://arxiv.org/abs/2610.36745
作者: Seungjae Han,Joshua Yedam You,Soi Kim,Minho Eom,Jinho Park,Seunghee Park,Byeongwook Lee,Young-Gyu Yoon
类目: Neurons and Cognition (q-bio.NC); Machine Learning (cs.LG); Signal Processing (eess.SP)
*备注: Accepted at NeurIPS 2026
Abstract:Individual neurons exhibit diverse molecular, morphological, and electrophysiological properties that shape their distinct roles in neural circuits. While these multimodal features are largely shared across individuals, they remain an underutilized source of information in analyzing neural population dynamics. We posit that these shared biological profiles provide a common reference for relating neural dynamics across subjects and recording sessions. Our framework leverages this structure by routing neurons into latent groups based on their biological profiles and analyzing population dynamics through group-level representations. Because these assignments are anchored in biological properties that are consistent across subjects, the learned organization supports transfer to unseen animals without retraining or neuron-wise correspondence. We evaluate our framework on whole-brain calcium imaging in larval zebrafish and large-scale electrophysiology in the mouse visual cortex. Across both species and recording modalities, the framework delivers state-of-the-art performance in both neural decoding and manifold analysis: it achieves superior decoding of behavior and stimulus labels while better preserving the temporal consistency and local geometric structure of latent trajectories on unseen data. Our experiments demonstrate that not only does providing biological features improve baseline performance, but our proposed framework achieves even stronger results by using these properties to explicitly structure the population rather than treating them as auxiliary inputs. Together, these results suggest that organizing neural populations by their multimodal biological features offers a promising framework for generalizing population dynamics across subjects and recording sessions.
[LG-110] Algorithmics for Graphical Separation Criteria
链接: https://arxiv.org/abs/2610.36502
作者: Moisés Chavira-Flores,Sebastian Weichwald,Leonard Henckel
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注: Accepted at PGM 2026, PMLR, Vol. 343
Abstract:Statistical graphical models are representations of the dependences between the components of a random vector. They encode these dependences via graphical separation criteria. We consider the algorithmic problem of checking separation statements for the various graph classes and separation criteria that have been developed by the community. We establish that all criteria we study can be checked by modular variations of reachability. For all separation criteria but \sigma -separation, this yields a linear-time algorithm. This is effectively the optimal attainable time since loading the graph itself is linear-time in the number of edges in the graph. Using the CIfly framework for deploying modular variations of reachability, we also provide R and Python code implementing our algorithms. We evaluate empirical runtime and scaling behavior in a simulation study.
[LG-111] Second-order Sinkhorn Geometry and sqrtn-Inference
链接: https://arxiv.org/abs/2610.36488
作者: Tristan Luca Saidi,Florian Gunsilius,Gonzalo Mena
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Statistics Theory (math.ST)
*备注:
Abstract:The Sinkhorn divergence, a debiased form of entropic optimal transport, has been widely used in various scientific disciplines due to its computational and statistical tractability. Unlike the Wasserstein distance, however, the Sinkhorn divergence is not a metric and it does not naturally admit a formal Riemannian structure. These properties are central to many applications of optimal transport methods that utilize interpolation, barycentric averaging, parallel transport or other geometric operations. Recent work addresses these shortcomings by constructing a formal Riemannian tangent space and a metric tensor on the space of probability measures that quadratically approximates the Sinkhorn divergence. In this work, we extend this framework with two major threads of contributions. For a suitable class of finite-smoothness cost functions, we prove Hadamard differentiability of the Sinkhorn metric tensor and we derive the Levi-Civita covariant derivative in this geometry; we then use the covariant derivative to obtain the geodesic and parallel transport equations, each of which amount to an ODE on a fixed Reproducing Kernel Hilbert Space (RKHS) with smooth coefficients. Our second contribution is to linearize the solutions to the geodesic and parallel transport equations using standard tools from Banach space ODE theory. We show that under standard regularity conditions these linearizations imply \sqrtn -limiting Gaussian processes centered at population values for the plugin estimators of the exponential map, the logarithmic map, parallel transport and the induced Sinkhorn distance. Moreover, we show that the nonparametric bootstrap is consistent for all objects mentioned. To the best of our knowledge, these are the first \sqrtn --limit theorems and bootstrap consistency results for canonical geometric operations in a transport-based geometry on the space of probability measures.
[LG-112] HELICS: Biobank-scale Conditional Synthetic Genome Generation via Latent Flow Matching NEURIPS2026
链接: https://arxiv.org/abs/2610.36361
作者: Ahmad Abdel-Azim,Xihong Lin
类目: Genomics (q-bio.GN); Machine Learning (cs.LG)
*备注: NeurIPS 2026
Abstract:Individual-level genotype data underpin genetic discovery, disease-risk modeling, and therapeutic development, yet many critical settings remain data-limited, including underrepresented ancestries, rare or imbalanced disease cohorts, and restricted-access biobanks. Conditional synthetic genome generation would support privacy-aware benchmarking, simulation, and data augmentation, but requires learning an ultra-high-dimensional discrete distribution over hundreds of thousands to tens of millions of correlated single-nucleotide polymorphisms (SNPs), or sites of DNA variation. We introduce HELICS: High-fidelity gEneration via Latent Interpolant Conditional Synthesis, a phasing-free latent flow matching framework for genome-wide SNP-level generation from biobank-scale genotype data. HELICS tokenizes chromosomes with convolutional autoencoders, reducing 600K SNPs to 4,096 continuous latent tokens, and trains a transformer parameterized conditional flow matching model over the concatenated multi-chromosome latent representation. Trained on over 417K genomes in the UK Biobank, HELICS generates genome-wide common-variant synthetic cohorts without haplotype phasing and achieves stronger fidelity to real genotypes than existing haplotype-based simulators, including high allele-frequency agreement in external evaluation and superior preservation of local and long-range SNP correlation structure. Conditioning on ancestry and genetic liability summaries enables targeted generation across genetic risk profiles; in perturbation analyses, by increasing the conditioned disease trait liability, HELICS induces SNP-level dosage changes in generated genomes that are highly correlated with published association weights. HELICS provides a scalable generative modeling framework for SNP-level genome data and a path toward conditional synthetic cohorts for data-limited genetic studies.
[LG-113] Balanced Active Subspaces for Optimal Control via Trajectory Optimization
链接: https://arxiv.org/abs/2610.36342
作者: Emmanuel S. Ameh,Samuel E. Otto
类目: Optimization and Control (math.OC); Machine Learning (cs.LG); Dynamical Systems (math.DS); Numerical Analysis (math.NA)
*备注:
Abstract:The curse of dimensionality associated with solving the Hamilton-Jacobi-Bellman equation for the value function remains a key obstacle for real-time optimal feedback control of high-dimensional nonlinear dynamical systems. This is especially severe in aerospace applications such as relaminarizing unsteady turbulent flows and recovering from aerodynamic stall where nonlinear dynamics play a significant role. We introduce Balanced Active Subspaces for Optimal Control (BASOC), a model reduction framework which seeks to find directions in state space along which the value function is most sensitive, and the distribution of states exhibits maximum variance along optimal trajectories. BASOC leverages trajectory optimization and the Pontryagin minimum principle to obtain state and co-state (value function gradient) data along locally optimal trajectories for constructing state and gradient covariance matrices. Balancing these matrices provides an oblique projection for approximating the value function, which can also be used to construct reduced-order models and surrogate feedback control laws. We show that for linear-quadratic optimal control problems, BASOC becomes equivalent to a snapshot method for initial condition balancing of the optimally controlled closed-loop realization. The framework is demonstrated on a nonlinear three-dimensional system where POD and balanced truncation are known to perform poorly, and on the nonlinear complex Ginzburg-Landau equation to demonstrate feasibility for larger-scale systems.
[LG-114] FASST: A Compact Stateless Atmosphere Model for Long-Term Ocean Coupling NEURIPS2026
链接: https://arxiv.org/abs/2610.36297
作者: Pratham Aggarwal,Duncan Watson-Parris,William E. Chapman
类目: Atmospheric and Oceanic Physics (physics.ao-ph); Machine Learning (cs.LG)
*备注: Accepted at Tackling Climate Change with Machine Learning: workshop at NeurIPS 2026
Abstract:Coupled climate simulations are computationally expensive, largely because of the atmosphere. We introduce FASST, a coupled climate model that retains the prognostic POP2 ocean while replacing the atmosphere and sea ice with lightweight U-Net emulators. The atmospheric emulator is diagnostic, predicting surface fluxes from the evolving ocean-ice state every 6 simulated hours, while the sea-ice emulator advances the ice state monthly. For 140 years of unforced integration, FASST maintains a stable global-mean climate without SST restoring, with post-transient global-mean SST of 291.30 \pm 0.22\mathrmK compared with 291.25 \pm 0.10\mathrmK in CESM 2 piControl. FASST also captures interannual Niño- 3.4 variability, although with weaker amplitude than CESM ( 0.62\mathrmK versus 1.02\mathrmK ). These results demonstrate that a stateless learned atmosphere can sustain stable, free-running coupled climate simulations over century-scale integrations.
[LG-115] Approximation Families and the Statistical Efficiency of Denoising Score Matching
链接: https://arxiv.org/abs/2610.36205
作者: Brian Miyatake
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Statistics Theory (math.ST)
*备注: 45 pages, no figures
Abstract:We study score matching over finite-dimensional approximation spaces generated by gradients of eigenfunctions of the infinitesimal generator of a forward diffusion. We compare polynomial approximation under one-dimensional Ornstein-Uhlenbeck flow with trigonometric-polynomial approximation under heat flow on the torus. In the Ornstein-Uhlenbeck setting, we show that for every non-Gaussian initial distribution uniformly comparable to the standard Gaussian, the asymptotic score-estimation variance grows exponentially along a subsequence of polynomial degrees. Averaging the score-matching loss under the diffusion semigroup, which we call operator-informed score matching (OISM), reduces the variance at each fixed degree but does not eliminate this exponential instability at sufficiently small times. A local minimax lower bound shows that the same high-degree dependence arises when estimating the polynomial projection of the score, rather than being an artifact of OISM. In contrast, under heat flow on the torus, OISM over trigonometric polynomials admits finite-sample control of the score error along the forward diffusion and yields the classical minimax rate for estimating sufficiently smooth densities under expected Kullback-Leibler loss.
[LG-116] Anytime Valid Confidence Bands for Distribution Functions
链接: https://arxiv.org/abs/2610.36149
作者: Bror Hultberg,Dave Zachariah,Antônio H. Ribeiro
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:Nonparametric inference often requires characterizing an unknown function, such as a cumulative distribution function (CDF), across its entire domain rather than a specific point. We develop confidence bands for distribution functions of a data stream that are valid simultaneously over all times and all points, even when each observation may depend arbitrarily on the past. We begin by deriving such bands for the time-averaged conditional CDF and show how their widths are determined by the depth of a binary search tree. We then show that this result is an instance of a framework through which one can construct valid confidence bands uniformly over time and over an arbitrary index set for any indexed family of time-averaged distribution functions. We illustrate the framework on importance-weighted and right-censored CDFs, as well as multivariate CDFs.
[LG-117] Symmetry Breaking Is the Price of Finite Capacity
链接: https://arxiv.org/abs/2610.36138
作者: Maggie Chlon,Leon Chlon
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:If rotating a problem never changes its answer, one might expect its best model to be rotation-invariant; we show that this fails, predictably and by an exact amount, when capacity is limited. A symmetry groups a model’s directions into blocks of equally strong directions, a symmetric model must use each block whole, and when the budget cuts a block every optimal model breaks the symmetry. For a Gaussian problem with a k -unit bilinear predictor, the log-loss price of symmetry is exactly (U_k-I_k)/(2\sigma^2) , the signal of the best k directions minus that of the best whole blocks. Because an attention head compares tokens through a similarity of rank at most its head dimension, whatever computation precedes it, the same price governs heads fitted to a symmetric similarity by squared error, and heads trained this way on three-dimensional directions break symmetry at exactly the predicted head dimensions, with the predicted rotation sensitivity (RMSE 0.004). Under softmax attention, flatter attention restores the predicted symmetric head where tested, while with peaked attention broken heads beat every whole-shell symmetric head at 16 of 20 head dimensions. We give a procedure that separates capacity-driven breaking from other causes.
[LG-118] SuperJEPA: Pretraining with a Supervision-Augmented Joint-Embedding Predictive Architecture for High Energy Physics
链接: https://arxiv.org/abs/2610.36121
作者: Dylan Sheldon Rankin,Ho Fung Tsoi
类目: High Energy Physics - Experiment (hep-ex); Machine Learning (cs.LG)
*备注: 20 pages, 3 figures. Submission to SciPost
Abstract:Foundation models often obtain powerful results through very large unlabeled training datasets that dwarf available labeled datasets in size, but high energy physics already provides large samples of high-fidelity labeled simulation. We therefore ask whether self-supervised objectives improve over supervised pretraining alone when large labeled samples already exist. We introduce SuperJEPA, a joint-embedding predictive architec- ture augmented with a supervised objective, and find improved downstream jet-tagging performance across multiple tasks, model sizes, and pretraining dataset sizes. The gains persist for both frozen and fine-tuned encoders, motivating predictive self-supervision in future large-scale particle-physics foundation models.
[LG-119] Learning on-the-fly quasi-stationarity diagnostics from low-dimensional surrogates
链接: https://arxiv.org/abs/2610.36078
作者: Noé Blassel
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Probability (math.PR)
*备注: 91 pages, 18 figures, 3 tables
Abstract:Accelerated dynamics methods, such as parallel replica dynamics, require estimates of the decorrelation time, after which a process trapped in one of its metastable states is distributed according to the quasi-stationary distribution (QSD). In a method proposed by Binder, Lelièvre and Simpson, decorrelation times are defined by coupling a particle approximation of the non-linear conditioned probability flow of the process with MCMC-inspired diagnostics. We explore an alternative to MCMC diagnostics, based on coarse-graining of non-linear probability flows and simulation-based inference. We first develop a coarse-graining framework for killed Markov jump processes conditioned on survival, proving uniform-in-time, non-asymptotic estimates approximating the non-linear probability flow of the true process by the non-linear flow of a surrogate process, defined with a collective variable. The framework does not assume reversibility of the true process, and also gives bounds which scale explicitly in an adiabatic regime. Motivated by these results, we then construct a quasi-stationarity diagnostic using simulation-based inference: a recurrent neural network (RNN), trained on particle approximations of non-linear effective flows, estimates posterior probabilities of decorrelation. The learned diagnostic is then transferred to collective-variable traces of high-dimensional particle approximations. We investigate the behavior and robustness of this diagnostic on a set of numerical experiments, comparing it to the MCMC diagnostic of Binder et al.\ on both synthetic examples and a molecular system, in and out of its training prior distribution. When its prior assumptions hold, the RNN diagnostic outperforms the MCMC diagnostic in ROC analysis, is less conservative, fails less often, and has an interpretable decision threshold. Comments: 91 pages, 18 figures, 3 tables Subjects: Machine Learning (stat.ML); Machine Learning (cs.LG); Probability (math.PR) Cite as: arXiv:2610.36078 [stat.ML] (or arXiv:2610.36078v1 [stat.ML] for this version) https://doi.org/10.48550/arXiv.2610.36078 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Noé Blassel [view email] [v1] Mon, 26 Oct 2026 18:03:40 UTC (2,919 KB) Full-text links: Access Paper: View a PDF of the paper titled Learning on-the-fly quasi-stationarity diagnostics from low-dimensional surrogates, by No’e BlasselView PDFHTML (experimental)TeX Source view license Current browse context: stat.ML prev | next new | recent | 2026-10 Change to browse by: cs cs.LG math math.PR stat References Citations SciXGoogle Scholar Semantic Scholar export BibTeX citation Loading… BibTeX formatted citation loading… Data provided by: Bookmark checked="checked"class=“labs-tab-input”> Bibliographic Tools Bibliographic and Citation Tools Bibliographic Explorer Toggle Bibliographic Explorer (What is the Explorer?) Connected Papers Toggle Connected Papers (What is Connected Papers?) Litmaps Toggle Litmaps (What is Litmaps?) scite.ai Toggle scite Smart Citations (What are Smart Citations?) Code, Data, Media Code, Data and Media Associated with this Article alphaXiv Toggle alphaXiv (What is alphaXiv?) Links to Code Toggle CatalyzeX Code Finder for Papers (What is CatalyzeX?) DagsHub Toggle DagsHub (What is DagsHub?) GotitPub Toggle Gotit.pub (What is GotitPub?) Huggingface Toggle Hugging Face (What is Huggingface?) ScienceCast Toggle ScienceCast (What is ScienceCast?) Demos Demos Replicate Toggle Replicate (What is Replicate?) Spaces Toggle Hugging Face Spaces (What is Spaces?) Spaces Toggle TXYZ.AI (What is TXYZ.AI?) Related Papers Recommenders and Search Tools Link to Influence Flower Influence Flower (What are Influence Flowers?) Core recommender toggle CORE Recommender (What is CORE?) Author Venue Institution Topic About arXivLabs arXivLabs: experimental projects with community collaborators arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv’s community? Learn more about arXivLabs. Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?) mathjaxToggle(); We gratefully acknowledge support from our major funders, member institutions, , and all contributors. About Help Contact Subscribe Copyright Privacy Accessibility Operational Status (opens in new tab)
[LG-120] he proton structure one mode at a time: a Neural Tangent Kernel analysis of a DIS fit
链接: https://arxiv.org/abs/2610.36034
作者: Amedeo Chiefa,Luigi Del Debbio,Richard Kenway
类目: High Energy Physics - Phenomenology (hep-ph); Machine Learning (cs.LG)
*备注: 70 pages, 37 figures
Abstract:We study the training dynamics of a feed-forward neural network (NN) through the analysis of the neural tangent kernel (NTK) in a phenomenologically relevant fit of parton distribution functions (PDFs) from deep inelastic scattering (DIS) data. The exact integration of the gradient flow derived in a previous single-flavour analysis is generalised to the multi-flavour case, yielding an analytical perspective on the relationship between data and methodology during training. The Shapley value recently used in the context of PDFs is derived within our statistical framework for observables that depend linearly on the PDFs, and we outline a geometric interpretation of training. We expose the role of gradient-based optimisation algorithms and of the parametrisation in the regularisation of the ill-posed inverse problem that affects the determination of PDFs from experimental data, demonstrating that the regularisation amounts to the incremental inclusion of NTK modes, each learned at the timescale set by the corresponding eigenvalue. We find that only a few leading modes of the NTK are necessary to reconstruct the target PDFs, while the noise of the data is carried by the sub-leading ones learned at later stages of training. Finally, we consider the impact of the initialisation on the training dynamics, and recover a transient behaviour observed in the previous analysis and in the machine learning (ML) literature. Beyond sharpening our understanding of the training dynamics, our framework can be applied to other parametrisations (such as fixed functional forms) and ML applications in physics, as well as to stress tests in the presence of inconsistent data.
[LG-121] Causal DAG-Guided Reinforcement Learning for Long-Term Physical Activity Maintenance: Using Prior Data to Inform Future Deployment
链接: https://arxiv.org/abs/2610.36019
作者: Xueqing Liu,Young Hyun Cho,Bhanu Teja Gullapalli,Predrag Klasnja,Susan Murphy
类目: Applications (stat.AP); Machine Learning (cs.LG)
*备注:
Abstract:Digital health interventions can provide scalable, personalized support for physical activity (PA), yet sustaining PA over the long-term remains difficult. Because affective experiences with PA are important for activity sustainment, we developed an online reinforcement learning (RL) system for the forthcoming ADAPT factorial trial that personalizes twice-daily activity suggestions aimed at improving affective associations with PA. This setting poses several challenges: participant-specific data are scarce relative to the candidate state space, important states and rewards are partially observed, interventions are frequent while the primary outcome is measured weekly, and activity suggestions can both improve PA-related experiences and increase intervention burden. We address these challenges using an expert-specified causal directed acyclic graph as an integrated RL design tool, guiding state construction, estimation of partially observed states, delayed credit assignment, and representation of beneficial and adverse intervention pathways. We further construct a longitudinal simulator calibrated from a prior micro-randomized trial to compare candidate RL designs under realistic state evolution, missingness, and participant heterogeneity before deployment. Together, this work illustrates how causal domain knowledge and prior trial data can make long-horizon, participant-specific RL more practical for digital health interventions targeting sustained behavior change.
[LG-122] Anchored Stein Variational Gradient Descent Algorithm for Non-smooth Sampling
链接: https://arxiv.org/abs/2610.36018
作者: Thanh Dang,Matthew X. Burns,Jiaming Liang
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:We introduce a Stein variational framework for sampling from a target density p\propto e^-U when the target potential U is possibly non-smooth. The key idea is to introduce a smoother reference potential U_0 and the anchor weight a=e^U-U_0 . The resulting anchored Stein variational gradient descent (SVGD) algorithm is able to target p\propto e^-U without computing the gradient of U and include standard SVGD as a special case. In particular, we allow the target potential U to be locally Lipschitz while the reference potential U_0 is continuously differentiable, along with other suitable assumptions. Under this low regularity regime, we establish a Kullback-Leibler first variation formula for anchored perturbations, identify the corresponding steepest descent direction in the reproducing kernel Hilbert space and prove a nonasymptotic convergence bound for the anchored SVGD algorithm in terms of an anchored kernelized Stein discrepancy. Finally, we also provide guidance on constructing a suitable reference potential U_0 based on U . Subjects: Machine Learning (stat.ML); Machine Learning (cs.LG) Cite as: arXiv:2610.36018 [stat.ML] (or arXiv:2610.36018v1 [stat.ML] for this version) https://doi.org/10.48550/arXiv.2610.36018 Focus to learn more arXiv-issued DOI via DataCite
[LG-123] Geometric priors for single-patch neural Calabi–Yau metrics on the quintic: a conservative fixed-ψ Ablation study
链接: https://arxiv.org/abs/2610.35920
作者: D Yang Eng
类目: Algebraic Geometry (math.AG); Machine Learning (cs.LG)
*备注:
Abstract:We investigate how geometric priors shape neural approximations of the complex Monge–Amp`ere equation on the quintic Calabi–Yau threefold. This fixed- \psi diagnostic compares four single-patch ans"atze: a coordinate SIREN potential model, a direct-metric Cholesky baseline, an algebraic section ansatz, and a section-plus-neural hybrid. In a 96-run learning-rate and frequency archive, section-based models reach validation \sigma=4.14 – 4.16% , whereas SIREN and Cholesky remain at 41.29% and 44.27% . Multi-seed controls use a shared affine- \eta evaluator and five fresh validation clouds. Stable section settings give 3.84 – 4.77% , compared with 28.46 – 30.59% for Cholesky and 34.73 – 38.13% for SIREN. Hybrid projective features do not improve best-case accuracy, but they stabilise the stressed regime. They limit \sigma to 4.26 – 4.41% , versus 4.45 – 205.45% for raw coordinates. Within this protocol, section-based geometric structure is the main driver of accuracy; coordinate-only models plateau at substantially higher error, and raw hybrid flexibility can amplify instability. We provide a projective-scaling derivation, document protocol asymmetries, and identify projectively consistent hybrid corrections as the next step.
[LG-124] Reconstructing the Human Development Index at the Individual Level from Survey Microdata: Development and Validation of a Methodology
链接: https://arxiv.org/abs/2610.35893
作者: Tcharlies Bachmann Schmitz(PX Center, Brazil),Natalia Veronez da Cunha(Universidade do Planalto Catarinense, UNIPLAC),Isabelle Ferreira(Universidade do Planalto Catarinense, UNIPLAC)
类目: Applications (stat.AP); Machine Learning (cs.LG); General Economics (econ.GN)
*备注: 12 pages, 3 figures. Replication notebook and data archived at this https URL
Abstract:This study proposes and validates a methodology for reconstructing the Human Development Index (HDI) from individual-level survey microdata, enabling disaggregated estimates for demographic subpopulations while maintaining coherence with the official HDI. Data from 40,185 respondents in 21 countries from the European Social Survey (ESS7, 2014) were analysed, combined with World Bank indicators for life expectancy, gross national income per capita, and expected years of schooling. The health dimension was estimated from national life expectancy, adjusted by self-rated health; income, by a lognormal multiplier calibrated by the Gini coefficient; and education, by combining observed and expected years of schooling. The dimensions were aggregated by geometric mean, weighted by sampling weights, with confidence intervals estimated via the delta method. Among the 39,706 individuals with valid data (98.8%), the reconstructed HDI showed high agreement with the official 2014 HDI, with a mean absolute error of 0.0099 and a Pearson correlation of 0.917. Consistent differences across subgroups evidenced the methodology’s convergent validity and its capacity to produce disaggregated estimates of human development, expanding the HDI’s potential for analyses of socioeconomic and demographic inequalities.
附件下载
