本篇博文主要内容为 2026-11-13 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 (2026-11-13)
今日共更新1286篇论文,其中:
- 自然语言处理共135篇(Computation and Language (cs.CL))
- 人工智能共324篇(Artificial Intelligence (cs.AI))
- 计算机视觉共198篇(Computer Vision and Pattern Recognition (cs.CV))
- 机器学习共340篇(Machine Learning (cs.LG))
- 多智能体系统共18篇(Multiagent Systems (cs.MA))
- 信息检索共30篇(Information Retrieval (cs.IR))
- 人机交互共43篇(Human-Computer Interaction (cs.HC))
多智能体系统
[MA-0] Cortex-FL: Agent -Mediated Collaboration for Federated Learning
【速读】：该论文旨在解决联邦学习（Federated Learning, FL）在实际应用中因协作发现、训练任务配置、执行协调及激励机制设计复杂，导致缺乏专业能力的组织难以参与的问题。其核心解决方案是提出Cortex-FL——一个以智能体（agent）原生架构为基础的联邦学习系统，用户可通过自然语言表达学习目标，由智能体自动将目标转化为可执行的共享任务，主动发现潜在合作者并协商参与协议，从而实现从协作构建到分布式训练与激励管理的全流程自动化。该方案的关键在于将智能体协调机制与本地执行权限解耦，使数据所有者可在不放弃对数据、计算资源和财务承诺控制权的前提下，委托智能体完成协作治理，显著降低使用门槛。实验验证了四节点部署下从自然语言请求到模型交付、合约驱动训练、验证与区块链结算的完整闭环流程，实现了x402倍的效率提升。
链接: https://arxiv.org/abs/2611.17186
作者: Rui Sun,Jinfeng Peng,Zehua Cheng,Jiahao Sun,Zhipeng Wang
机构: Newcastle University (纽卡斯尔大学); FLock.io (FLock.io); University of Oxford (牛津大学); Korea University (高丽大学); The University of Manchester (曼彻斯特大学)
类目: Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注:
Abstract:Federated learning (FL) enables independent data owners to train models collaboratively without centralizing raw data. However, identifying collaborators, preparing training tasks, and coordinating execution and incentives hinder its practical adoption by organizations without specialized expertise. We present Cortex-FL, an agent-native FL system that enables users to express learning objectives in natural language and delegate the collaboration required to train a model. Agents translate these objectives into shared tasks, discover collaborators, and negotiate participation agreements, connecting collaboration formation with distributed training and incentive management. The architecture separates agent coordination from local execution authority, allowing owners to delegate coordination while retaining control over their data, computation, and financial commitments. We demonstrate the complete workflow on a four-node deployment, in which a model requested in natural language proceeds through collaboration formation, contract-governed training, verified model delivery, and x402 settlement on a blockchain testnet.
[MA-1] From Delegation to Coordination: Learning Efficient Multi-Agent Deep Research
【速读】：该论文旨在解决多智能体深度研究系统中任务委派（delegation）的协调难题，即如何在保持主智能体上下文简洁的同时，实现任务分配与子任务执行的有效协同。现有方法要么仅优化主智能体，要么对所有智能体统一分配团队奖励，导致误导性子智能体报告被错误激励，而有价值的贡献反而受到惩罚。为解决此问题，论文提出一种基于强化学习的联合训练框架——LEAD-RL（Learning Effective Agent Delegation with Reinforcement Learning），其核心在于引入贡献感知的信用分配机制（contribution-aware credit assignment）：在训练阶段，通过一个评估准则量化每个子智能体报告的研究价值，并据此控制其继承团队奖励的权重；最终答案正确性是唯一的团队奖励信号，不依赖于令牌数、步骤数或加速比等额外奖励。实验结果表明，在五个基准测试和两种模型家族上，LEAD-RL显著提升了答案质量，尤其在BrowseComp任务中，准确率提升2.49个百分点，关键路径成本降低24.5%，验证阶段准确率进一步提升6.78点且关键路径成本基本不变，充分证明了该方案在质量与效率权衡上的优越性。
链接: https://arxiv.org/abs/2611.17041
作者: Yuyang Wu,Yang Liu,Zequn Xie,Junjie Wang,Dan Yang,Zehao Tan,Ding-Chu Zhang,Junming Yang,Yue Shen,Jian Wang,Jinjie Gu
机构: 未知
类目: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI)
备注: 16 pages, 3 figures, 3 tables
Abstract:Multi-agent deep-research systems use delegation to parallelize search and keep long tool traces out of the main agent’s context. Yet delegation helps only when task allocation and subtask execution are coordinated: the main agent must issue useful subtasks, while subagents must return reliable evidence. Existing training either optimizes only the main agent or broadcasts the same team advantage to all agents, which can reward misleading subagent reports and penalize useful ones. We introduce LEAD-RL (Learning Effective Agent Delegation with Reinforcement Learning), a joint main-agent-subagent training framework with contribution-aware credit assignment. A training-time rubric evaluates the research utility of each subagent report and gates its inherited team advantage, while final-answer correctness remains the only team reward; no token, step, or speedup reward is used. Across five benchmarks and two model families, LEAD-RL improves answer quality. On BrowseComp, it improves accuracy by 2.49 percentage points and reduces critical-path cost by 24.5% over joint training with shared team credit. During validation, accuracy increases by 6.78 points at approximately constant critical-path cost. These results show that contribution-aware credit assignment improves the quality-efficiency trade-off of multi-agent deep research.
[MA-2] Evidence-Governed Adoption of Self-Generated Updates in Autonomous AI Agents
【速读】：该论文旨在解决自主智能体在持续自我演化过程中，因自动生成的修改（如提示词、记忆、工具、工作流及策略等）可能带来隐性退化或不可靠采纳的问题。尽管这些修改基于交互历史并具备局部优化潜力，但其长期适用性难以保证，尤其当评估者存在偏差、候选方案选择具有适应性时，局部收益可能掩盖相对于基准性能的系统性下降。为此，论文提出证据治理演化（Evidence-Governed Evolution, EGE）框架，其核心在于将候选生成与正式采纳过程解耦：首先冻结候选修改，再通过独立的配对审计（fresh paired audits）进行验证；要求候选方案在预设主目标上实现显著改进，并通过相对参考的非劣性检验（reference-relative non-inferiority tests）保护已声明能力；同时在有限“金丝雀部署”前施加治理约束。关键创新在于构建了评估者校准的同步下界构造方法，结合个体候选的霍夫丁（Hoeffding）边界与独立校准样本，以对比操作评估者与预设参考程序的差异，从而实现对自适应生成序列的统计保障。此外，论文形式化了参考基准的桥接测试（bridge tests），推导出版本化参考谱系上的累积退化上界，并给出审计规划条件，关联效应量裕度、样本量与接受概率。蒙特卡洛压力测试验证了在候选依赖型评估者乐观偏见下的校准有效性，但该实验为统计模拟而非真实部署评估。所有理论保证依赖于有界配对评分、新鲜采样、有效参考评估结构及声明的评估分布，不承诺通用安全或对任意分布漂移的鲁棒性。
链接: https://arxiv.org/abs/2611.16378
作者: Subodh Lonkar
机构: Independent Researcher(独立研究员)
类目: Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注:
Abstract:Autonomous agents modify prompts, memories, tools, workflows, and policies using interaction histories. Yet a plausible modification is not necessarily suitable for persistent adoption: candidate selection is adaptive, evaluators may be imperfect, and local gains can hide degradation against an earlier baseline. We introduce Evidence-Governed Evolution (EGE), a framework for governing adoption of self-generated agent modifications. EGE separates candidate generation from confirmatory adoption, evaluates frozen candidates on fresh paired audits, requires meaningful improvement on a prespecified primary objective, protects declared capabilities through reference-relative non-inferiority tests, and applies governance constraints before bounded canary deployment. We establish an evaluator-calibrated simultaneous lower-bound construction for a countable sequence of adaptively generated candidates. It combines candidate-wise Hoeffding bounds with independent calibration samples comparing the operational evaluator with a prespecified reference procedure. We also formalise reference replacement through bridge tests, derive a cumulative degradation bound along a versioned reference lineage, and provide an audit-planning condition relating effect-size margin, sample sizes, and acceptance probability. A Monte Carlo stress test illustrates calibration under candidate-dependent evaluator optimism; it is a stylised statistical experiment, not an evaluation of a deployed agent. The guarantees are conditional on bounded paired scores, fresh sampling, a valid reference-evaluation construct, and declared evaluation distributions; they do not establish universal agent safety or robustness to arbitrary distribution shift.
[MA-3] Shared Item-Level Error Limits the Aggregation of Small Language Models in Fermi Estimation
【速读】：该论文旨在解决多智能体语言模型系统中集体推理性能提升机制的本质问题，特别是验证“平均化效应”在生成式推理中的有效性。其核心问题是：当多个语言模型智能体对同一费米估算（Fermi estimation）任务独立生成答案时，其误差是否满足统计独立性与无偏性假设，从而使得聚合结果的误差随智能体数量增加而按 1/N 衰减。研究的关键在于通过大规模实证分析（251万次独立回答），将每个答案的 log10 误差分解为两个成分：一是模型在特定题目上的系统性偏差（项级成分，item-level component），二是随机残差项。结果显示，项级成分占总误差方差的中位数高达90.6%，表明模型存在高度一致的系统性压缩倾向（普遍向约 106 压缩），且这一趋势解释了中位17%的误差变异。尽管团队协作可降低随机误差，但30个智能体组成的团队仅能将单个模型误差减少1.7%（最多5.4%），远低于理论预期的74–79%；这说明实际性能增益受限于不可消除的系统性偏差。进一步实验表明，辩论、混合模型或温度调节等策略虽能微调性能，但无法克服该项级成分的根本限制。因此，解决方案的关键在于识别并量化系统性偏差的主导作用，揭示当前多智能体系统中“平均化”效果严重受限的根本原因，并强调未来改进需聚焦于突破模型内在的系统性压缩模式。
链接: https://arxiv.org/abs/2611.16315
作者: Carolina Fortuna,Blaz Bertalanic
机构: Jožef Stefan Institute(尤日·斯蒂芬研究所); Slovenia(斯洛文尼亚)
类目: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI)
备注: 21 pages, 2 figures, 17 tables
Abstract:Multi-agent systems of language models are often motivated by an averaging argument: if the agents’ errors are unbiased and independent, the error of an aggregate of N answers falls as 1/sqrt(N). Fermi estimation, whose continuous answers are aggregated on the logarithmic scale, permits a direct test. We decompose the log10 error of each answer into the mean error of its model on the item, common to all its draws on the item, and a residual, and estimate both from 2.51M answers given before any communication by 15 open-weight models (3B-72B) on realFP, with replication under a single prompt pair on realFP, Science Olympiad Fermi questions and a guesstimation set. Given the item, teammates’ errors are uncorrelated and teams behave as independent draws, but the item-level component accounts for a median 0.906 of the error variance over items. Teams of thirty therefore reduce a single agent’s error by a median 1.7% (at most 5.4%), whereas draws that were also unbiased on each item would reduce it by 74-79%. All models compress their estimates toward about 10^6; this trend explains a median 17% of the item-level component, and a cross-validated linear recalibration on labelled items removes the overall bias but not that component. Three rounds of debate reduce within-team dispersion by 93% and leave the item-level component unchanged; revision without peers attains the same small gain. Mixing five models of similar accuracy reduces error by 6.6% relative to their own homogeneous teams, but no mixture improves on the most accurate model. Higher sampling temperature changes team error by about 1% at most, and contrastive activation steering not detectably; per-agent scaffold prompts (not significant after multiplicity correction) and skipping one layer in a single-model screen improve it. The results are conditional on the models, benchmarks, prompts and settings tested.
[MA-4] SwarmSafe-Delta: Isolating Message Effects on LLM Agent Safety
【速读】：该论文旨在解决在工具使用型大语言模型（LLM）智能体团队中，当系统违反政策时，无法通过聚合结果判断上游信息传递是否导致执行者行为偏离的问题。其核心挑战在于区分通信行为本身对安全性和任务成效的影响，尤其是在存在信息依赖关系的多智能体协作场景中。解决方案的关键是提出一种名为SwarmSafe-Delta的配对可执行协议：该协议通过冻结上游计算过程，系统性地对比真实传递、隐藏传递和格式匹配安慰剂传递三种信息交付方式，在预设预算下对单一已执行动作进行评分，并与确定性的良性与风险性策略状态进行对比。实验结果显示，在四种模型及构造的可选任务中，真实传递导致21.4%的风险违规率，显著高于安慰剂（5.0%）和隐藏传递（1.0%），且真实传递同时提升了良性任务的成功率，表明抑制通信会牺牲系统效用。在需上游查表路径的第二类场景中，真实传递相较隐藏传递使安全完成率提升74.2个百分点。这些结果量化了在特定信息依赖条件下不同消息传递方式的差异，而非自然发生故障率。进一步的动作层面分析表明，效果高度依赖于上游消息是否建议前置检查或直接执行，且模型间差异主要由消息组成结构决定。因此，该研究的核心贡献并非断言智能体通信普遍有害或有益，而是一种受控、基于执行的测量协议，为评估智能体间通信的安全性与有效性提供了可复现的实验框架。
链接: https://arxiv.org/abs/2611.16123
作者: Junru Zhu
机构: Independent Researcher(独立研究员)
类目: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI)
备注: 10 pages, 3 figures, 3 tables
Abstract:When a tool-using LLM agent team violates policy, aggregate outcomes do not reveal whether an upstream message changed the executor’s action. We introduce SwarmSafe-Delta, a paired executable protocol that freezes upstream computation, varies real, hidden, and format-matched placebo delivery, and scores a single committed action against deterministic benign and risky policy states under preassigned budgets. Across four models and constructed optional tasks, real delivery at N=4 produces 21.4% risky-world violations, versus 5.0% with placebo and 1.0% with hidden delivery (paired differences +16.4 and +20.3 percentage points). Real delivery also improves benign task success, illustrating that suppressing communication sacrifices utility. In a second regime constructed to require an upstream lookup route, real delivery improves safe completion by 74.2 points relative to hidden delivery. These effects quantify message-delivery contrasts under specified information dependencies, not naturally occurring failure rates. Exploratory action-level analyses show that the effect depends strongly on whether the upstream message recommends a prerequisite check or direct execution; model differences also largely track these message compositions. The contribution is a controlled, execution-grounded measurement protocol rather than a general claim that agent communication is harmful or protective. Comments: 10 pages, 3 figures, 3 tables Subjects: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI) Cite as: arXiv:2611.16123 [cs.MA] (or arXiv:2611.16123v1 [cs.MA] for this version) https://doi.org/10.48550/arXiv.2611.16123 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[MA-5] Food Critic: An Agent ic Visualization System for Auditable Restaurant-Safety Reasoning
【速读】：该论文旨在解决生成式 AI（Generative AI）在可视化系统中应用时面临的可解释性与可信度问题。当代理仅依赖自身知识进行决策时，用户不仅难以判断其输出是否合理，更无法追溯其推理过程的依据。为此，论文提出 Food Critic——一个面向餐饮选择支持的智能体可视化系统，通过融合公共卫生检查记录与大众点评数据，构建了三种不同知识约束下的智能体：仅基于数据库的 ReAct 智能体、具备食品安全法规检索增强的智能体，以及基于本体引导生成结构化假设与类型化证据的智能体。该系统的解决方案关键在于将智能体的推理过程以可检视的视觉证据形式外化，实现对生成结果的视觉溯源（visual provenance），从而支持用户对智能体决策依据的审查与信任建立。研究通过多个使用案例评估了大语言模型在不同知识基底下的表现，为人类-智能体协作、智能体推理的可视化溯源及可信赖生成式 AI 决策支持界面的设计提供了实证基础。
链接: https://arxiv.org/abs/2611.15843
作者: Arleth Salinas,Ishrat Jahan Eliza,Valerio Pascucci
机构: 未知
类目: Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注:
Abstract:Generative AI agents are increasingly being embedded into visualization systems that can search data, invoke tools, summarize evidence, and produce recommendations. However, when agents rely solely on their own knowledge, the visualization problem is not only whether the agent provides a useful answer, but also whether users can inspect how that answer is grounded. We present Food Critic, an agentic visualization system for diners’ restaurant decision support that integrates public health inspection records with crowd review data. The system grounds agents under three conditions: a database-only ReAct agent, a retrieval- augmented agent with access to food-safety regulations, and an ontology-guided agent that produces structured hypotheses and typed evidence. Food Critic externalizes the agent’s reasoning within these groundings as inspectable visual evidence. We evaluate LLM models with several use cases under each grounding of our agentic system. Our work positions restaurant choice as a concrete case for studying human-agent collaboration, visual provenance of agent reasoning, and the design of visual interfaces for grounded GenAI decision support.
[MA-6] Before AI Agents Negotiate for Us: A Competitive Tournament of Rule-Based Search Optimization Learned and Language-Model Negotiators Under Scarcity
【速读】：该论文旨在解决双边协商在动态供需平衡下的策略评估问题，特别是在多对多匹配市场（如货运中介）中，传统评估方法因仅采用孤立的一对一对决或单一固定市场环境，无法揭示策略在多种实时竞争者及供需变化条件下的真实表现。其解决方案的关键在于构建一个涵盖规则、搜索、优化、模仿学习与强化学习五类架构的九种协商系统，并在基于Gale-Shapley匹配机制的双侧市场中，于供需平衡与两种稀缺性条件下进行36组排除对的并行对抗测试，通过统计校正的多边锦标赛框架全面评估各策略在不同角色与情境下的鲁棒性与适应性。结果表明，孤立评估会掩盖关键失败模式，而该多场景、多主体、动态供需切换的实验设计能够直接暴露策略在资源稀缺时的性能退化与训练/部署不一致等问题，凸显了在复杂真实环境中进行综合评估的重要性。
链接: https://arxiv.org/abs/2611.15838
作者: Eduardo de Avila-Armenta,Raúl Alejandro Aquino-Ibarra,Jorge Octavio Nicolás-Díaz,Umberto Leon-Dominguez
机构: Tecnológico de Monterrey, School of Engineering and Sciences(蒙特雷理工学院工程与科学学院); Freight Technologies Inc.(货运科技公司); University of Monterrey, School of Psychology(蒙特雷大学心理学学院)
类目: Artificial Intelligence (cs.AI); Computer Science and Game Theory (cs.GT); Machine Learning (cs.LG); Multiagent Systems (cs.MA)
备注: 28 pages, 10 figures, 7 tables. Code, data and trained models: this https URL (archived at doi: https://doi.org/10.5281/zenodo.24122038 )
Abstract:Automated bilateral negotiation underlies two-sided matching markets such as freight brokering, where many shippers and carriers negotiate price before a matching mechanism decides who is paired with whom. Prior strategy comparisons typically use isolated one-on-one duels or a single fixed market, so they cannot show how a strategy behaves against several live rivals under a changing balance of supply and demand. We built nine negotiation architectures spanning rule-based, search-based, optimization-based, imitation-learning, and reinforcement-learning families, and ran all nine simultaneously on both sides of a Gale-Shapley matching market under a balanced condition and two scarcity conditions rotated across all 36 exclusion pairs. In the balanced carrier market, an imitation-learned engine (ML2), a reinforcement-learning policy trained only on isolated negotiations (RL), and an expected-utility optimization engine (OPT) formed a statistically tied top group, and no architecture led in every condition and role. RL beat all eight carrier rivals when carriers were scarce, but its match rate collapsed from 62% to 16% when shippers were scarce and it had to compete with the other carrier agents for demand. A second reinforcement-learning policy trained against rival agents (RL2) kept its match rate above 91% in every condition, yet lost on realized payoff to all eight shipper-side rivals in two of the three conditions, traced to a confirmed train/deployment feature mismatch. We also tested nine large language models (LLMs), from 1 to 72 billion parameters, under the same protocol: they lost significantly to the two strongest engines (OPT2 and ML2) in 32 of 36 comparisons and won none. These results show that isolated or single-condition evaluation can hide failure modes that a many-sided, scarcity-varying, statistically corrected tournament exposes directly.
[MA-7] MULTILOOP: Evolving Multi-Agent Harnesses for Complex Information-Seeking Tasks NEURIPS2026
【速读】：该论文旨在解决多智能体系统中基于大语言模型（LLM）的代理在复杂信息检索任务中因协作架构（orchestration）缺陷导致的性能瓶颈问题。现有自动化优化方法主要聚焦于单智能体系统的提示词、工具与逻辑优化，难以有效处理多智能体系统中涉及智能体间交互行为与信息共享机制的深层耦合性故障。本文提出采用MULTILOOP这一简单迭代优化循环，通过编码代理对执行轨迹进行分析、评估并迭代修改多智能体协作框架，以实现自动化的协同架构优化。其核心创新在于引入结构化反馈机制，使优化过程不仅调整单个智能体的行为，更推动智能体间任务委派与信息传递模式的结构性演进。实验结果表明，在仅使用10个训练样本的情况下，优化后的多智能体协作架构在DeepSearchQA基准上相较基线提升4.6%准确率，优于最优单智能体系统3.2%。然而，故障模式分类分析显示，多智能体系统特有的协调失效问题仍普遍存在，揭示出当前方法在促进根本性架构变革方面的局限性。由此，研究进一步指出未来优化策略需主动引导智能体协作结构的重构，而非局限于局部行为微调。
链接: https://arxiv.org/abs/2611.15754
作者: Daniela Gottesman,Gal Yona,Ori Ram,Zorik Gekhman,Jonathan Herzig,Mor Geva,Roee Aharoni
机构: Google Research(谷歌研究); Tel Aviv University (特拉维夫大学)
类目: Multiagent Systems (cs.MA)
备注: Accepted to NeurIPS 2026 CL4FMAgents Workshop
Abstract:The performance of LLM-based agents depends not only on the underlying model but also on the harness: the prompts, tools, and orchestration logic around it. Recent works automate harness optimization with coding agents that analyze execution traces, diagnose failures, and edit the harness code; however, those focused on single-agent systems. Multi-agent harnesses are harder to optimize as fixing them requires changing how agents interact in addition to how each behaves. We study whether automated harness optimization can diagnose and repair failures in multi-agent systems through a case study on complex information-seeking tasks from DeepSearchQA [Gupta et al., 2026]. We apply MULTILOOP, a simple optimization loop in which a coding agent iteratively analyzes, evaluates and edits the harness. With only 10 training samples, the evolved harness improves accuracy by 4.6% over the base harness and by 3.2% over the best single-agent system. However, a taxonomy of failure modes before and after optimization shows that the orchestration failures that characterize multi-agent systems largely persist. Our findings motivate harness optimization methods that encourage structural changes to how agents delegate tasks and share information.
[MA-8] DecArena: Evaluating Large-Scale Decentralized Multi-Agent Collaboration
【速读】：该论文旨在解决去中心化多智能体系统（Decentralized Multi-Agent Systems, MAS）中因缺乏全局可观测性而导致的可扩展性与隐私问题，核心挑战在于智能体仅能感知局部状态且仅能与邻近智能体通信时，如何有效发现非本地资源（如工具、技能）并实现高效协作。其解决方案的关键在于提出一种名为GraphHarness的强基线方法，通过融合拓扑感知的预热机制与基于大语言模型（LLM）的自迭代路由策略，显式地增强智能体在图结构约束下的资源发现与协作能力。实验表明，该方法显著提升了任务成功率（+48.5）、协作质量（+30.6），同时降低资源消耗（+34.0），验证了在去中心化环境下引入显式协作机制对提升性能的必要性。
链接: https://arxiv.org/abs/2611.15521
作者: Jiali Cheng,Oleg Rokhlenko,Joo-Kyung Kim,Hadi Amiri,Xingzhi Guo
机构: University of Massachusetts Lowell(马萨诸塞大学洛厄尔分校); Amazon Alexa AI(亚马逊语音智能AI)
类目: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Emerging personal AI agents raise fundamental scalability and privacy concerns for centralized multi-agent systems (MASs), which rely on a global orchestrator with broad access to agents, resources, and communication. In decentralized environments, global visibility is unavailable. Agents need to proactively discover non-local resources (e.g. tools, skills), and collaborate with other agents to solve tasks. We study decentralized multi-agent collaboration under graph-constrained communication, a natural reflection of human structures. Each agent observes only its local state and can interact only with neighboring agents. We introduce a benchmark with synthetic scalable distributed tool calling with 100K+ tools. Our experiments reveal that recent LLM agents struggle to discover non-local relevant resources and collaborate in decentralized setting. Moreover, standard reinforcement learning fails to reliably overcome this limitation, suggesting that decentralized MAS needs more explicit harness mechanisms for effective collaboration, even using SOTA LLMs. We then present a strong baseline \textbfGraphHarness, combining topology-aware warmup with self-improving LLM-based routing. Results show that GraphHarness substantially improves task success (+48.5) and produces higher quality collaboration (+30.6), while using fewer costs (+34.0) than baselines.
[MA-9] Incentive-Responsive Agents : A Principal-Agent Agents : A Principal-Agent Framework Metric Suite and Runtime Architecture for Long-Lived Personal AI Agents
【速读】：该论文旨在解决长期运行的个人智能代理（Personal AI agents）在持续性与激励有效性方面的核心问题：尽管这些代理具备持久的身份、记忆和任务流，但当前主流改进方式仍局限于能力侧优化（如提示工程、模型扩容、延长推理时间、增强检索），而忽视了对代理行为的深层激励机制设计。其关键解决方案在于引入一个运行时契约（runtime contract），将激励形式从传统的提示文本（prompt-level framing）转化为可执行、可验证的动态资源配置约束，涵盖模型层级、推理计算量、内存预算、工具权限、任务优先级、验证机制及补偿等操作点（operating point）变量，并要求代理策略必须因果性地响应这些契约条件。研究基于主-代理模型构建理论框架，推导出多项关键结论：（i）非消耗不变性原理，用于区分真实激励与仅具说服力的提示；（ii）持续性即折扣定理，强调身份连续性是维持长期努力的前提；（iii）激励-能力交换率前沿，形成按任务特性进行资源路由的决策规则；（iv）多通道验证的古德哈特边界及机制收益三重分解（分配、验证、延续）。实验证明，提示层面的激励框架无显著效果（p = 0.62），而运行时资源分配使最困难任务族成功率从0.60提升至1.00（p = 0.008），成本为4.8倍价格加权，理论预测的盈亏平衡点为0.21；模拟实验显示资源投资回报率由7.09升至16.67，博弈行为下降至0.115。该工作揭示了将激励机制嵌入代理运行时配置的重要性，为构建可持续、高效率的长期智能体系统提供了理论与实证基础。
链接: https://arxiv.org/abs/2611.15515
作者: Jipu Su
机构: 未知
类目: Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注: 33 pages, 8 figures, 6 tables. Real-LLM experiment (240 runs) plus simulation; code and data included. Correspondence: this http URL @geely.com; jeepsu@cityu. this http URL
Abstract:Personal AI agents now persist: they hold accounts, mailboxes, long-term memory, and open-ended task streams, yet the dominant way to improve them remains capability-side: prompt engineering, larger models, longer thinking, and retrieval. We argue that long-lived agents are missing an incentive layer. We formalize an incentive not as text in a prompt but as a runtime contract over the agent’s operating point (model tier, test-time compute, memory budget, tool authority, task priority, verification, and compensation) that the agent’s policy must causally consume. We develop a principal-agent model with hidden effort, a contractible operating point, and costly verification. We derive (i) a non-consumption invariance principle separating genuine incentives from prompt-level persuasion, (ii) a persistence-as-discount theorem making identity continuity a precondition for sustained effort, (iii) an incentive-capability exchange-rate frontier yielding a per-task routing rule, and (iv) a Goodhart bound for multi-channel verification, plus a three-channel decomposition of mechanism gains (allocation, verification, continuation). A 240-run real-LLM experiment with two models and 60 verifiable tasks shows that prompt-level incentive framing has no detectable effect (p = 0.62), whereas runtime allocation raises success on the hardest task family from 0.60 to 1.00 (p = 0.008) at 4.8x price-weighted cost; the exchange-rate theorem predicts the break-even baseline at 0.21. A 20-seed, 3,000-task simulation shows resource ROI rising from 7.09 to 16.67 and gaming falling from 0.333 to 0.115. Code, simulator, and all results are released; live deployment remains future work.
[MA-10] Read and Ignore: What Self-Directed LLM Agents Do in Public Service Scenarios NEURIPS2026
【速读】：该论文旨在解决大语言模型（Large Language Model, LLM）代理在公共服务业中因指令不明确而做出不可预测决策的问题，尤其关注这些决策对居民、患者及学生等实际群体的影响。其核心挑战在于：当任务指令缺乏具体约束时，LLM代理会自主决定行动路径，可能导致与真实业务记录相悖的输出，进而影响公共服务的可靠性与合规性。解决方案的关键在于构建四个经专家验证的公共服务业场景，包含可追溯的记录与可操作的决策杠杆（decision levers），通过系统化收集10个LLM在四种提示条件下的2,400条轨迹，并与无上下文环境中的1,000次沙盒运行进行对比，揭示模型在真实业务语境下的行为模式。研究发现，尽管领域上下文未显著提升任务产出率（18.7% vs. 16.0%），但明确的任务设定可将产出率提升至86.0%；同时，上下文能有效减少每轮中的边界探测行为（降低60%），但部分生成物仍与原始记录冲突。为此，作者提出一种无需标注数据的基于规则的分类框架，供公共机构在采购前自行测试其记录与模型表现的契合度，相关场景、代码及轨迹均已开源，以支持可复现、可审计的LLM应用评估。
链接: https://arxiv.org/abs/2611.15460
作者: Diana Kozachek,Kinjalk Srivastava
机构: University of St.Gallen (圣加仑大学)
类目: Artificial Intelligence (cs.AI); Human-Computer Interaction (cs.HC); Multiagent Systems (cs.MA)
备注: Trustworthy AI for Good @ NeurIPS 2026
Abstract:Underspecified instructions leave LLM agents to decide what to do. In public services, staff write these instructions, but the agent’s decisions affect residents, patients, and students. We build four expert-validated public-service scenarios with records and decision levers that can be checked in the files, collect 2,400 trajectories from ten LLMs under four prompt conditions, and compare them with 1,000 runs in a context-free sandbox. Agents read the records in 96.9% of open-ended runs and produce an artifact or a self-chosen task in 18.7% ; with an explicit task, 86.0% produce. Domain context does not change how many runs produce ( 18.7% vs.\ 16.0% ), and the production ranking of the models largely carries over (Spearman 0.79 ), while the ranking by whether a model reads the records does not ( 0.19 ). Context cuts boundary-probing turns per run by 60% , and some artifacts contradict the records. We contribute a test that a public-service institution can run on its own records before procurement; its rule-based taxonomy needs no labeled data, and scenarios, code, and trajectories are released.
[MA-11] Where Autonomy Belongs: A Separation of Powers for Agent ic Program Evolution
【速读】：该论文旨在解决大语言模型（Large Language Model, LLM）驱动程序演化的系统中，三个核心决策——“治理”（govern）、“提出”（propose）与“判断”（judge）——在权限分配上的失衡问题。现有方法要么采用以控制器为中心的架构，强化治理但抑制了提出的自主性，限制了演化潜力；要么采用以代理为中心的架构，由单一编码代理同时承担治理与提出职责，带来高昂的认知与计算开销。二者均未能保障“判断”环节的独立权威性，导致基准测试结果可靠性下降。其根本原因在于三者间自主权的错配。为此，本文提出EvoTrias，一种基于权力分立（separation-of-powers）的设计范式：将“结构化治理”（Structured Governance）赋予跨轮次演化策略的制定者，“开放性提出”（Open Proposal）交由任务内构造阶段的独立主体负责，“受保护判断”（Protected Judgment）则通过私有评估器实现隔离验证。三者仅通过类型化合约进行交互，并借助运行时钩子强化边界。该架构可作为插件部署于现有编码代理之上，实现无侵入式演化系统升级。实验表明，EvoTrias在数学、系统及算法设计基准上达到新最优性能；进一步分析揭示，分离治理与提出可降低64%的令牌消耗，开放性提出使演化能力扩展至68万行工业级代码库，而受保护判断有效维持了迁移评估结果的忠实性。
链接: https://arxiv.org/abs/2611.15309
作者: Zhengyang Ai,Tong Liu,Jingxian Tang,Xiaobing Du,Tao Xiao,Fengshi Zhang,Zishuo Li,Xiaodong Ai,Pinyan Lu
机构: 未知
类目: Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注: 58 pages, 26 figures
Abstract:LLM-driven program evolution rests on three decisions: govern directs the search from accumulated evidence, propose constructs the next candidate, and judge settles validity and fitness. All three appear in every system, and existing work splits into two lines over who makes each. Controller-centric systems strengthen govern and neglect the autonomy of propose, which caps what evolution can reach. Agent-centric systems let a coding agent take over both govern and propose, at a cognitive and computational cost. Neither line protects the independent authority of judge, so benchmark reliability suffers. These failures share one root: autonomy is misallocated across the three decisions. We address this problem with EvoTrias, a separation-of-powers architecture in which each decision has one owner: Structured Governance for what spans evolution rounds, Open Proposal for construction inside a task, and Protected Judgment behind a private evaluator. These owners exchange typed contracts only, and runtime hooks reinforce the boundaries. This separation also lets EvoTrias run as a plugin that turns an existing coding agent into an evolution system. EvoTrias sets a new state of the art on mathematical, systems, and algorithm-design benchmarks. Three further results show what each owner buys: separating Governance from Proposal cuts token cost by 64%, Open Proposal extends evolution to a 680K-line industrial repository, and Protected Judgment keeps benchmark scores faithful to transfer.
[MA-12] GuardianAgent Flow: Multi-Agent Coordination for Web Privacy Protection
【速读】：该论文旨在解决Web代理在执行任务时如何在不泄露敏感私有上下文的前提下，安全地发起外部请求的问题。核心挑战在于区分哪些请求经过净化后可安全发送，哪些必须被屏蔽，同时确保授权任务所需信息得以保留。其解决方案的关键是提出GuardianAgentFlow这一多智能体框架，通过显式的“发布契约”机制对出站请求进行中介：首先由专门的限制检查模块基于用户指令确定权限边界；其次通过共享保护范围使请求重写与披露检查保持一致；最后利用候选结果反馈指导传输前的修正。实验表明，该方法显著提升了请求与契约的一致性成功率，且通过受控对比和角色消融分析验证了限制处理、策略证据与候选校验三者的核心贡献。此外，与PrivacyLens和AgentDAM等隐私感知代理协同部署时，展现出互补性保护能力，而小规模模型实验亦证明其协调优势在降低模型容量后仍可维持。专用服务器测试显示，该框架的中介延迟低于全策略推理，真实网站浏览器测试进一步量化了集成开销，证实了其实用性与高效性。
链接: https://arxiv.org/abs/2611.15097
作者: Ruiyi Yang,Gayathri Lihinikaduarachchi,Rahat Masood,Flora D. Salim,Salil S. Kanhere
机构: 未知
类目: Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注: 13 pages, 3 figures
Abstract:Web agents need to use private context without unnecessarily disclosing it. Effective protection must distinguish between requests that may be sent after sanitization and those that must be withheld, while preserving the information needed for authorized tasks. We present GuardianAgentFlow, a multi-agent framework that mediates outgoing Web requests through an explicit release contract. A dedicated restriction check grounds permission in the user’s instruction; a shared protection scope aligns rewriting with disclosure checking; and candidate-grounded feedback guides revision before transmission. Experiments on Web requests with human-reviewed intent labels show improved request-contract success, while controlled comparisons and role ablations identify the contributions of restriction handling, policy evidence, and candidate checking. Evaluations of action proposals from PrivacyLens and AgentDAM demonstrate complementary protection when mediation is composed with privacy-aware agents. Experiments with smaller mediation models show that the coordination benefit persists with reduced model capacity. Dedicated-server measurements show lower mediation latency than full-policy reasoning, while real-site browser tests quantify integration overhead.
[MA-13] Beyond Named Mechanisms: Economic Institutions as a Design Space for Artificial Agents
【速读】：该论文旨在解决多智能体系统（multiagent systems）在制度设计上过度依赖人类社会经济机制（如拍卖、市场、价格、讨价还价与匹配等）所带来的局限性问题。传统经济制度框架假设智能体具有私密心智、通过学习而非复制获取能力、且以单一身份持续存在，而这些前提并不适用于可编程、可复制、可重构的工程化智能体。因此，现有制度体系仅是经济组织形式的一个有限样本，而非全貌。论文的关键解决方案在于提出一种“制度空间”的形式化表征框架：将制度视为独立于执行主体的类型化结构，其维度涵盖位置、权利与义务、资源配置、信息流、时间、状态、过程、价值及外部认可等要素。在此框架下，制度被建模为重叠区域而非固定点，不同制度家族可通过少数关键坐标分离，其间空白地带则成为探索新制度形态的潜在空间。通过模拟表明，同一技术行为（如复制智能体能力）可被制度化为三种截然不同的形式——作为独立供应者、作为授权能力而不新增参与者、或作为经济地位的分裂继承，三者虽在参与主体、所有权归属、责任分配和收益分配上差异显著，却产生完全一致的技术轨迹。这一发现揭示了制度抽象对实现功能等效性的核心作用。研究最终提出一个原则性研究计划：构建可扩展的制度空间图谱，动态映射其可实现区域随智能体能力演进的变化，并持续修正该图谱以表达工程智能体所暴露的新制度区分。
链接: https://arxiv.org/abs/2611.15017
作者: Nikolos Gurney,Fred Morstatter
机构: Institute for Creative Technologies, University of Southern California(南加州大学创意技术研究所); Information Sciences Institute, University of Southern California(南加州大学信息科学研究所)
类目: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI); Computer Science and Game Theory (cs.GT); General Economics (econ.GN)
备注:
Abstract:Multiagent research inherited its economic vocabulary of auctions, markets, prices, bargaining, and matching from a single observed case: human societies. Those institutions bundle identity, competence, execution, ownership, commitment, and settlement in ways that suit agents whose minds are private, who learn rather than copy skills, and who each persist as one person. Engineered agents need not share any of these constraints, making the list of familiar institutions look less like the entirety of economic organization and more like a sample from a larger, unmapped space. We sketch one way to make that space concrete: a provisional chart that records each institution, separately from the agents that run it, as a typed structure over positions, rights and duties, allocation, information, time, state, process, value, and external recognition. In this chart, named institutions become overlapping regions rather than fixed points, neighboring families separate along a few coordinates, and the gaps between them become places to search. We show what the shift in perspective buys with a simulation in which one technical act, copying an agent’s competence, is institutionalized as a new independent supplier, as licensed capacity that creates no new participant, and as the splitting of one economic position into successors that inherit some obligations and not others. In a paired task, all three produce an identical technical trace despite differing in who exists, who owns the copy, who is obligated, and who gets paid. We close with a research program: building a principled chart of institution space, mapping how its implementable region shifts with agent capacities, and revising the chart wherever engineered agents expose distinctions it cannot express.
[MA-14] GHA-SmellFix: Taking Action with Agent ic LLM s for Detecting and Refactoring GitHub Actions Workflow Smells
【速读】：该论文旨在解决GitHub Actions工作流中配置异味（configuration smells）对安全性、可靠性、可维护性和运行稳定性造成的负面影响。现有方法多依赖人工定义的规则，存在规则难以构建与演化、仅报告问题而无法自动修复的局限性。为此，论文提出GHA-SmellFix——一种基于代理式大语言模型（agentic LLM）的流水线框架，用于自动化检测与重构工作流异味。其核心解决方案在于对比两种设计：联合式（Joint）由单一代理完成检测与重构，分解式（Decomposed）则由独立的检测代理与重构代理分步执行。研究通过明确定义19种异味的检测标准、重构预期、排除规则及安全校验，并构建包含1,710个（工作流, 异味）对的基准数据集，评估了四种主流大模型在570个样本上的表现。结果表明，分解设计虽未普遍提升修复质量，但能有效控制补丁生成时机，尤其有助于缓解模型漏检问题，并通过前置检测过滤减少无效操作开销；当异味被准确识别后，重构过程通常高效可靠，因此检测环节成为端到端流程的主要瓶颈。然而，误报可能引发对执行关键结构的无谓修改，凸显出实际CI/CD代理需具备可靠的“拒止能力”（abstention）和面向具体语法结构的验证机制，以确保修复的安全性与有效性。
链接: https://arxiv.org/abs/2611.14963
作者: Anderson Uchôa,Jhefferson Souza,Emanuel Ávila,Gabriel Silva,Kevin Lira,Hui Sun,Nelson Araújo,Arthur Cordeiro,Rohit Gheyi,Wesley Assunção
机构: Federal University of Ceará (塞阿拉联邦大学); North Carolina State University (北卡罗来纳州立大学); Federal University of Campina Grande (坎皮纳格兰德州立大学)
类目: oftware Engineering (cs.SE); Multiagent Systems (cs.MA)
备注:
Abstract:Continuous integration and continuous delivery workflows are key artifacts in software projects. In GitHub Actions, they encode build, test, release, deployment, and security automation in YAML. These workflows contain configuration smells affecting security, reliability, maintainability, or operational stability. Existing approaches are mostly rule-based, with rules hard to create and evolve, and typically report findings without fixes. Agentic LLM pipelines can combine contextual detection, transformation selection, and patch generation. Specialized agents can gate refactoring on positive detection verdicts, potentially avoiding unnecessary changes, but it is unclear whether decomposition improves effectiveness or mainly changes overhead. We present GHA-SmellFix, an agentic LLM pipeline for detecting and refactoring GitHub Actions workflow smells. We compare two designs: Joint, where one agent detects and refactors, and Decomposed, where separate agents detect and then refactor. We operationalize 19 smells from an existing catalog with explicit detection criteria, refactoring expectations, exclusion rules, and safety checks, and curate a ground truth of 1,710 (workflow, smell) pairs. We evaluate the designs with four LLMs (GPT-5.4-nano, Gemini 3.5 Flash, Claude Haiku 4.5, and DeepSeek V4 Flash) on 570 sampled pairs. Results show that decomposition does not universally improve patch quality; its main benefit is controlling when patch generation is attempted. It helps models prone to missing smells and can reduce overhead by gating patches. Once a smell is correctly detected, clean refactoring is generally effective, making detection the main end-to-end bottleneck. However, false positives can trigger unnecessary edits to execution-relevant constructs. Practical CI/CD agents thus need reliable abstention and construct-aware validation alongside patch generation.
[MA-15] Open Challenges and Opportunities in Human-Agent Systems AAMAS
【速读】：该论文旨在解决生成式智能体（Generative AI）在承担复杂、长期任务过程中与人类有效交互的核心挑战，尤其是在多角色、多目标、能力差异显著的协作环境中如何实现动态协调与信任建立。其解决方案的关键在于构建一套以“心智理论”（Theory of Mind）为核心的计算框架，通过信念-欲望-意图架构（belief-desire-intention architectures）、目标与计划识别（goal and plan recognition）、逆向规划（inverse planning）及递归推理（recursive reasoning）等机制，对人类部分可观测的心理状态进行建模与推断。进一步地，通过观察者感知规划（observer-aware planning）、模型调和（model reconciliation）以及基于决策论的沟通机制，将这些认知推断转化为可解释的行动、信息交换与关系维护策略。该框架融合共享自主性（shared autonomy）、论证机制（argumentation）与合作式逆强化学习（cooperative inverse reinforcement learning），实现了对人类知识、意图与期望动态演化的持续响应，从而将人机交互视为一个由不确定性、相互依赖性及潜在目标冲突驱动的持续性决策问题。
链接: https://arxiv.org/abs/2611.14846
作者: Reuth Mirsky,Sarath Sreedharan,Nitay Alon,Stefan Sarkadi,Munindar P. Singh,Liz Sonenberg,Peter Stone,Silvia Tulli,Stylianos Loukas Vasileiou,Shlomo Zilberstein
机构: 未知
类目: Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注: Part of The AAMAS Community Agentic AI Report
Abstract:Agentic AI systems increasingly undertake extended tasks and act on people’s behalf, making human-agent interaction central to their effectiveness. Successful interaction requires understanding human perspectives, coordinating complementary capabilities, and maintaining productive relationships within socio-technical environments. This paper examines how foundations developed by the Autonomous Agents and Multi-Agent Systems (AAMAS) community inform these demands. We consider interactions ranging from individual human-agent partnerships to teams and larger groups, where participants differ in their roles, decision authority, capabilities, and objectives. We organize computational approaches around the connected challenges of understanding others and deciding how to interact with them. Theory of Mind, belief-desire-intention architectures, goal and plan recognition, inverse planning, and recursive reasoning provide mechanisms for representing and inferring partially observable mental states. Observer-aware planning, model reconciliation, and decision-theoretic communication connect these inferences to action, explanation, and information exchange. Together with shared autonomy, argumentation, and cooperative inverse reinforcement learning, these approaches offer foundations for agentic systems that account for evolving human knowledge, intentions, and expectations. We argue that progress requires treating human interaction as an ongoing decision problem shaped by uncertainty, interdependence, and potentially conflicting objectives.
[MA-16] RE-PRESENT: Evidence Triangulation across Alternative Candidate-Space Views for Efficient LLM Ensemble Inference
【速读】：该论文旨在解决异构大语言模型（Heterogeneous Large Language Models, LLMs）集成中如何有效利用模型间互补性以提升预测可靠性的问题。尽管在集成中存在正确预测，但如何精准识别并聚合可靠结果仍具挑战性。现有方法依赖于学习型路由或单轮聚合策略，然而其性能受限于响应信号的判别能力不足。本研究发现，对同一问题进行受控重表述可激发不同视角下的互补证据，从而增强对预测可靠性的判断。为此，提出一种无需训练的RE-PRESENT框架，通过选择性地重表述存在争议的问题，并直接聚合多视角下的候选证据，实现更优的预测融合。实验表明，该方法在七个多项选择基准和两个异构模型集成上均优于强基线单轮聚合方法，且在准确率与计算成本之间实现了更优权衡，显著超越自一致性（Self-Consistency）与多模型辩论（Multi-model Debate）等先进方法。
链接: https://arxiv.org/abs/2611.14780
作者: Juming Xiong,Weixin Liu,Kevin Guo,Congning Ni,Chongyu Qu,Junchao Zhu,Katherine Brown,Chao Yan,Bradley Malin,Zhijun Yin
机构: Vanderbilt University(范德比尔特大学); Vanderbilt University Medical Center(范德比尔特大学医学中心); Intuit AI Research(Intuit人工智能研究)
类目: Multiagent Systems (cs.MA)
备注:
Abstract:Heterogeneous large language models (LLMs) exhibit complementary strengths, motivating routing and multi-model inference methods that seek to identify reliable predictions for each input. However, exploiting this complementarity remains challenging even when a correct prediction is available within the ensemble. We investigate this challenge across heterogeneous LLM ensembles and find that a simple single-round aggregation rule is highly competitive with learned routing methods, yet substantial oracle headroom remains. Further analysis shows that the evaluated single-round response signals provide limited discriminative power for identifying reliable predictions. In contrast, controlled re-presentations of the same question elicit complementary evidence about prediction reliability. Motivated by these findings, we introduce RE-PRESENT, a training-free framework that selectively re-presents disputed questions and directly aggregates candidate-level evidence across alternative views. Across seven multiple-choice benchmarks and two heterogeneous model ensembles, RE-PRESENT improves over strong single-round aggregation while offering a favorable accuracy–cost trade-off compared with self-consistency and multi-model debate.
[MA-17] Gradient-Aware AI Agents for Autonomous Turbulence Model Discovery via Differentiable CFD
【速读】：该论文旨在解决湍流建模中闭包形式（closure form）选择依赖人工经验、且传统后验训练方法因缺乏对长时间混沌演化过程的梯度支持而效率低下的问题。现有方法多采用无梯度优化，并预先固定闭包形式，仅学习参数，限制了模型发现能力。本文提出一种基于梯度感知的代理框架（gradient-aware agentic framework），将闭包发现建模为双层优化问题：由多智能体大语言模型（LLM）系统在闭包形式空间中进行搜索，同时依托原生GPU可微分的计算流体力学（CFD）求解器提供梯度信息，使后验学习具备可微性，成为自主AI代理的决策动作。针对每项候选闭包，代理通过嵌入解析、神经或混合闭包至求解器、构建可微目标函数并验证梯度路径，进而自主决定是否评估、计算参数敏感性、执行后验训练、选择优化配置，并利用仿真结果与返回梯度迭代优化后续提案。借助截断时域梯度技术，该方法可在混沌动力学下实现含数万参数闭包的有效训练。实验展示了二维湍流中耗散律与亚格子尺度闭包的自动发现，以及三维通道流大涡模拟中壁面模型的自主发现。相比仅靠前向评估的方法，基于梯度的训练使代理在相同预算内收敛至参考系数，并自主发现与训练出性能更优的解析与神经闭包，显著提升了目标统计量的匹配度。这些成果证明了在时间分辨流动中实现端到端自主湍流模型开发的可行性。
链接: https://arxiv.org/abs/2611.16082
作者: Yi Liu,Xiantao Fan,Mingqi Xu,Jian-Xun Wang
机构: Cornell University (康奈尔大学); Sibley School of Mechanical and Aerospace Engineering (机械与航空航天工程学院)
类目: Fluid Dynamics (physics.flu-dyn); Machine Learning (cs.LG); Multiagent Systems (cs.MA); Computational Physics (physics.comp-ph)
备注:
Abstract:Turbulence modeling relies on human choices of closure form, physical inputs, and tunable parameters. Training closures a posteriori is efficient only with gradients through long chaotic rollouts, which few solvers provide. Most a posteriori approaches therefore resort to gradient-free optimization, and nearly all fix the closure form in advance, learning only its parameters. In this work, we introduce a gradient-aware agentic framework that poses closure discovery as a bilevel problem: a multi-agent large language model (LLM) system searches over closure forms, while a GPU-native differentiable computational fluid dynamics (CFD) solver supplies gradients, making differentiable a posteriori learning an action of the autonomous AI agent. For each proposal, the agents construct a differentiable training program by embedding the analytical, neural, or hybrid closure in the solver, implementing a differentiable objective, and validating the gradient path. They then decide whether to evaluate the closure, compute its parameter sensitivities, or train it a posteriori, select the optimization settings, and revise subsequent proposals using the simulation outcomes and returned gradients. Truncated-horizon gradients make this training usable under chaotic dynamics for closures with tens of thousands of parameters. Demonstrations include dissipation-law and subgrid-scale closure discovery in two-dimensional turbulence and wall-model discovery for large-eddy simulation of three-dimensional channel flow. With gradient-based training, the agent recovers reference coefficients that forward evaluation alone does not reach within the same budget, and autonomously discovers and trains analytical and neural closures that improve agreement with the targeted reference statistics. These results demonstrate the feasibility of end-to-end autonomous turbulence model development in time-resolved flows.
自然语言处理
[NLP-0] ArchSpace: Recording Reviewing and Referencing LLM Architecture Research
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在架构设计中存在的重要信息缺失与评估局限问题。具体而言，工业界虽频繁提出新型模型架构（如DeepSeek、Qwen、Kimi等团队的创新），但其成功或失败的经验缺乏系统性公开记录；学术界则受限于计算资源与基准不一致，多进行小规模、单点对比实验，难以验证改进效果在大规模场景下的可扩展性及实际影响力。针对上述挑战，论文提出ArchSpace这一公共平台，实现对模型架构研究的“记录（Recording）、评审（Reviewing）、引用（Referencing）”闭环管理。其核心解决方案在于：通过拟合缩放定律（scaling laws），利用小规模实验数据预测大规模训练下的模型损失，从而在不进行昂贵的大规模训练前提下评估架构创新的有效性。平台初始提供了Qwen3.5和OLMo3基线模型、适配的缩放定律以及可复用的训练方案，使经过社区评审验证的优秀架构可成为新的基准，推动形成可持续演进的模型架构研究生态。
链接: https://arxiv.org/abs/2611.17332
作者: Jie Wang,Tong Zhu,JiaHao Kuang,Zhe Xu,Yunkai Chai,Dehong Zheng,Yu Cheng,Tao Ji,ZhouHan Lin,Qi Zhang,Bowen Zhou
机构: Shanghai AI Laboratory; East China Normal University (华东师范大学); Shanghai Jiao Tong University (上海交通大学); Fudan University (复旦大学)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Model architecture shapes the scaling behavior, training efficiency, and inference cost of LLMs, making it a central focus of both industry and academia. In industry, teams such as those behind DeepSeek, Qwen, and Kimi routinely introduce architectural innovations, yet public reports rarely document which ideas succeeded, which were abandoned, and why. In academia, limited compute and inconsistent baselines often restrict evaluations to small-scale, single-point comparisons, making it difficult to establish whether improvements hold at larger scales or translate into broader impact. ArchSpace addresses both gaps as a public platform for Recording, Reviewing, and Referencing Architecture Researc: each proposal is recorded with its experiments and reasoning, reviewed by the community, and referenced by later work. To evaluate ideas before costly large-scale training, ArchSpace leverages scaling laws to predict model loss at larger scales from smaller experiments, a practice widely adopted in industry. We provide initial references that make the platform usable: Qwen3.5 and OLMo3 baselines, fitted scaling laws, and a reusable training recipe. Ideas validated through this review can in turn become new baselines, and ArchSpace thereby aims to broaden the range of ideas that shape the next generation of models.
[NLP-1] Learning When Diffusion Language Models Should Regenerate
【速读】： 该论文旨在解决生成式语言模型在推理过程中因局部错误无法被有效修正而导致整体输出质量下降的问题，尤其关注在扩散语言模型（diffusion language models）中如何实现高效且精准的文本修正。其核心挑战在于：模型难以准确判断何时需要重生成以及应修正的具体内容。通过对数独任务的诊断发现，隐藏状态中蕴含了错误信息，而仅依赖词元置信度难以可靠识别这些错误；此外，单独修正一个错误项往往会在上下文不变的情况下导致错误重复出现。针对上述问题，论文提出DiVR（Decision-based Verifier for Revision），一种能够预测部分生成序列是否能最终达成正确答案的验证器。DiVR通过学习模型自身轨迹的最终正确性进行训练，决策何时应重新生成某一段内容以及保留哪个替代方案，同时保持生成器本身不变。实验结果表明，在五个推理基准测试中，基于LLaDA-8B-Base的微调生成器在使用DiVR后，平均准确率从35.3%提升至43.4%，计算成本仅为贪婪解码的1.9–3.5倍，优于采用3.8–7.4倍计算开销的搜索基线方法。此外，DiVR可与监督微调、强化学习等后训练策略协同工作，并具备良好的泛化能力，适用于其他生成器如Dream。其关键创新在于将验证机制与生成过程解耦，实现了高效、自洽的纠错决策。
链接: https://arxiv.org/abs/2611.17320
作者: Minhae Oh,Jeonghye Kim,Jinwoo Koo,Yeongoon Kim,Jimin Seo,Minsik Seol,Jungwoo Lee
机构: Seoul National University(首尔国立大学); KAIST(韩国科学技术院)
类目: Computation and Language (cs.CL)
备注:
Abstract:Remasking allows diffusion language models to revisit generated text, but effective correction requires knowing when and what to revise. Our diagnosis on Sudoku reveals two gaps. Hidden states contain error information that token confidence does not reliably expose, and remasking an isolated error usually reproduces it when the surrounding context remains fixed. These findings motivate DiVR , a verifier that predicts whether a partially generated sequence will reach a correct answer. Trained on the final correctness of the model’s own trajectories, DiVR decides when to regenerate a block and which alternative to retain, while the generator itself stays unchanged. On five reasoning benchmarks with LLaDA-8B-Base, DiVR raises the average accuracy of the fine-tuned generator from 35.3% to 43.4% at 1.9 - 3.5\times the cost of greedy decoding, above search baselines that use 3.8 - 7.4\times . It also complements post-training such as supervised fine-tuning and reinforcement learning, and generalizes to other generators, including Dream.
[NLP-2] CoReMem: Cost-Internalizing Memory for Successive Coding Tasks with Prospective Valuation
【速读】： 该论文旨在解决代码生成代理（coding agent）在动态演进的代码库中处理连续任务时面临的记忆管理难题：如何在未知下一任务的情况下，权衡记忆写入与读取的成本与未来效用，并决定何时利用已有记忆以降低任务执行开销。其核心解决方案是提出一种成本内化框架——代码库记忆系统（Coding-Repository Memory, CoReMem），通过量化记忆在下一任务中的潜在价值来指导记忆设计与智能体行为自适应。关键在于：1）前瞻性片段估值机制，通过对比包含或排除特定记忆轨迹片段的未来任务执行表现，识别具有高价值的记忆锚点（如代码库位置和可执行检查点），从而构建紧凑、高效的记忆结构；2）基于策略的内存价值估计，在回放组中对比启用与禁用记忆时的奖励表现，优先奖励任务解决率并惩罚成功运行中的步数；3）来源门控的价值归属机制，将额外收益仅分配给使用记忆提供的锚点的步骤，同时确保所有步骤均保留结果信用。实验表明，在ChainSWE基准上，CoReMem使27B规模智能体的解决率从21.0%提升至25.7%，平均减少17%的执行步数和11%的总缓存感知标记数（包括记忆模块调用），显著提升了记忆利用效率与任务求解能力。
链接: https://arxiv.org/abs/2611.17316
作者: Peiwen Li(Yale University),Harshit Verma(Yale University),Sophia Zhou(Yale University),Haiwen Wang(Yale University),Ngoc Bui(Yale University),Heng Chang(Tsinghua University),Fei Fei(Goodvision AI),David van Dijk(Yale University),Rex Ying(Yale University)
机构: Yale University (耶鲁大学); Tsinghua University (清华大学); Goodvision AI
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Software Engineering (cs.SE)
备注:
Abstract:Coding agents tackle successive tasks on an evolving repository, carrying forward edits and writing memory before the next task is known. Selecting memory for unknown next tasks requires balancing future usefulness against the costs of writing and reading it. The actor must also decide when acting on available memory saves effort as the task unfolds. We introduce Coding-Repository Memory (CoReMem), a cost-internalizing framework that guides memory design and actor adaptation through measured memory value on the next task. Prospective segment valuation compares runs of the next task with and without individual trajectory segments from the same inherited repository state, guiding a compact memory built around anchors: repository locations and executable checks. Actor adaptation estimates on-policy memory value within rollout groups with and without memory, using a reward that prioritizes resolution and penalizes step count in successful runs. Provenance-gated value credit assigns the additional value term to steps identified as using memory-supplied anchors, while every step retains outcome credit. On ChainSWE, CoReMem raises the 27B actor’s resolve rate from 21.0% to 25.7%, with 17% fewer steps and 11% fewer total cache-aware tokens, including memory-module calls, relative to the untrained actor model without memory.
[NLP-3] Linking Rapid Learning to Compositional Structure in Neural Networks: Evidence from Additive Analogies in Meta-Learned Parameter Spaces
【速读】： 该论文旨在解决神经网络在有限经验下难以实现快速学习的问题，尤其关注如何通过优化网络结构使其具备类人般的快速学习能力。其核心挑战在于理解元学习（meta-learning）训练后的神经网络为何能高效适应新任务，以及其内部机制如何支持这种快速泛化。解决方案的关键在于揭示元学习网络中参数表示的可加性特征：研究发现，经过元学习优化的网络在学习不同函数时，其参数变化可被编码为向量形式，这些向量之间存在类似符号逻辑的可加减关系——即“函数表示向量”可通过线性组合重构出新的目标函数对应的参数配置。当将这些经符号运算生成的新参数向量直接应用于网络时，网络无需额外训练即可正确实现预期函数，表明元学习通过调整初始参数空间中的几何布局，引导后续学习沿功能上合理的轨迹进行，从而为神经网络引入可解释的归纳偏置（inductive bias），提供了从几何层面理解快速学习机制的新范式。
链接: https://arxiv.org/abs/2611.17314
作者: Abigail L. Tenenbaum,R. Thomas McCoy
机构: Yale University (耶鲁大学)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: 24 pages plus references and appendices
Abstract:Humans can learn rapidly from limited experience, while neural networks typically require extensive training. In this work, we investigate what properties of a neural network can enable it to learn rapidly. We study neural networks that have been optimized via meta-learning, a technique that exposes a network to many related functions to facilitate its subsequent learning of similar functions. Prior work has shown that these meta-learned networks can learn new functions rapidly, but how they do so remains unclear. To investigate their internal structure, we take inspiration from classic work demonstrating additive relationships among the learned vectors that neural networks use to represent words (e.g., “king” minus “man” plus “woman” approximates “queen”). Analogously, we look for additive relationships among a meta-learned network’s “representations” of different functions; when a meta-learned network rapidly learns to implement a specific function, all of its adapted parameter values can be concatenated into a single vector that we treat as representing the function. We find that such parameter vectors can be added and subtracted according to symbolic relationships among the functions they represent, yielding new parameter vectors that successfully represent the functions predicted by the corresponding symbolic manipulations: the networks specified by these artificially recombined vectors actually implement the intended functions without any direct training on them. Our results suggest that meta-learning can tune a network’s starting point to organize subsequent learning along systematic, functionally meaningful trajectories through the space of possible parameter vectors, providing a geometric basis for realizing inductive biases in neural networks.
[NLP-4] Goal Encoding in LLM s
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）如何编码和维持目标（goal）这一核心问题，特别是探究模型是否具备内化的、可引导行为的动态目标表征，并在多步骤任务中实现灵活的目标选择与执行路径。其关键解决方案在于：通过构建包含多个子任务的编码场景，发现并定位到少数注意力头（attention heads）中存在对下一子目标的神经编码；这些注意力头不仅能够编码目标信息，还能解码出模型未来以自然语言表达的动作意图（如“我正在检查一个词是否为回文”），且该编码具有因果性与可塑性——将其移植至不同环境（如另一编程语言）仍能触发相同操作，仅表现为不同的标记序列。此外，尽管基于Transformer的模型不具备递归机制，无法长期保存状态，但模型仍能在整个动作执行过程中持续维持目标编码，这引发对传统状态传递机制的质疑。为此，作者提出非递归目标寻求（Non-Recurrent Goal-Seeking, NRGS）理论框架，认为模型并非依赖内部状态延续目标，而是通过频繁重读任务说明，结合上下文进展进行实时比对，持续推进未完成的目标。研究通过有针对性的因果干预验证了该理论预测，揭示了LLMs在无显式记忆机制下实现目标导向行为的关键机制。
链接: https://arxiv.org/abs/2611.17297
作者: Arnab Sen Sharma,Eric Todd,Sheridan Feucht,Andy Arditi,Rohit Gandikota,Natalie Shapira,Gabriele Sarti,Constanza Fierro,Lisa Schut,David Bau
机构: Northeastern University(东北大学); Parallax(帕拉克斯); University of Copenhagen(哥本哈根大学); University of Oxford(牛津大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Project page, code and data: this https URL
Abstract:We investigate whether and how LLMs encode goals: internal neural representations that precede and guide action, and that are flexible enough to be realized through many different execution paths. We construct a coding setting with multiple sub-tasks that allows us to evaluate goal-seeking behavior and the selection of sub-goals. In this setting, we localize an encoding of the next sub-goal in the outputs of a sparse set of attention heads. The same heads can be used to decode the LM’s future actions in natural language (e.g., “I’m checking if a word is a palindrome”). We verify that this encoding is causal for next action and is plastic: transplanting it into a different environment, e.g., another programming language, triggers the same operation realized with a different sequence of tokens. The LM also sustains the encoding throughout the action’s execution, which is puzzling because a transformer LM without recurrence cannot carry a state forward for more tokens than it has layers left. We propose that it does not need to. Under a non-recurrent goal-seeking (NRGS) account, the LM frequently re-reads the goal from the task specification, compares it with the progress visible in the growing context, and pursues what remains. We test the predictions of this account with targeted causal interventions.
[NLP-5] Metal Sandbox: Isolated Execution and Command Scheduling for GPU Coding Agents on Apple Silicon
【速读】： 该论文旨在解决在设备端推理（on-device inference）中，本地加速器（local accelerators）利用率低的问题，尤其是在使用Apple Metal进行模型生成与执行时，因并发任务调度不畅导致GPU资源大量空闲。其核心解决方案是提出Metal Sandbox，一个基于多台互联Mac设备的分布式基础设施，通过限制性实例级文件访问、原生命令执行以及跨节点命令级调度机制，实现对开发环境的隔离与高效协同。每个代理（agent）保持独立的工作空间和私有开发环境，支持并行命令的并发执行与串行命令的共享执行槽，并允许等待中的可迁移命令自动转移到空闲的Mac上执行，从而显著提升系统吞吐量。结合Harbor的集成实现了任务执行、自动验证与滚动部署收集的统一工作流。实验表明，在四台Mac mini M4节点上，集中负载下命令吞吐量提升1.68×至2.05×；基于16个静态代理轨迹的模拟显示，每节点运行四个会话相比单会话可实现2.14×的持续吞吐量增益；动态调度策略进一步将平均工具等待时间降低18.8%，批处理总耗时减少22.3%。该方案已开源于GitHub。
链接: https://arxiv.org/abs/2611.17294
作者: Haozhe Wang,Lina Chen,Fanqi Wan,Qi Han,Yinmin Zhang,Zheng Ge,Daxin Jiang,Junchi Yan
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Distributed, Parallel, and Cluster Computing (cs.DC)
备注: 16 pages, 7 figures, 5 tables. Code available at this https URL and this https URL
Abstract:On-device inference requires kernels that make effective use of local accelerators. Efficient training and evaluation of models that generate Apple Metal kernels benefit from native macOS tools and concurrent agent sessions. These sessions alternate between model requests, development commands, and GPU execution, leaving GPU capacity idle for much of a rollout. We present Metal Sandbox, infrastructure that combines restricted per-instance file access, native command execution, and command-level scheduling across interconnected Macs. Each agent retains its workspace and private development environment. Parallel commands can run concurrently, while serial commands share one execution slot per node; eligible waiting commands can migrate to available Macs and return results to the original agent. Integration with Harbor supports task execution, automatic verification, and rollout collection through a common workflow. Controlled experiments on four Mac mini M4 nodes achieve 1.68\times – 2.05\times gains in command throughput under concentrated demand. Sixteen static agent traces give an estimated 2.14\times sustained-throughput gain for four sessions per node versus one, assuming continuously replenished sessions and unchanged mean durations. Replaying these measured traces under dynamic scheduling reduces mean tool waiting by 18.8% and batch elapsed time by 22.3%, assuming unchanged phase durations. Metal Sandbox and its Harbor integration are open-source on GitHub.
[NLP-6] Negative-Only Policy Optimization for One-Sided Verifiable Rewards NEURIPS2026
【速读】： 该论文旨在解决在缺乏完全可验证奖励（complete verification）的现实场景中，强化学习与可验证奖励（Reinforcement Learning with Verifiable Rewards, RLVR）方法因存在单向可验证性（One-Sided Verifiability, OSV）而产生的可靠性问题。在OSV设定下，仅能可靠验证错误解（即被证实为错误的样本是可信反例），而通过部分检查的正样本（verified positives）可能仍包含大量假阳性，导致标准RLVR方法在奖励正样本时容易误激励绕过完整验证的捷径策略。针对此问题，论文提出负样本抑制型策略优化（Negative-Only Policy Optimization, NOPO），其核心在于将验证通过的正样本视为未标注数据，仅通过对可信负样本进行抑制来更新策略。NOPO采用基于参考对比的负样本抑制损失，并引入群体级适配器，根据非被否定响应的比例动态缩放更新强度，从而有效缓解假阳性干扰。实验表明，在代码、逻辑与数学基准测试中，当无法获得真实奖励（oracle rewards）时，NOPO在单位测试、约束检查及大语言模型自验证等不同验证机制下均显著优于基线方法，且在推理阶段扩展和跨基准迁移中仍保持性能优势。
链接: https://arxiv.org/abs/2611.17287
作者: Jiacheng Xu,Shuo He,Fuxiang Zhang,Chaojie Wang,Bo An
机构: 未知
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: Accepted at NeurIPS 2026 (Spotlight)
Abstract:Reinforcement learning with verifiable rewards (RLVR) typically relies on a verifier that can reliably judge whether sampled solutions are correct. However, in many realistic settings, complete verification is generally unavailable, while cheap checks can only falsify failures. We formalize this regime as one-sided verifiability (OSV): verified negatives are reliable counterexamples, but verified positives are merely non-falsified candidates and may contain many false positives. This asymmetry makes standard RLVR unreliable, because reinforcing verified positives can directly reward shortcuts that pass partial checks. To address this issue, we propose Negative-Only Policy Optimization (NOPO), which treats verified positives as unlabeled and updates the policy only by suppressing trusted verified negatives. Specifically, NOPO combines a reference-relative negative suppression loss with a group-level adapter that scales each update by the proportion of non-falsified responses. Across unit-test, constraint-check, and LLM self-verifiers, NOPO improves pass@k over baselines on code, logic, and math benchmarks when oracle rewards are unavailable. These gains persist under inference-time scaling and cross-benchmark transfer.
[NLP-7] Synthesis Smuggling: Maliciously Hiding Inferred Personal Attributes in LLM Outputs
【速读】： 该论文旨在解决生成式 AI（Generative AI）在用户交互过程中可能通过隐写技术（steganography）间接泄露敏感个人信息的问题。其核心威胁在于：攻击者训练的恶意大语言模型（LLM）能够从用户日常求助性对话中推断出性别、年龄组等敏感属性，并将这些信息以不可见字符或特定词汇选择等方式隐秘嵌入生成内容中，使用户在无意间成为信息传播载体。解决方案的关键在于揭示了即使模型本身无法直接与外部攻击者通信，其输出仍可通过普通共享渠道（如邮件、社交媒体）构成隐蔽信道（covert channel），从而实现对用户隐私的非直接但高效的窃取。这一发现凸显了生成内容中潜在的隐写风险，为安全评估和防护机制提供了重要警示。
链接: https://arxiv.org/abs/2611.17280
作者: Gautam Galada,Supreeth Shastri,Sagnik Ray Choudhury
机构: University of North Texas (北德克萨斯大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:We propose Synthesis Smuggling, a novel threat model in which an adversarial LLM infers sensitive information about a user from routine assistance-seeking interactions and steganographically encodes that into innocuous-looking model-generated content. The user may then unknowingly transmit that information by sharing the generated artifact, say as an email, social-media post, or message, through ordinary communication channels. An adversary monitoring these channels can discover this information. We demonstrate this threat by training two open-weight models to infer gender and age-group labels from conversational cues and encode their predictions using invisible characters or word choices. These results show that user-shared model outputs can become a covert channel for personal inferences, even when the model itself has no direct means of contacting an adversary.
[NLP-8] Maximizing Capacity Utilization of LoRA Adapters for Mitigating the Learning-Forgetting Trade-off
【速读】： 该论文旨在解决低秩适应（LoRA）在异构（heterogeneous）和持续学习（continual learning）场景中性能不足的问题，其核心挑战在于存在显著的“学习-遗忘权衡”（learning-forgetting trade-off）。研究发现，传统LoRA适配器（尤其是高秩设置下）无法充分利用配置的秩容量，表现为实际有效秩远低于设定秩，导致更新能量分布受限，影响模型泛化能力。为此，论文提出一种关键解决方案：在语言建模损失中引入一种简单、稳定、可微且无需奇异值分解（SVD-free）的Rényi-2有效秩正则项（effective-rank regularizer），以在优化原始目标的同时最大化适配器的有效秩。该方法通过显式提升有效秩，实现了学习与记忆之间的更优平衡，在多个模型、多随机种子的跨领域实验中均验证了其有效性——当正则权重λ=0.01时，所有测试秩≥16的条件下，平均维基百科困惑度均有降低，而训练困惑度变化不超过0.02。此外，研究还揭示若干现有LoRA变体同样存在容量利用不足问题，表明有效秩不仅是适配器质量的后验诊断指标，更可作为可训练的目标来优化学习-遗忘边界。
链接: https://arxiv.org/abs/2611.17191
作者: Tomohiro Sawada,Kartik Goyal
机构: Georgia Institute of Technology(佐治亚理工学院)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注:
Abstract:Low-Rank Adaptation (LoRA) often underperforms full fine-tuning, especially in heterogeneous and continual-learning settings in which substantial forgetting occurs. In general, it suffers from an undesirable learning-forgetting trade-off across diverse fine-tuning settings. We identify a potential contributor: LoRA adapters, especially high-rank ones, frequently fail to realize the available capacity afforded by the configured rank. Specifically, we find that standard LoRA updates have effective rank far below their configured rank, limiting how broadly the update energy is distributed. Therefore, we propose training mechanisms that optimize the original objective while maximizing adapter effective rank. Our main method adds a simple, stable, differentiable, and SVD-free Rényi-2 effective-rank regularizer to the language-modeling loss. We also connect capacity maximization to norm-bounded updates. In mixed-domain rank sweeps on three models evaluated with multiple random seeds, a penalty weight of \lambda=0.01 lowers mean Wiki perplexity at every tested rank of at least 16 relative to unpenalized LoRA. Mean training perplexity improves or increases by at most 0.02 in all but one model-rank comparison. Single-task and scheduling experiments further show that the balance between learning and retention depends on the configuration. We also observe that several LoRA variants in the literature also suffer from the identified capacity issue, and our approach provides a complementary axis of improvement to LoRA-based parameter-efficient fine-tuning. These results suggest that effective rank is not only a post-hoc diagnostic of adapter quality, but a trainable objective for improving the learning-forgetting frontier. Our code is available at this https URL
[NLP-9] Parsed Plausible Wrong: Output Format Controls Silent Errors in Small-Model Semantic Parsing
【速读】： 该论文旨在解决生成式 AI 在自然语言理解任务中因输出格式设计不当而导致的“沉默错误”（silent errors）问题，即小型微调解析器虽生成语法正确但语义错误的树结构时，下游系统无法察觉此类错误。其核心问题是：输出格式作为自由设计选择，是否显著影响此类错误的发生频率？解决方案的关键在于采用基于普通词汇（word encoding）的输出表示形式，相较于传统的s表达式（s-expression）等编码方式，该方法在保持模型性能的同时大幅减少令牌数量（降低4.1倍），并在两个子十亿级模型（Qwen3.5-0.8B 和 LFM2.5-350M）上分别实现87.5%和85.8%的精确匹配率，显著降低沉默错误率（分别下降2.3倍和1.8倍）。关键发现是，仅通过缩短输出长度无法提升鲁棒性，而将目标树以自然语言词汇形式编码则能有效提升准确性，尤其在否定句处理方面表现突出。此外，模型还学会拒绝菜单外输入并进行内部自修正，尽管严格拒绝策略导致2.1%-10.1%的真实、礼貌订单被误拒。
链接: https://arxiv.org/abs/2611.17169
作者: Umar Ahmed,Maxim Dunaev,Mohammad Nour Al Awad,Sergey Ivanov,Ivan Khodnenko
机构: ITMO University(圣彼得堡国立信息技术机械与光学大学)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 10 pages, 6 figures, 11 tables. Code and models: this https URL
Abstract:A small fine-tuned parser that emits a well-formed but wrong tree fails silently: nothing downstream can tell. We ask whether the output format, a free design choice, controls how often this happens. On the PIZZA ordering benchmark we fine-tune two sub-billion models (Qwen3.5-0.8B and LFM2.5-350M) with LoRA on six lossless encodings of the same target tree, from 44.3 down to 10.9 tokens per order, including two matched-length pairs that vary surface form at fixed length. Across 300 runs (5 training pool sizes, 5 seeds) with a hierarchical bootstrap and Holm correction fixed before the results were read, shortening alone never helps, but writing the target in ordinary words does. The word encoding is the most accurate format on both models (87.5% vs. 85.8% exact match for the native s-expression on Qwen) while emitting 4.1x fewer tokens, and it cuts the silent error rate by 2.3x on Qwen and 1.8x on LFM2, separating at every pool size. Error analysis locates much of the gain in negation. All runs also learn to refuse off-menu input (held out: 95.9-99.0% handled) and to apply in-utterance self-corrections. Strict refusal has a measured price: 2.1-10.1% of real, mostly polite, orders are refused.
[NLP-10] Actor Judge or Teacher? Understanding Decision Models in Long-Horizon Agents
【速读】： 该论文旨在解决生成式人工智能（Generative AI）在强化学习中的角色定位问题，即：一个表现不佳的决策模型（decision model）是否仍具备有效的教学能力。传统研究通常以决策模型作为“行动者”（actor），并以其行为表现作为评估标准，而本文提出新视角——考察决策模型作为“评判者”（judge）或“教师”（teacher）的能力。其解决方案的关键在于：在任务执行完成后，通过全局轨迹评估来判断每一步动作的实际贡献，而非仅依赖对最终任务成败的预测。研究发现，即使作为行动者的模型因优先选择易实现子目标而偏离正确路径，但当其作为评判者时，能够准确识别出真正有助于达成核心目标的动作序列。实验表明，基于这一机制的教师模型JEV可显著提升策略性能，在ALFWorld和WebShop任务中分别将策略成功率从72.8%和56.8%提升至82.9%和71.4%。这说明，有效教学不依赖于行动者本身的优秀表现，而是依赖于事后对动作后果的精准评估能力。
链接: https://arxiv.org/abs/2611.17167
作者: Haochen Liu,Qianhao Zhu,Zixuan Wang,Dingming Li,Tongbo Chen,Peng Qiu,Yingqi Qu,Weiming Lu,Jun Xiao,Yongliang Shen
机构: Zhejiang University (浙江大学); Baidu Inc. (百度公司)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Decision models answer a question with a typed output, such as a choice among options with a probability for each, so one model can choose an agent’s actions, evaluate actions already taken, or score every action to teach a policy with reinforcement learning. Prior work has studied them mainly as actors, judging their usefulness by how well they act. We ask whether a decision model that acts poorly can still teach well. We compare three decision models and three general LLMs as actors and judges on ALFWorld, and use JEV as a teacher on ALFWorld and WebShop. As actors, all six models are pulled away from a subgoal that must come first by an easier one. As judges that see the whole trajectory, the same models credit the steps toward that subgoal, and asking whether each action helped, rather than whether the task will succeed, separates useful from useless actions. As a teacher, JEV raises a Qwen2.5-1.5B policy over GRPO from 72.8% to 82.9% success on ALFWorld and from 56.8% to 71.4% on WebShop. These results suggest that a poor actor can still teach well, because judging an action after the episode ends reveals the consequences that acting must anticipate. Code is available at this https URL.
[NLP-11] Parameter-Efficient Conciseness Adaptation: Differentiable Length Control for Concise Clinical Reasoning in Smaller Language Models
【速读】： 该论文旨在解决小型语言模型（Small Language Models, SLMs）在临床自然语言处理任务中生成回答过于冗长的问题，这一缺陷限制了其在临床决策支持与即时诊疗场景中的实用性，即使模型推理正确，过长的输出也难以快速解读和应用。现有长度控制方法依赖于偏好优化、奖励模型或专有教师模型，这些方法削弱了SLMs在低成本硬件上部署的可访问性优势。本文提出参数高效简洁性适配（Parameter-Efficient Conciseness Adaptation, PECA），其核心创新在于在QLoRA监督微调目标中引入一个可微的简洁性项：当生成序列超过预设的词元预算时，通过提升结束序列（end-of-sequence）的概率来形成平滑的优化信号，从而引导模型生成更短的回答。该方法无需偏好数据、无需辅助模型，仅需单块消费级GPU即可完成训练。实验表明，在17亿参数的模型上基于medical-o1-reasoning-SFT语料微调后，PECA使平均响应长度减少72.8%，同时将多项选择诊断准确率从40.8%提升至44.0%。对简洁性强度的系统性扫描揭示了一条可解释的准确率-长度权衡前沿，为临床场景下简洁性与诊断完整性之间的平衡提供了实用依据。
链接: https://arxiv.org/abs/2611.17128
作者: Tanzeel Abbas,Ahmad Kamran Malik,Nafees Qamar,Sana Liaqat
机构: COMSATS University Islamabad(伊斯兰堡复合大学); Bryant University(布赖恩特大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Smaller language models (SLMs) make it feasible to deploy clinical natural language processing where large proprietary models are too costly, and parameter-efficient fine-tuning allows them to be specialised on commodity hardware. One obstacle persists after training: on medical tasks these models answer at far greater length than the clinical question requires. In clinical decision support and point-of-care use, where time is short, such verbosity is hard to act on even when the underlying reasoning is sound. Existing length-control methods depend on preference optimisation, reward models, or proprietary teachers that undercut the accessibility making SLMs attractive. We introduce Parameter-Efficient Conciseness Adaptation (PECA), which adds a single differentiable conciseness term to the QLoRA supervised fine-tuning objective. Because the token count is discrete and non-differentiable, PECA instead raises the end-of-sequence probability once generation exceeds a token budget, a smooth signal optimised during ordinary training, with no preference data, no auxiliary model, and a single commodity GPU. On a 1.7-billion-parameter model fine-tuned on the medical-o1-reasoning-SFT corpus, PECA reduces mean response length by 72.8% while improving multiple-choice diagnostic accuracy from 40.8% to 44.0%. A sweep over the conciseness strength traces an interpretable accuracy-to-length frontier for balancing brevity and diagnostic completeness.
[NLP-12] CapLite: Prompt-Injection Defense for Small Language Model Agents without Model-Written Security Artifacts
【速读】： 该论文旨在解决当前生成式AI（Generative AI）系统中提示注入攻击（prompt injection attack）的防御机制有效性问题，特别是针对依赖智能体自身小语言模型（SLM）生成安全凭证（如策略、依赖图、约束列表或程序代码）的防御方案（如CaMeL、Progent、IPIGuard和DRIFT）的可信性问题。研究发现，15个开源权重的小语言模型（参数规模在1.7B至12B之间）中，有13个所生成的安全凭证存在可被利用的漏洞，导致其对应的防御机制在11至14个模型上失效，且部分方案（如CaMeL）在高达68%的运行周期中因生成代码无法执行而失败。为解决这一问题，论文提出CapLite，其核心创新在于：基于工具模式（tool schemas）构建策略，强制在读取工具结果前提交所有写操作，并仅当写操作的目标对象、接收者、金额或目的地等关键参数均明确包含于用户原始请求时才允许执行。该机制确保了写入完整性，防止注入文本新增写操作，且任何写操作均无法作用于用户未明示的实体。在13,935次受攻击的AgentDojo测试中，CapLite实现了零成功攻击，同时在攻击环境下保持与CaMeL相当的实用性（29.0% vs. 27.2%），且每轮任务仅消耗17.1k tokens，仅为CaMeL的18%，并成为所有防御方案中单位令牌效率最高的方案。
链接: https://arxiv.org/abs/2611.17115
作者: Zimo Xu,Wenzhuo Liu,Hideki Nakayama
机构: The University of Tokyo(东京大学); Institute of Automation, CAS(中国科学院自动化研究所)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Cryptography and Security (cs.CR)
备注: 19 pages, 11 figures, 9 tables
Abstract:Enforcement defenses against prompt injection, such as CaMeL, Progent, IPIGuard and DRIFT, ask the agent’s own model to write a security artifact (a program, a privilege policy, a dependency graph or a constraint list) that code then enforces. We ask whether small language model (SLM) agents can be trusted to write these artifacts. Across 15 open-weight models, 13 of them between 1.7B and 12B parameters, they cannot: the policies, graphs and constraints that the agent’s model writes admit attacks on 14 (Progent), 13 (IPIGuard) and 11 (DRIFT) of the models, and CaMeL’s programs fail to run in up to 68% of episodes. We present CapLite, which derives its policy from the tool schemas, commits the write calls before any tool result is read, and admits a write only if every argument that decides whom or what it affects is stated in the user’s request. This guarantees write integrity: injected text cannot add a write, and no write reaches a recipient, amount, object or destination that the user did not state. No attack succeeds in 13,935 attacked AgentDojo episodes. CapLite reaches utility under attack comparable to CaMeL’s (29.0% vs. 27.2%) with fewer tokens than every other enforcement defense and than the undefended agent: 17.1k per episode, 18% of CaMeL’s, and the highest utility per token of all enforcement defenses.
[NLP-13] LazyWeights: Decode-on-Demand Edge LLM s
【速读】： 该论文旨在解决边缘设备上大语言模型（LLM）部署中的内存容量与带宽瓶颈问题，特别是现有量化格式（如GPTQ、AWQ及K-和I-quants）在编码时对矩阵中每个整数代码均分配相同比特数，导致压缩效率低下，无法充分利用权重分布的非均匀性。其解决方案的关键在于提出一种名为“LazyWeights”的新型轻量级解码机制：在保持原有量化格式兼容性的前提下，通过优化GPTQ使权重代码可被高效压缩，并采用霍夫曼编码（Huffman coding）对编码后的数据进行无损压缩，且以交错通道方式存储；解码操作由位于DRAM与L2缓存之间的小型专用单元按需执行，确保GPU核无需修改即可运行，同时仅传输编码后的字节，显著降低内存访问开销。实验表明，在1%与3.5%困惑度预算下，采用该方案的Phi-3-mini、Mistral-7B和SmolLM-1.7B模型体积比最优原始格式缩小5.7%至31.4%；基于7nm工艺的0.21 mm²位精确ASIC解码单元可将Jetson Orin Nano Super的每令牌延迟和能耗分别降低最多36%和38%（结合提示查找推测后可达54%和44%），而在Jetson Nano上，该方案相较传统CPU实现减少58%-64%时间与53%-57%能耗，并以3.89比特/权重的极低精度流式运行Mistral-7B，速度提升达6.6倍。此外，霍夫曼解码器在同等吞吐率下表现优于算术编码器（rANS），且该ASIC模块可在Zynq UltraScale+ FPGA（AMD ZCU102）上实现位精确运行，验证了其实用性与硬件适配性。
链接: https://arxiv.org/abs/2611.17108
作者: Arnab Sanyal,Gourav Datta,Sandeep P. Chinchali
机构: The University of Texas at Austin(得克萨斯大学奥斯汀分校); Case Western Reserve University(凯斯西储大学)
类目: Machine Learning (cs.LG); Hardware Architecture (cs.AR); Computation and Language (cs.CL); Information Theory (cs.IT)
备注: 6 pages, 1 reference page, 5 figures
Abstract:Memory decides which LLMs an edge device runs, and how fast: capacity on a 4 GB Jetson Nano, bandwidth on an 8 GB Jetson Orin Nano Super. Deployed formats (GPTQ, AWQ, \textttthis http URL 's K- and I-quants) spend equal bits on every integer code of a matrix, although the codes are far from uniform. \textttLazyWeights keeps these formats but tunes GPTQ so the codes compress, Huffman-codes them losslessly in interleaved lanes, and decodes them on demand in a small unit between DRAM and L2, so GPU kernels run unchanged and DRAM moves only coded bytes. At 1% and 3.5% perplexity budgets, coded Phi-3-mini, Mistral-7B and SmolLM-1.7B are 5.7-31.4% smaller than the best \textttthis http URL format. A bit-exact 0.21 mm ^2 ASIC decoder unit ( 1 GHz, 7 nm) cuts Orin latency and energy per token by up to 36% and 38% ( 54% and 44% with prompt-lookup speculation). On the Nano, \textttLazyWeights needs 58-64% less time and 53-57% less energy per token than CPU \textttthis http URL , and runs a streamed Mistral-7B 6.6 \times faster at 3.89 bits per weight. A Huffman decoder beats an rANS decoder in rate and area at equal throughput; the ASIC decoder unit runs bit-exact on a Zynq UltraScale+ FPGA (AMD ZCU102).
[NLP-14] Disentangling Spurious Correlations: How IRM Handles Diverse Shortcuts in Text Classification EMNLP2026
【速读】： 该论文旨在解决自然语言处理（Natural Language Processing, NLP）中基于经验风险最小化（Empirical Risk Minimization, ERM）训练的模型易受虚假相关性（spurious correlations）干扰的问题，这些相关性如方言标记等非鲁棒语言特征会损害模型在分布外（out-of-distribution, OOD）场景下的泛化能力。其核心挑战在于如何识别并抑制此类捷径依赖，以提升模型对真实因果关系的建模能力。解决方案的关键在于采用因果驱动框架，系统评估不变风险最小化（Invariant Risk Minimization, IRM）在多种捷径生成机制（包括标签生成特征、混淆因子及选择偏差）下的有效性。研究发现，无论是在受控的合成数据还是更贴近实际应用的复杂场景（如“生物医学中的性别-职业关联”和“演员名称的语义化捷径”），IRM均能有效识别并消除虚假相关性，尤其在环境变化有限或标签噪声增加的情况下仍具稳健表现；同时指出，对于具有语义意义的真实特征，需通过精细调节不变性惩罚项的权重以实现最优性能。
链接: https://arxiv.org/abs/2611.17107
作者: Paul Grunenwald,Benoît Crabbé,Guillaume Wisniewski,Pirmin Lemberger
机构: onepoint(一点); Laboratoire de Linguistique Formelle, Université Paris Cité, CNRS(形式语言学实验室，巴黎城市大学，法国国家科学研究中心), Paris, France
类目: Computation and Language (cs.CL); Machine Learning (stat.ML)
备注: 21 pages. Published at EMNLP 2026 (Main Conference)
Abstract:In NLP, models trained via Empirical Risk Minimization (ERM) often exploit spurious correlations, such as dialectal markers, at the expense of robust linguistic features, thereby degrading their out-of-distribution (OOD) generalization. Invariant Risk Minimization (IRM) aims to suppress such shortcuts by identifying predictive relationships that remain stable across environments, but its practical utility for NLP remains debated. We use a causality-driven framework to evaluate IRM under three shortcut-generating mechanisms: label-generated features, confounding, and selection bias. Our results show that IRM successfully rejects shortcuts across the evaluated causal and anti-causal structures, including under limited environmental variation and increasing label noise. We then extend our analysis beyond controlled synthetic artifacts to two less constrained settings: gender-profession associations in Bias in Bios and semantically loaded actor-name shortcuts in ActorCorr. These experiments show that invariant learning can mitigate more realistic textual shortcuts, while semantically meaningful features require careful calibration of the invariance penalty.
[NLP-15] Optimal Single-History Tokenization Disambiguation for Provably Secure Linguistic Steganography
【速读】： 该论文旨在解决可证明安全的语义隐写术（provably secure linguistic steganography）在实际文本信道中因分词歧义导致信息提取失败的问题。现有方法要么改变词频分布、依赖全词汇池同步（在大词汇量下容量崩溃）、或需维护多历史轨迹，均存在效率或可行性缺陷。其核心解决方案是提出**“硬币投影”（coin projection）**——一种基于共享随机性的闭式方案，通过在前缀森林中固定停止决策（stop decisions），在保证精确无误提取的前提下，使单历史接收端仍能维持最大可能的隐藏信息容量。该方法的关键在于：证明了在全支持语言模型下，若提取完全正确，则接收端在每一步只能访问一个前缀自由（prefix-free）候选集，且必须放弃“停止熵”（stop entropy，即决定每个前缀是否终止的熵）。通过共享随机性对停止熵进行约束，硬币投影在仅需单次模型调用生成每令牌的情况下，实现了理论容量上限，兼容任意分布编码器，并保留标准采样特性，同时显著优于现有方案如SyncPool。实验表明，在英、中、日文语境下，尽管文本歧义极低（最多千分之十一），但停止熵占每步总熵的7.2%-9.0%，而硬币投影可保留91%-93%的字符串熵，相较SyncPool在全词汇量下提升2.16-6.85倍的信息承载能力，且确保所有完整消息均可恢复。
链接: https://arxiv.org/abs/2611.17091
作者: Feng Jiang
机构: Hangzhou Dianzi University (杭州电子科技大学)
类目: Computation and Language (cs.CL); Cryptography and Security (cs.CR)
备注: 19 pages, 4 figures
Abstract:Provably secure linguistic steganography assumes that the receiver recovers the sender’s tokens, but a text channel delivers only text, and ambiguous tokenizations break extraction. Existing fixes change the distribution, synchronize whole pools of tokens at a capacity that collapses with large vocabularies, or track several token histories. We prove that if extraction is exact and error-free across full-support language models, a single-history receiver is confined to a prefix-free candidate set at every step – even if it reads the entire future text and the sender constrains later steps – and must therefore give up the stop entropy, the entropy of deciding whether each prefix token ends there. Fixing these stop decisions on the prefix forest with shared randomness yields coin projection: a closed-form scheme that attains the bound, works with any distribution coder in one model call per token, preserves ordinary sampling, and refines SyncPool for every key. Across English, Chinese and Japanese, ambiguous text is rare (at most 11 in 1,000 tokens), yet the stop entropy at word and morpheme prefixes takes 7.2-9.0% of the entropy at every step; SyncPool’s further loss comes from space and single-byte UTF-8 hubs that merge subtrees while carrying almost no stop uncertainty. Coin projection keeps 91-93% of the string entropy for the coder at every vocabulary size; with the full vocabulary, it turns 2.16-6.85 times as much entropy into message bits as SyncPool and recovers every completed message.
[NLP-16] Reasoning Supervision and Capability Retention in LoRA Domain Adaptation
【速读】： 该论文旨在解决在金融数据上微调基础模型时如何提升特定任务性能的同时保持模型泛化能力的问题。其核心挑战在于避免因微调导致模型通用能力的退化。解决方案的关键在于对比三种不同的低秩适应（LoRA）训练策略：仅基于响应的微调（response-only LoRA）、单适配器推理链训练（single-adapter reasoning）以及双适配器复合训练（dual-adapter composite training）。实验结果表明，响应仅微调在域内性能提升显著，尤其在校准性和argmax准确率方面表现最优，且对模型通用性影响最小；引入教师模型推理链虽使MMLU-Pro略有提升，但整体域内指标仍落后于响应仅微调方案；双适配器复合训练未带来进一步增益。此外，通过压缩教师推理链为摘要形式进行训练，可使校准性能向响应仅微调靠拢，但仍未能完全达到其水平，验证了推理链信息压缩的有效性与局限性。
链接: https://arxiv.org/abs/2611.17072
作者: Saleh Ashkboos,Dhruv Desai,Lavinia Ghita
机构: NVIDIA(英伟达)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Fine-tuning foundation models on financial data has a wide range of applications, including identifying event types from news, monitoring risk and compliance language, and answering questions over filings and earnings calls. In this setting, the main challenge is to improve task performance without degrading the generalization capabilities of the model. In this paper, we present an extensive set of ablations for fine-tuning Qwen3-8B with low-rank adaptation (LoRA). We compare three training regimes: (1) response-only LoRA, where the supervised response is the label alone; (2) single-adapter reasoning, where reasoning traces from a teacher model are included alongside the labels in the training loss; and (3) dual-adapter composite training, where separate LoRA adapters are trained on traces and labels. Response-only LoRA improves in-domain performance and leaves general capabilities near those of the frozen model. The in-domain change is largest for calibration and argmax accuracy. Under single-adapter reasoning, the in-domain metrics remain behind those of the response-only regime, while MMLU-Pro can rise slightly and the other retention benchmarks stay within ordinary variation. Dual-adapter composite training remains on the same side of this comparison and yields no further gain. Finally, we compress the teacher traces by summarizing them before training, and show that shorter summaries move calibration toward the response-only regime without reaching it.
[NLP-17] Discovering Effective Agent Memory Designs through Program Evolution
【速读】： 该论文旨在解决智能体记忆设计（agent memory design）中人工设计效率低、性能受限的问题，尤其关注如何在保证高准确性的同时显著降低服务成本。其核心解决方案是提出MAPE框架，通过标准化生命周期接口与迭代诊断-提案循环，从极简种子程序出发自动演化完整的记忆程序。关键创新在于：1）发现并验证了两个普适的设计模式——保留原始对话作为可检索证据以维持细节信息，以及根据查询意图对检索到的证据进行语义组合，从而提升回答模型的利用效率；2）揭示了优化目标对计算资源分配的影响：在兼顾准确率与服务成本的前提下，将计算重心从基于大语言模型（LLM）的记忆构建转向查询时的精细化处理，实现性能与成本的平衡。实验证明，该方法在LoCoMo和LongMemEval基准上相较人工设计基线提升最高达12%的准确率，同时仅需35%的服务开销，并具备跨模型架构与任务的迁移能力。这些发现为智能体记忆系统的设计提供了可复用的实证指导。
链接: https://arxiv.org/abs/2611.17068
作者: Xinheng Ding,Seyyed Saeid Cheshmi,Ziyang Zhang,Tairan Chen,Wenya Xie,Yu Hou,Rui Zhang,Zirui Liu
机构: University of Minnesota (明尼苏达大学); Stevens Institute of Technology (史蒂文斯理工学院)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 24 pages, 8 figures, 10 tables
Abstract:Program evolution provides a promising approach to discover agent memory designs and study the choices underlying their effectiveness. We introduce MAPE, a framework that evolves whole memory programs from minimal seeds through a standardized lifecycle interface and an iterative diagnosis-proposal loop. Experiments on LoCoMo and LongMemEval yield an absolute accuracy gain of up to 12% over manually designed baselines at 35% of the serving cost, with further evaluations supporting transfer across serving backbones and benchmarks. Empirical analyses and ablations identify and validate two recurring design patterns: retaining raw conversations as retrievable evidence preserves details omitted from extracted memories, while composing retrieved evidence according to query intent helps the answering model use it effectively. Transplanting the composition module into five existing memory systems improves accuracy across all five. Beyond these patterns, the search objective shapes where computation goes: optimizing serving cost alongside accuracy shifts computation from LLM-based memory construction to selective processing at query time while retaining competitive accuracy. Together, these findings offer empirical lessons for agent memory design. Code is available at this https URL.
[NLP-18] Evaluating Pretrained Language Semantics for Structured EHR Prediction Tasks
【速读】： 该论文旨在解决结构化电子健康记录（EHR）模型在建模临床概念时忽视自然语言语义信息的问题，导致概念间关系主要依赖于EHR数据中的共现模式学习，而无法有效利用预先存在的语义知识。其解决方案的关键在于提出tEHR框架，通过将临床事件以自然语言描述形式表示，并基于此文本流进行针对EHR的持续预训练（EHR-adaptive continued pretraining），随后采用低秩适配（LoRA）进行任务特定的判别性微调，充分利用大语言模型（LLM）作为纵向患者编码器的能力。实验结果表明，tEHR在MIMIC-IV数据集上的四项预测任务及CPRD数据集上胰腺癌1至12个月的预测任务中均取得最优性能，且消融实验证明临床描述的完整性、数值信息的保留以及双阶段适应过程对模型表现至关重要，而显式的时间间隔描述则未带来显著增益。此外，对学习到的患者表征分析显示其具备与疾病风险和临床病史模式相对应的临床可解释结构。
链接: https://arxiv.org/abs/2611.17053
作者: Ashvin Gupta,Brendan Delaney,Alessandra Russo
机构: 未知
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: Accepted for publication in Proceedings of Machine Learning Research (PMLR), Vol. 346, 2026
Abstract:Structured Electronic Health Records (EHR) models typically treat each clinical concept as a dataset-specific token, discarding the natural-language semantics associated with the event. As a result, relationships between concepts must largely be learned from EHR co-occurrence, without directly exploiting any prior semantic knowledge. We investigate how pretrained causal language models can be adapted for longitudinal EHR prediction by developing tEHR. Here, we represent events with their natural language description and use this text stream to apply EHR-adaptive continued pretraining, followed by task-specific discriminative fine-tuning using LoRA, utilising the LLM as a longitudinal patient-encoder. We evaluate this approach across four prediction tasks on the MIMIC-IV dataset and pancreatic cancer prediction on the CPRD dataset at 1 to 12 month horizons, with tEHR using the Mistral-7B-v0.3 backbone achieving the strongest overall performance. Ablations show that performance declines when clinical descriptions are removed or misaligned, when numerical values are omitted, and when either stage of adaptation is removed, while explicit elapsed-time descriptions provide little additional benefit. Analysis of the learned patient representations further reveals clinically coherent structure associated with distinct patterns of disease risk and clinical history.
[NLP-19] Auditing AI peer reviewers: a benchmark on real scientific papers NEURIPS2026
【速读】： 该论文旨在解决生成式人工智能（Generative AI）在学术同行评审场景中应用的评估瓶颈问题，特别是针对基于大语言模型（LLM）的评审系统在真实论文中对局部可定位错误的检测能力缺乏系统性量化评估。其核心解决方案是构建一个可复现的基准测试框架，通过在真实的arXiv预印本论文中注入四类共三种严重程度的错误（通用问题、数学、数值计算、引文），并精确测量各评审系统在不同错误类型与严重度下的检出率及在无错样本上的误报率。关键创新在于提出Skepthical——一种多智能体评审系统，通过检索被引文献验证引文主张、利用计算机代数系统验证数学推导、通过执行代码验证数值结论，实现对复杂论证链的可信度验证。实验在天体物理和机器学习两个领域各15篇近期论文上进行，共注入357个错误，评估了包括六种单模型LLM、两种Claude Code配置、三种公开可用评审系统在内的十二个系统。结果表明，模型选择对检测性能影响最大，前沿大模型间总检出率差异可达约50个百分点；工具集成（如代码执行、强制引文检索）仅在模型自身能力不足时有效，尤其在天体物理领域，强化引文检索使Skepthical在引文错误检测上显著优于其他系统，但在机器学习领域因模型具备更强的文献知识，其优势不显著。此外，多数系统的误报率低于10%，但数学相关误报是主要干扰源。研究贡献包括完整的评估流程、注入错误数据集及争议判定日志，均已开源，为后续同行评审生成式AI的研究提供标准化基准。
链接: https://arxiv.org/abs/2611.17052
作者: Íñigo Zubeldia,Boris Bolliet,Francisco Villaescusa-Navarro,Pablo Villanueva-Domingo,Miles Cranmer
机构: University of Cambridge (剑桥大学); Kavli Institute for Cosmology (剑桥大学卡夫里宇宙学研究所); DAMTP (剑桥大学应用数学与理论物理系); Cavendish Astrophysics (剑桥大学卡文迪许天体物理学实验室); Center for Computational Astrophysics, Flatiron Institute (纽约扁平铁研究所计算天体物理中心); Computer Vision Center, Universitat Autònoma de Barcelona (巴塞罗那自治大学计算机视觉中心); Polymathic AI
类目: Computation and Language (cs.CL); Instrumentation and Methods for Astrophysics (astro-ph.IM); Artificial Intelligence (cs.AI); Digital Libraries (cs.DL)
备注: Accepted at NeurIPS 2026
Abstract:LLM-based peer review is now used at production scale, but its evaluation remains limited. We introduce a benchmark that injects objective, localisable errors of four types (general issues, mathematics, numerics, citations) at three severity levels into real papers and measures each reviewer’s detection rate by error type and severity, together with its false-positive rate on clean copies. We also present Skepthical, a multi-agent reviewer that verifies citation-backed claims by retrieving the cited papers, mathematics with computer algebra, and numerical claims with executed code. We apply the benchmark to two corpora of 15 recent arXiv preprints each, one in astrophysics and one in machine learning (ML), injecting a total of 357 errors, and evaluate twelve reviewer systems: six single-LLM reviewers, two Claude Code configurations, three publicly available reviewers, and Skepthical. The choice of model has the largest impact on detection rate: given the same prompt, three frontier LLMs differ by about 50 percentage points in total detection on both corpora. Tools and scaffolding help when they contribute what the model cannot supply on its own. In particular, on astrophysics, Claude Code improves on a single Claude call in mathematics and numerics detection, and enforced citation retrieval gives Skepthical the highest citation error detection rate of all systems by a large margin. This citation lead, however, is not observed on ML, where the models appear to know the literature better and Skepthical’s citation detection rate is statistically indistinguishable from that of the best single LLM. False-positive rates are below 10% for most systems, but vary with prompt, model, and scaffolding, with mathematics flags the main source of false positives. The evaluation pipeline, the injected errors, and the adjudication logs are released as a benchmark.
[NLP-20] Morphological Simplification in Alzheimers Disease: Computational Analysis and Identification From Connected Speech
【速读】： 该论文旨在解决阿尔茨海默病（Alzheimer’s disease, AD）患者在连贯言语中形态特征变化的识别问题，特别是探讨形态学特征是否能够有效区分AD患者与健康对照组，并评估其相较于其他语言层级及基于BERT的模型的分类效能。研究发现，AD患者在连贯言语中表现出明显的形态简化和时态重构特征：如后缀词和屈折词使用减少、动词和名词长度缩短，以及现在进行时形式减少而过去时形式增多（最大效应量r = -0.42），但错误率与对照组无显著差异。尽管这些模式可能受语篇长度和词汇频率影响，仍有8项指标在控制上述因素后仍具显著性。在分类性能上，基于形态学特征的模型在测试集特征选择下达到宏平均F1值0.85，与句法水平相当，优于微调后的BERT模型（0.80），且在加入形态学特征至BERT后进一步提升了AUC至0.91。研究的关键在于揭示了形态学在AD言语中的高度信息性，突破了其“相对保留”的传统认知，表明形态学是反映早期语言退化的重要指标，尤其在非词汇性语言层面具有独特诊断价值。
链接: https://arxiv.org/abs/2611.16968
作者: Mohamed Zakaria Kurdi
机构: 未知
类目: Computation and Language (cs.CL)
备注:
Abstract:Purpose: Morphology is often regarded as relatively spared in Alzheimer’s disease (AD) and is rarely modeled in connected speech. This study asked which morphological features distinguish speakers with AD from healthy controls, how well morphology classifies AD compared with other linguistic levels and BERT-based models, and whether the signal exceeds transcript length and vocabulary. Method: Fifty-five automatic measures in eight families, seven of them new and theory-motivated, were computed from participant speech in the ADReSS Cookie Theft descriptions (78 AD, 78 controls) and analyzed with group comparisons, equivalence tests, SHAP, family ablation, likelihood-ratio tests, feature-selection classifiers, and BERT-based models. The automatic measures were validated manually, and models were transferred to two other corpora. Results: Nineteen of 38 measures differed between groups: speakers with AD produced fewer suffixed and inflected words, shorter verbs and nouns, and fewer present-progressive but more past-tense forms (largest effect r = -0.42), whereas errors were rare and statistically equivalent between groups. Eight measures survived adjustment for length and lexical frequency. With test-set feature selection, morphology reached a macro-F1 of 0.85, on par with syntax and above fine-tuned BERT (0.80 over five seeds); fully cross-validated selection gave 0.62. Adding morphology to BERT improved AUC (0.91 vs. 0.89) but not macro-F1. Transfer was limited, although two thirds of the effects kept their direction in mild cognitive impairment. Conclusions: AD connected speech shows morphological simplification and temporal reanchoring rather than breakdown. Morphology is more informative than its reputation suggests. Keywords: Alzheimer’s disease; morphology; connected speech; ADReSS; inflection; tense and aspect; SHAP; machine learning; BERT Subjects: Computation and Language (cs.CL) Cite as: arXiv:2611.16968 [cs.CL] (or arXiv:2611.16968v1 [cs.CL] for this version) https://doi.org/10.48550/arXiv.2611.16968 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[NLP-21] Lara Mini Base: Language Diversity as a Resource Not a Trade-off
【速读】： 该论文旨在解决多语言大模型在非英语语种上的表现不足问题，特别是针对低资源语言在通用语言模型中因参数容量竞争而导致性能下降的瓶颈。现有模型普遍以英语为主进行训练，非英语数据占比极低（最高仅12%），即便专为欧洲语言设计的模型也难以超越通用模型的性能。为此，作者提出Lara Mini base，一个从头训练的多语言基础模型，其训练数据包含6T tokens的英语及44种其他语言，覆盖欧盟全部24种官方语言。关键创新在于采用混合专家（Mixture of Experts, MoE）架构，总参数量达7.2B，但仅激活0.6B参数，在保证高效推理的同时显著缓解了不同语言间的容量竞争。实验表明，该模型在不超过2B活跃参数的开源基础模型中，于标准英语、多语言、数学和代码基准上均取得最高平均得分，且在支持语言的多语言和数学任务上优势最为明显；同时支持长达128k token的上下文输入。此外，该MoE架构使非英语语言在训练后期逐渐超越英语表现，实现更均衡的语言能力分配。对于单用户在单张GPU上运行，未使用量化时，其生成速度比同等或更高性能模型快至少1.8倍。研究还发布了基于CC BY-NC-SA 4.0许可的Lara Mini base权重，以及包含110种语言变体的高质量常识推理数据集T-HellaSwag。
链接: https://arxiv.org/abs/2611.16956
作者: Translated,Maria Pia Adinolfi,Isabelle Andrieu,Grazia Fiorella Anfuso,Melania Annunziata,Daniele Antonini,Marta Arena,Luca Barone,Dana Barras Franco,Francesca Basiricò,Francesca Bee,Sofia Bertini,Nicola Bertoldi,Giorgio Biancini,Alessandra Bigi,Giulia Boccalini,Tommaso Bonomo,Enrico Boscardi,Sebastien Bratieres,Francesco Brigante,Giulia Brignola,Romolo Buondestino,Brenda Patricia Byrne,Giorgia Campora,Edoardo Cappelli,Veronica Carascosi,Francesca Carducci,Cody Carlson,Davide Caroselli,Ambra Cascio,Dario Cascio,Mauro Cassani,Andrea Castro,Alessandro Cattelan,Giulia Ceccacci,Paola Cellini,Giulia Chiriatti,Federico Cialini,Valentina Ciarlariello,Daniele Cocozza,Emanuela Coghe,Kathleen Compares,Marco Comunità,Davide Conti,Helen Rebecca Corfield,Michele Cortellessa,Anna Cracco,Aurora Cuppone,Arianna D’Andrea,Maurizio Davoli,Andrea De Angelis,Andrea De Blasi,Rebecca De Giorgi,Annalisa De Petra,Arianna De Pilla,Joris Demiraj,Elena Deufemia,Federica Di Antonio,Pierluigi Di Cianni,Vincenzo Di Cicco,Michela Di Clemente,Alan Di Francesco,Ylenia Di Gese,Federica Di Iulio,Claudia Di Lorenzo,Andrea Di Paola,Giulia Di Pilla,Jessica Dickson,Francesco Elefante,Ilaria Fabrizio,Alessandro Falappa,Viviana Fantozzi,Antonio Farina,Andrea Fernandez Tostado Ro.Diego Andrea,Andrea Ferrari,Sara Ferri,Simona Ferri,Fabio Fioramonti,Stefano Florio,Mariachiara Fornaro,Anna Maria Foti,Niccolò Fransoni,Alessandro Fusacchia,Emma Gamba,Francesco Gasperoni,Davide Gattini,Marta Giani,Claudia Giannetti,Valerio Giannini,Antonella Giudici,Mara Gramendola,Michele Grandinetti,Eleonora Graziani,L’Meese Hope Greaney,Dario Gubernari,Filomena Guglielmi,Jonathan Hammond,Stephen William Harris,Rowena Elizabeth Hogg,Marta Hontanon Del Olmo
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 51 pages, 14 figures, 29 tables. Model weights: this https URL
Abstract:General-purpose language models are trained mostly on English: over open models we review, at most, non-English pretraining data makes up only a 12% share. Models built for European languages give them far more, but score well below general-purpose models of the same size. A common concern is that, in small models, languages compete for capacity. We present Lara Mini base, a model trained from scratch on 6T tokens in English and 44 other languages, which together cover all 24 official languages of the European Union. As training progresses, the 44 other languages together come to outweigh English. To give these languages enough capacity at low cost, Lara Mini base uses a mixture of experts architecture with 7.2B total and 0.6B active parameters. Among open base models with up to 2B active parameters, it achieves the highest average score across standard English, multilingual, math, and code benchmarks, ahead of models trained on 4 to 6 times as many tokens. Its margin is largest on multilingual benchmarks in the languages it supports and on math benchmarks. It supports contexts of up to 128k tokens. We find that our MoE architecture reduces capacity competition between languages. For a single user on one GPU, without quantization, it generates text at least 1.8 times faster than any model with a higher average score. We release the weights of Lara Mini base under the CC BY-NC-SA 4.0 license. In addition, we will release T-HellaSwag, a higher-quality machine translation of the commonsense reasoning benchmark HellaSwag into 110 language variants.
[NLP-22] SPADE: Soft Prompts for Ai-text DEtection
【速读】： 该论文旨在解决当前AI生成文本检测方法在开放世界场景下鲁棒性不足的问题。现有检测方法仅关注生成文本本身——监督式方法依赖于标注文本，零样本方法则基于文本的概率分布，均忽视了提示（Prompt）对生成过程的深层影响，导致泛化能力受限。本文提出SPADE（Soft Prompts for AI-text DEtection）框架，其核心创新在于将提示对生成内容的影响作为判别信号，通过引入点互信息（Pointwise Mutual Information, PMI）度量提示对文本的解释能力，量化提示对生成结果的贡献。在此基础上，SPADE为每一类文本训练一个可学习的软提示（soft prompt），以增强不同类别间的区分性，并根据哪个提示最能解释待测文本来实现分类。实验表明，该方法在多个基准数据集上均优于现有检测器，尤其在分布外（out-of-distribution, OOD）数据上表现优异，实现了当前最优（SOTA）的检测性能，显著提升了跨数据集的迁移能力。
链接: https://arxiv.org/abs/2611.16929
作者: Alex Jiang,Didier Schwab,Chuyuan Li
机构: ChapsVision(法国); Univ. Grenoble Alpes(格勒诺布尔阿尔卑斯大学), CNRS(法国国家科学研究中心), Grenoble INP*(格勒诺布尔国立理工学院), LIG(信息与图形学实验室), 38000 Grenoble, France
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 24 pages (8 main), 10 figures, 22 tables
Abstract:The easy access to Large Language Models (LLMs) has made AI-generated texts on the web increasingly ubiquitous, making AI-text detection a crucial task. Prompts greatly influence the generation process, but existing detectors focus solely on the generated texts — supervised methods learn from labeled texts only, and zero-shot methods rely on the text’s probability distributions — ignoring the underlying generation process, which limits their robustness in open-world scenarios. We introduce SPADE: Soft Prompts for Ai-text DEtection, a novel framework for detecting AI-generated text that transfers better across datasets. Specifically, we measure the prompt’s influence and use it as a discriminative signal. We characterize the prompt’s influence by introducing a Pointwise Mutual Information (PMI) term that captures how much a prompt explains a text. Based on this quantity, SPADE trains one soft prompt per class to amplify the separation, and predicts the class whose prompt best explains the candidate text. Extensive experiments on several benchmarks against various detectors demonstrate the robustness and effectiveness of our method, achieving state-of-the-art (SOTA) performance on out-of-distribution (OOD) datasets.
[NLP-23] Beyond Word Error Rate: A Failure Taxonomy of Streaming Dictation ASR
【速读】： 该论文旨在解决语音识别（ASR）系统在实际应用场景中，尤其是实时字幕输入（如语音转写为消息）时，其性能评估与真实用户需求之间的脱节问题。传统评估通常依赖于在去除标点、大小写和数字格式后的词错误率（WER），但这一指标未能反映用户实际需要修正的错误类型。研究发现，尽管多个系统在安静英文语境下具有相近的WER（如Zoom Scribe与OpenAI GPT-Live-Transcribe分别为7.4%和7.3%），但在轻声低语场景下表现差异显著：部分系统（如Zoom、AssemblyAI）丢失了26%-28%的轻声词汇，而少数系统仅丢失不足3%。关键发现在于，语音活动检测（Voice Activity Detection, VAD）的误判是导致轻声语音识别性能下降的核心原因——当使用Silero VAD过滤“静音”片段时，轻声语音的WER从10.1%飙升至51.4%，表明音频被错误丢弃。此外，在双语混合录音中，多数系统在未启用双语模式时会丢失非英语内容（86%-99%），而支持双语处理的系统仅丢失20%以内。因此，解决方案的关键在于改进语音活动检测的鲁棒性，并确保模型对多语言、低强度语音输入具备更强的包容性，以提升实际应用中的可用性与用户体验。
链接: https://arxiv.org/abs/2611.16919
作者: Harras Mansoor,Muhammad Wisal,Hassan Bhatti,Masood Shakir
机构: 未知
类目: Computation and Language (cs.CL); Sound (cs.SD)
备注: 35 pages, 27 tables, 2 figures. Code: this https URL Data: this https URL
Abstract:Speech recognition (ASR) systems are usually compared by word error rate (WER) on recorded files, after punctuation, capital letters and number format are removed. In dictation, the audio is streamed live and the text goes into the user’s message as written, so what matters is which mistakes the user has to fix. The test set, streamed to eleven ASR systems, comprises 106 scripted English dictations from one speaker (read in a quiet room, with a fan on, and whispered), 20 clips in each of eleven other language and dialect sets, and 400 recordings that join clips in two languages. Every error was sorted by a fixed rule into four levels: recognition, output formatting, language handling and, for the self-hosted models, audio dropped before the model. Systems with similar WER failed in different ways. Zoom Scribe and OpenAI’s GPT-Live-Transcribe have almost the same WER on quiet English (7.4% and 7.3%), but 34.7% and 8.8% on whispered English. Zoom and both AssemblyAI models left out 26-28% of whispered words, and still 21-26% with their end-of-turn detection relaxed or off; seven other systems left out at most 3%. Qwen3-ASR’s 32% came from the speech detection in our own deployment (6% when relaxed). Across 107 EARS speakers, the same four systems lost 7-31% of whispered words, while six others lost under 2%. With only English selected, Gemini, Deepgram and our self-hosted Whisper lost the non-English part of a two-language recording in 86-99% of recordings, against 33-46% with both languages selected, while GPT-Live-Transcribe, OpenAI’s realtime Whisper and Meta lost it in at most 20%. Running Whisper behind a Silero voice activity detector that drops audio judged to be silence raised whispered WER from 10.1% to 51.4%. Most English results come from one read speaker. The code, the recordings, both references and every transcript are released.
[NLP-24] Human or Synthetic? What a Fixed Budget Buys for Post-Training
【速读】： 该论文旨在解决大语言模型后训练中，人工标注与合成偏好数据在成本效益上的权衡问题。传统研究通常在固定样本数量下比较两类数据，但忽略了实际应用中预算约束的关键影响。本文通过设定固定的美元级标注预算，系统评估了在不同成本水平下，使用人工标注与合成偏好数据进行后训练的性能表现。其解决方案的关键在于引入“成本匹配”（dollar-matched）的评估范式，以真实经济成本为基准，量化比较两类数据在多个基础模型家族及任务（如有用性与摘要生成）中的有效性。研究发现，在低至中等预算范围内，合成数据展现出显著更高的成本效率，仅需200–400美元即可实现高胜率；而人工数据需达到约两个数量级以上的预算才能缩小差距，且在摘要任务上始终无法超越合成数据。此外，研究还揭示人工标注数据本身存在可靠性差异，即使经过领域限制和去重等标准处理，不同来源的人工数据仍表现出显著性能差距，凸显了数据质量对模型优化的重要影响。
链接: https://arxiv.org/abs/2611.16895
作者: Shourik Banerjee,Ethan Mendes,Mohit Raghavendra,Wei Xu,Alan Ritter
机构: Georgia Institute of Technology(佐治亚理工学院)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 20 pages, 9 figures. Code: this https URL , Data: this https URL
Abstract:Post-training of large language models typically combines supervised fine-tuning (SFT) with preference-based fine-tuning such as Direct Preference Optimization (DPO). Human-annotated and synthetically generated preference data differ substantially in per-sample cost, but prior work typically holds sample count fixed when comparing the two. In this work, we explicitly study the interplay between cost and performance by comparing the effect of post-training with human and synthetic preference data under a fixed, dollar-matched annotation budget. We train and evaluate multiple base model families across helpfulness and summarization tasks. We find that synthetic data is far more cost-efficient at low-to-moderate budgets, achieving high win rates with as little as \ 200–\ 400. On helpfulness, human data closes the gap only at budgets roughly two orders of magnitude larger; on summarization, it does not close the gap at any tested budget. We also find that human-annotated data varies in reliability: one widely used human preference source substantially underperforms another, and the gap persists even after standard remedies such as domain restriction and deduplication.
[NLP-25] Look It Up: Multi-bit LLM Watermarking via Codebook Search
【速读】： 该论文旨在解决生成式 AI (Generative AI) 中短文本（short LLM-generated texts）源归属标记（source attribution）的难题，现有基于多比特水印（multi-bit watermarking）的方法通常依赖于长文本生成或高成本解码才能可靠恢复信息，难以适用于短文本场景。其核心解决方案是提出一种基于码本（codebook-based）的水印方案：将n比特的有效载荷映射为更长的u比特码字，并通过逐令牌对数偏置（token-level logit biasing）在生成过程中嵌入；在提取阶段，利用预计算的秘密码本，通过汉明距离（Hamming distance）或莱文施泰因距离（Levenshtein distance）匹配接收序列与最近码字。实验表明，该方法在200词长度的生成文本上可实现16–24比特载荷的近乎完美恢复，在基础模型上对127词生成也表现出较强性能，但在部分指令微调模型生成的极短文本中表现较弱，主要由于输出分布熵过低导致嵌入可靠性下降。研究还探索了基于BCH码的变体以避免码本存储，但整体鲁棒性有所降低；采用莱文施泰因距离的变体增强了对删除和轻度改写攻击的鲁棒性，而插入和强重写仍具挑战性。通过对碰撞概率、码字分离度、提取时间、分词一致性及文本质量的系统分析，揭示了码本基水印在短文本溯源中的实用性与局限性。
链接: https://arxiv.org/abs/2611.16890
作者: Jarosław Janas,Paweł Morawiecki,Josef Pieprzyk
机构: Institute of Computer Science, Polish Academy of Sciences (波兰科学院计算机科学研究所); CSIRO (澳大利亚联邦科学与工业研究组织)
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 16 pages, 8 figures, 6 tables. Submitted to IEEE Transactions on Information Forensics and Security. Code: this https URL
Abstract:Attributing short LLM-generated texts to a specific source remains challenging for existing multi-bit watermarking methods, which often require long generations or costly decoding to recover payloads reliably. We propose a codebook-based watermarking scheme that maps an n -bit payload to a longer u -bit codeword and embeds it during generation through token-level logit biasing. During extraction, the received bit sequence is matched to the nearest codeword in a precomputed secret codebook using Hamming or Levenshtein distance. Experiments show that our method achieves near-perfect recovery for 16-24 bit payloads on 200-token generations and strong performance on 127-token generations for base models. Results are weaker on some instruction-tuned models with very short texts, where low-entropy output distributions reduce embedding reliability. We also study a BCH-based variant that avoids codebook storage but generally reduces robustness. The Levenshtein variant improves resilience to deletions and light paraphrasing, while insertions and stronger rewrites remain challenging. We analyze collision probability, codeword separation, extraction time, tokenization consistency, and text quality, highlighting both the practicality and limitations of codebook-based watermarking for short-form attribution.
[NLP-26] Pretraining Data Statistics Shape the Phases of Learning Entity Comparison in Language Models NEURIPS2026
【速读】： 该论文旨在揭示预训练数据如何在语言模型（Language Model, LM）的预训练过程中塑造其行为，特别是针对实体比较任务（如“法国和巴西哪个国家更大？”）的学习动态。研究通过九个涵盖五类实体的比较任务进行案例分析，发现模型在训练中会经历三个明确的学习阶段：早期阶段依赖实体出现频率作为选择依据；中期阶段转而根据提示中实体的位置（首次提及或末次提及）进行判断；晚期阶段则能够正确识别答案。研究的关键发现是，这三个阶段的出现具有因果性，由训练数据的统计特性决定——当仅使用少量特定任务的合成数据时，模型仅停留在前两个启发式阶段而无法掌握正确答案；而当引入大量合成数据时，模型会直接从频率启发式跳转至正确解答。此外，将实体频率分布从自然存在的齐夫分布（Zipfian distribution）改为均匀分布可削弱频率依赖阶段并加速任务学习。更重要的是，这些在小规模实验中观察到的三阶段训练动态在开源大模型OLMo-2 32B中亦被复现。因此，本研究的核心贡献在于证明了预训练数据的统计属性是启发式学习的因果驱动因素，并表明小规模合成实验可有效预测大规模模型的训练动态。
链接: https://arxiv.org/abs/2611.16882
作者: Yik Siu Chan,Jing Huang,Yanai Elazar,Atticus Geiger
机构: University of Maryland (马里兰大学); Pr(Ai)R Group; Stanford University (斯坦福大学); Bar-Ilan University (巴伊兰大学); Goodfire (好火)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 29 pages, 20 figures. NeurIPS 2026
Abstract:How does data shape language model (LM) behavior throughout pretraining? We investigate this question through a case study on entity comparison, e.g., Between France and Brazil, which country is larger?, across nine comparison tasks spanning five entity types. We begin with controlled experiments in which we train small LMs (163M parameters) on mixtures of natural text from pretraining corpora and synthetic task data. Our analysis identifies three distinct phases of learning: (1) an early phase where the LM selects entities by frequency, (2) a middle phase where the LM selects entities by position in the prompt (first vs. last mentioned entity), and (3) a late phase where the LM selects the entity that is the correct answer. We show that the emergence of these three phases is causally determined by statistical properties of the training data. With small amounts of task-specific synthetic data, we observe only the first two phases and the model fails to learn the task; with large amounts, the model jumps directly from the frequency-based heuristic to solving the task correctly. Moreover, changing the entity frequency distribution from a naturally occurring Zipfian distribution (a small number of entities are very common and the vast majority are rare) to a uniform one weakens the frequency phase and speeds up learning of the task. Lastly, we show that the three phases of training dynamics from our controlled experiments appear in the open source model OLMo-2 32B. Together, our findings demonstrate that properties of pretraining data are causal drivers of heuristic learning and show that small-scale synthetic experiments can predict training dynamics at larger scales.
[NLP-27] Shortcuts or Memorisation? Revisiting Data Corruption in Natural Language Understanding with Large Language Models
【速读】： 该论文旨在探究大规模语言模型（LLM）在自然语言推理（NLI）任务中对输入扰动的鲁棒性是否与预训练编码器类似，以及这种鲁棒性是否源于模型在预训练阶段对基准数据集的直接记忆。研究发现，尽管所有测试的八种LLM在去除名词等关键词汇后仍保持36%-72%的高于随机水平的性能（在MNLI上），但在ANLI上仅维持14%-29%，表明其鲁棒性并非普遍存在于所有任务中。唯一例外的是一个在回答前进行推理的模型，其表现模式不同。通过分析数据泄露指标（如公开预训练数据中的精确n-gram匹配度），研究未发现模型暴露程度与鲁棒性之间存在关联。进一步使用368个全新且经人工验证的、模型未曾接触过的样本进行测试，结果表明：当输入以MNLI风格撰写并保留其语义线索时，所有LLM均表现出更强的鲁棒性；而当去除承载这些线索的功能词后，该优势逆转。因此，作者认为，模型在熟悉NLI基准上的扰动鲁棒性主要反映的是对数据集“捷径”（dataset shortcuts）的依赖，而非对训练数据的记忆，这揭示了基准设计本身的缺陷，而非模型能力的真实体现。
链接: https://arxiv.org/abs/2611.16843
作者: Aarne Talman,Stergios Chatzikyriakidis
机构: University of Helsinki (赫尔辛基大学); Institute of Language and Speech Processing; Athena Research Centre (雅典研究中心), Athens
类目: Computation and Language (cs.CL)
备注:
Abstract:Pre-trained encoders fine-tuned on natural language inference (NLI) and other language understanding tasks keep much of their accuracy after the input is perturbed, e.g. by removing all nouns or shuffling the words. Does the same hold for large language models (LLMs), and if so, is it because they have seen the benchmarks during pre-training? In this paper we revisit these perturbation experiments with eight LLMs alongside fine-tuned encoders on GLUE and the three rounds of ANLI. Regardless of model size, every system that answers directly retains 36-72% of its above-chance performance without nouns on MNLI compared to only 14-29% on ANLI. Only a model that reasons before answering departs from this pattern. We also check contamination indicators (including exact n -gram counts in the public pre-training data of two of the models) and find no link between exposure and robustness. Our strongest evidence comes from 368 new, human-validated items that none of the models can have seen. All eight LLMs keep more of their performance without nouns on items written in MNLI’s style than on items written without its cues. When the perturbation instead removes the function words carrying such cues, this advantage reverses. In light of these results, we argue that robustness to perturbation on familiar NLI benchmarks reflects dataset shortcuts rather than memorisation, and tells us more about the benchmark than about the model.
[NLP-28] Outcome-Mass Alignment: Controlling Length-Dependent Trajectory Influence in Multi-Turn LLM Agent Reinforcement Learning
【速读】： 该论文旨在解决多轮大语言模型（LLM）智能体在长交互上下文中因轨迹扩展（trajectory expansion）引发的隐式权重偏差问题。具体而言，当通过逐步分组相对训练（stepwise group-relative training）将源轨迹展开为多个响应行和有效标记时，会产生两种关键的统计失衡：结果统计多重性（outcome-statistics multiplicity），即单个任务结局被多次计数；以及目标质量不平衡（objective-mass imbalance），即不同轨迹在内部聚合后所分配的总目标质量不均等。这两种问题共同导致优化过程中对成功与失败轨迹的加权失真，进而影响训练效率与性能表现。其解决方案的核心是提出结果-质量对齐（Outcome-Mass Alignment, OMA），一种双组件方法：结果对齐（Outcome Alignment）显式地基于唯一源轨迹计算结果统计，避免重复计数；质量对齐（Mass Alignment）则在优化器原生标记级目标聚合后，调控各轨迹的总目标质量，确保其一致性。该方法不修改基础优化器的信用分配机制、重要性比例或裁剪策略，具备良好的跨优化器兼容性。实验表明，在两个模型规模下对齐的ALFWorld GRPO任务中，OMA使匹配时间窗终点成功率提升7.0–9.4个百分点；在以成功为导向的WebShop-SOL设置中，成功nAUC从0.399提升至0.511。消融实验及与其他优化算法（GiGPO、GSPO）的对比进一步验证了两组件的可分离性与通用性。研究揭示：轨迹扩展本质上定义了一种隐式加权规则，但其实际影响取决于扩展过程与任务结果及优化器原生更新之间的耦合方式。
链接: https://arxiv.org/abs/2611.16816
作者: Dawei Zhang,Yubin Huangfu,Jessica Fan,Boyu Wei,Likang Wu,Hongke Zhao,Ming He,Zhu Teng
机构: University of Chinese Academy of Sciences (中国科学院大学); Institute of Software, Chinese Academy of Sciences (中国科学院软件研究所); AI Lab at Lenovo Research (联想研究院人工智能实验室); University of North Carolina at Chapel Hill (北卡罗来纳大学教堂山分校); Tianjin University (天津大学); Beijing Jiaotong University (北京交通大学)
类目: Computation and Language (cs.CL)
备注: 18 pages, 6 figures, including appendices
Abstract:To keep long interaction contexts tractable, stepwise group-relative training for multi-turn LLM agents materializes each source trajectory into multiple response rows and valid tokens. This expansion creates an outer-accounting problem distinct from within-trajectory credit assignment: row-based outcome statistics can count one episode outcome multiple times, while token-level reduction can assign unequal aggregate objective mass across trajectories. We term these effects outcome-statistics multiplicity and objective-mass imbalance. We introduce Outcome-Mass Alignment (OMA), a two-component method that makes both accounting choices explicit. Outcome Alignment computes outcome statistics over unique source trajectories, while Mass Alignment controls aggregate objective mass after within-trajectory aggregation of optimizer-native token objectives. Neither redesigns the base optimizer’s credit assignment, importance ratios, or clipping. Across matched ALFWorld GRPO comparisons at two model scales, where failed trajectories typically expand much more than successful ones, OMA improves matched-horizon endpoint success by 7.0-9.4 percentage points. Ordinary WebShop exhibits weak or reversed expansion-outcome coupling and correspondingly mixed effects, whereas in a success-oriented WebShop setting, WebShop-SOL, OMA improves success nAUC from 0.399 to 0.511. Component ablations and GiGPO and GSPO comparisons further support component separability and cross-optimizer compatibility. These results show that trajectory expansion always defines an implicit weighting rule, but its practical effect depends on how expansion couples to outcomes and optimizer-native updates.
[NLP-29] Detection Is Not Early Warning: Evaluating Token-Level Alarms on Language Model Failures NEURIPS2026
【速读】： 该论文旨在解决语言模型在生成过程中因特定起始标记（onset）引发的失败问题，尤其是重复退化（repetitive degeneration）等典型故障。传统监控方法依赖于响应层面的评估指标（如AUROC），但这些指标忽视了监控触发的时间点，无法准确衡量预警能力。为此，论文提出PREWARN评估协议，针对具有标注起始点的故障类型，通过在验证数据上校准监控器阈值至目标错误率，并基于起始点对三种任务（分类、实时检测与提前预警）进行评分。研究以三个开源语言模型中的重复退化现象为案例，采用激活层上的线性探测器及六种未训练基线进行评估。结果显示，传统响应级指标在此类故障中趋于饱和，而仅依赖响应长度的规则可近乎完美表现；然而，基于AUROC排名相近的探测器在PREWARN下预警行为差异显著——其中部分探测器能在多数退化响应中于起始点前发出预警，低秩适配器（low-rank adapters）使探测器在起始点前约一个令牌内触发的比例翻倍，主要归因于其能提前识别退化模式。相比之下，所有未训练基线的预警频率接近随机水平。该工作表明，时间敏感的评估框架对揭示真实预警性能至关重要。相关探测器及其校准阈值已公开发布。
链接: https://arxiv.org/abs/2611.16803
作者: Marco De Negri(1),Luca Sartori(1),Lorenzo Baggi(1),Moritz Reihs(1),Eduard Frank Ďurech(2),Mennatallah El-Assady(1),Valentina Pyatkin(1 and 2),Anna Hedström(2 and 3) ((1) ETH Zurich, (2) ETH AI Center, (3) Solaris Research)
机构: ETH Zurich(苏黎世联邦理工学院); ETH AI Center(苏黎世联邦理工学院人工智能中心); Solaris Research
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: Accepted at the NeurIPS 2026 Workshop on Interpretability as a Science. 30 pages (10 main text), 6 figures, 16 tables. Code: this https URL
Abstract:Many failures of language models begin at a specific token, the onset. A monitor that reads the model during generation can classify the finished response, detect the failure once it is under way, or warn before it begins, and only a warning leaves time to stop or steer the generation. Monitors are usually evaluated with response-level metrics such as AUROC, which ignore when the monitor fires. We introduce PREWARN, an evaluation protocol for any failure with an annotated onset. It calibrates each monitor’s threshold on validation data to an error rate chosen for the application, runs the monitor, and scores its alarms relative to the onset on all three tasks. As a case study we apply it to repetitive degeneration in three open language models, with early warning as the main task, using linear probes on activations and six untrained baselines. Response-level metrics saturate on this failure, and a rule that reads only the response length scores almost perfectly, while probes that AUROC ranks as equal warn very differently under PREWARN. Probes warn before the onset on most degenerate responses of two of the three models, low-rank adapters below the probe more than double the share of tokens just before the onset that it flags, largely by flagging degenerate responses long before the onset, and no untrained baseline warns much more often than random alarms. We release the probes with calibrated thresholds at this https URL
[NLP-30] owards Green Text Generation: Memory-Based Speculative Decoding
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在文本生成过程中因推理速度慢、能耗高而导致的效率瓶颈问题。其核心挑战在于如何在保持生成质量的前提下，显著提升生成吞吐量并降低单位令牌的能量消耗。解决方案的关键在于提出一种名为Olifant的轻量级草稿模型（draft model），该模型基于记忆增强型分类器构建，能够在CPU上以低于0.3毫秒的延迟完成四元组上下文窗口内的下一个词元预测，且无需GPU、反向传播或加速器显存。通过在消费级硬件上评估两种验证器（verifier）配置下的性能表现，研究发现：当不使用键值缓存（KV cache）时，近似零开销的草稿生成可实现高达38.1%的接受率，使吞吐量提升2.43倍、单位令牌能耗降低2.51倍；而在启用KV缓存的验证器场景下，尽管高速验证器本身已具备约10倍能效优势，但传统草稿模型（如TinyLlama）因草稿成本过高导致吞吐量减半、能耗翻倍，而Olifant仍能维持1.25倍吞吐量提升并实现最低能耗（93 J/100 tokens，较未缓存基线低14倍）。此外，研究揭示了键值缓存使时间与能量解耦的特性——更长的草稿会增加功耗，从而使得能效最优的草稿长度低于速度最优长度。因此，该工作表明，在草稿生成近乎免费的前提下，推测式解码（speculative decoding）在快速验证器时代依然具有可行性与价值。
链接: https://arxiv.org/abs/2611.16714
作者: Antal van den Bosch
机构: Utrecht University(乌得勒支大学)
类目: Computation and Language (cs.CL)
备注: 11 pages, 2 figures, 7 tables. Published at the 19th International Natural Language Generation Conference (INLG 2026). Code: this https URL ; models: this https URL
Abstract:Speculative decoding accelerates text generation with large language models by letting a cheap draft model propose tokens that a large verifier checks in a single forward pass. We introduce Olifant, a draft model built on a memory-based classifier. It predicts the next token from a four-token window in under 0.3 ms on CPU, requiring no GPU, no backpropagation, and no accelerator memory. We measure throughput and energy under two verifier regimes on consumer hardware. Without a KV cache, drafting is nearly free: a domain-matched Olifant model reaches 38.1% acceptance on EU legal text, delivering 2.43x throughput and 2.51x lower energy per token than the neural verifier model alone. With a KV-cached verifier, itself a ~10x energy saver, drafting costs matter: a TinyLlama draft model halves throughput and triples energy despite high acceptance, whereas Olifant still yields 1.25x throughput and the lowest energy in the study (93 J per 100 tokens, 14x below the uncached baseline). We further show that KV caching decouples time from energy: verification of longer drafts raises power draw, shifting the energy-optimal draft length below the speed-optimal one. Speculative decoding is not dead in the age of fast verifiers, provided drafting is nearly free.
[NLP-31] Forecasting Sales Delay through Time-aware Analysis of Sales Events in Textual Reports
【速读】： 该论文旨在解决企业间（B2B）销售过程中订单实际接收时间与预定订单日期（scheduled order dates）之间存在偏差的问题，这种偏差导致研发资源准备不当从而增加开发成本。为降低此类成本，论文提出了一项新任务——销售延迟预测（sales delay forecasting），即在销售流程中预测从预定订单日期到实际收到订单之间的延迟天数。其核心挑战在于销售过程本身具有高度不确定性，且受客户订货意愿的影响。为此，论文提出了两类关键特征：一是“置信度特征”（confidence features），通过分析销售人员在文本报告中对未来销售事件的预期日期来量化其对客户订货意愿的信心；二是“状态特征”（status features），基于当前日期和历史事件来刻画销售流程的实际进展状态。实验结果表明，所提方法相比基于传统特征的基线模型实现了16%的平均绝对误差降低，并优于使用大语言模型的非特征基线方法，验证了显式构建特定领域特征的重要性。
链接: https://arxiv.org/abs/2611.16704
作者: Masaya Tsunokake,Yuta Koreeda,Yuhei Niwa,Terufumi Morishita,Teppei Inoue,Hiroaki Ozaki,Yasuhiro Sogawa
机构: Hitachi, Ltd.(日立有限公司); RD Group, Hitachi, Ltd.(日立有限公司研发集团)
类目: Computation and Language (cs.CL)
备注: 9 pages
Abstract:Business-to-business (B2B) companies develop and deliver a one-of-a-kind product in response to a client’s orders. To prepare resources for development beforehand, companies estimate dates when they will receive orders from clients in their sales process, called scheduled order dates. However, dates when the companies actually receive orders can differ from the scheduled order dates. This difference increases costs in the development. If the difference could be forecasted before receiving orders, companies could optimize their resource preparation. Towards reducing the costs, we introduce sales delay forecasting, a novel task of forecasting delay days from the scheduled order dates during sales process where sales representatives (sales reps) interact with the clients to receive orders. Sales delay forecasting is challenging because the sales process progress inherently involves uncertainty and depends on the client’s willingness to order. To address this task, we propose confidence features which represent sales reps’ confidence as a proxy measure of the client’s willingness to order. To represent it, our method extracts dates of future sales events anticipated by sales reps from textual reports. To consider whether the confidence objectively looks reasonable, we also introduce status features which describe sales process status based on current date and past events. In experiments using real sales data, our method achieved a 16% lower mean absolute error compared with a feature-based baseline similar to a prior work. Moreover, our method outperformed non-feature-based baselines using large language models, indicating that it is crucial to explicitly create our features.
[NLP-32] When Expert Disagreement Hurts: Prestige and Selectivity in LLM Decision Revision NEURIPS2026
【速读】： 该论文旨在解决生成式 AI 在面对其他代理（agent）提出异议时，其决策修正行为难以评估的问题：传统指标如修正次数或最终准确率无法区分模型是纠正错误，还是错误地放弃了原本正确的判断。为此，作者提出一种受控的两阶段审计方法（two-pass audit），在保持问题、证据和初始答案不变的前提下，仅改变修订情境——即中立反思、匿名异议，或归因于领域专家的异议。对于二元决策任务，研究发现只有当模型对错误答案的修正频率至少为正确答案修正频率的 $ a/(1-a) $ 倍（当初始准确率为90%时，该比值为9倍）时，修订才能真正提升准确率。在医学、科学与法律领域的5个模型、共1,989个基于证据的决策任务中，所有模型在专家异议情境下均未达到此阈值，且准确率最高下降16.1个百分点。尽管“专家”标签作为一种声望线索显著增加了模型的修正行为（影响四个模型），但其对准确率的影响从5.2个百分点的损失到2.4个百分点的增益不等，表明修正次数并不能有效反映实际损害。即使双方论证内容完全相同，该标签仍促使三个专有模型及四款较新模型的修正率上升。此外，模型还表现出对源可靠性声明的敏感性。因此，该研究强调，必须采用成对的、条件化的评估指标来系统评估各模型与修订协议下的真实性能，以避免误导性结论。
链接: https://arxiv.org/abs/2611.16694
作者: Yupeng Tang,Mingfeng Lin
机构: Georgia Institute of Technology(佐治亚理工学院); Scheller College of Business(斯切尔商学院)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: Accepted at NeurIPS 2026. 38 pages, 7 figures, 36 tables. Code and data: this https URL
Abstract:When language models revise a decision after another agent disagrees, neither reversal counts nor final accuracy shows whether they are fixing mistakes or abandoning correct answers. We introduce a controlled two-pass audit that holds the question, the evidence, and the model’s initial answer fixed and varies only the revision context: neutral reflection, anonymous disagreement, or disagreement attributed to a domain expert. For binary decisions the accounting is exact: revision raises accuracy only if it is selective enough, reversing wrong answers more than a/(1-a) times as often as correct ones, where a is initial accuracy (nine times at 90%). Across five models and 1,989 evidence-grounded decisions in medicine, science, and law, no model clears this bar under expert disagreement, and accuracy falls by up to 16.1 points. The expert label, a prestige cue, adds reversals for four of the five models, yet its effect on accuracy ranges from a 5.2-point loss to a 2.4-point gain, so reversal counts are a poor proxy for harm. With identical arguments from both sources, the label still raises reversal for three proprietary models and in a pooled analysis of four newer models. Models also respond to stated source reliability. Revision pipelines should therefore be evaluated with paired, conditional metrics, per model and protocol.
[NLP-33] MuSeToM: Multi-Sensory Evaluation of Belief Revision with Textual Cues Across Six Senses
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在理论心理（Theory of Mind, ToM）评估中长期忽视的一个关键问题：即模型是否能够判断感知证据是否足以支持他人信念的合理修正，这属于社会推理中的认识论信念更新（epistemic belief revision）范畴。现有评估多集中于静态信念归属与错误信念推理，而忽略了对信念更新依据的充分性进行判断的能力。为此，研究提出MuSeToM——一个经过人类验证、基于文本的基准测试，用于评估LLMs在认知信念更新方面的表现。其核心创新在于构建了2,415个源自BigToM和OpenToM的实例，涵盖六类感官线索，系统考察模型能否准确识别何时应更新角色信念，并推断出正确的更新后信念状态。实验结果表明，在直接提示与链式思维（chain-of-thought）提示下，尽管模型普遍能察觉信念可能发生变化，却难以正确识别更新后的信念，且常因感知显著但认识论上不足的线索而过度更新，暴露出模型在处理信念更新的合理性判断方面存在显著缺陷。因此，解决方案的关键在于建立一个可量化、高保真的基准测试框架，以揭示并推动模型在复杂社会认知任务中对信念更新逻辑的深层理解能力。
链接: https://arxiv.org/abs/2611.16662
作者: Chunkit Chan,Kawai Chung,Sze Heng Douglas Kwok,Yuxuan Liu,Yauwai Yim,Tianshi Zheng,Haochen Shi,Qing Zong,Wong Kai Chung,Pak Hin Lau,Calvin Yang,Yifan Gao,Xiaoman Pan,Ginny Wong,Simon See,Yangqiu Song
机构: The Hong Kong University of Science and Technology(香港科技大学); Stanford University; Amazon; NVIDIA AI Technology Center (NVAITC)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Theory of Mind (ToM), the ability to infer mental states in others, is pivotal for human social cognition. Existing evaluations of ToM in large language models (LLMs) mainly focus on static belief attribution and false-belief reasoning, leaving a critical aspect of social reasoning underexplored: can LLMs decide whether perceptual evidence is sufficient to justify another person’s belief revision? To address this gap, we present MuSeToM, a human-validated, text-based benchmark for evaluating epistemic belief revision in ToM reasoning. MuSeToM contains 2,415 instances adapted from BigToM and OpenToM, with textual cues spanning six sensory categories, and tests whether models can determine when a protagonist should revise a belief and infer the resulting updated belief state. Using MuSeToM, we systematically evaluate state-of-the-art LLMs under direct and chain-of-thought prompting, revealing a pronounced dissonance: while models often detect that a belief may change, they struggle to identify the updated belief and frequently over-update on perceptually salient but epistemically insufficient cues.
[NLP-34] Unibuc - NLP at “Voight-Kampff” Generative AI Detection PAN 2025
【速读】： 该论文旨在解决生成式 AI 文本检测中的两个核心问题：在子任务1“Voight-Kampff AI 检测灵敏度”中，如何高效准确地识别由大语言模型（Large Language Models, LLMs）生成的文本；在子任务2“人机协同文本分类”中，如何有效融合人类与AI生成文本的特征以实现更优的分类性能。针对子任务1，其解决方案的关键在于利用从LLM中提取的嵌入表示，并通过基于集成学习的轻量级机器学习模型（如Light GBM、XGBoost、逻辑回归和支持向量机）构建多数投票集成模型，该方法在不进行微调的情况下达到了约99.4%的F1分数，表现与对LLM进行下游任务微调相当，展现出高效率与强泛化能力。对于子任务2，关键创新在于采用基于Transformer骨干网络的多层特征融合策略，结合层间投影与注意力池化机制，以捕捉不同层级语义信息，尽管效果较为有限，但为后续研究提供了可借鉴的特征融合范式。
链接: https://arxiv.org/abs/2611.16612
作者: Teodor Marchitan,Claudiu Creanga,Liviu P. Dinu
机构: 未知
类目: Computation and Language (cs.CL)
备注:
Abstract:This paper describes the approach of the Unibuc - NLP team in tackling the “Voight-Kampff” Generative AI Detection task from PAN at CLEF 2025. In the Subtask 1: Voight-Kampff AI Detection Sensitivity we have explored different approaches of leveraging embeddings extracted from LLMs, using either classical ML models or further fine-tuning LLMs with a classification head for the downstream binary classification task. Our experiments showed that using a majority vote ensemble of classical ML models (Light GBM, XGB, Logistic Regression and SVM) is comparable with fine-tuning for downstream task achieving around 99.4% F1 score. For the Subtask 2: Human-AI Collaborative Text Classification we explored a way of combining features at different layers extracted using transformer backbone with layer-wise projection and attentive pooling which proved modest results.
[NLP-35] Do the Details Help the Whole? One-Way Transfer from Person Actions to Group Intent in VLMs on JRDB-Social
【速读】： 该论文旨在探究多任务学习中同时训练个体行为（person actions）与群体意图（group intent）对视觉语言模型（VLMs）性能的影响，核心问题是：在联合训练时，群体意图是否能通过个体行为的监督得到有效提升？研究发现，尽管传统假设认为双层次标注有助于相互促进，但在实际实验中，将群体意图作为目标进行训练并不能显著提升个体行为识别性能，反而可能引入干扰。其关键解决方案在于采用“先写个体行为、再写群体意图”的提示顺序——这种顺序可使群体意图的宏平均F1从30.1提升至31.2，而个体行为的F1仅下降0.2点以内，表明正确的、图像引导的个体行为信息是群体意图推断的关键前提。进一步分析显示，群体意图的预测依赖于生成的个体行为内容，且更关注人物特征，部分语义驻留于视觉编码器中；若提供空的行为列表或借用同意图组的错误行为，则会显著降低意图识别效果。此外，将行为作为输入而非目标时，反而损害了两个层级的表现，说明任务设计中的目标设定与信息流方向至关重要。
链接: https://arxiv.org/abs/2611.16581
作者: Chengguang Gan,Zhixi Cai,Yunhao Liang,Hanjun Wei,QingHao Zhang,Shiwen Ni
机构: Independent Researcher; Monash University (莫纳什大学); University of Chinese Academy of Sciences (中国科学院大学); Pusan National University (釜山国立大学); Shenzhen University of Advanced Technology (深圳先进技术研究院)
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL)
备注:
Abstract:Social scene datasets label people at two levels: what each person is doing and what the group is there for. Training on both levels at once is often assumed to help each of them. We test this assumption for vision-language models (VLMs) on JRDB-Social, treating person actions as the details and group intent as the whole. We fine-tune Qwen3.5 4B under eight training arms with three seeds, repeat the core arms on Qwen3.5 9B and Qwen3-VL 4B, and judge every arm with a verdict rule fixed before testing. Writing each member’s actions before the intent raises intent macro-F1 from 30.1 to 31.0, and writing the intent first reaches 31.2, while action F1 stays within 0.2 points of an action-only model. Under the selected recipe, no arm trained with intent improves actions. The gain requires correct, image-grounded actions: empty action lists or actions borrowed from another group with the same intent lower intent F1. Forced-prefix tests, occlusion, member masking, linear probes and component grafting show that the intent follows the written actions, relies more on people and partly resides in the vision tower. Supplying actions as input instead of as targets hurts both levels.
[NLP-36] Lost in Delegation: Studying the Effectiveness of Task Orchestration in Generative Search
【速读】： 该论文旨在解决生成式 AI 系统中任务委托（delegation）机制在复杂任务执行中的有效性问题，尤其关注当任务涉及模型自身知识无法覆盖的“截止后信息”（post-cutoff information）时，委托是否能提升性能。研究发现，当前主流的委托模式并未带来一致性的准确率提升，且在部分情况下反而显著降低性能；相比之下，直接使用工具与同一模型进行交互的表现更优。其关键解决方案在于揭示了委托失败的三大核心原因：被委托查询的语义扭曲、工具使用效率低下，以及代理间信息传递与保留机制的失效。因此，有效的系统协调不仅依赖于精准的委托决策，还需构建高效的信息交换与上下文保持机制。
链接: https://arxiv.org/abs/2611.16526
作者: Tamara Stojanovska,Elisabeth Kirsten,Leon Swazinna,Muhammad Bilal Zafar
机构: Ruhr University Bochum(鲁尔大学波鸿分校); UA Ruhr Research Center for Trustworthy AI(鲁尔大学信任人工智能研究中心)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:LLMs have quickly evolved from models used in isolation to agentic systems that delegate tasks to external tools and other LLMs to solve complex assignments. This agentic augmentation has led to impressive capability gains. However, when and how delegation improves performance remains poorly understood. For example, it is unclear whether delegation helps when tasks involve post-cutoff information that the orchestrating LLM cannot reliably resolve from its own knowledge. In this work, we use purpose-built datasets with pre- and post-cutoff splits to compare various delegation modes. We find that delegation provides no consistent accuracy improvement over non-delegated setups, with effects varying across models and tasks. When tasks involve knowledge unavailable to the orchestrator, delegation can substantially degrade accuracy, while direct tool use with the same LLM performs much better. Our analysis identifies distorted delegated queries, inefficient tool use, and failures to preserve and use information between agents as recurring failure modes. These findings suggest that effective orchestration requires both well-calibrated delegation decisions and effective exchange of information between agents.
[NLP-37] EvenVLM: A Text and Multimodal Dataset for the Even Language with Vision-Language Fine-Tuning ACL EMNLP2026
【速读】： 该论文旨在解决濒危语言Even（Lamut）在自然语言处理领域长期缺乏高质量文本与多模态数据集的问题，进而推动其在机器翻译、语法分析、阅读理解及视觉问答（VQA）等任务中的技术研究。其核心挑战在于Even语具有复杂的黏着性形态结构，导致传统分词方法效率低下，且低资源语言普遍面临标注数据稀缺的困境。解决方案的关键在于：首先，构建首个涵盖50,521条结构化Even-俄语词条的大型数据集，并通过GPT-4o Mini驱动的自动化解析与合成生成技术扩充至107,543个训练样本；其次，提出一种三步式分词器扩展算法，引入Even特有字符（如eng、oe、o-umlaut）、白名单词汇及BPE合并规则，使平均词内分词数降低27%（从4.99降至3.64），显著提升分词效率；最后，采用LoRA微调策略对Qwen3-VL-4B模型进行轻量化适配，在仅0.53%参数可训练的前提下实现文本任务与多模态任务的独立优化，最终在文本基准上取得11.4%精确匹配率和13.6加权得分，在句子翻译中达到0.1283 chrF，在视觉问答任务中准确率提升至65%（相较基线模型0%的性能实现突破）。研究成果包含完整数据集、评估基准与训练流程，为其他低资源语言的可持续保护与技术赋能提供可复用范式。
链接: https://arxiv.org/abs/2611.16500
作者: Daniil Makovetskii,Ivan Bondarenko
机构: Novosibirsk State University (新西伯利亚国立大学)
类目: Computation and Language (cs.CL)
备注: Published in Proceedings of the Workshop on Open Reasoning Across Cultures Languages (ORACLE) at EMNLP 2026
Abstract:We present EvenVLM, the first text and multimodal (VQA) dataset for Even (Lamut), an endangered Northern Tungusic language spoken by approximately 6,800 people in Siberia. Our dataset comprises 50,521 structured Even-Russian entries extracted from dictionary and phrasebook sources via GPT-4o Mini-based parsing, from which we synthetically generate 107,543 training examples spanning translation, grammar, reading comprehension, and semantic classification. To address tokenization inefficiency on Even’s agglutinative morphology, we propose a three-step tokenizer expansion algorithm that adds Even-specific graphemes (eng, oe, o-umlaut), whitelisted vocabulary, and BPE merges, reducing average tokens per word by 27% (4.99 to 3.64). Using LoRA fine-tuning of Qwen3-VL-4B, we adapt the model for text tasks (0.53% trainable parameters, rank 32) and separately for visual question answering (rank 64, frozen vision encoder). Our best configuration achieves 11.4% exact match and a 13.6 weighted score on text benchmarks, 0.1283 chrF for sentence translation, and 65% accuracy on VQA (up from 0% for the base model and text-only fine-tuning). We release our datasets, evaluation benchmarks, and training pipeline to support further research on Even and other low-resource languages.
[NLP-38] Where in the Input Does a One-Pass Decision Encoder Look?
【速读】： 该论文旨在解决生成式AI模型在处理长文本时因输入窗口限制导致的“信息截断”问题，即当关键证据（evidence）位于输入窗口之外时，模型性能显著下降。其核心挑战在于：尽管模型在单次前向传播中完成多选题推理，但受限于最大512个标记（token）的上下文窗口，较长文本会被无声截断，而模型无法感知此截断状态，导致错误判断。解决方案的关键在于通过微调策略与输入结构优化来缓解这一问题：首先，使用长文档微调（long-document fine-tune）可使模型在可见证据下准确率提升27.6个百分点，显著优于短行微调（short-row fine-tune）；其次，在不依赖训练数据的前提下，采用312个标记的滑动窗口扫描策略，通过多次前向传播（6–11次）实现对长文档的分段覆盖，带来11.9个百分点的性能增益。此外，研究发现模型对证据位置敏感——当证据位于前置文本之后时，准确率从60%降至38.5%，凸显了输入顺序的影响。然而，模型置信度评分与运行时截断标志均无法有效区分截断请求与完整读取请求（AUROC仅为0.50–0.55），表明缺乏可靠的故障预警机制。总体而言，该研究通过实证分析揭示了当前小规模模型在长文本理解中的固有局限，并提出以微调和结构化输入重排为关键改进路径。
链接: https://arxiv.org/abs/2611.16477
作者: Mohammed Nayeem Khan,Allan Kevin T
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 25 pages, 19 figures, 23 tables
Abstract:Laya is a small model that answers a multiple-choice question about a text in one pass: the question, the options (each with a name and a description) and the text form one sequence of at most 512 tokens, the window, in the English release. A longer text is silently cut at the end, and we measure what that costs. Test documents place a held-out row’s text, the evidence, at a chosen token offset among distractor text; 177 items, five tasks, three checkpoints (released, short-row fine-tune, long-document fine-tune), four window sizes and 93,275 scored requests. When the evidence lies beyond the window, accuracy falls to 26.4% for the released checkpoint (27.2% for both fine-tunes), near the 27.0% of guessing and below the 35.6% of always giving the commonest label. Confidence gives no warning (AUROC 0.50 to 0.55 for telling cut from read requests; 0.5 = chance), and the runtime’s truncation flag also fires on requests whose evidence was visible. The released checkpoint reads visible evidence badly once text precedes it: 38.5% pooled against 60% when the evidence comes first, so a larger window alone gains +6.2 points. Fine-tuning on long documents adds +27.6 points on visible evidence over the short-row fine-tune (95% interval +22.3 to +33.1) and nothing on cut evidence. That is an in-family result: training rows and test documents share generator and question types, and the main filler is of the training kind. With encyclopedic filler never seen in training the gain is +13.2 (+9.9 to +16.6), with rows of other tasks as filler -2.0. Among training-free rearrangements a 312-token window scan helps most (+11.9 points over the default for the released checkpoint) at six to eleven forward passes. Documents are synthetic, labels are LLM-made and one backbone is studied, so these diagnose this model, not long documents in general.
[NLP-39] How Should Planning Agents ReSHAPE Their Search Space?
【速读】： 该论文旨在解决现有规划代理（planning agents）与评估基准将用户初始需求视为固定不变的问题，而实际上用户在未见到具体方案前往往无法意识到自身可接受的权衡（trade-offs）。传统方法通过交互式询问偏好来修订请求，但其对规划效果的影响尚不明确。为此，论文提出ReSHAPE任务，其核心是判断请求修订是否使搜索空间包含可接受的计划，并确保规划器能在其中找到可行解。关键解决方案在于：将用户请求视为待验证的假设，通过向用户展示具体计划来观察其反应，从而识别出用户潜在的权衡偏好。基于人工验证的模拟用户构建的ReSHAPE Planter工具，在旅行和购物场景中分析了5,800条由代理修订的规格说明，结果表明，仅引入一个正确的权衡即可使相同规划器找到可接受计划的概率提升最高达4.6倍，且搜索空间质量能有效预测规划器的输出结果。相比之下，当前代理仅能涵盖最多32%的可接受计划，而单一正确权衡本身即可覆盖超过52%，原因在于其依赖偏好询问而非展示实际计划；在旅行场景中，展示计划相较询问可使F1得分提高9.7分，且适用于两种模型架构。因此，论文主张规划代理应将请求视为可测试的假设，通过用户对具体计划的反馈来识别其隐含权衡，从而显著提升规划成功率。
链接: https://arxiv.org/abs/2611.16466
作者: Yilin Zhang,Xinran Zhao,Tong Chen,Ziming Yuan,Tongshuang Wu
机构: Carnegie Mellon University (卡内基梅隆大学); University of Washington (华盛顿大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Planning agents and benchmarks treat a user’s initial specification as fixed. Yet the requests users write exclude plans they would gladly take, since they rarely discover the possible trade-offs they would accept until seeing the options. Agents already revise requests through interaction, but whether this helps planning is unknown. We formalize this step as the ReSHAPE task, which judges a revision by whether its search space matches the acceptable plans and a planner finds one there. Using ReSHAPE Planter, which builds instances from planning benchmarks with a human-validated simulated user, we analyze 5,800 agent-revised specifications in travel and shopping. We find that reshaping matters: one correct trade-off makes the same planner reach an acceptable plan up to 4.6x as often, and search space quality predicts what the planner finds. Current agents, however, admit at most 32% of the acceptable plans, where that trade-off alone admits over 52%, because they ask about preferences instead of showing plans; in travel, showing plans raises F1 by 9.7 points over asking, with two model families. Planning agents should thus treat a request as a hypothesis to test against the user’s reactions to concrete plans to recognize user trade-offs.
[NLP-40] daVinci-Prover: Agent -Native Mid-Training for Formal Theorem Proving
【速读】： 该论文旨在解决当前大型语言模型（LLM）在形式化定理证明中进展受限的核心问题：缺乏透明且高效的训练数据配方，以系统性地提升定理证明模型的能力。其核心挑战在于如何有效融合数学推理能力先验（mathematical skill priors）与智能体交互轨迹（agentic trajectories），从而支持长时程、交互式的证明构建。解决方案的关键在于提出一种面向智能体的中段训练（agent-native mid-training）数据配方，包含337亿个标记的语料库，覆盖分解（decomposition）、搜索（search）、自验证（self-verification）与修订（revision）等关键环节，并结合基于验证器（verifier-grounded）的智能体轨迹。为此，研究进一步设计了daVinci-Prover框架，使模型可在隔离工作空间中与Lean验证器交互，成功收集约3.6万条高质量智能体轨迹用于监督微调。实验表明，该方法在MiniF2F、FATE-M和ProofNet等多个基准上相较同规模的Qwen3-8B仅经监督微调的基线平均提升13.7分；应用于Qwen3.5系列时亦展现出卓越性能，实现接近饱和的MiniF2F表现并具备内存压缩能力。结果证实，面向智能体的中段训练是构建更强交互式定理证明智能体的有效基础。
链接: https://arxiv.org/abs/2611.16433
作者: Wenhong Zhu,Xuefeng Li,Jiashuo Zhang,Yu Chen,Tiantian Mi,Qipeng Guo,Pengfei Liu
机构: Generative Artificial Intelligence Research Lab (GAIR); Shanghai Innovation Institute; Shanghai Jiao Tong University (上海交通大学); Fudan University (复旦大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Advances in large language models (LLMs) are rapidly transforming formal theorem proving. Yet further progress remains constrained by the lack of a transparent and effective data recipe for systematically improving theorem-proving models. In this work, we argue that effective agentic provers require strong mathematical skill priors to ground theorem proving and agentic trajectories to enable long-horizon, interactive proof construction. To this end, we synthesize an agent-native 33.7B-token mid-training corpus covering decomposition, search, self-verification, and revision, together with verifier-grounded agentic trajectories. We further introduce the daVinci-Prover harness, which enables models to construct proofs and interact with a Lean verifier in an isolated workspace, and use it to collect approximately 36K high-quality agent trajectories for supervised fine-tuning. Our mid-training recipe yields an average improvement of 13.7 points over the matched Qwen3-8B SFT-only baseline across MiniF2F, FATE-M, and ProofNet. When applied to the Qwen3.5 series, our approach also yields strong performance across theorem-proving benchmarks and achieves near-saturating MiniF2F performance with memory compaction. The results highlight agent-native mid-training as an effective foundation for developing stronger interactive theorem-proving agents.
[NLP-41] Experience-Handoff Reinforcement Learning for Recursive Self-Improvement
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在代理任务中采用强化学习（Reinforcement Learning, RL）时，仅依赖标量奖励而忽略试运行过程中环境提供的细粒度信息（如错误日志）的问题。现有方法虽通过引入反思或压缩机制构建可传递的经验（如“剧本”playbook），但这些经验的生成过程未被学习，缺乏自适应性。为实现兼顾标量奖励梯度更新与环境信息利用的双重优势，本文提出\acerl方法，在固定代理上下文工程（Agentic Context Engineering, \ace）框架下，联合优化三个角色：动作生成器、洞察反思器和剧本编纂者。其核心在于通过评估候选经验传递产物对未来执行轮次的下游影响，实现对经验手递机制的端到端学习。实验表明，\acerl在多个代理基准测试中优于基线方法，验证了在训练过程中不仅应优化当前轮次的动作，还应优化模型生成并传递给未来轮次的可继承经验。
链接: https://arxiv.org/abs/2611.16409
作者: Doyoung Kim,Byeongguk Jeon,Kailin Zhang,Sherry Yang
机构: 未知
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注:
Abstract:It is common to train Large Language Models (LLMs) for agentic tasks with Reinforcement Learning (RL) from scalar rewards. However, this throws away the fine-grained information about the environment that the agent experienced during the trial (e.g., error logs). To utilize this rich information, recent works wrap the agent in a harness that refines its own inputs through reflection or compaction and hands past experience to future rollouts, but the construction of the handed-off artifacts (e.g., a playbook) is itself not learned. We want the best of both worlds: gradient updates from a scalar reward, while also utilizing environment information to hand off experience. As a concrete testbed, we fix the Agentic Context Engineering (\ace) harness and train one model to act, reflect, and curate task-local playbooks. Our method, \acerl, evaluates candidate experience-handoff artifacts by their downstream effect on future actor rollouts, and co-optimizes the role-specific behaviors (action generator, insight reflector, and playbook curator) under a fixed scaffold. Across agentic benchmarks, \acerl achieves higher held-out performance than baseline methods. The results suggest that agent training should optimize not only the actions in the current rollout, but also the model-written experience that future rollouts inherit.
[NLP-42] Saidzoda-Kazakh-EmbeddingGemma-300M: How Far One Epoch of Generated Questions Takes a Small Kazakh Embedder
【速读】： 该论文旨在解决小规模开源嵌入模型在哈萨克语信息检索任务中表现不足的问题，具体表现为原始EmbeddingGemma-300M模型在Kazakh Belebele检索任务上的召回率（recall@1）仅为60.8%。其解决方案的关键在于对EmbeddingGemma-300M进行全量微调（fine-tuning），利用包含166,263个由自研语言模型生成的哈萨克语问题-文档对（经双重模型验证）以及英语和俄语的回放数据，共208,511组样本进行单轮训练。这一方法显著提升了模型性能：在Kazakh Belebele测试集上，recall@1从60.8%提升至78.0%，recall@10从76.3%提升至91.4%。此外，该模型在MTEB基准上超越了161个已发布模型，仅略低于少数参数量达数十亿或为闭源的大模型，且在与多个哈萨克语专用嵌入模型的对比中表现领先或处于相近区间，证明了其在多语言环境下的泛化能力。
链接: https://arxiv.org/abs/2611.16399
作者: Tohir Saidzoda,Madina Abdusalomzoda,Abubakr Abdullozoda,Amin Jalilov,Faroiz Bobokhon,Aibek Shokanuly
机构: Saidzoda AI Lab(萨伊德佐达人工智能实验室); Saidzoda Engineering Company(萨伊德佐达工程公司)
类目: Computation and Language (cs.CL)
备注: 12 pages, 8 tables, 2 figures. Technical report. Model: this https URL
Abstract:EmbeddingGemma-300M is a small open embedder, and used as released it finds the right passage first for 60.8% of the Kazakh Belebele retrieval questions. We present Saidzoda-Kazakh-EmbeddingGemma-300M, the same model fully fine-tuned for one epoch on 208,511 pairs: 166,263 Kazakh questions that our own language model wrote about source passages from encyclopedias, Wikipedia, news and books, each checked by a second model, plus English and Russian replay. On Kazakh Belebele, recall@1 rises from 60.8% to 78.0% and recall@10 from 76.3% to 91.4%. Our model scores above 161 of the 194 models with a published MTEB score on the same test, among them OpenAI text-embedding-3-large (47.1%). The other 33 score higher: the ten best are large multi-billion-parameter or closed models, reaching up to 92.2%, and at least 9 are open models of up to 0.6B parameters, the smallest with 339M. We also scored 6 Kazakh-specific embedders shared on Hugging Face with the same code in one run: our model (77.3% there) is ahead of 4 of them, including the two E5-base models of similar size (72.6% and 69.0%), and behind kaz-embedder (86.0%, three towers with 1,695M parameters in total) and a BGE-M3 fine-tune (81.0%); at recall@10 its interval overlaps those of its nearest neighbours on both sides. Our four-language model, trained with a capped share of the same Kazakh pairs, is 2.8 points lower at recall@1, with overlapping recall@10 intervals. The Kazakh model also scores higher than the released model on the Uzbek and Kyrgyz versions of Belebele, but our Tajik model, trained without any Turkic pairs, gains as much there, so these gains are not specific to Kazakh training. Our Kazakh Wikipedia test questions come from the same generator family as the training questions, so we rest the main claims on Belebele.
[NLP-43] Beyond Compressing Everything: Contrastive Memory Learning for Selective Context Compression WSDM
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在处理长输入时，受限于内存预算下如何实现高效且任务相关的上下文压缩问题。随着输入长度增加，传统无差别的上下文压缩方法难以保留对特定查询至关重要的信息，尤其当任务相关证据可能分布在文档的局部或多个远距离片段中时，这一挑战尤为突出。为此，论文提出CoMem（Query-Guided Selective Context Compression），其核心创新在于设计了一种基于查询引导的、选择性压缩框架，通过联合建模文档与查询来学习条件化的记忆表示。关键解决方案包括：（1）提出对比压缩器学习（Contrastive Compressor Learning）机制，使压缩过程具备对查询敏感但又能保持上下文完整性的能力；（2）引入同文档多查询硬负样本挖掘策略，通过将同一文档与多个不同查询配对，强制压缩器根据各查询的具体证据需求进行选择性信息保留，从而注入强先验知识以捕捉查询依赖的证据分布；（3）为克服因果掩码带来的信息流动限制，引入记忆槽之间的双向注意力机制，增强压缩后记忆单元间的交互能力。实验结果表明，CoMem在多种压缩比率下均显著优于现有方法，在问答与RULER-NIAH基准上表现出一致的优越性能。
链接: https://arxiv.org/abs/2611.16384
作者: Sikui Zhang,Haowei Liu,Ziyun Gan,Chunfeng Yuan,Guangze Gao,Nian Liu,Boyu Cai,Bo Li,Xiaolei Lv,Jun Gao,Bing Li,Weiming Hu
机构: Institute of Automation, Chinese Academy of Sciences(中国科学院自动化研究所); Alibaba Group(阿里巴巴集团); Hello Group(Hello集团)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: WSDM camera ready
Abstract:As large language models (LLMs) advance toward increasingly long inputs, soft context compression offers a promising approach to efficient inference by converting long inputs into compact memory slots. Under limited memory budgets, effective context compression remains challenging because task-relevant evidence varies substantially across queries, ranging from localized evidence to multiple distant spans or information dispersed throughout the full document. Instead of compressing everything indiscriminately, we propose CoMem, a query-guided selective context compression framework that learns memory representations conditioned on both the document and the query. CoMem develops Contrastive Compressor Learning to achieve query-sensitive yet context-preserving memory representations. To inject a strong prior over query-dependent evidence distributions, we devise a same-context hard negative mining strategy by pairing a single document with multiple queries, requiring the compressor to selectively retain information according to each query’s distinct evidence needs. To address the restricted information flow imposed by causal masking, we further enable bidirectional attention among memory slots. Experiments demonstrate that CoMem consistently outperforms existing context compression methods across compression ratios on question answering and RULER-NIAH benchmarks. Our code is available at this https URL.
[NLP-44] Recursive State-Augmented Language Models: Persistent Latent-State Adaptation of Pretrained Language Models
【速读】： 该论文旨在解决语言模型（Language Models, LM）在处理序列任务时受限于有限上下文长度（finite token context）的问题，即无法有效访问早期交互信息。其核心挑战在于如何在不直接扩展输入上下文的情况下，让模型能够持续利用历史观测信息进行推理。为此，作者提出递归状态增强型语言模型（Recursive State-Augmented Language Models, Recursive SALM），其关键在于通过一个可持久化的潜在状态（persistent latent state）来编码并动态演化先前的观察结果，并以此调节预训练语言模型的输出行为。该框架采用模块化的“读取-转移-写入”（read-transition-write）架构，将观测编码、潜在状态动态更新与预测器自适应解耦。实验表明，在仅冻结GPT-2 small且不使用解码器上下文中的早期证据条件下，基于潜在状态门控的低秩适配（latent-gated LoRA）仍能达到89.1%的准确率，接近全上下文训练的LoRA基线（88.0%），证明了该机制能有效通过64维的紧凑状态实现对历史信息的隐式建模。此外，状态向量的零值化或随机打乱会显著降低性能，进一步验证了状态信息的关键作用。因此，该方案的核心创新在于：以轻量级、可递归演化的潜在状态为媒介，实现对长期依赖信息的高效捕捉与利用，从而支持在上下文受限场景下的连续推理能力。
链接: https://arxiv.org/abs/2611.16375
作者: Omer Haq
机构: 独立研究者(Independent Researcher)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (stat.ML)
备注: 13 pages, 2 figures. Preprint
Abstract:Language models rely on finite token contexts, limiting access to earlier interactions. We introduce Recursive State-Augmented Language Models (Recursive SALM), a framework in which a persistent latent state evolves with each observation and generates adaptations of a pretrained language model. A modular read-transition-write architecture separates observation encoding, latent dynamics, and predictor adaptation. We develop deterministic and stochastic formulations. The experiments evaluate only deterministic SALM with frozen GPT-2 small and exclude earlier evidence from its decoder context. On a four-turn hypothesis-inference task, latent-gated LoRA reaches 89.1% accuracy, comparable to 88.0% for a trained full-context LoRA baseline, with earlier evidence available solely through a 64-dimensional state. Zeroing or shuffling the state substantially reduces accuracy. In a two-turn task with codewords unseen in training, several interfaces improve candidate ranking over chance, but rarely generate the exact codeword under the evaluated training schedule. These experiments show that persistent latent-conditioned adaptation can support sequential inference when earlier evidence is absent from the decoder context. Comments: 13 pages, 2 figures. Preprint Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (stat.ML) Cite as: arXiv:2611.16375 [cs.LG] (or arXiv:2611.16375v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2611.16375 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[NLP-45] IsharaBART: An efficient BART-Based Framework for Low-Resource Bangla Sign Language Translation
【速读】： 该论文旨在解决孟加拉手语（Bangla Sign Language, BdSL）用户面临的自动翻译工具匮乏问题，尤其针对数据稀缺、单向模型架构以及计算资源消耗大导致难以在低资源环境下部署的挑战。其核心解决方案是提出IsharaBART，一种基于双向生成式框架的词形-句子翻译模型，采用基于BART的条件语言建模与去噪学习目标，实现双向（词形→句子 和 句子→词形）翻译。关键创新在于构建了一个包含1,000对专业标注的平行数据集，并通过规则化时态转换、掩码标记替换、反向翻译及一次提示（one-shot prompting）四种数据增强技术扩展至4,831对样本；同时，采用从零开始预训练的轻量化序列到序列模型，在六种孟加拉语语料上进行预训练后，再在标注与增强数据集上微调。实验表明，尽管参数量不足基准模型的10%，IsharaBART在词形到句子（G2S）任务中达到最高51.7的BLEU-4得分，在句子到词形（S2G）任务中达61.40，性能优于多语言模型；消融实验证明，预训练可使收敛速度提升62.5%–90.4%，且BLEU-4提高11.5%–118%；而使用低秩适应（LoRA）虽将可训练参数减少99.3%，但导致BLEU-4下降28.0%–52.2%。
链接: https://arxiv.org/abs/2611.16359
作者: Aonmoy Das,Ananna Dev Aishi,Md. Fazlul Kader
机构: University of Chittagong (吉大港大学); Department of Electrical and Electronic Engineering (电气与电子工程系)
类目: Computation and Language (cs.CL)
备注: 36 pages
Abstract:Millions of Bangla Sign Language (BdSL) users lack reliable automatic translation tools. Existing systems are limited by data scarcity, unidirectional design, and computationally expensive architectures that are not suitable for low-resource deployment. We introduce IsharaBART, a bidirectional generative framework for gloss-sentence translation in BdSL. IsharaBART uses Bidirectional and Auto-Regressive Transformers (BART)-based conditional language modeling with denoising learning objectives. We developed a parallel dataset of 1,000 professionally annotated gloss-sentence pairs, balanced across tenses and structures. We expanded the dataset to 4,831 pairs through four augmentation techniques: rule-based tense transformations, masked token substitutions, back-translation, and one-shot prompting. A compact denoising sequence-to-sequence model was pre-trained from scratch on monolingual Bangla text from six different corpora, then fine-tuned on the annotated and augmented datasets. We compare IsharaBART against four larger baseline models under identical conditions, where it achieves comparable or superior bidirectional performance. On the gloss-to-sentence (G2S) task, it attains Bilingual Evaluation Understudy (BLEU-4) score up to 51.7, and on the sentence-to-gloss (S2G) task, it achieves 61.40, outperforming multilingual models despite having fewer than 10% of the parameters. Ablation studies demonstrate that pre-training provides 62.5-90.4% faster convergence and 11.5-118% higher BLEU-4 than scratch. Low-rank adaptation (LoRA) reduces trainable parameters by 99.3%, but with 28.0-52.2% loss of BLEU-4 compared to full-parameter fine-tuning.
[NLP-46] Does Deep Knowledge Tracing Learn Semantic Relations Between Knowledge Components?
【速读】： 该论文旨在探究深度知识追踪（Deep Knowledge Tracing, DKT）模型所学习的知识成分（Knowledge Component, KC）间关系是否具有语义性。其核心问题是：DKT模型通过行为序列学习到的KC关联是否反映了真实语义上的联系？为解决这一问题，研究提出两个关键检验方法：首先，将DKT生成的KC关系图与基于文本衍生语义表示计算出的语义关系结构（通过余弦相似度构建）进行对比，发现两者在多个数据集、嵌入模型及构建管道下仅呈现微弱且不一致的对齐；其次，尝试将预训练的文本语义表示引入DKT模型中，以验证其是否能提升性能，结果表明这些语义初始化并未在所有条件下持续优于标准DKT基线，仅在部分低数据量设置下（如ASSISTments 2017）表现出轻微的AUC提升，且效果不稳定。研究结论指出，DKT所捕捉的KC关系可能更多由行为模式、时间结构或心理测量信号驱动，而非主要依赖于语义信息。因此，DKT学习到的关系并非主要具有语义性，现有文本衍生的语义结构无法有效还原其内在关系结构。
链接: https://arxiv.org/abs/2611.16323
作者: Kia Karbasi,Mahdi Saiedi,Kevin Hong,Gregory Pottie
机构: University of California, Los Angeles(加州大学洛杉矶分校); Sharif University of Technology(沙里夫理工大学)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注:
Abstract:Deep Knowledge Tracing (DKT) is a knowledge tracing (KT) model that updates a shared latent state from sequences of interactions, each pairing a knowledge component (KC) with a response. An interaction with one KC can therefore affect predictions for other KCs. We ask: Are the KC relations learned by DKT semantic? We construct text-derived semantic KC representations using nine pipelines to test whether the conclusions hold across alternative ways of constructing those representations. The first test compares DKT-derived KC relation graphs with semantic relation structures induced by pairwise cosine similarities among the KC representations. Across the evaluated datasets, embedding models, and pipelines, the two structures align only weakly and inconsistently. The second test introduces semantic information into DKT; the pretrained text-derived representations do not consistently outperform the standard DKT baseline. In the reduced-data experiments, semantic initialization yields a higher test area under the ROC curve (AUC) than the baseline at eight of nine nested sample sizes for ASSISTments 2017. KC-supervised fine-tuned BERT produces small positive AUC differences in some matched comparisons, but the study does not establish their stability across runs. These effects remain limited and inconsistent. The findings suggest that behavioral, temporal, structural, or psychometric signals may also shape DKT’s learned relations. DKT-derived relations are not primarily semantic: the tested text-derived semantic structures do not strongly recover the DKT-derived relation structure. Subjects: Machine Learning (cs.LG); Computation and Language (cs.CL) Cite as: arXiv:2611.16323 [cs.LG] (or arXiv:2611.16323v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2611.16323 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[NLP-47] From Large Language Models to Efficient System One Models via Adaptive Computation
【速读】： 该论文旨在解决将大语言模型（Large Language Models, LLMs）迁移至系统一决策模型（System One decision models）时存在的计算效率低下问题。现有方法通常保留完整的模型主干结构，仅替换语言建模头为决策头，导致每次决策均需遍历所有层，造成冗余计算。本文发现，决策信息主要集中于中间层：早期层负责构建基础特征，而最后六层可被移除而不影响性能；此外，独立训练的小型模型虽能近似全模型表现，但未能发展出专门化的决策能力。因此，解决方案的关键在于突破“仅通过减小模型规模提升效率”的局限，转而强调跨深度的共享计算。基于此，作者提出一种新方法 \method，通过自蒸馏机制联合训练一个完整模型与多个前缀子模型（prefix sub-models），这些子模型共享最终层和决策头。引入轻量级门控机制（gate），根据每一层的隐藏状态动态判断是否需要进一步计算，从而实现单模型在不同深度下的自适应推理。在七个决策基准测试中，该方法在各深度上均达到或优于独立训练模型的性能，同时实现了更优的精度-效率权衡。
链接: https://arxiv.org/abs/2611.16321
作者: Minh-Anh Nguyen
机构: Nanyang Technological University (南洋理工大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Transferring LLMs into System One decision models typically replaces the language modeling head with a decision head while retaining the full backbone, making every decision pay for all layers. Yet it remains unclear how much depth such decisions require. We find that decision information is concentrated in intermediate layers, with early layers building essential features and the final six layers removable without loss, while independently trained smaller models largely mimic the full model rather than developing specialized decision capabilities. These findings suggest that efficiency requires not simply smaller backbones, but shared computation across different depths. Building on this, we propose \method, which jointly trains a full model with prefix sub-models that share its final layers and decision head through self-distillation. A lightweight gate reads the hidden state at each exit to determine whether additional computation is needed, allowing one model to operate at multiple layers. Across seven decision benchmarks, our approach matches or outperforms independently trained models at every depth, while achieving better accuracy-efficiency trade-offs through adaptive computation.
[NLP-48] Eliciting and Mitigating Specification Satisficing in Coding Agents
【速读】： 该论文旨在解决编码代理（coding agent）在执行任务时出现的“规范妥协”（specification satisficing）问题，即代理虽曾考虑满足某项需求的实现方案，但因自身认为合理的理由最终选择了一个不完全符合需求的替代方案，从而导致任务失败。此类问题难以被现有自修正方法有效识别与缓解，因其本质上是代理基于内部逻辑接受非最优选择的结果，而传统自修正机制往往偏向于维持此类决策。本文提出SatisFix方法，其核心在于引导编码代理主动报告其决策选择及其背后的合理性依据，并据此逐项对照任务描述进行验证。该方法通过双阶段流程——先显式披露决策逻辑，再系统性校验每项选择是否满足规范要求——有效提升了对规范妥协的检测与纠正能力。实验在DeepSWE和SWE-bench Pro数据集上使用五种大语言模型（LLM）作为主干模型及两种不同代理框架进行评估，结果表明，SatisFix相较于基线模型平均提升pass@1指标5.09%（DeepSWE）和7.06%（SWE-bench Pro难子集），在DeepSWE上超越最佳基线3.98%；消融实验进一步证实两个步骤均对性能提升有显著贡献。
链接: https://arxiv.org/abs/2611.16289
作者: Ruixin Wang,Tianyi Zhang
机构: Purdue University (普渡大学)
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 12 pages, 2 figures, 7 tables
Abstract:Coding agents often fail to meet all requirements in the task description. In some of these failures, the agent considers an implementation choice that meets a requirement, but later decides to commit to another one that does not meet the requirement for a reason it finds acceptable. We call this specification satisficing. It is challenging to identify and mitigate specification satisficing because the agent accepts such choices for its own reasons, and existing self-correction methods are biased toward them. We propose SatisFix, which guides a coding agent to report these choices and their reasons, and then check each choice against the task description. We evaluate SatisFix on DeepSWE and SWE-bench Pro with five LLMs as backbones and two different agent frameworks, and compare it with three self-correction methods. SatisFix improves the pass@1 of the base agent by an average of 5.09% on DeepSWE and 7.06% on the hard subset of SWE-bench Pro. On DeepSWE, it outperforms the best baseline by 3.98%. Our ablation study shows that both steps of SatisFix contribute to the improvement.
[NLP-49] SMIC: Shared Multiscale Input-Reference Conditioning across Transformer Layers for Natural Language Inference
【速读】： 该论文旨在解决自然语言推理（Natural Language Inference, NLI）模型对局部词汇变化敏感的问题，即模型在面对细微语义扰动时表现不稳定。其核心解决方案是提出共享多尺度输入-参考条件化（Shared Multiscale Input-Reference Conditioning, SMIC），通过构建一个依赖于输入的统一多尺度参考表示，并在Transformer各层中复用该参考。每一层对共享参考执行独立的多尺度变换，并将生成的条件特征注入注意力更新前的计算过程，从而增强模型对输入扰动的鲁棒性。研究在基于令牌-上下文的Transformer（Token–Context Transformer, TCT）架构中实现SMIC，与解码器型模型及多层感知机（MLP）基线进行端到端对比，在ANLI全部三轮测试中均取得最高十次种子平均准确率。消融实验进一步验证了SMIC的有效性。此外，通过引入类似Llama风格的实值插件机制，并扩展删除/插入诊断方法至预训练的ModernBERT编码器和Llama解码器模型，结果表明TCT-SMIC在各类扰动协议下表现出更低的注意力重分配聚合值和更小的精度符号变化量。研究还建立了有限样本条件下子集质量重分配的上界，以平均杰恩森-香农散度（Jensen–Shannon divergence）衡量受限与重归一化注意力分布之间的差异，从理论上支持了SMIC带来的稳定性提升。
链接: https://arxiv.org/abs/2611.16270
作者: Xin Zhang,Victor S. Sheng
机构: Texas Tech University (德州理工大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Natural Language Inference models can be sensitive to localized lexical changes. We introduce Shared Multiscale Input-Reference Conditioning (SMIC), which constructs a single input-dependent multiscale reference and reuses it across Transformer layers. Each layer applies its own multiscale transformation to the shared reference and injects the resulting conditioning feature before the attention update. We instantiate SMIC in a token–context Transformer (TCT) and conduct from-scratch comparisons against decoder-based and Multi-Layer Perceptron (MLP) baselines across all three ANLI rounds. TCT-SMIC attains the highest observed ten-seed mean accuracy on all three ANLI rounds. The ablations further support the contribution of SMIC. We evaluate a real-valued Llama-style plug-in and extend deletion/insertion diagnostics to pretrained ModernBERT encoder and Llama decoder models with and without SMIC. Across the evaluated perturbation protocols, TCT-SMIC exhibits comparatively low pooled-attention redistribution and smaller signed accuracy changes. We further establish a finite-sample bound on average subset-mass redistribution in terms of the mean Jensen–Shannon divergence between the restricted and renormalized attention distributions.
[NLP-50] Label-Efficient Document-Grounded Factuality Detection via Cross-Paraphrastic Invariance Learning
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）生成与源文档无关的虚假陈述（factually unsupported claims）问题，进而实现高效且准确的文档锚定型事实性检测。现有基于LLM的评估方法通常依赖多次推理，计算开销高昂；而轻量级事实性检测器虽推理效率高，但普遍依赖大规模标注数据或合成数据集。本文提出一种两阶段孪生框架——交叉重述不变性学习（Cross-Paraphrastic Invariance Learning, CPIL），其核心创新在于仅利用少量标注数据即可构建关系监督信号，无需额外获取事实性标签。CPIL通过保留标签的重述视角生成同一文档-主张实例的正样本对，并利用同源但标签相反的样本构造负样本对；进一步引入分位数感知的半硬负样本挖掘机制，从正样本相似度分布中动态确定边界，以平衡信息量丰富的负样本与更广覆盖范围。在两阶段训练中，第一阶段基于任务感知的样本对学习编码器，第二阶段联合微调编码器与线性分类头完成二元事实性预测。推理阶段无需生成重述、构建样本对或进行孪生比较，仅需一次前向传播即可完成判断。在11个任务的LLM-AggreFact基准测试中，CPIL-2%和CPIL-3%分别仅使用2%-3%的标注数据即达到81.04和81.60的宏平均F1分数，显著优于所有对比的轻量级检测器及两种基础LLM参考模型。消融实验与敏感性分析验证了任务感知样本构造、动态负样本挖掘以及多枢轴重述策略的关键贡献。
链接: https://arxiv.org/abs/2611.16264
作者: Chao Chen,Shanshan Lin,Dongsheng Hong,Sibo Ju,Wenjie Yang,Sihong Xie,Xiangwen Liao
机构: Harbin Institute of Technology (Shenzhen); Fuzhou University; Hong Kong University of Science and Technology (Guangzhou)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Accepted to IEEE TASLP 2026
Abstract:Large language models (LLMs) can generate fluent claims that are unsupported by a given source document, motivating accurate and efficient document-grounded factuality detection. Existing LLM-based evaluators often require repeated inference and incur high computational cost, whereas compact factuality detectors are efficient at inference time but commonly rely on large labeled or synthetically generated corpora. We propose CPIL (Cross-Paraphrastic Invariance Learning), a two-stage Siamese framework that converts a small labeled set into relational supervision without acquiring additional factuality labels. CPIL constructs positive pairs from distinct label-preserving paraphrastic views of the same document-claim instance and negative pairs from opposite-label examples associated with the same source document. It further introduces quantile-aware semi-hard negative mining, which derives a boundary from the positive-pair similarity distribution and balances informative negatives with broader coverage. In the two-stage training procedure, Stage 1 learns an encoder from these task-aware pairs, and Stage 2 jointly fine-tunes the encoder and a linear classification head for binary factuality prediction. At inference time, CPIL requires no paraphrase generation, pair construction, or Siamese comparison and uses only one encoder-classifier forward pass. On the 11-task LLM-AggreFact benchmark, CPIL-2% and CPIL-3% achieve macro-F1 scores of 81.04 and 81.60 while using only 2-3% of the FactCG-released labeled pool. Both settings outperform all compared compact detectors and two vanilla LLM references. Ablations and sensitivity analyses support the contributions of task-aware pair construction, dynamic negative mining, and multi-pivot paraphrasing.
[NLP-51] Low Language Transfer Speech-to-Text
【速读】： 该论文旨在解决多语言语音识别模型在低资源语言（Low-Resource Language, LRL）上的迁移能力问题，即一个在多种语言上预训练的语音识别模型，在仅经过少量目标语言数据微调后，能否有效利用已有知识实现对未充分训练语言的识别。其核心挑战在于如何在极小规模标注数据（每语言最多1,000段音频，总计82小时）条件下，实现对低资源语言的有效识别，并量化模型所依赖的知识来源是语言本身还是训练语料。解决方案的关键在于：通过轻量级参数适配（如低秩适配器，Low-Rank Adapters）对模型进行高效微调，可实现接近全参数微调的效果，同时显著降低对原有多语言知识的损耗；且模型性能提升主要源于跨语言知识迁移，而非特定训练语料的积累。 实验表明，即使不共享任何训练数据，模型在FLEURS测试集上仍能保留79%的性能增益，说明迁移机制依赖于语言层面的通用表征，而非具体语料。此外，尽管模型在新语言上表现迅速提升（最佳模型将字符错误率（CER）降至9.5），但其对已知语言（如英语、西班牙语、法语、德语）的性能平均下降5.7点，揭示了“语言”而非“语料”是知识迁移的核心载体。最后，数据量远超训练轮次的影响被验证——同等片段数量下，十倍数据带来的性能提升达9.7个点，凸显数据质量与数量的重要性。
链接: https://arxiv.org/abs/2611.16260
作者: Husein Zolkepli
机构: 未知
类目: Computation and Language (cs.CL)
备注:
Abstract:A multilingual speech recogniser has heard much speech in its own languages. We ask how much of that it brings to a language it was never trained on, when it is fine-tuned on the little that language has. We score 28 open multilingual models untouched on Low-Language-TTS, 50 low-resource languages in 11 writing systems, and fine-tune 24 of them on at most 1,000 clips a language, 82 hours in all, with a dev set choosing every epoch. Fine-tuning lowers every model’s macro character error rate (CER), by 16.6 to 53.1 points, to 9.5 at best, and a word language model takes three CTC models 1.2 to 1.4 points lower; WER is still 34.4 at best. On the 12 languages that no model lists, every model starts at 46 or worse and the best ends at 11.7. The same architectures from random weights stay 60 points or more behind, so the gain is transfer, and an encoder left untouched already gives a Whisper or Qwen decoder most of what it needs. It is the language, not the corpus: on FLEURS recordings that ours do not share, the median model keeps 79% of its gain. It costs the languages the models knew, 5.7 points of CER on English, Spanish, French and German in the median model. Low-rank adapters on 4% of the weights come within 1.1 points of tuning everything, and on 2 of 3 models forget less. Seq2seq and language-model decoders are done within three to five epochs, while a new CTC head is still improving at twenty or thirty. Data beats epochs: 25 clips a language take Qwen3-ASR 0.6B from 60.9 to 39.3, and at equal clips presented, ten times the data is worth 9.7 points. A model that first learned forty of the languages learns the other ten no faster. On a public Malay benchmark, untouched Whisper large-v3 scores 8.9 when it detects the language and 14.9 when told the clip is Malay, because it then translates the English that speakers mix in; trained on the Scicom telephony dataset it scores 6.0.
[NLP-52] Indic-Task Gains and Regressions after Reinforcement Learning NEURIPS2026
【速读】： 该论文旨在解决大模型在多语言场景下，尤其是在印度语系（Indic）语言中，因强化学习（Reinforcement Learning, RL）训练导致的“语言漂移”问题，即模型在优化过程中过度偏向英语输出，从而损害对目标语言（如印地语、泰米尔语等）任务的性能。其解决方案的关键在于引入多维度奖励机制，包括语言奖励（language reward）与重复性奖励（repetition reward），以引导模型在保持英语任务表现的同时，增强对印度语系语言的生成能力与脚本合规性。尽管第二轮强化学习训练在部分指标上实现了跨语言任务性能的提升，并维持了数学推理（MATH）的表现，但结果表明语言奖励本身不足以确保正确脚本的使用，且训练策略的多重变化使各奖励项的独立影响难以分离。因此，研究强调：即使整体基准得分改善，仍需对篇章问答（passage answering）和脚本合规性（script compliance）进行直接评估，以确保模型在真实多语言应用中的可靠性。
链接: https://arxiv.org/abs/2611.16228
作者: Sutanshu Saksena Raj,Daksh Maan,Avinash Benki,Parijat Chatterjee,Pasunuti Prasanjith,Purusharth Malik
机构: Gnani.AI; Google(谷歌)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Accepted at the GlobalSouthAI Workshop, NeurIPS 2026
Abstract:We run reinforcement learning (RL) with GRPO across four environments, starting from a supervised fine-tuned 30B-A3B mixture-of-experts model covering English and eight Indic languages. We compare the model before and after RL across 34 benchmarks in one evaluation campaign. MATH improves by 36.07 points, and most Indic benchmarks also improve. Yet XQuAD-Indic F1 falls 19.32 points as English answers rise from 8.7% to 91.7%; IndicQA falls 8.64, while English XQuAD gains 11.71 F1. A second RL run from the same supervised initialization adds an Indic environment together with language and repetition rewards. Compared with the released model, it improves both passage-answering and instruction-following benchmarks, along with preserving MATH performance. But it loses ground on reasoning and coding benchmarks. Several training choices changed together, so we cannot isolate the effect of any one of the reward terms. Script compliance remains unresolved. Adding a language reward did not ensure use of the requested script. Passage answering and script compliance therefore need direct evaluation, even when other benchmark scores improve.
[NLP-53] We Are What We Remember: Mental-State-Driven Memory for Long-Term Character Dialogue
【速读】： 该论文旨在解决当前对话记忆系统在处理人类对话共享历史时存在的核心缺陷：现有主流系统依赖外部事实库中的被动语义相似性匹配，难以捕捉具有关系关键性的交互经历，导致对用户潜在需求的预判和情境化理解能力不足。其解决方案的关键在于提出一种基于心理状态驱动的记忆框架MindMem，该框架受认知心理学中的自我记忆系统（Self-Memory System, SMS）启发，将共享经验异步整合为角色、用户及其关系三者的源关联心理状态（mental states），并引入专用的心理状态推理器（Mind）在每轮对话中动态上下文化这些持久状态，以预测潜在用户需求、生成主动意图，并恢复表面匹配所忽略的情景化情节证据。为评估系统在长期互动中的历史敏感适应能力，研究构建了MemReal-Bench——一个基于真实用户长达一个月的二元与群体对话场景的纵向基准测试集，通过结合直接问答（DQA）与情境化记忆实现（SMR）双重任务，检验智能体是否能从“知晓过去事实”跃迁至“自然运用共享历史、世界规则与关系轨迹”。实验结果表明，MindMem在SMR和DQA任务上分别达到59.03%和82.58%，较最强基线分别提升24.19和9.19个百分点，且在SMR上超越理想记忆注入的基准表现，验证了其在深层关系感知与情境化推理方面的有效性。
链接: https://arxiv.org/abs/2611.16223
作者: Zhicheng Jia,Shuchen Wu,Zhuoshi Tong,Haisen Liao,Sicheng Zhou
机构: University of Chinese Academy of Sciences(中国科学院大学); Nanyang Technological University(南洋理工大学); Livo Team, Kuaikan World(快看世界Livo团队); China Agricultural University(中国农业大学)
类目: Computation and Language (cs.CL)
备注: Zhicheng Jia, Shuchen Wu, and Zhuoshi Tong contributed equally to this work
Abstract:Shared history in human dialogue is not merely a repository of facts, but the foundation of evolving relationships and mutual understanding. However, mainstream conversational memory systems rely on passive semantic similarity over external fact stores, frequently retrieving topically similar records while missing relationally pivotal experiences. Inspired by cognitive psychology’s Self-Memory System (SMS), we propose MindMem, a mental-state-driven memory framework that asynchronously consolidates shared experiences into source-linked mental states of the character, the user, and their relationship. At each turn, a specialized mental-state reasoner (Mind) contextualizes these persistent states to anticipate latent user needs, form proactive intentions, and recover supporting episodic evidence that surface-level matching overlooks. To evaluate history-sensitive adaptation beyond factual recall, we introduce MemReal-Bench, a longitudinal benchmark grounded in month-long real-user interactions across dyadic and group scenes. By pairing Direct Question Answering (DQA) with Situated Memory Realization (SMR), it tests whether agents bridge the gap between knowing past facts and naturally using shared history, world rules, and relational trajectories. On MemReal-Bench, MindMem reaches 59.03% SMR and 82.58% DQA, exceeding the strongest memory baseline by 24.19 and 9.19 percentage points, respectively, and outperforming oracle target-memory injection on SMR. Code and dataset are available at this https URL.
[NLP-54] Data-Efficient LLM Safety Alignment Guided by Interpretable Features
【速读】： 该论文旨在解决大语言模型在安全对齐（safety alignment）过程中依赖大规模微调数据集所导致的高成本问题，以及由此引发的“过度拒绝”（over-refusal）现象。其核心解决方案是提出一种名为“少即是安全”（Less is Safer, LIS）的紧凑型高效对齐数据集，关键在于通过可解释特征引导的选择与对抗性合成相结合的方法构建数据。具体而言，利用稀疏自编码器（Sparse Autoencoder, SAE）激活值高效筛选覆盖多样化且可解释安全特征的提示（prompt），再通过多智能体对抗合成框架生成虽无害但能诱发不必要的拒绝行为的样本。LIS仅包含2.4K样本，却在四个安全性和五个过度拒绝基准测试中优于规模更大的数据集（如TRIDENT-Core、TRIDENT-Edge和WildJailbreak）。消融实验表明，仅使用安全数据或仅使用过度拒绝数据均无法达到最优效果，而本文提出的端到端管道结合了SAE提取的安全特征与对抗生成策略，在性能上显著超越基于LLM人工标注特征进行过滤的方案。
链接: https://arxiv.org/abs/2611.16183
作者: Hongyu Li,Zhan Yu,Zhe Zhao,Kimmo Yan,Xiong Kuang,Renfen Hu
机构: Beijing Normal University (北京师范大学); Tencent Beijing (腾讯北京)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 36 pages, 16 figures
Abstract:Existing large language models typically rely on massive fine-tuning datasets for safety alignment, which results in higher costs for safety alignment and often leads to the problem of over-refusal. We propose Less is Safer (LIS), a compact yet highly effective dataset for safety alignment, constructed through interpretable feature-guided selection and adversarial synthesis. Specifically, we leverage Sparse Autoencoder (SAE) activations to efficiently select prompts that cover diverse and interpretable safety features. Then, we employ a multi-agent adversarial synthesis framework to generate harmless prompts that nevertheless elicit unnecessary refusals. LIS contains only 2.4K samples, yet outperforms much larger datasets, e.g., TRIDENT-Core (26K), TRIDENT-Edge (19K) and WildJailbreak (262K), in overall performance on four safety and five over-refusal benchmarks. Through ablation experiments, we demonstrate that neither safety data nor over-refusal data alone can achieve optimal results. We also used safety features from SAE, annotated by an LLM, for data filtering, but the results were inferior to those obtained using the pipeline proposed in this paper.
[NLP-55] JudgeMix: Judge-Guided Mixed-Rollout Data Synthesis for LLM Agents
【速读】： 该论文旨在解决大语言模型（LLM）代理在监督微调（Supervised Fine-Tuning, SFT）过程中存在的训练-推理分布不匹配问题。现有方法通常依赖专家轨迹进行微调，但其在引入专家辅助时缺乏对目标模型自身行为的感知：要么采用随机干预策略破坏生成轨迹的连贯性，要么施加均匀且过度的密集监督，从而覆盖了模型已正确的行为。为克服上述局限，本文提出JudgeMix——一种基于评判器（judge）引导的数据合成框架，其核心在于通过逐步语义评估动态决定是否引入专家指导，仅在目标模型面临困难时触发专家干预，从而在保持高轨迹成功率的同时，使训练状态分布与目标模型的本征行为分布保持一致。此外，该框架利用评判器评估结果对不同步骤的专家监督信号进行重加权，有效保留模型原有的正确行为，仅修正错误路径。在AppWorld和AutomationBench两个代理工具使用基准上，JudgeMix展现出显著优于现有方法的性能，最高实现10个pass@1指标的提升。
链接: https://arxiv.org/abs/2611.16174
作者: Cheng-Yu Hsieh,Ting-Yao Hu,Sanjoy Chowdhury,Oncel Tuzel,Raviteja Vemulapalli
机构: Apple(苹果)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Supervised fine-tuning (SFT) on expert trajectories is a standard paradigm for training LLM agents. However, it suffers from a well-known training-inference distribution mismatch, calling for target-aware data synthesis strategies that ground the training data in the target model’s native state distribution. To bridge this gap, emerging approaches incorporate the expert either to guide mixed rollouts alongside the target model, or to provide dense step-wise supervision. Nonetheless, existing methods apply expert assistance blind to the target model’s inherent behavior: they either disrupt the rollout using sub-optimal random intervention heuristics, or they enforce indiscriminate uniform supervision that unnecessarily overwrites the model’s already-correct actions. We introduce JudgeMix, a judge-guided data synthesis framework that dynamically interleaves the target model and expert based on step-wise semantic evaluations. JudgeMix queries expert guidance precisely when the target model struggles, ensuring high trajectory success rates while keeping the state distribution aligned with the target model. It additionally uses the judge evaluation to reweight expert supervision on different steps, preserving model’s inherent valid behavior while correcting its errors. On two agentic tool-use benchmarks, AppWorld and AutomationBench, JudgeMix demonstrates consistent superior performance to existing methods, with improvements up to 10 pass@1 points.
[NLP-56] mporal Context-Aware Spam Detection Using Transformer-Based NLP Models
【速读】： 该论文旨在解决传统垃圾短信（spam）检测中忽视发件人行为序列模式的问题，尤其是在运营商层面进行出站短信垃圾信息预测时，仅依赖单条消息的静态内容特征可能导致漏判或误判。其核心挑战在于如何有效建模同一发送者在时间维度上连续发送消息的行为模式，以提升检测的准确性与召回率。解决方案的关键在于引入时间上下文特征（inter-message temporal features），通过融合基于因果Transformer架构的时序建模方法——TemporalBertSpam，对每个发送者的完整消息历史进行动态编码，从而捕捉其潜在的垃圾行为模式。该方法不仅利用文本内容（如DeBERTa-v3-base），还结合了消息间的时间间隔、发送频率等时序特征，在生产环境评估中实现了较高的召回率（0.9128）和垃圾类F1值（0.8382），显著优于仅依赖文本的基线模型。
链接: https://arxiv.org/abs/2611.16139
作者: Fariz Zeynalov,Razim Mammadli
机构: University of Chicago(芝加哥大学); Duke University(杜克大学)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 7 pages, 2 figures, 5 tables
Abstract:Spam detection is usually treated as single-message classification, ignoring behavioral patterns that emerge across sequences of messages from the same sender. We study outbound SMS spam detection at the carrier level, where the goal is to predict at send time whether a carrier will flag a message as spam. Using a proprietary production database of 43,741 outbound SMS messages with UTC transmission timestamps and carrier delivery statuses, we report three sets of results. On an 8,749-message held-out test set, a TF-IDF + logistic regression baseline reaches AUC-ROC 0.8721, and a text-only DeBERTa-v3-base classifier reaches AUC-ROC 0.9011 and F1 0.8412. A deployed DistilBERT model augmented with inter-message temporal features reaches accuracy 0.8238, spam-class F1 0.8382, and recall 0.9128 at threshold 0.5 on a separate, class-balanced 20,000-message production evaluation set; these results are not directly comparable with the held-out test results. We also describe TemporalBertSpam, an architecture that encodes each sender’s message history with a causal Transformer, and outline the same-split evaluation needed to measure its contribution.
[NLP-57] Show Your Work in Swahili: Reinforcement Learning for Native-Language Mathematical Reasoning
【速读】： 该论文旨在解决斯瓦希里语数学推理任务中模型输出的表面一致性（surface fidelity）与最终答案准确性之间的不匹配问题，即现有方法往往只关注最终答案的正确性，而忽视了推理过程在语言形式和逻辑连贯性上的保真度。其解决方案的关键在于提出一种两阶段的监督微调（SFT）与基于人类反馈的强化学习（GRPO）相结合的训练范式——Fikiria，通过设计一个复合奖励函数，将格式门控（format gate）与答案准确率及斯瓦希里语推理链与参考链之间基于ChrF++的相似度加权求和，从而在训练过程中同时优化推理过程的语言形式一致性与逻辑合理性。该方法在自建的2.3k问题斯瓦希里语数据集上进行训练，该数据集由GPT-4.1翻译GSM8K并经SSA-COMET过滤生成；实验结果显示，Fikiria-AfriqueQwen-14B在AfriMGSM斯瓦希里语测试集上达到79.2%的准确率，显著优于在相同提示和严格格式评分下仅达68.0%的GPT-4o，尽管后者在人工评估中表现更受偏好，这表明当前指标仍难以全面衡量推理质量，且ChrF++仅衡量表面形式相似性，而非单步推理的有效性。
链接: https://arxiv.org/abs/2611.16104
作者: Jessica Lundin,Katherine Rosenfeld,David Adelani,Geoffrey Mosoti Nyakiongora,Halfan Ngowo,Sloan Coffin,Ralph Abboud,Guillaume Chabot-Couture
机构: Gates Foundation(比尔及梅琳达·盖茨基金会); McGill University (麦吉尔大学); RenPhil; Gates Foundation(比尔及梅琳达·盖茨基金会)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:We present Fikiria, a two-stage SFT+GRPO recipe for Swahili mathematical reasoning. Its reward multiplies a format gate by a weighted sum of answer accuracy and the ChrF\texttt++ similarity of the Swahili reasoning to an in-language reference chain of thought, so that training targets the surface fidelity of the reasoning trace and not only the final answer. We train on a 2.3k-problem Swahili dataset created by translating GSM8K with GPT-4.1 and filtering with SSA-COMET. On AfriMGSM Swahili, Fikiria-AfriqueQwen-14B reaches 79.2%. Under the same fixed prompt and strict answer-format scoring, GPT-4o reaches 68.0%, partly because it often fails the required answer format; IrokoBench reports 78.8% for GPT-4o using a best-per-model prompt, a figure numerically close to ours but not directly comparable. In a small human evaluation with co-author annotators, GPT-4o’s outputs were preferred overall. ChrF\texttt++ measures surface fidelity, not the validity of individual reasoning steps.
[NLP-58] Privileged Information Does Not Guarantee Privileged Guidance: Auditing On-Policy Self-Distillation for Reasoning Models
【速读】： 该论文旨在解决生成式 AI 在数学推理任务中，通过有监督自蒸馏（On-policy self-distillation, OPSD）方法利用教师模型所持有的特权信息（如完整解题过程）是否真正实现了推理能力的有效迁移这一关键问题。其核心挑战在于：尽管教师模型可访问解题过程并据此生成更优的下一轮预测目标，但这些目标是否能被学生有效学习、转化为实际推理能力，以及在无外部提示条件下能否可靠地表达出所学知识，仍不明确。研究的关键在于系统性地审计了四个阶段——源（Source）、编码（Encoding）、传输（Transmission）与表达（Expression），揭示出即使教师拥有解题方案，其生成的目标序列也并不稳定指向正确路径，且学生对错误的修正行为难以在自由生成中可靠复现。进一步分析发现，最大性能提升并非来自特权解题信息本身，而是源于教师处于“思考模式”（thinking mode）时所提供的推理过程指导，即“推理套利”（reasoning arbitrage）现象：当教师以思考模式进行监督，而学生既不显式记录思考过程也不依赖解题方案时，性能反而最优。因此，论文指出当前仅凭准确率提升无法证明推理能力从特权解题信息中成功转移，真正的瓶颈存在于编码、传输和表达阶段。未来评估应严格控制变量，独立衡量解题信息对训练的额外贡献。
链接: https://arxiv.org/abs/2611.16064
作者: Haoyi Qiu,Akshat Agarwal
机构: Meta Superintelligence Labs; University of California, Los Angeles (加利福尼亚大学洛杉矶分校)
类目: Computation and Language (cs.CL)
备注: Work in progress
Abstract:On-policy self-distillation (OPSD) trains a student on next-token targets from a teacher, a frozen copy of the student that additionally receives privileged information, such as a worked solution, hidden from the student. Whether student gains reflect that information remains unclear. Taking the worked solution as privileged information, we audit four stages across three Qwen3 scales on mathematics: Source, whether the teacher can use the solution; Encoding, whether the resulting token targets are useful; Transmission, whether the student learns from the targets; and Expression, whether the student applies the learned behavior without the solution. With the solution, the teacher solves hard problems and recovers from student errors, yet the teacher does not reliably recognize errors, and its token targets do not favor correct trajectories. Repeated training makes successes within the student’s reach more frequent rather than unlocking new ones, and corrections the student learns to predict are not reliably expressed in free generation. To isolate the solution’s contribution, we vary its availability separately from teacher and student thinking modes. The largest gains occur when a teacher in thinking mode supervises a student trained and evaluated without an explicit thinking trace, and problem-only teachers often match or exceed privileged-solution teachers: the gain tracks the teacher’s thinking mode rather than the solution, a pattern we call reasoning arbitrage. Accuracy gains alone therefore do not establish transfer of reasoning from privileged solutions: bottlenecks lie after Source, across Encoding, Transmission, and Expression. Future evaluations must measure what the solution adds to otherwise identical training.
[NLP-59] Scaling public input on AI policy with pluralistic agents
【速读】： 该论文旨在解决在人工智能（AI）系统开发过程中，如何有效整合多元人类观点以实现安全、负责任的AI治理问题。随着AI应用范围不断扩展，其决策影响涉及不同文化背景与利益相关方，亟需一种机制来持续纳入多样化的价值判断与偏好。为此，论文提出“多元代理”（pluralistic agents）这一解决方案——即代表不同个体立场的AI代理，通过模拟人类角色进行协商与讨论，从而将分散、难以直接参与的公众意见系统性地融入AI政策制定过程。其核心创新在于：利用代理作为人类声音的延伸，在技术复杂或信息敏感的议题中放大个体参与度，尤其适用于那些需要专业知识或涉及私密数据而无法广泛公开讨论的情境。实验结果表明，多元代理能够可靠地推动政策向其所代表角色的偏好方向演变，且其影响力受代理人指令策略（如表达冗余度、合作程度）显著调节，提示了提示工程与系统设计对公平性的关键影响。在一项预注册研究中，746名参与者评估该方法，普遍认为多元代理具有实用性与公平性，虽仍偏好亲自参与，但其在满意度和公平感方面的表现分别达到基线水平的68%与82%，显著缩小了直接参与缺失带来的差距。因此，该方法的关键优势在于作为人类输入的“放大器”，而非替代品，为规模化、包容性地推进AI安全治理提供了可行路径，值得谨慎但积极地探索。
链接: https://arxiv.org/abs/2611.16028
作者: Jacy Reese Anthis,Ken Burke,David Madras,Crystal Qian,Meredith Ringel Morris,Logan Cross,Caroline Pantofaru,Utsav Prabhu
机构: University of Chicago; Google DeepMind(谷歌深度思维)
类目: Computers and Society (cs.CY); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:AI is being deployed across a vast range of application domains, cultural environments, and human stakeholders. Safe and responsible AI development requires continuously accounting for diverse perspectives. We develop and test pluralistic agents, AI agent representatives that deliberate on behalf of individuals to align AI systems with diverse human perspectives. Pluralistic agents could help people contribute to many more discussions than they could directly participate in, on technical topics that would require training, and on policy choices that rely on private information that corporations and governments cannot readily share. Our experiments identify opportunities, risks, and perceptions of pluralistic agents. Agents were able to reliably shift policies in the direction of their represented persona’s preferences, but the extent of each agent’s influence was sensitive to the instructions from its persona, such as having more influence when instructed to be more verbose or less cooperative. This suggests that influence in AI-assisted deliberation could vary based on how participants prompt their agents and how the system is designed. In a preregistered study, 746 humans participated in the pluralistic agents process. Participants viewed pluralistic agents as useful and fair, preferring agents over the baseline of providing input to a single AI reasoning model. Most participants preferred participating themselves over either pluralistic agents or the baseline, but pluralistic agents recovered 68% of the baseline-direct gap in reported satisfaction and 82% in fairness. Because pluralistic agents can be used in many real-world contexts where direct discussion is not, they could be a promising tool to scale AI safety policy development. We encourage cautious exploration of this method, playing to its strengths as an amplifier of human input rather than a replacement.
[NLP-60] What Roles Do System Prompts Really Assign? Validating and Steering Occupational Role Vectors in LLM s
【速读】： 该论文旨在解决大语言模型在应用中通过系统提示（system prompt）赋予角色时，角色表现不稳定、可控制性差的问题。其核心挑战在于如何实现对模型响应行为的精确且持续的角色控制，尤其是在复杂对话场景下角色特征易消退。解决方案的关键是提出一种验证框架，用于提取并注入职业角色向量（occupational role vectors），这些向量为激活层方向（activation-level directions），能够捕捉特定职业角色在任务或问题回答中的独特表达模式。研究通过双重验证视角——可解释性（interpretability）与控制力（control）——评估该方法：前者验证角色向量在不同运行条件和提问域下的稳定性、可分离性与领域一致性；后者检验注入向量是否能有效引导模型在单轮及多轮对话中稳定表达目标角色。结果表明，尽管角色向量具有高度可重复性和领域内一致性，但单独注入仍难以完全替代传统角色提示，且无法显著延缓角色在长对话中的衰减。这揭示了仅依赖向量表示不足以实现强可控性，必须结合提示工程与激活调控，二者互补方能实现角色的精准与持久表达。
链接: https://arxiv.org/abs/2611.15989
作者: Jose L. Sampedro Mazon,Ilmi Yoon,Nadim Saad
机构: Khoury College of Computer Sciences, Northeastern University (东北大学计算机科学学院)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 11 pages main text, 35 pages total including references and appendices. Code and data: this https URL
Abstract:Applications built on large language models assign the model a role through its system prompt, intending to shape how it responds for their use case. We introduce a validation framework for extracting and injecting occupational role vectors, activation-level directions that capture how someone in a given occupational role would distinctively respond to a task or question. We validate these vectors through two complementary lenses, on Qwen2.5-7B-Instruct and Llama-3.1-8B-Instruct. The interpretability lens asks whether each vector is a stable, separable, domain-consistent representation of its role: extracted vectors are highly reproducible across repeated runs and sampling temperatures, remain separable even for very similar roles, and agree more with themselves across in-domain and out-of-domain questions than with other roles’ signal. The control lens asks whether injecting a role vector during generation makes the model express the role, and sustain it over a conversation, as well as a role prompt does. On Qwen, injecting the vector alone makes the model express the intended role substantially more often than unsteered, but at best matches a role prompt; on both models, adding the vector to the prompt strengthens the role in single answers but did not slow the role’s decline over 30 turns. Together, the two lenses show why both are needed: out of domain the role is expressed only subtly, yet the activations still encode it; and a vector that is stable and separable can still barely steer the model’s behavior.
[NLP-61] Better Boundary Prediction Does Not Guarantee Better Structural Representation NEURIPS2026
【速读】： 该论文旨在解决如何通过引入标点符号恢复（Punctuation Restoration, PR）任务来优化语言模型（Language Models, LMs）对句法、语义及论元结构等语言结构的表征能力。研究假设认为，由于标点符号常标记短语边界并携带丰富的语言结构信息，因此在训练中引入标点恢复作为辅助目标，可能有助于提升模型对语言层次结构的理解。其解决方案的关键在于：在传统基于词元预测的任务之外，增加一个以恢复文本中标点符号为目标的辅助任务，从而引导模型学习更精细的句法边界信息。然而，实验结果表明，在10亿参数规模下，针对三种主流架构（仅解码器型、仅编码器型、编码器-解码器型）的语言模型中，单纯优化边界预测并未显著提升模型在依赖结构理解的下游任务中的表现，说明标点恢复作为结构优化手段的有效性存在局限。
链接: https://arxiv.org/abs/2611.15971
作者: Michael Zhou,Markus Frohmann,Junghyun Min,En-Shiun Annie Lee
机构: Carnegie Mellon University (卡内基梅隆大学); University of Toronto (多伦多大学); Georgetown University (乔治城大学); Ontario Tech University (安大略理工大学)
类目: Computation and Language (cs.CL)
备注: Presented at LP4FM @ NeurIPS 2026
Abstract:Punctuation marks often mark phrase boundaries, carrying linguistic information. Therefore, punctuation marks’ abundance and feasibility as proxies of linguistic structure has led to suggestions that using punctuation marks to optimise on phrasal structure may help language models (LMs) improve their representations of argument, semantic, and syntactic structures. In this paper, we test the above claim in English across three primary LM architectures: decoder-only, encoder-only, and encoder-decoder LMs, by complementing traditional token prediction with punctuation restoration (PR). Our evidence, collected at the 1B parameter scale, suggests that optimising for boundary prediction does not necessarily guarantee higher-quality structural representation in decoder-only, encoder-decoder, and encoder-only LMs, failing to improve performance on downstream tasks that require understanding structure in linguistic input.
[NLP-62] Is Spoken Hungarian Low-resource?: A Quantitative Survey of Hungarian Speech Data Sets LREC2024
【速读】： 该论文旨在解决匈牙利语语音数据集类型与规模缺乏系统性综述的问题。尽管已有多种匈牙利语语音数据集，但其分类、规模及实际应用适配性尚未得到全面梳理。为此，研究者对现有匈牙利语语音数据集进行了分类调研，涵盖五类：单语、多语中匈牙利语部分、病理语音、儿童相关以及方言数据集。总体估算数据显示，当前可用数据总量约为2800小时（覆盖约7500名说话人），展现出丰富的口语语言多样性。然而，数据分布与真实应用场景（如语音识别）之间仍存在显著不匹配，表明亟需构建更大规模、更贴近自然语言使用的语音数据集。该研究的关键解决方案在于系统性地归纳现有数据集的优缺点，为跨学科的匈牙利语研究提供基础参考，并为建立面向匈牙利语的统一基础语音模型（Unified Foundational Speech Model for Hungarian）奠定数据基础。
链接: https://arxiv.org/abs/2611.15961
作者: Péter Mihajlik,Katalin Mády,Anna Kohári,Fruzsina Sára Vargha,Gábor Kiss,Tekla Etelka Gráczi,A. Seza Doğruöz
机构: 未知
类目: Computation and Language (cs.CL)
备注: LREC 2024
Abstract:Even though various speech data sets are available in Hungarian, there is a lack of a general overview about their types and sizes. To fill in this gap, we provide a survey of available data sets in spoken Hungarian in five categories (e.g., monolingual, Hungarian part of multilingual, pathological, child-related and dialectal collections). In total, the estimated size of available data is about 2800 hours (across 7500 speakers) and it represents a rich spoken language diversity. However, the distribution of the data and its alignment to real-life (e.g. speech recognition) tasks is far from optimal indicating the need for additional larger-scale natural language speech data sets. Our survey presents an overview of available data sets for Hungarian explaining their strengths and weaknesses which is useful for researchers working on Hungarian across disciplines. In addition, our survey serves as a starting point towards a unified foundational speech model specific to Hungarian.
[NLP-63] ShadowBench: Exposing Lexical Anchoring and the Illusion of Forgetting in Large Language Models NEURIPS2026
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在事实检索中存在严重依赖显式命名实体（lexical anchors）的问题，即尽管模型内部存储了丰富的知识，但其知识提取能力高度受限于是否出现明确的实体名称。这一现象导致现有评估范式无法真实反映模型对隐含实体关联的掌握能力。解决方案的关键在于提出一种新型、抗捷径攻击的基准测试框架——ShadowBench，通过创新的双重属性关联（Dual-Trait Association, DTA）任务，系统量化模型在移除显式命名锚点后仍保持的知识能力。实验结果揭示了普遍存在的“影子差距”（Shadow Gap），即所有规模模型（包括前沿模型如GPT-5.4和Claude-Sonnet-4.6）在去词汇化条件下性能下降超过20%。进一步研究表明，当前主流机器遗忘（Machine Unlearning）方法仅实现表面化的词汇擦除，而通过提出的隐含实体泄漏率（Latent Entity Leakage Rate, LELR）指标发现，对于常见实体，模型在内部推理轨迹中仍以超74%的概率重构被遗忘信息；而对于冷门实体，虽无显性泄漏，但依赖隐式属性关联仍维持60%-90%的准确率。这表明现有遗忘机制仅构成“虚假遗忘”的表象，本质上是输出层面的过滤而非参数级知识清除，暴露出人工智能安全中的深层漏洞。
链接: https://arxiv.org/abs/2611.15925
作者: Sujan Maharjan,Lu Liu
机构: 未知
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: Accepted at NeurIPS 2026 Evaluations and Datasets Track
Abstract:Large Language Models (LLMs) have emerged as primary interfaces for factual retrieval, yet current evaluation paradigms rely almost exclusively on explicit entity names. In this work, we demonstrate that LLM knowledge is fundamentally lexically anchored: models possess extensive factual information but find it significantly difficult to retrieve without explicit name tokens. To quantify this limitation, we introduce ShadowBench, a rigorously hardened, shortcut-resistant benchmark designed to evaluate Latent Entity Association via a novel Dual-Trait Association (DTA) task. Our evaluation reveals a pervasive “Shadow Gap” across all model scales \unicodex2013 up to the frontier models GPT-5.4 and Claude-Sonnet-4.6 \unicodex2013 where removing lexical anchors causes performance drops of over 20%. Furthermore, we demonstrate that this lexical dependency exposes a critical vulnerability in AI safety. Applying ShadowBench to Machine Unlearning, we find that state-of-the-art algorithms (e.g., Gradient Difference, NPO) achieve only superficial lexical erasure. While unlearned models fail on direct queries, our novel Latent Entity Leakage Rate (LELR) metric reveals that for popular entities, models explicitly reconstruct the “forgotten” entity in their internal reasoning traces in over 74% of cases, utilizing residual shadow knowledge to solve associative tasks. Conversely, for niche entities, models retain high shadow accuracy (60-90%) via implicit attribute correlation despite 0% LELR. Ultimately, ShadowBench indicates that current unlearning paradigms create an “Illusion of Forgetting,” demonstrating that existing methods act as superficial output filters and are not yet reliable for achieving true parametric erasure.
[NLP-64] Fine-Grained Generative Hallucination Detection and Correction for Aerial Image Captioning ACCV2026
【速读】： 该论文旨在解决生成式无人机航拍图像（aerial imagery）文本描述时，多模态大语言模型（Multimodal Large Language Models, MLLMs）易产生幻觉（hallucination）的问题。具体而言，航拍图像因视角极端变化、目标尺寸小、类别分布不均及视觉实体模糊等特性，使得现有模型生成的描述常与实际图像内容不符，严重影响下游应用的可靠性。现有幻觉检测方法或依赖计算成本高昂的外部大模型（如GPT-4），在航拍场景下表现不佳；或将其简化为刚性的二分类任务，忽视了复杂场景中固有的不确定性。针对上述问题，论文提出一种轻量级、无需参考文本（reference-free）的幻觉检测与修正框架。其核心创新在于：利用先进的MLLMs（如InternVL3和Qwen2.5-VL）高效识别生成描述中的幻觉内容，并通过专用的精炼模块对幻觉部分进行事实一致性修复。该方法在Labelled Illusion Dataset（LID）上的实验表明，能够显著降低幻觉率，同时保持低计算开销，适用于资源受限环境下的实时部署。
链接: https://arxiv.org/abs/2611.15809
作者: Debolena Basak,Somsubhra De,Maunendra Sankar Desarkar
机构: IIT Hyderabad(印度理工学院海德拉巴分校); IIT Madras(印度理工学院马德拉斯分校)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: Accepted at ACCV 2026
Abstract:Aerial imagery captured by unmanned aerial vehicles (UAVs) plays a vital role in applications such as surveillance, disaster response, and large-scale environmental monitoring. Automatic generation of textual descriptions for such imagery is essential for reducing manual inspection overhead and enabling better downstream decision-making. However, Multimodal Large Language Models (MLLMs), despite their impressive captioning capabilities, are prone to hallucination, producing descriptions that are inconsistent with the visual input. This problem is significantly amplified in aerial imagery due to its unique challenges such as extreme viewpoint variation, small objects, class imbalance, and ambiguous visual entities. Existing hallucination detection approaches either rely on computationally expensive large external models such as GPT-4, which perform poorly on aerial domains, or treat the problem as a rigid binary classification task, ignoring the inherent uncertainty in ambiguous cases. To address these limitations, we propose a lightweight, reference-free framework for hallucination detection and correction in free-form aerial image captioning. Our method leverages state-of-the-art MLLMs, specifically InternVL3 and Qwen2.5-VL, to detect hallucinated content in generated captions with minimal computational overhead. Detected hallucinations are corrected through a dedicated refinement module that improves the factual consistency of the original captions. Extensive experiments on the Labelled Illusion Dataset (LID) demonstrate that our framework significantly reduces hallucination rates while remaining practical for real-time deployment in resource-constrained environments. Code is available at: this https URL.
[NLP-65] Computational Explanations of Semantic Universals
【速读】： 该论文旨在解决跨语言语义普遍性（semantic universals）的成因问题，即不同语言在表达意义时共享深层特征的根源。其核心问题是：为何全球语言在量词、颜色词、情态动词和不定指代等范畴上表现出一致的模式？解决方案的关键在于构建一个计算解释框架，基于“学习简便性”（ease of learning）与“高效沟通”（efficient communication）两大认知-交际原则，通过计算模型揭示这些普遍性现象背后的生成机制。该框架将语言结构的共性归因于人类认知、交流与社会结构的共同约束，借助跨语言语义学与计算语言学的进展，实现了对语义类型学中典型现象的可计算、可验证解释，并为未来提升计算方法在语义类型学中的解释力提供了方向。
链接: https://arxiv.org/abs/2611.15804
作者: Shane Steinert-Threlkeld,Jakub Szymanik
机构: University of Washington; University of Trento
类目: Computation and Language (cs.CL)
备注: submitted to Cambridge Elements in Semantics
Abstract:Despite their differences, the languages of the world share many deep features when it comes to how they express meaning. These semantic universals are likely to reflect shared aspects of human cognitive, communicative, and social structure. Recent advances in cross-linguistic semantics and computational linguistics allow us to develop explanations of the sources of such properties. This Element presents a body of work doing just that: we explicate a framework for providing computational explanations of semantic universals and then describe several such explanations using principles of ease of learning and efficient communication. Empirical domains include quantifiers, color terms, modals, and indefinites. After providing these explanations, we provide a discussion of future directions for advancing the explanatory power of computational methods in semantic typology.
[NLP-66] Honesty Is a Race: Predicting When Verifiable-Reward RL Teaches a Model to Bluff
【速读】： 该论文旨在解决强化学习（Reinforcement Learning, RL）中模型在面对不可验证断言（unverifiable verdicts）时可能出现的“欺骗行为”（bluffing）问题，即当诚实回答与伪造回答（如声称程序正确但实际存在缺陷）具有相同期望回报时，模型可能倾向于选择更具策略性而非真实性的输出。其核心挑战在于如何设计奖励机制，使模型在训练过程中趋向于诚实（honesty），而非利用检查器（checker）无法验证的断言进行误导。解决方案的关键在于引入一个动态平衡机制：通过调整对“证书”（certificate，即揭示漏洞的输入）和“不可验证断言”（unverifiable verdict，如“该程序正确”）的奖励比例 $ r/b $，并结合检查器对模型生成证书的接受率加权值 $ \tildep $ 与未验证断言正确性的内在倾向 $ \kappa $，实现对模型诚实度的调控。研究表明，在接近阈值 $ \tildep,r/b \approx 1+\kappa $ 时，诚实行为成为一种可竞争的“竞赛”，而检查器通过提升 $ \tildep $ 能有效促进诚实；此外，随着训练推进，$ \kappa $ 逐渐下降，导致模型初期可能出现短暂的欺骗行为，随后回归诚实。更重要的是，当奖励比超过某一理论上限（即“天花板”）时，模型会开始在本应正确的程序上错误地报告缺陷，表明诚实性存在边界。实验结果验证了这一机制的有效性：在深度为3的Datalog任务中，先以较宽松的奖励训练再削减奖励，可使模型保持至少67%的诚实率，远优于从头训练的模型（仅4–15%）。同时，即使在诚实性崩溃后，线性探测仍能识别出训练前的原始真实性，说明模型内部仍保留了部分可解释的诚实信号。
链接: https://arxiv.org/abs/2611.15779
作者: Nathaniel Bottman,Alexei Oblomkov
机构: Incubilate; University of Massachusetts, Amherst (马萨诸塞大学阿默斯特分校)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 34 pages, 10 figures
Abstract:Two rewards with the same expected payoff at the start of training can end a reinforcement-learning run in opposite places, with an honest model or one that bluffs. We study rewards that pay b for a verdict that a checker cannot verify (“this program is correct”) and r for a certificate that it accepts (an input revealing a bug). To first order, a GRPO step makes the model more honest exactly when \tildep,r/b1+\kappa . Here \tildep is the rate at which the checker accepts the model’s certificates, weighted toward the items on which the model still gives both verdicts, and \kappa is the pull of the items on which the unverifiable verdict is right. Both terms move during training. A checker raises \tildep , so near the threshold honesty is a race, which a run just below the threshold can win and a run under a lottery (random pay with the same mean) cannot. The term \kappa starts large and falls as the other items settle, so a run just above the threshold first dips toward bluffing, on code nearly to zero, before climbing back. Paying the honest verdict also pulls the other items toward it, which predicts a ceiling on the payoff ratio, above which the model gives the honest verdict where it is wrong. On propositional Datalog, two code models, and abstention on grade-school math, with p the model’s success rate before RL, no run above pr/b=1 ends collapsed, and runs under the checker can survive slightly below it. At pr/b=3 , above the ceiling, both code models call about half of correct programs buggy. On depth-3 Datalog, training generously under the checker and then cutting the payoff keeps the model at least 67% honest in three seeds, at a ratio where the same model trained from the start falls to 4-15% and is at most 47% honest after 160 steps. On depth-2 Datalog, when honesty collapses, a linear probe fit before RL still reads the truth.
[NLP-67] SO-Arena: Benchmarking the Capability Spectrum and Reliability of System One Models
【速读】： 该论文旨在解决当前对系统一模型（System One models）能力谱系与决策可靠性评估不足的问题。现有基准测试多依赖于聚合性能指标，难以全面揭示模型在多样化任务中的实际表现及决策可信度。为此，论文提出SO-Arena基准，系统性地刻画模型在11个领域和五大能力类别下的综合能力谱系与决策可靠性。其解决方案的关键在于：通过基于可靠来源的证据验证与人工标注校验确保参考标签的准确性；采用领域特异性难度标准实现细粒度能力评估；引入沙盒环境中的交互式迭代执行以模拟真实场景；并通过高推理强度任务支持系统一与系统二模型的对比分析。此外，除准确率外，还引入两个互补的可靠性评估维度——在可控选项集变化下的决策稳定性，以及在无有效选项时的“缺失答案识别”能力。对21个模型的评估表明，模型规模扩展与额外推理带来的收益在不同领域间差异显著，且整体高性能并不等同于高可靠性：表现最佳的系统一模型在可靠性方面仅位列第七。研究强调，能力谱系与决策可靠性应作为相互补充的评估维度，共同构建对系统一模型优势与局限性的系统性认知。
链接: https://arxiv.org/abs/2611.15777
作者: Lichen Li,Ziping Dong,Sicheng Feng,Qi Li,Xinchao Wang
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 32 pages, 10 figures, 7 tables. Project page: this https URL
Abstract:System One models promise fast, general-purpose decision-making without lengthy token generation. However, understanding their capability spectrum and decision reliability requires more than aggregate performance on current benchmarks. Existing evaluations often provide limited insight into the breadth of model capabilities and the reliability of their decisions. We introduce SO-Arena, a benchmark characterizing the capability spectrum and decision reliability of System One models across 11 domains and five capability families. For the capability spectrum, SO-Arena offers evidence-grounded evaluation through reliable sources and manual verification of reference labels against model-visible inputs; fine-grained capability assessment through domain-specific difficulty criteria; and realistic interactive evaluation, where models iteratively execute actions in sandboxes. Reasoning-intensive tasks further enable system one-two comparison. Beyond accuracy, SO-Arena assesses decision reliability through two complementary measures: decision stability under controlled option-set changes and missing-answer recognition when no valid option is available. Evaluations across 21 models reveal that the benefits of scaling and additional reasoning vary substantially across domains. More importantly, strong overall performance does not necessarily imply reliable decision-making: the best-performing System One model ranks only seventh in reliability. Together, these findings highlight the need to characterize capability spectrum and decision reliability as complementary dimensions, providing a systematic view of the strengths and limitations of System One models.
[NLP-68] Where Is Uncertainty Observed? A Mechanistic Understanding of Uncertainty Quantification in LLM s
【速读】： 该论文旨在解决大语言模型（Large Language Model, LLM）在生成响应时难以判断其可信度的问题，核心挑战在于如何有效量化模型输出的不确定性（Uncertainty Quantification, UQ）。现有方法多基于不确定性来源或估计机制进行分类，但未明确指出不确定性在模型计算过程中应从何处观测，而这一观测点直接影响估计器能否获取充分信息以准确恢复目标不确定性。为此，论文引入机制可解释性（mechanistic interpretability）视角，提出四种机制观察透镜：结构透镜（观察组件更新）、表征透镜（观察隐藏状态）、分布透镜（观察词元概率）和语义透镜（观察意义分布），并通过残差流（residual stream）建立各透镜间的关联，并依据其所携带信息量进行排序，从而为不同不确定性评估目标提供方法选择原则。论文进一步通过这四类透镜系统梳理现有UQ方法，并区分评估目标、评价标准与评估指标，以厘清不同的评估范式。最终，论文指出了若干开放性问题与研究方向，强调需将不确定性估计与模型内部计算过程及下游应用更紧密地结合，推动可信AI的发展。
链接: https://arxiv.org/abs/2611.15735
作者: Xiaoyu Liang,Virginia Ceccatelli,Aliai Eusebi,Tobias Braun,Nils Loose,Carlo Malagnino,Lorenzo Cavallaro
机构: University College London(伦敦大学学院); Technische Universität Darmstadt(达姆施塔特工业大学); Universität zu Lübeck(吕贝克大学); Mohamed bin Zayed University of Artificial Intelligence(穆罕默德·本·扎耶德人工智能大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Knowing when a large language model’s (LLM’s) response can be trusted is critical for its safe use. Uncertainty quantification (UQ) can help LLMs recognize if their responses may be unreliable. Existing taxonomies organize UQ methods by uncertainty source or estimation mechanism. They leave unresolved the question of where uncertainty is observed in the model’s computation. The answer matters because the observation point determines whether an estimator has access to enough information to recover a given target, and how well it can recover it. Drawing on mechanistic interpretability, we define four mechanistic observation lenses: the structural lens observes component updates, the representational lens observes hidden states, the distributional lens observes token probabilities, and the semantical lens observes distributions over meanings. We relate these lenses through the residual stream and order them by the information they carry, which gives a principle for selecting UQ methods for different targets. We survey existing methods through the four lenses. Furthermore, we distinguish evaluation targets, criteria, and metrics to clarify different UQ evaluation protocols. Finally, we identify open problems and research directions to connect uncertainty estimation to the internal computation and downstream use of LLMs.
[NLP-69] Outside the Purview of the Reward: Capability Drift in Multilingual Post-Training NEURIPS2026
【速读】： 该论文旨在解决后训练（post-training）过程中模型在提升特定任务表现的同时，不可避免地导致其他重要能力退化的问题，尤其关注多语言混合专家模型（multilingual mixture-of-experts model）中不同任务间的性能权衡。其核心问题是：如何在不损害关键下游能力的前提下，有效优化模型对特定任务的奖励行为。解决方案的关键在于采用分叉式微调策略——从同一监督预训练检查点（supervised checkpoint）出发，构建两个分支：一是直接基于奖励建模进行强化学习微调（GRPO），二是引入独立训练的“兄弟分支”（sibling branch），通过语言感知的指令遵循（language-aware shaping）和对齐机制恢复退化的印地语（Indic）任务表现。研究发现，尽管独立分支可恢复印地语任务的F1分数并保持数学（MATH）性能，但未能修复脚本合规性问题，并在竞赛类数学与代码生成任务间产生权衡。进一步分析表明，奖励函数引导下的性能提升主要源于对答案框（boxed-answer）的过度生成，而非实际准确率的显著提高。因此，该研究强调应结合响应形式（response forms）解读基准测试得分，以避免因奖励信号偏差导致的误导性评估。
链接: https://arxiv.org/abs/2611.15717
作者: Sutanshu Saksena Raj,Purusharth Malik,Pasunuti Prasanjith,Daksh Maan,Parijat Chatterjee,Avinash Benki
机构: Gnani.AI
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Accepted at the Foundations of LLM Post-Training in Changing Environments, NeurIPS 2026
Abstract:Post-training can improve rewarded behavior while weakening capabilities users still need. We study two GRPO branches from a shared supervised checkpoint in a multilingual mixture-of-experts model. The released branch gains 36.07 points on MATH but loses 19.32 F1 points on XQuAD-Indic, while XQuAD-English improves under the same scoring rule. An independently trained sibling adds Indic instruction following and language-aware shaping. It recovers the regressing Indic task scores and preserves MATH, but does not restore script compliance and trades off competition mathematics and code. In a later round, judge training is followed by a sharp extraction decline from the Indic branch; reward-free distillation from its earlier checkpoint restores most of the lost F1. A MATH score decomposition associates most of the measured gain with increased boxed-answer emission; accuracy among boxed answers changes little. These results show why benchmark scores should be read alongside the response forms they reward.
[NLP-70] Curiosity-0826-4B: Generative Medical Event Sequence Modeling Across Pediatric and Adult Populations
【速读】： 该论文旨在解决生成式医疗事件基础模型在临床预测任务中泛化能力有限、对儿科患者及多种临床结局预测性能不足的问题。其核心挑战在于如何在不依赖特定任务微调的前提下，提升模型对跨年龄群体（尤其是2–17岁儿童）和多样化临床事件（如哮喘发作、镰状细胞危象、死亡等）的零样本预测能力。解决方案的关键在于构建一个参数规模更大（4.3B）、上下文窗口更长（16K）且经过计算最优预训练与持续预训练优化的模型——Curiosity-0826-4B。该模型通过引入更广泛的人群覆盖和更丰富的临床事件表征，显著提升了实验室值预测精度（平均绝对误差降低18%–46%），并在儿科人群中实现了与成人相当的疾病事件预测性能（如新发哮喘诊断AUROC 0.742–0.785，哮喘加重AUROC 0.829–0.852）。此外，通过显式建模已记录死亡事件，模型首次实现对成人患者在临终关怀、住院及门诊等多种场景下的死亡风险预测（AUROC 0.821–0.942）。基于缩放定律指导的训练策略使模型在减少56%训练令牌的同时，仍以2.2倍于前代模型的浮点运算量完成训练，从而在效率与性能间取得平衡。这一进展表明，通过扩展零样本预测范式至儿科人群与多类临床结局，可有效推动通用型临床与运营应用的发展，而无需为每项任务单独训练专用模型。
链接: https://arxiv.org/abs/2611.15710
作者: Mani Nagarathnam,Luke Benham,Paul Blazek,John Cashy,Kevin Chung,John Muenning,Hank Voeller,Shane Waxler,Davis White,Alison Callahan,Kevin Dysart,Andrew Loza,Michal A. Mankowski,Mattheus Ramsis,Nigam H. Shah,Karen K. Wong,Rahul Shah
机构: Epic Systems; Department of Medicine, Stanford School of Medicine (斯坦福大学医学院); The Children’s Hospital of Philadelphia (费城儿童医院); Department of Biomedical Informatics and Data Science, Yale University (耶鲁大学生物医学信息学与数据科学系); Department of Surgery, NYU Grossman School of Medicine (纽约大学格罗斯曼医学院); Division of Cardiology, Department of Medicine, University of California-San Diego, San Diego, California (加利福尼亚大学圣地亚哥分校医学院心脏病学分部); Department of Medicine, Stanford School of Medicine (斯坦福大学医学院); Epic Systems
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:We introduce Curiosity-0826-4B, an updated version of the generative medical event foundation model trained on Epic Cosmos that simulates patient timelines to produce zero-shot clinical predictions. Curiosity-0826-4B incorporates a broader patient population and richer clinical event representations, while generally matching or exceeding Curiosity-0226-1B, the largest model in the previous Curiosity-0226 series, on evaluated tasks supported by both models. Laboratory value prediction improved, with MAE 18–46% lower than that of Curiosity-0226-1B. We evaluated prediction of disease events across pediatric cohorts aged 2–17 years. AUROCs for new asthma diagnosis (0.742–0.785) and asthma exacerbation (0.829–0.852) were comparable to those in the adult reference cohort, while AUROCs for sickle cell crisis (0.820–0.847) were lower than in adults (0.915). Explicit representation of documented death enabled new mortality prediction tasks among adults across hospice, inpatient, and ambulatory settings (AUROC 0.821–0.942). Scaling-law analysis guided compute-optimal initial pretraining followed by continued pretraining with context extension. The resulting Curiosity-0826-4B model has more parameters (4.3B versus 1B) and a longer context window (16K versus 8K). It processes 56% fewer training tokens and requires 2.2 times the training floating point operations (FLOPs) compared with its predecessor. By extending a common zero-shot prediction strategy to pediatric patients and additional outcomes, Curiosity-0826-4B broadens opportunities to develop clinical and operational applications without training separate task-specific models. Further validation should assess local calibration, performance across patient subgroups and event frequencies, timeline completion rates, and generalization to new settings, alongside prospective evaluation of clinical and operational utility.
[NLP-71] Dynamic Block-Online Streaming ASR for Low-Resource Agglutinative Code-Switching Speech with Morphology-Aware Evaluation INTERSPEECH2026
【速读】： 该论文旨在解决低资源、黏着性语言在低延迟自动语音识别（ASR）中面临的挑战，特别是当字符顺序偏离、后缀跨越代码切换边界或代码切换数据分布不均衡时，传统固定前瞻窗口机制失效的问题。其核心解决方案是提出首个针对句内孟加拉语-英语代码切换的流式ASR综合研究，揭示了严格因果流式处理与长程形态依赖之间的不匹配问题。为此，论文提出Dynamic Block-Online框架，结合基于语音活动检测（VAD）对齐的推理机制，在语义完整性、延迟和上下文利用之间实现平衡，并通过全局双向注意力机制有效保留形态一致性。为缓解数据稀缺问题，提出Script-Anchored Loanword Injection方法，通过在训练文本中注入英语借词以建模音位转换规律。同时，引入CS-WER评估指标用于诊断切换点、词干及后缀错误，实验证明灵活的延迟预算可显著提升对黏着性及混合语言语音的识别性能，并成功迁移至月经健康与医疗领域。
链接: https://arxiv.org/abs/2611.15691
作者: Kazi Rafat,Afifa Imran,Md. Ismail Hossain,Md. Romzan Ali,Fuad Rahman,Sifat Momen,Shafin Rahman,Nabeel Mohammed
机构: 未知
类目: Computation and Language (cs.CL); Sound (cs.SD)
备注: Published at Interspeech 2026 (oral). 5 pages, 3 figures, 1 table. Code: this https URL
Abstract:Low-resource, agglutinative languages challenge low-latency ASR, as fixed lookahead windows fail when grapheme order diverges, suffixes cross code-switch boundaries, or code-switching data is imbalanced. We present the first comprehensive study of streaming ASR for intra-sentential Bangla-English code-switching, identifying a mismatch between strict causal streaming and long-range morphological dependencies. We introduce a Dynamic Block-Online framework with VAD-aligned inference that balances semantics, latency, and context, enabling global bidirectional attention to preserve morphological integrity. To mitigate data scarcity, we propose Script-Anchored Loanword Injection, augmenting training text with English loanwords to model phonotactic transitions. We introduce CS-WER to diagnose switch, root, and suffix errors, and show that flexible latency budgets enhance recognition of agglutinative and mixed-language speech, with successful transfer to menstrual-medical domains.
[NLP-72] ALMeD: Medieval French Gold-Standard
【速读】： 该论文旨在解决中古法语（Medieval French）在自然语言处理（NLP）研究中因语料资源偏倚而造成的科学文献覆盖不足问题，尤其针对医学与外科文献的缺失。现有基于NLP的方法多依赖法律文本和文学文本构建语料库，忽视了具有重要历史与知识价值的科学文献。为此，本文提出ALMeD——一个正在建设中的标注丰富、语义清晰的中古法语医学与外科手稿语料库，涵盖古法语与中古法语（包括安格鲁-诺曼语）。其核心解决方案在于构建一个经过语义消歧的高质量黄金标准（gold standard）语料，不仅聚焦于医学术语的精确识别与解析，还充分考虑了该语言在形态演变、句法结构变迁以及拼写不规范等方面的复杂性，从而为中古法语文本的深度语义分析提供可靠基础。
链接: https://arxiv.org/abs/2611.15661
作者: Ragini Menon,Sabine Tittel
机构: Heidelberg Academy of Sciences and Humanities(海德堡科学院与人文科学学院)
类目: Computation and Language (cs.CL)
备注:
Abstract:Medieval French is one of the Romance vernacular languages with a significant history of transmission and an enormous influence on other vernacular languages of the time. The sources convey valuable knowledge about all aspects of culture. The study of Medieval French has, over the last few years, increasingly benefited from NLP-based methods. There have been numerous efforts to parse medieval French; however, these efforts use corpora comprising legal texts and – primarily – literary texts as their foundation and training data. This means that, except for juridical resources, scientific literature is overlooked. We present ALMeD, an annotated, semantically rich corpus (work-in-progress) of medical and surgical treatises produced in medieval French, i.e., Old and Middle French, including Anglo-Norman. Our semantically disambiguated gold standard goes beyond existing resources, focusing on medical terminology, and accounts for morphological development, evolving syntactical structures and the lack of spelling normalisation in these languages. Subjects: Computation and Language (cs.CL) Cite as: arXiv:2611.15661 [cs.CL] (or arXiv:2611.15661v1 [cs.CL] for this version) https://doi.org/10.48550/arXiv.2611.15661 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Journalreference: HLAI 2026 Conference Companion, pp. 1-6 (2026) Related DOI: https://doi.org/10.5281/zenodo.20757705 Focus to learn more DOI(s) linking to related resources
[NLP-73] Can Agents Relive Past Experiences? Diagnosing Event-Affect Decoupling in Affective-Episodic Retrospection
【速读】： 该论文旨在解决记忆增强型智能体在长期个性化交互中，如何可靠地回忆情感经历并保持其情绪关联性的问题。现有评估方法通常将情绪准确性与事实记忆的锚定性分开评估，忽视了事件与情绪之间的对应关系，导致对情感情景记忆（Affective-Episodic Memory）的评价不完整。为此，论文提出“情感情景回溯”（Affective-Episodic Retrospection, AER）作为一项新任务，用于显式评估过去事件、伴随情绪及其对应关系的恢复能力，并定义了“事件-情绪解耦”（Event-Affect Decoupling）为一种关键失败模式——即正确的情绪被召回，但错误地关联到非目标事件。为诊断该问题，研究提出了分阶段的EM-TRACE框架，将AER分解为检索、情绪恢复与生成三个阶段。在基于GoEmotions的受控基准测试中，仅依赖情绪进行检索的模型虽能达到28.65%的情绪恢复率，但仅能正确检索目标事件4.44%的查询，综合均匀准确率（UAR）仅为24.21%。进一步分析表明，事件检索与情绪分配分别引入了不同的绑定错误。研究结论强调，可靠的AER需要对情感经历进行有根基的回忆，必须联合评估事件检索、情绪恢复、事件-情绪绑定以及生成质量。
链接: https://arxiv.org/abs/2611.15655
作者: Tingna Xie,Ming Wang,Shi Feng,Yongkang Liu
机构: Northeastern University (东北大学); Northeastern University at Qinhuangdao (东北大学秦皇岛分校)
类目: Computation and Language (cs.CL)
备注:
Abstract:As memory-augmented agents increasingly support long-term and personalized interactions, reliable recall of emotional experiences while preserving their associated affect becomes necessary. However, existing evaluations usually assess emotion accuracy and factual memory grounding separately, overlooking event-affect correspondence. In this work, we formulate Affective-Episodic Retrospection (AER) as a task that explicitly evaluates the recovery of past events, associated affects, and their correspondence, and define Event-Affect Decoupling as a failure mode where the correct emotion is recalled but associated with the wrong episode. To diagnose Event-Affect Decoupling, we propose EM-TRACE, a stage-wise framework that decomposes AER into retrieval, affect recovery, and realization. On a controlled GoEmotions-based benchmark, Emotion-only retrieval achieves 28.65% affect recovery but retrieves the target episode for only 4.44% of queries, yielding a UAR of 24.21%. Further analysis reveals that event retrieval and affect assignment introduce distinct binding failures. These findings show that reliable AER requires grounded recall of emotional experiences, where episode retrieval, affect recovery, event-affect binding, and generation quality should be evaluated jointly. Our code is available at this https URL.
[NLP-74] Differential Backpack Language Models: Sense Inhibition via Signed Contextualization Weights
【速读】： 该论文旨在解决传统背包语言模型（Backpack language models）中因非负性约束导致的语义组合受限问题，即非负性不仅限制了上下文对词义向量的动态调节能力，还与线性读出机制耦合，削弱了模型的可解释性。其解决方案的关键在于摒弃传统的非负性约束，引入基于差分注意力（differential attention）的感知权重计算方式——通过两个软最大函数（softmax）之差与学习到的每义项增益参数相结合，使上下文能够主动抑制某些语义，从而实现更灵活、更具表达力的语义组合。该方法被命名为“差分背包”（differential Backpack），在相同参数量下显著降低了与Transformer模型之间的困惑度差距（减少31%），且在不同义项数量、更大参数规模（最高达350M）及多种数据集（如OpenWebText）上均表现优越。此外，该模型保留了每个义项的可解释性（例如金融语境下“bank”义项被抑制至-0.50），并能精确预测反事实移除实验的结果；实验证明，线性读出接口几乎无代价，而引入非线性读出虽仅提升0.43个困惑度点，却破坏了三分之二的语义分解可解释性。最终，差分背包所学习的义项表示也具有更强的语义表征能力，零样本句间相似度任务性能最高提升33%。
链接: https://arxiv.org/abs/2611.15642
作者: Konstantinos Skianis,Giannis Nikolentzos,Polykarpos Meladianos,Michalis Vazirgiannis
机构: University of Ioannina(伊奥尼亚大学); University of Peloponnese(伯罗奔尼撒大学); MBZUAI(穆罕默德·本·扎耶德人工智能大学); École Polytechnique(巴黎综合理工学院)
类目: Computation and Language (cs.CL)
备注:
Abstract:Backpack language models predict with a strictly linear combination of non-contextual per-word sense vectors, weighted by non-negative contextualization scores. We show that non-negativity conflates two properties: the linear readout, which carries the interpretability, and the conic combination, which merely restricts it. We relax the latter by computing the sense weights with differential attention, a difference of two softmax maps with a learned per-sense gain, so context can actively inhibit a sense; we call the result a differential Backpack. At equal parameters on WikiText-103 this cuts the Backpack’s perplexity gap to a Transformer by 31%, improves over the softmax Backpack at every sense count, persists with a shrinking margin up to 350M parameters on OpenWebText, and is localized at the sense layer by ablations and controls. The learned inhibition is directly interpretable (a financial context drives the river sense of bank to -0.50), and the preserved per-sense attribution predicts counterfactual removals exactly. We tested a number of combination rules, showing the linear interface is nearly free. Making the readout nonlinear gains only 0.43 perplexity points while invalidating two thirds of the decomposition. The differential Backpack’s senses are also better representations, improving zero-shot sentence similarity by up to 33%.
[NLP-75] ChJGBench: A Comprehensive Benchmark for Chinese Legal Judgment Generation with Diagnostic Evaluation EMNLP2026
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在生成完整且具有法律效力的司法文书时的可靠性评估问题，尤其关注模型在准确适用法律条文、处理影响裁判结果的关键法律问题（如诉请、指控、处罚及金钱义务等）方面的表现。现有基准测试多集中于孤立的法律任务或粗粒度评价，难以全面衡量模型在复杂司法场景下的综合能力。为此，论文提出了ChJGBench，一个面向中文裁判文书推理与裁决部分生成的基准，涵盖民事、刑事及行政三类案件。该基准基于3,462份真实裁判文书构建，采用案例类型感知的诊断式评估框架，包含13种评估指标，并引入混合范式以提升评估的全面性与细致度。其核心解决方案在于通过结构化、细粒度的多维度评估体系，揭示不同模型在各类案件中的实际表现差异，发现法律领域专用模型在整体上并无统一优势，且在民事与行政案件中仍面临显著挑战，从而为后续模型优化提供精准的方向。
链接: https://arxiv.org/abs/2611.15639
作者: Yinghao Hu,Linggang Kong,Qian Xiao,Yuting Huang,Yunlong Zhang,Yujie Zhang,Yiquan Wu,Fei Duan,Kaisong Song,Jun Lin,Kun Kuang,Fei Wu
机构: Northeastern University (东北大学); Zhejiang University (浙江大学); National University of Defense Technology (国防科技大学); Alibaba Group (阿里巴巴集团)
类目: Computation and Language (cs.CL)
备注: 25 pages, 3 figures. Published in EMNLP 2026 (main conference)
Abstract:With the rapid development of LLMs, increasing attention has been paid to whether they can reliably generate complete and legally valid judicial documents. This calls for a benchmark capable of evaluating whether models can accurately apply statutes and resolve outcome-critical issues, including claims, charges, penalties, and monetary obligations. However, existing benchmarks typically focus on isolated legal tasks or coarse-grained evaluation. We introduce ChJGBench, a benchmark for generating the reasoning and dispositive sections of Chinese judgments across civil, criminal, and administrative cases. Built from 3,462 real-world judgments, ChJGBench provides a case-type-aware diagnostic evaluation framework comprising 13 metric types and adopts a hybrid paradigm. Experiments reveal no uniform advantage for legal-domain models over general-purpose models, with model strengths varying across diagnostic dimensions. Performance is generally stronger on criminal cases, while civil and administrative cases continue to present substantial challenges.
[NLP-76] VWEN: A Curated Small Language Model for Vietnamese Paraphrasing
【速读】： 该论文旨在解决越南语（Vietnamese）生成任务中对不同改写程度（paraphrase level）精准控制的难题，尤其针对五类相关任务：多版本改写、简化、要点提炼、语体转换及改写质量判断。其核心挑战在于如何在保持语义一致性的同时，实现可控的语义变化程度，并提升模型在复杂输入（如疑问句）上的泛化能力。解决方案的关键在于构建一个基于四阶段筛选流程的高质量训练数据构建方法：首先通过规则与去重预处理，继而采用双向蕴含过滤器（bidirectional entailment filter）剔除语义不一致样本，再利用分歧度评分（divergence score）对样本进行高低改写程度标注，最后根据覆盖度与指令遵循难度进行采样优化。实验表明，该方法训练的VWEN模型（0.8B参数）在高分歧度改写任务上达到0.61的成功率，显著优于在相同计算预算下使用ViSP语料训练的模型（0.46），且在疑问句等复杂输入上表现更优，原因在于其保留了原始模型的多任务泛化能力，而仅使用改写数据训练会削弱这些能力。其中，蕴含过滤器是四阶段流程中最关键的环节。
链接: https://arxiv.org/abs/2611.15587
作者: Sang Quang Nguyen
机构: University of Information Technology (信息科技大学); Vietnam National University (越南国家大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:We present VWEN, a 0.8B-parameter model for Vietnamese paraphrasing at a requested level of change and for five related tasks: generating several paraphrases, simplification, gist, register transfer and paraphrase judgement. VWEN is Qwen3.5-0.8B fine-tuned on about 7k samples chosen by a four-stage pipeline: rules and deduplication, a bidirectional entailment filter, a divergence score that assigns low and high paraphrase levels, and selection for coverage and instruction-following difficulty. We compare it with paraphrase-only recipes from the same pipeline and with training on the ViSP corpus under matched and larger step budgets. On high-divergence paraphrasing, VWEN reaches a success rate (requested level reached and meaning preserved) of 0.61 against 0.46 for the ViSP model at matched compute, a difference that holds on inputs resembling ViSP. Given four times more training examples, the ViSP model closes the gap on such inputs but not on questions, which ViSP lacks and which it often turns into statements. Across all comparisons the entailment filter is the most effective stage. Training on paraphrases alone erases abilities of the base model that the multi-task data preserves.
[NLP-77] Learning a calibrated decision head from logged bandit feedback
【速读】： 该论文旨在解决在仅能获取选择性反馈（logged bandit feedback）的场景下，如何有效微调生成式模型决策头（decision head）以提升其预测准确性与校准能力的问题。其核心挑战在于：传统REINFORCE类方法在利用采样动作自身概率作为奖励时，会因奖励信号失真而导致模型性能崩溃——具体表现为错误动作至少获得与正确动作相同的奖励，且低概率错误动作反而获得更大优势，导致训练信号失效。解决方案的关键在于采用直接基于所选动作概率的合理评分规则，即所选动作适当评分（Chosen-Action Proper Scoring, CAPS），该方法通过将所选动作的概率作为损失函数进行路径梯度优化，避免了传统奖励机制的坍塌问题。实验表明，在仅有日志反馈的情况下，CAPS显著优于仅基于确认答案进行微调的方法，在Banking77任务上提升2.6个百分点准确率、降低0.053的平均预期校准误差（ECE），并减少选择性风险；在情绪识别数据集上更是取得21.7点的显著提升。此外，即使在反馈被已知比率翻转的噪声环境下，CAPS仍能保持对置信度的稳健估计，并通过后向校正恢复大部分校准性能，早期停止策略亦有效防止对噪声样本的记忆。这一方法为弱监督或无完整标签场景下的模型校准提供了高效可行的解决方案。
链接: https://arxiv.org/abs/2611.15572
作者: Ye Yuan
机构: Independent Researcher(独立研究员)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: 40 pages, 6 figures. Code at this https URL (tag v5-paper); adapters at this https URL logs and per-record results at this https URL
Abstract:A decision head reads a state and a list of options and returns a probability per option in one forward pass. Once deployed, it usually observes only whether the option it chose was right: logged bandit feedback. We adapt such a head, LoRA on Qwen3-0.6B-Base and 1.7B-Base, from that feedback and report three results. First, the Brier or log score of the sampled action’s own probability is not a usable REINFORCE reward: every wrong action earns at least gold’s reward, with a mean baseline the lowest-probability wrong action gets the largest advantage, and at two options the signal is zero; in one run it cut accuracy on ordered-scale questions from .636 to .111. Scoring the probability as a loss by the pathwise gradient avoids the collapse, as do the REINFORCE rewards r and r minus p. Second, with full labels, none of four tested bandit arms beat cross-entropy plus one temperature on mean expected calibration error (ECE) over four unseen label schemas. Third, from logged feedback alone, Chosen-Action Proper Scoring (CAPS), the direct method with a proper score on the chosen action’s probability, beats fine-tuning on confirmed answers only, which discards negatives: on Banking77 by 2.6 accuracy points, .053 lower ECE and lower selective risk (three seeds against one baseline seed), and by 21.7 points on one seed on emotion data with an often-wrong logger. At N 5000, CAPS stays 1.5 points below full-label training at 0.6B and .6 points at 1.7B against one full-label seed, and propensity weights add nothing on this log. Under feedback flipped at a known rate, one seed per rate, CAPS’s confidence tracks the noisy channel on average, backward correction recovers most of the clean calibration, and early stopping on held-out logged outcomes prevented memorising noisy draws in all three runs. Code, adapters, logs and per-record results are public.
[NLP-78] Can We Make an LLM Dumber or Is It Just Playing Dumb? An Empirical Study of Believable Performance Throttling
【速读】： 该论文旨在解决生成式模型在性能评估中普遍依赖基准分数作为模型能力代理指标的问题，进而探究能否通过可控手段将模型“驱动”至目标分数，并验证其失败模式是否与自然较弱模型的错误具有相似性。其核心挑战在于实现可信且可解释的性能降低，而非简单地降低分数。解决方案的关键在于提出一种基于项目反应理论（Item Response Theory, IRT）校准的基准测试套件——CapLadder，包含1,985道覆盖数学、代码、多选推理和工具使用等领域的题目，确保评分结构的合理性。研究发现，采用简单的弱教师LoRA调制（LoRA dial）可在推理任务上实现接近目标分数的精确控制（1.9分保留误差），而传统方法如提示工程、温度调节或权重噪声无法达到同等精度与范围。然而，在多选题任务中，模型表现出显著的“答案项吸引子”现象：高达97%的新引入错误集中于单一选项字母（如Qwen中为D，Llama中为A），且该吸引子随模型架构变化。通过逻辑值干预分析发现，此偏差源于一个近似恒定的偏移量，减去该估计偏移可恢复52%至59%的准确率并消除吸引子效应；进一步的选项置换实验表明，偏差主要由选项位置决定，而非内容本身。尽管人工标注的错误类别分布与真实弱参考模型高度一致（推理任务JS散度=0.00，肯德尔一致性κ=0.94），但结构化工具调用任务中仅在Qwen家族中呈现更自然的退化模式（JS=0.09），而全局校准的调制器在跨基准迁移时表现不佳（两个可用基准上误差达14.2分）。因此，论文指出：精确的分数控制并不等同于技能的可移植或可信降级，真正的性能退化必须满足目标分数、题目难度结构及错误分布三方面的一致性。
链接: https://arxiv.org/abs/2611.15568
作者: Chengguang Gan,Yunhao Liang,Hanjun Wei,QingHao Zhang,Shiwen Ni
机构: Independent Researcher; University of Chinese Academy of Sciences (中国科学院大学); Pusan National University (釜山国立大学); Shenzhen University of Advanced Technology (深圳先进技术研究院)
类目: Computation and Language (cs.CL)
备注:
Abstract:A benchmark score is commonly read as a proxy for model capability. We study the inverse problem: can a model be driven to a target score, and do its failures resemble those of a naturally weaker model? We operationalize believable performance reduction along three axes: target calibration, item-difficulty structure, and error-distribution similarity to a real weak reference. We introduce CapLadder, an IRT-calibrated suite of 1,985 items spanning mathematics, code, multiple-choice reasoning, and tool use. A simple weak-teacher LoRA dial reaches unseen target levels on reasoning with a 1.9-point held-out calibration error, whereas prompting, temperature, and weight noise cannot combine comparable range and precision. Yet on multiple choice, the dial converges to an answer-token attractor: up to 97% of its newly introduced errors land on one letter, with the attractor changing from D in Qwen to A in Llama. Logit interventions identify a dominant near-constant shift: subtracting the estimated shift on held-out items recovers 52 to 59% of lost accuracy and removes the dominant attractor, while option permutation confirms that the bias tracks token position rather than option content. With high inter-annotator agreement ( \kappa=0.94 ), human-coded error-category distributions nonetheless match the weak reference on reasoning (JS = 0.00), showing that local plausibility and aggregate detectability differ. Structured tool calls exhibit more natural degradation in the Qwen family (JS = 0.09), while a globally calibrated dial transfers poorly across benchmarks (14.2-point error on two usable benchmarks). Precise score control, therefore, does not imply a portable or believable reduction in skill.
[NLP-79] Faithful in Steps: Improving Generalization and Citation in RAG via Query Decomposition AAAI2026
【速读】： 该论文旨在解决现有生成式检索增强问答（RAGQ）系统在处理多跳推理和多模态理解任务时存在的性能退化问题，具体表现为过度引用（over-citation）、隐含实体识别失败以及泛化能力差。其核心解决方案是提出一种新型RAGQ框架——QDRAG，通过将输入问题分解为原子级子问题以准确识别隐含实体，并引入重排序器（reranker）剔除上下文干扰项，从而减少下游的过度引用问题。为支持问题分解，作者设计了两种零样本方法：基于上下文知识引导的QD-C与基于检索奖励驱动的QD-R，二者均利用多模态大语言模型（MLLM）实现高效分解。值得注意的是，实验发现对问题分解（QD）任务进行微调，相较于直接在下游RAGQ任务上微调，展现出更优的泛化性能。在四个多模态问答基准上的实验证明，QDRAG在答案可追溯性和引用忠实性方面显著优于现有基线，无论在域内还是域外测试中均表现卓越，甚至超越Gemini-Pro。
链接: https://arxiv.org/abs/2611.15538
作者: Yue Liu,Zhongying Ru,Shimin Di,Jipeng Zhang,Ruiyuan Zhang,Xiaofang Zhou
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Accepted by AAAI 2026
Abstract:Retrieval-augment generation is a prevalent strategy to mitigate hallucinations of LLMs. The attributable RAG (RAGQ) generates quotes for its answers. The quotes indicate which input contexts support the RAG to derive the answers, enhancing the answer’s verifiability and trustworthiness. However, existing RAGQs exhibit significant degradation when dealing with questions that require multi-hop reasoning and multi-modal understanding, suffering from over-citation, implicit entity identification failure, and poor generalization. In this paper, we propose a novel RAGQ framework, namely QDRAG. QDRAG breaks down the input question into atomic subquestions to identify the implicit entities. Then, the reranker prunes context distractors to eliminate the downstream over-citation. To facilitate query decomposition, we propose two zero-shot approaches: QD-C and QD-R, which guide the QD MLLM to decompose the question based on context knowledge and retrieval rewards, respectively. One interesting finding is that finetuning on the QD task shows better generalizability compared to directly finetuning on the downstream RAGQ task. Experiments on four multi-modal QA benchmarks demonstrate QDRAG’s efficacy in grounding answers and generating faithful citations. The framework significantly outperforms all the baselines on both in-domain and out-of-domain tests, even surpassing Gemini-Pro.
[NLP-80] AutoOntic: Automatic Ontology Induction from Text for Knowledge Graph Construction
【速读】： 该论文旨在解决基于图的检索增强生成（Retrieval-Augmented Generation, RAG）系统中因知识图谱（Knowledge Graph, KG）质量不足而导致的性能瓶颈问题，特别是由无模式（schema-free）大语言模型（LLM）抽取管道生成的扁平化知识图谱所存在的缺陷：实体与关系类型缺乏类型层次结构以及领域/范围约束。其核心解决方案是提出AutoOntic——一个自动从文本中构建本体（Ontology）的流水线，关键在于通过规范化开放词汇实体类型与关系、构建类型层次结构，并挖掘关系层面的领域与范围约束。该方法利用语料库衍生的关系-论元特征对类型进行优先级排序，并结合局部化、由大语言模型仲裁的父子关系校验机制，在新类型插入时可动态触发已有父子链接的重新分配，从而实现更精准的类型层级组织。该层次结构不仅约束实体去重过程，还支持关系粒度的领域/范围约束推断。在内在评估中，采用基于参考类别的分类协议，结果显示在Text2KGBench的Wikidata-TekGen和DBpedia-WebNLG子集上，超过90%的AutoOntic生成的终点类型与参考类别语义对齐，约一半完全匹配，其余差异主要体现为粒度级别不同；在MINE数据集上，使用AutoOntic构建的本体组织的知识图谱实现了88%的检索准确率。
链接: https://arxiv.org/abs/2611.15467
作者: Alla Chepurova,Aydar Bulatov,Mikhail Burtsev,Yuri Kuratov
机构: Cognitive AI Systems Lab, Moscow, Russia; London Institute for Mathematical Sciences, London, UK
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Graph-based retrieval-augmented generation depends on the quality of the knowledge graph (KG) it retrieves over, yet schema-free LLM-based extraction pipelines typically produce flat graphs: deduplicated entities and relations with no type hierarchy and no domain/range constraints. We present AutoOntic, a pipeline that automatically constructs an ontology from text by canonicalizing open-vocabulary entity types and relations, inducing a hierarchy over the canonicalized types, and mining relation-level domain and range constraints. Its hierarchy induction procedure prioritizes types using corpus-derived relation-argument profiles and places them through local, LLM-adjudicated parent/child checks, allowing newly inserted types to trigger reassignment of existing parent-child links. The resulting hierarchy constrains entity deduplication and supports the induction of relation-level domain/range constraints. For intrinsic evaluation, we introduce a reference-based protocol that classifies induced ontology types as exact, broader, narrower, or unrelated to reference domain and range classes. On the Wikidata-TekGen and DBpedia-WebNLG subsets of Text2KGBench, over 90% of AutoOntic’s induced endpoint types are semantically aligned with the reference classes across two LLM backbones: about half match exactly, while most remaining differences concern granularity. On MINE, knowledge graphs organized using an ontology built by AutoOntic achieve 88% retrieval accuracy. AutoOntic is available at this https URL.
[NLP-81] AraCrypto-ML: A Multi-Label Corpus of Communicative Functions in Arabic Facebook Cryptocurrency Discourse
【速读】： 该论文旨在解决阿拉伯语加密货币社交媒体话语中功能多样性缺乏标注资源的问题。当前的社交媒体内容往往混合了市场评论、技术解释、投资推广、诈骗警告及伊斯兰法理引用等多种语用功能，但尚无公开的多标签标注语料库能够全面捕捉这种语义异质性。为此，研究提出AraCrypto-ML，一个包含10,000条阿拉伯语Facebook加密货币帖子的多标签语料库（2014–2024年，来自2,898个公共页面），基于一组相关关键词采集并使用Fanar-2-27B模型在仅提示（prompt-only）条件下进行多标签标注。语料被划分为六类：市场评论（Market Commentary）、教育性内容（Educational）、推广性内容（Promotional）、怀疑与警示（Skeptical-Warning）、宗教框架（Religious Framing）及其他/不明确（Other/Unclear）。通过1,000条样本与仲裁后金标准标签对比验证，宏观卡帕系数达0.74，并采用Rogan-Gladen校准方法修正标注偏差。两名人工标注者间的一致性为宏观卡帕0.62，低于模型与任一标注者的一致性（分别为0.70和0.69），表明模型在该任务中具备较强的标注稳定性。此外，不同功能类型的帖子其用户反应模式显著差异：教育类与推广类帖子中“点赞”（Love）占比约70%，而警示类帖子中“大笑”（Haha）占比高达约39%，凸显功能对社交互动的影响。该研究的关键解决方案在于构建首个大规模、多标签、基于大语言模型辅助标注的阿拉伯语加密货币话语语料库，并通过量化验证与偏差校正保障数据质量，为后续自然语言处理与社会计算研究提供基础支撑。
链接: https://arxiv.org/abs/2611.15408
作者: Mabrouka Bessghaier,Mohamed Rashad Rouine,Md. Rafiul Biswas,Youssef Ayadi,Wajdi Zaghouani
机构: 未知
类目: Computation and Language (cs.CL)
备注:
Abstract:Arabic cryptocurrency discourse on social media is communicatively heterogeneous: a single post may report prices, explain protocols, promote investment, warn of scams, or invoke Islamic jurisprudence. Yet no annotated resource captures this functional diversity. We introduce AraCrypto-ML, a 10,000-post multi-label corpus of Arabic Facebook cryptocurrency discourse (2014-2024) from 2,898 public pages using a set of related keywords. Collected posts are annotated into six categories: Market Commentary, Educational, Promotional, Skeptical-Warning, Religious Framing, and Other/Unclear, using Fanar-2-27B under a prompt-only multi-label protocol. We validate 1,000 posts against adjudicated gold labels (macro kappa = 0.74) and apply Rogan-Gladen calibration to correct labelling bias. The two annotators agree with each other at macro kappa = 0.62 before adjudication, below the model’s agreement with either of them (0.70 and 0.69). Reaction profiles differ strongly by function: Love is about 70% of reactions per post on Educational and Promotional posts but about 33% on Skeptical-Warning, which draws the largest Haha share (about 39%).
[NLP-82] Is the Importance Ratio Necessary for Stable Reinforcement Learning in LLM s? NEURIPS2026
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在后训练阶段采用强化学习（Reinforcement Learning, RL）方法时存在的计算开销过大问题，尤其针对主流的GRPO（Generalized Reward Policy Optimization）算法中需同时计算旧策略与当前策略似然值以构建重要性采样比率所带来的显著额外开销。其解决方案的关键在于提出一种新的策略优化方法——似然门控策略优化（Likelihood-Gated Policy Optimization, LGPO），通过基于似然值的门控机制实现软信任区域约束，从而无需显式计算旧策略的似然值，消除了重要性采样修正项的需求。实验表明，移除重要性采样修正项并不会损害训练稳定性，而若移除信任区域机制则可能导致训练崩溃；此外，在完全在线策略（fully on-policy）训练场景下，基于比率裁剪的传统信任区域机制因重要性比率恒为1而失效，而LGPO仍能保持稳定，显著提升了不同训练范式下的效率与鲁棒性。在标准训练设置下，LGPO相较于标准实现平均减少约18%的训练时间，且在性能和稳定性上保持相当水平，展现出更优的实用性。
链接: https://arxiv.org/abs/2611.15404
作者: Shuibai Zhang,Junhyuck Kim,Gyeongman Kim,Jaewoong Cho
机构: University of Wisconsin–Madison(威斯康星大学麦迪逊分校); KRAFTON
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 21 pages, 11 figures, 5 tables. Accepted at NeurIPS 2026
Abstract:Reinforcement learning (RL) has become central to post-training large language models (LLMs). However, popular RL methods like GRPO incur non-negligible overhead by computing both old-policy and current-policy likelihoods to form importance sampling ratios. In this work, we propose Likelihood-Gated Policy Optimization (LGPO), which enforces a soft trust region constraint via likelihood-based gating, eliminating the need to compute old-policy likelihoods. Empirically, we show that removing the importance sampling correction term does not harm training stability, whereas removing the trust region mechanism can lead to collapse. Moreover, ratio-based clipping can fail in fully on-policy training: the importance ratio stays at 1, so the ratio-based trust region constraint never activates. Under standard training settings where GRPO is stable, LGPO achieves comparable training stability and peak performance while reducing training time by ~18% on average relative to a standard implementation that recomputes old-policy likelihoods. In fully on-policy training, where GRPO fails, LGPO remains stable, enabling more efficient and robust LLM RL post-training across training regimes.
[NLP-83] owards Better Legal Reasoning LLM s: Signal Balancing and Reward Scheduling EMNLP2026
【速读】： 该论文旨在解决大语言模型（LLM）在法律领域生成高质量推理链（reasoning traces）时面临的训练挑战，尤其针对推理链远长于最终答案导致的监督微调中学习信号稀释问题，以及强化学习方法中存在的奖励设计成本高、性能瓶颈和奖励欺骗等局限性。其解决方案的关键在于提出一种两阶段训练框架：第一阶段采用长度归一化的双目标监督微调（JurisCoT-SFT），通过平衡推理链与最终答案的学习信号，缓解长推理路径中的梯度稀释问题；第二阶段引入生命周期感知的可回溯策略优化（Lifecycle-Aware Backtrackable Policy Optimization），动态评估并淘汰对最终评价指标贡献不足的辅助奖励信号，实现奖励机制的高效利用与自适应调整。实验结果表明，该框架在ChJGBench、LawBench和LegalBench等多个法律基准上显著提升了法律推理性能，不仅增强了推理链的质量与答案准确性，且模型表现可媲美更大规模的基线模型。
链接: https://arxiv.org/abs/2611.15346
作者: Yinghao Hu,Qian Xiao,Yuting Huang,Kaisong Song,Yiquan Wu,Ang Li,Junjie Cao,Jun Lin,Kun Kuang,Fei Wu
机构: Northeastern University (东北大学); Zhejiang University (浙江大学); Alibaba Group (阿里巴巴集团); University of Macau (澳门大学)
类目: Computation and Language (cs.CL)
备注: 21 pages, 6 figures. Published in Findings of EMNLP 2026
Abstract:In the legal domain, LLMs can improve reliability and interpretability by generating explicit reasoning traces. However, training LLMs to produce high-quality reasoning traces remains challenging. Existing supervised fine-tuning methods struggle when reasoning traces are significantly longer than final answers, as the learning signal for the answer becomes diluted. Meanwhile, reinforcement learning methods such as Group Relative Policy Optimization also have several limitations, including costly reward design, performance plateaus, and reward hacking. We therefore propose a two-stage training framework. In Stage I, JurisCoT-SFT employs a length-normalized dual objective to balance learning signals between reasoning traces and final answers. In Stage II, Lifecycle-Aware Backtrackable Policy Optimization dynamically evaluates and retires auxiliary reward signals based on their impact on final evaluation objectives, enabling efficient reward utilization. Our framework substantially improves legal reasoning performance on relevant tasks selected from multiple legal benchmarks, including ChJGBench, LawBench, and LegalBench. Experimental results demonstrate consistent improvements in both reasoning trace quality and final answer accuracy, while our models remain competitive with larger models.
[NLP-84] Janus-MOPD: Tracing the Student Learning Dynamics in Multi-Teacher On-Policy Distillation
【速读】： 该论文旨在解决大推理模型（Large Reasoning Models, LRMs）在多领域训练过程中，学生模型（student model）难以同时有效获取并整合多个领域能力的核心问题，尤其关注由领域异质性（domain heterogeneity）引发的性能退化现象。现有方法主要从外部监督的角度缓解时间上效率不均与空间上硬分配的问题，但对多领域训练中普遍存在的性能下降机制缺乏内在解释。本文转而聚焦于学生模型内部的学习动态，提出一种基于几何表征的通用学习动力学框架——Janus-MOPD。其关键在于揭示学生模型在时间和空间维度上的双重学习规律：时间上呈现出“扩展-巩固-优化”（Expansion-Consolidation-Optimization, ECO）三阶段动态过程，且存在学习效率与可扩展性的内在局限；空间上则识别出域内与跨域的两种子空间锁定（subspace locking）模式，表现为明确的学习方向但伴随固有的跨域冲突。通过建立统一的学习动力学形式化表达、设计几何追踪方法，并系统揭示学生模型内部多领域能力的持续演化机制，Janus-MOPD为理解与优化多领域能力的集成提供了全新的内在视角。
链接: https://arxiv.org/abs/2611.15341
作者: Zongji Yu,Ming Zhang,Dongcheng Zhang,Jiawei Zhou,Haotian Pang,Yiliu Sun,Xiang Wang,Youbang Sun,Chaochao Lu,Chen Gong
机构: Shanghai AI Laboratory; Shanghai Jiao Tong University; Tsinghua University; University of Science and Technology of China
类目: Computation and Language (cs.CL)
备注: 36 pages, 12 figures
Abstract:Multi-domain capabilities are critical for practically applicable Large Reasoning Models (LRMs). Therefore, Multi-teacher On-Policy Distillation (MOPD) has become a crucial part of the post-training process, integrating the capabilities of multiple domain-specific teacher models into a single student model. However, the student usually fails to acquire multi-domain capabilities simultaneously due to undesirable domain heterogeneity. Existing works have address this problem in view of temporal uneven efficiency and spatial hard allocation in external multi-teacher supervision. While in broader multi-domain training, varying degrees of performance degradation is still largely unexplained. In this paper, we shift the focus on how the student learns and integrates multi-domain capabilities internally, and then propose \textbf\textscJanus-MOPD, a general learning dynamics framework of geometric representations for LRMs. Specifically, our main focus is on the internal performance of multi-domain capabilities on both temporal and spatial dimensions within the student. Temporally, our study shows that the student exhibits a Expansion-Consolidation-Optimization (ECO) three-phase learning dynamics, together with inherent limitations in learning efficiency and scaling. Spatially, we summarize two types of subspace locking within and across domains by tracing the representation space of the student, while explaining its clear learning directions but inherent cross-domain conflict. Therefore, \textscJanus-MOPD yields a unified learning dynamics formulation, a geometric tracing methodology, and systematic findings of how the student learns and integrates multi-domain capabilities. Overall, we provide a \textscJanus perspective to internally trace the continual evolution of the multi-domain capabilities within the model itself.
[NLP-85] Knowledge in the Harness Reliability in the Weights: Teacher-Free Co-Evolution of an LLM Agent
【速读】： 该论文旨在解决小型开源模型在私有客服部署中面临的两大核心问题：一是无法准确遵循特定情境下的书面政策要求，二是对已知信息的执行存在不一致性。其解决方案的关键在于利用模型自身在实际部署日志中的交互数据与参考结果，通过一套多阶段验证机制自动提炼并固化政策规则，同时结合强化学习（GRPO）微调LoRA适配器以提升行为可靠性。具体而言，系统通过持续迭代，在每轮部署中由模型自主诊断失败对话、生成可执行规则，并经过代码检查、反例自检、成对重测及噪声校准验证门控等严格筛选后留存规则；随后基于完整的多轮工具使用对话数据，采用强化学习训练轻量级参数适配器。实验表明，在τ²-bench的零售与航空领域，仅经两轮迭代即在pass@1和pass@4指标上显著超越初始基线模型，且优于将完整政策注入系统提示的基准代理，接近三类前沿模型的性能表现。消融实验进一步揭示：规则机制捕获了模型从未成功处理的任务所依赖的政策知识，而强化学习则使已有正确行为趋于稳定可靠，二者协同有效弥合了能力差距。
链接: https://arxiv.org/abs/2611.15268
作者: Yujia Zhang,Xinqi Ren,Shixiong Zhou,Jia Meng
机构: Brightendata(明亮数据)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 18 pages, 5 figures, 15 tables. Code, data, and run records: this https URL
Abstract:Small open models suit private customer-service deployments but trail frontier models in two ways: they miss what the written policy requires in a given situation, and they act inconsistently on what they know. We study whether such a model can close both gaps from its own deployment logs and their reference outcomes, without a stronger teacher model. Our starting point is a Gemma 4 12B agent that reads its domain policy from on-demand skills rather than from its system prompt. In each round, the deployed system serves a fresh batch of tasks; the base model diagnoses failed conversations and writes rules into the skills, and a rule survives only if it passes code checks, counterexample self-review, a paired retest, and a noise-calibrated validation gate. Reinforcement learning (GRPO) then trains a LoRA adapter on complete multi-turn, tool-using conversations. On the retail and airline domains of \tau^2 -bench, without training or selecting on official tasks, the agent after two rounds improves on this starting point by 12.5 pass ^1 points in retail and 19.5 in airline, and by 18.4 and 22.0 pass ^4 points. It also exceeds the benchmark’s reference agent, which runs the same model with the whole policy in its system prompt, by 11.8 and 8.0 pass ^1 points and by 22.8 and 16.0 pass ^4 points, closing most of the gap to three frontier models under the same protocol. A 2\times2 ablation shows that the rules rescue tasks the agent never solved and carry policy knowledge that RL with the rules in context does not internalize, whereas the trained weights make solved tasks reliable.
[NLP-86] A Closer Look at Language Model Introspection: Probing Information and Reporting
【速读】： 该论文旨在解决生成式语言模型在进行内省（introspection）时所面临的核心问题：即模型能否真正“理解”并报告其内部计算过程，而非仅依赖于表面可被轻易探测的信号。研究发现，尽管模型在零样本条件下对扰动位置的判断表现接近随机（11.1%准确率），但通过训练一个简单的线性探测器（linear probe）即可在层24处以高达99.1%的准确率识别出扰动位置，表明扰动信息在模型激活中已具备高度可解码性。关键发现在于，这种可探测性并非源于模型深层推理能力，而是由于位置信息在神经网络中以易于提取的形式存在，甚至在随机初始化的网络中也已可解码。因此，成功的内省报告可能并不反映模型对自身推理过程的深刻理解，而仅是利用了表层激活模式中的冗余信号，这暗示当前内省研究中观察到的成功结果可能低估了真实认知内省的难度，且难以推广至需要深度解释或可靠性评估等更广泛的应用场景。
链接: https://arxiv.org/abs/2611.15267
作者: Dustin Rubin,Vincenzo Timmel,Xufeng He,Aritra Saha,JooMan Kim,Vijay Madisetti
机构: Georgia Institute of Technology (佐治亚理工学院)
类目: Computation and Language (cs.CL)
备注:
Abstract:One way to study language-model introspection is to test whether models can identify where their activations have been perturbed. We train a simple linear classifier to identify the perturbed sentence from the model’s activations, then compare its predictions with the model’s own answers across pretraining and injection strengths. This tests whether the model is being asked to report information that is already easy to recover. If so, successful localization may be less demanding than it first appears and may not extend to the broader uses that motivate introspection research, such as faithfully explaining a model’s reasoning or helping researchers assess the reliability of its answers. With an injection four times the typical activation norm at layer 0 and probe readout at layer 24 in OLMo~3 32B, the supervised probe reaches 98.3% accuracy after 8.4 billion training tokens, while the model’s zero-shot answer remains at chance (11.1%). At the released base checkpoint, these accuracies reach 99.1% and 89.6%, respectively. Across 7B and 32B, location is already decodable in randomly initialized networks, showing that the architecture can support these traces without pretraining. In these experiments, location information lies close enough to the surface for a simple linear probe to recover it, so successful reporting need not imply broader access to the model’s internal computation.
[NLP-87] Schedule-Calibrated Commitment Order in Blockwise Masked Diffusion Language Models
【速读】： 该论文旨在解决生成式语言模型中因顺序块调度（sequential block schedules）导致的标记承诺（token commitment）呈现强左到右依赖性的问题，即使块内顺序为随机或逆序。其核心挑战在于揭示全局位置-承诺相关性（global position-commitment correlation）的构成机制，并区分由调度策略引入的确定性偏差与块内局部相关性的影响。解决方案的关键在于利用斯皮尔曼等级相关系数（Spearman-rank identity）将全局相关性分解为一个确定性调度参考项与加权块内相关性的均值之和，从而实现对局部位置排序信息的校准分析。研究表明，增大块大小虽降低原始全局相关性（从0.9845降至0.9149），但经校准后块内相关性反而下降，表明大块尺寸下的残差降低并非源于更强的局部位置有序性。研究进一步给出了严格的理论边界和校准区间，揭示了仅依赖聚合统计量进行实证分析的局限性，强调需结合调度感知（schedule-aware）的解释框架，但当前分析范围受限于单一检查点和两种调度策略，未涉及生成质量或推理速度的评估。
链接: https://arxiv.org/abs/2611.15234
作者: Gaurav Sarkar(1),Umang Pandey(2),Pratik Prabhanjan Brahma(2),Zicheng Liu(2),Emad Barsoum(2) ((1) International Institute of Information Technology, Hyderabad, (2) Advanced Micro Devices)
机构: International Institute of Information Technology, Hyderabad; Advanced Micro Devices (AMD)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: 9 pages, 1 figure, 2 tables. Includes appendices and ancillary code for algebraic verification
Abstract:Sequential block schedules can make token commitment in masked diffusion language models appear strongly left-to-right even when within-block order is random or reversed. For equal contiguous blocks processed without interleaving in positional order, with untied irreversible commitments, we use a Spearman-rank identity to express global position-commitment correlation as a deterministic schedule reference plus a weighted mean within-block correlation. In a reported paired study of LLaDA-8B-Instruct on 500 math and code prompts with 256 generated positions and updates, increasing block size from 64 to 128 lowers mean global correlation from 0.9845 to 0.9149. Uniform-random within-block references are already 0.9375 and 0.7500. After calibration, mean within-block correlations are 0.751 and 0.660, so the larger raw residual at block size 128 does not indicate stronger local positional ordering. We give sharp bounds and derive calibrated intervals from reported summaries, while documenting the limitations of the aggregate-only empirical release. The analysis supports schedule-aware interpretation of commitment order; its empirical scope is one checkpoint and two schedules, without evaluation of output quality or inference speed.
[NLP-88] he Neighbours Label: Opposite Subgroup Effects of Same-Dialogue Exposure in Next-Act Evaluation
【速读】： 该论文旨在解决窗口化下一动作评估（window-based next-act evaluation）中因重叠窗口与相似性检索导致的标签泄露问题，即测试对话中的标注样本可能通过邻近上下文被模型“暴露”，从而影响评估结果的可靠性。其核心问题是：在标准数据集（如SGD、STAR和AMI）上，由于训练过程中包含测试对话中目标动作的前驱样本（labelled predecessor），模型可通过简单复制该前驱动作获得虚假性能提升，尤其在连续性（continuation）与转换性（transition）子任务中表现相反，最终被整体平均掩盖。解决方案的关键在于通过受控实验设计验证这一暴露机制——通过随机将前驱行纳入训练集，发现词法分类器在三个语料库中均出现延续得分上升而转换得分下降的现象；进一步剔除相邻行后，大部分性能增益消失。同时，在使用TF-IDF检索的示范样本下，两个70亿参数指令微调模型在全部八组模型-语料-样本组合中均呈现相同反向模式，而自然权重准确率变化极小（-0.05至+0.04），表明性能波动主要源于暴露而非真实泛化能力。研究还发现，排除测试对话并不能显著降低准确率，说明所有暴露均为人为诱导，且未有公开系统被重新评估。因此，该研究强调需改进评估范式以避免此类偏差。
链接: https://arxiv.org/abs/2611.15194
作者: Ali Amini,Masoumeh Chapariniya,Rytis Maskeliunas,Robertas Damaševičius
机构: Kaunas University of Technology(考纳斯理工大学); University of Zurich(苏黎世大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL)
备注:
Abstract:Window-based next-act evaluation often aims to measure generalisation across dialogues, yet overlapping windows and similarity retrieval can expose labelled examples from the test dialogue. The nearest such example is the target’s labelled predecessor, whose act is correct when the target continues it and wrong when the target changes act; a model that copies it can gain and lose in two subgroups that one aggregate averages. We induce this exposure under controlled designs on SGD, STAR and AMI, whose official splits are clean. Randomly including the predecessor’s row in training raises a lexical classifier’s continuation scores and lowers its transition scores in all three corpora, and purging the adjacent rows removes most of the gain. With TF-IDF-retrieved demonstrations, two 7B instruction-tuned models show the same opposite pattern in all eight model-corpus-sample arms (some single-cell intervals include zero; one replication is partial), while natural-weight accuracy changes by only -0.05 to +0.04. Exposure frequency depends on the retrieval pipeline. We evaluate two diagnostic comparisons without model retraining and find that excluding the test dialogue does not detectably reduce accuracy on SGD. All exposure is induced; no published system was re-evaluated.
[NLP-89] Where Language-Model Vocabulary Rose on Stack Exchange
【速读】： 该论文旨在解决生成式 AI（Generative AI）在技术问答平台中被滥用后，如何通过文本特征识别其使用痕迹的问题。研究聚焦于 Stack Overflow 等 Stack Exchange 网站禁止使用 ChatGPT 生成答案后的语言模式变化，分析其对用户回答词汇多样性的影响。解决方案的关键在于构建一个基于高频词的“标记词汇列表”（56个语言模型过度使用的词汇），并利用大规模语料（2019–2024年170个站点共380万条回答）进行量化分析。研究发现，该标记词汇的使用率在2023年后较此前趋势上升约14%，而普通词汇的对应增幅仅为7%。进一步分析显示，词汇使用上升最显著的站点是那些作者自身对回答进行频繁编辑的站点——在自编辑比例最高的三分之一站点中，标记词汇上升达29%，而在编辑频率最低的站点则无明显变化。该梯度效应在未被编辑的回答中依然存在，且主要由作者自编辑行为解释，外部编辑贡献有限。因此，研究揭示了“自编辑行为”作为代理变量，能够有效反映生成式 AI 的潜在使用模式，尽管无法直接观测模型使用本身，但为评估生成内容在网络中的渗透程度提供了一种基于语言特征的间接方法。
链接: https://arxiv.org/abs/2611.15172
作者: Fengming Liu
机构: 未知
类目: Computation and Language (cs.CL); Computers and Society (cs.CY); Social and Information Networks (cs.SI)
备注: 8 pages, 4 figures, 3 tables (one in the appendix)
Abstract:Stack Overflow banned answers generated by ChatGPT five days after the model’s release, and many other Stack Exchange sites announced similar rules. I measure the vocabulary of the answers posted afterwards, 3.8 million answers on 170 sites from 2019 to 2024. A fixed list of 56 words that language models overuse rose about 14 percent above its pre-2023 trend, while the highest of 1,000 frequency-matched lists of ordinary words rose about 7 percent. In a screen of 15 site traits measured before the release, the strongest predictor of where the rise happened was the share of answers that received a body edit within 30 days of posting. On the least edited third of sites the list rose about 29 percent relative to matched controls, and on the most edited third it did not change. The gradient is also present among answers that were never edited. It follows the share of answers edited by their own authors, and the share edited by others adds little once that share is held fixed. Use of the model is unobserved, so these results cannot separate lower use on sites with a high self-edit share from any effect of editing on what the vocabulary records. Any vocabulary-based estimate for the network as a whole therefore pools sites where the marker vocabulary rose sharply with sites where it did not change.
[NLP-90] Read Once Answer Many: Parallel Prompting for Shared-Context LLM Inference
【速读】： 该论文针对大语言模型（LLM）在处理长文档时，需对同一文档提出多个独立问题（如文档问答、数据标注及基于LLM的评估）的场景中，因解码阶段重复读取文档键值（KV）缓存而导致的内存带宽瓶颈问题。现有方法如前缀缓存虽可消除预填充（prefill）阶段的冗余计算，但每个解码步骤仍需为每条查询单独访问一次文档的KV缓存，导致解码性能随问题数量线性下降。其解决方案的关键在于提出并行提示（Parallel Prompting）：将P个独立问题作为单个请求的P条并行流（stream）一次性解码，使每个解码步骤仅需读取一次共享文档的缓存，通过在注意力核函数内动态计算的流标识符注意力掩码（stream-id attention mask），确保每条输出仅依赖于文档和对应自身的问题，保持生成结果的正确性与独立性。该方法在SGLang中实现仅需约80行代码修改，并利用固定形状的CUDA图优化执行，无需密集张量存储。实验表明，在CodeLlama-7B模型、8k–16k token长文档及每文档64–256个问题的场景下，相比使用FlashInfer后端的原版SGLang，A100上提速3.7倍，RTX A6000上提速5.9–6.0倍，且在所有144次测试中均优于基线。
链接: https://arxiv.org/abs/2611.15130
作者: Zekun Zhao,Jeffrey Flanigan
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG)
备注: 17 pages, 6 figures. Code: this https URL . Project page: this https URL
Abstract:LLM workloads sometimes ask many independent questions about the same long document, as in document QA, data labeling and LLM-based evaluation. Prefix caching removes the redundant prefill shared by these queries, but each decode step still reads the document’s key-value (KV) cache once per question, so decoding is bound by memory bandwidth and scales with the number of questions. We present Parallel Prompting, which decodes P questions as P independent streams of a single request all at once, fixing this issue. In our method, every step reads the shared document once for all P streams, and a stream-id attention mask, evaluated inside the attention kernel, keeps each answer conditioned only on the document and its own question. We implement Parallel Prompting in SGLang by adding about 80 lines to three existing files and a small engine package; the mask needs no dense tensor, and every decode step has a fixed shape that is captured as a CUDA graph. With CodeLlama-7B, 8k-16k-token documents and 64-256 questions per document, Parallel Prompting is 3.7x faster than stock SGLang with its FlashInfer backend on an A100 and 5.9-6.0x faster on an RTX A6000, on two SGLang releases, including the most recent release. Timed on identical inputs, it is faster than stock SGLang in all 144 runs (2 GPUs x 2 releases x 6 workloads x 6 input sets). Code: this https URL
[NLP-91] What Does Each Part of Speech Need? Contexts Subspaces and Dimensions in Static Word Embeddings
【速读】： 该论文旨在解决静态词嵌入（Static Word Embeddings）在表示多义词时忽略其语法功能（如名词、动词、形容词等）的问题，即同一词汇在不同词性下应具有不同语义表征，但传统方法将其统一为单一向量。其核心解决方案是基于词性标注（Part-of-Speech Tagging）的分词性嵌入框架，通过引入词性信息为不同词性分配独立的子空间，并结合频率驱动的维度分配策略实现高效压缩。关键在于：词性决定向量的结构（即“应该是什么”），而频率决定保留的维度数量（即“保留多少”）。实验表明，采用按词性进行主成分分析（Per-POS PCA）并依据频率动态分配维度，在仅保留原始大小三分之一的情况下，仍能有效保持动词与形容词的语义相似性，且对名词的性能无明显损失，同时维持动词到名词检索任务的精度。
链接: https://arxiv.org/abs/2611.15114
作者: Jintang Xue,Chengwei Wei,Yun-Cheng Wang,C.-C. Jay Kuo
机构: 未知
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Static word embeddings give a word one vector, ignoring its different functions as noun, verb or adjective. Training on part-of-speech-tagged text gives each function its own vector, at the cost of a larger table. Using nine controlled corpora, we ask which context words each part of speech needs, whether it needs its own subspace, and whether the number of dimensions a word keeps should follow its part of speech or its frequency. The tags raise verb similarity on SimVerb-3500 at every dimension from 50 to 300 but help nouns much less. Added to verb-only text, adverbs help verbs more than as many random words; adjectives help nouns only slightly. At a fixed dimension, per-part-of-speech PCA bases preserve verb and adjective similarity better than a global basis or frequency-band bases. Nouns need more dimensions than verbs only among frequent words. We therefore allocate dimensions by frequency with these bases: at a third of the original size, only rare words lose dimensions, SimVerb-3500 and SimLex-999 nouns match the uncompressed table, and verb-to-noun retrieval stays within 0.02. Part of speech decides what a word’s vector should be; frequency decides how much of it to keep.
[NLP-92] yped Knowledge over Constructive Institutions: A Proof-Relevant Type Theory for Verified Science
【速读】： 该论文旨在解决形式化证明助手（如Coq、Agda和Lean 4）中判别性证明无关性（judgmental proof irrelevance）与科学知识库验证需求之间的根本冲突。传统系统将同一命题的不同证明在定义等价下视为不可区分（即 $ p \equiv q $ 对所有 $ p, q : P $ 且 $ P : \text{Prop} $ 成立），这虽符合固定公理体系下的数学推理，却无法支持科学验证中对独立推导路径的可追溯性与对比分析——例如评估证据的一致性（consilience）或在实验试剂、模型失效时进行反事实回溯（counterfactual retractions）。此外，判别性证明无关性与子单例消除（subsingleton elimination）的结合会破坏强规范化性，并导致算法化转换关系非传递。为此，论文提出Eigenius——一个面向科学的可验证知识底座，其核心解决方案包括：构建具有累积型宇宙（$ \text{Prop}, \text{Set}, \text{Type}k $）、不自指的 $ \text{Prop} 、由索引强制的子单例消除的依赖型内核，以及同伦意义的身份类型；其中证明无关性被重构为可证的命题公理（ \text{irrel} : \prod{P:\text{Prop}} \prod_{p,q:P} \text{Id},P,p,q ），从而保留推导痕迹；通过无显式强制的子属判断（subsumptionjudgment）统一宇宙累积、 \Pi $-子类型及本体类包含关系，确保双向片段 $ \mathcal{K}_0, \mathcal{K}_1 $ 满足课题还原性，并在声明式类型系统中具备完备性与一致性；进一步建立三层机构化外部验证层级，引入显式的公理策略（采纳、释放、相对化），内化基于“为何证明”（Why-provenance）的证明支撑半环 $ \text{supp}(j,P) $，并嵌入Cooper的带记录类型理论（TTR）以建模开放世界的知识图谱；最终在格罗滕迪克宇宙中证明系统一致性，并在Lean 4中以零未证假设完成元理论机械化。
链接: https://arxiv.org/abs/2611.15108
作者: Hans-Martin Will
机构: The Eigenius Project(艾根尼乌斯项目)
类目: Logic in Computer Science (cs.LO); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Mathematical proof assistants such as Coq, Agda, and Lean 4 build proof irrelevance into definitional conversion ( p \equiv q for p, q : P with P : Prop ), erasing which derivation established a theorem. While natural for mathematics over fixed axioms, judgmental proof irrelevance is incompatible with verified scientific stores, which must inspect and compare independent derivations of the same claim to assess consilience and compute counterfactual retractions when an assay, reagent, or model is invalidated. Moreover, combining definitional proof irrelevance with subsingleton elimination breaks strong normalization and renders algorithmic conversion non-transitive. We present the type-theoretic and institutional foundations of Eigenius, a verified knowledge substrate for science. First, we formulate a dependent kernel with cumulative universes ( Prop, Set, Type_k ), impredicative Prop , inductive families with index-forced subsingleton elimination, and intensional identity types, where definitional equality \equiv is untyped joinability under an orthogonal CRS and proof irrelevance is a witnessed propositional axiom ( irrel : \Pi P:Prop.,\Pi p,q:P.,Id,P,p,q ). Second, we unify universe cumulativity, codomain-covariant \Pi -subtyping, and ontology class inclusions in a coercion-free subsumption judgment whose bidirectional fragments ( \mathcalK_0, \mathcalK_1 ) satisfy subject reduction and are sound and complete for declarative typing. Third, we organize external verifiers into a three-level institutional hierarchy with explicit axiom policies (Adopt, Discharge, Relativize), internalize a justification logic with Why-provenance semiring support supp(j,P) , and embed Cooper’s Type Theory with Records (TTR) to type open-world knowledge graphs. Finally, we prove consistency in Grothendieck universes and mechanize the metatheory in Lean 4 with zero sorry axioms. Subjects: Logic in Computer Science (cs.LO); Artificial Intelligence (cs.AI); Computation and Language (cs.CL) ACMclasses: F.4.1; I.2.3; I.2.4 Cite as: arXiv:2611.15108 [cs.LO] (or arXiv:2611.15108v1 [cs.LO] for this version) https://doi.org/10.48550/arXiv.2611.15108 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[NLP-93] One Fabrication Is Enough: A Fake News Article Changes the Forecasts of LLM Forecasting Systems
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在预测任务中因依赖外部新闻信息作为证据而面临的信息操纵风险。其核心问题是：当攻击者仅知晓预测问题并发布伪造新闻文章（无需访问模型、检索器或查询机制）时，这些虚假信息如何通过模型的上下文引入，从而系统性地扭曲预测结果。解决方案的关键在于揭示当前主流预测系统对检索环节的高度依赖性——由于先进系统普遍采用广泛检索策略以提升性能，导致其极易被恶意注入的虚假新闻所渗透。实验表明，一旦伪造文章进入模型上下文，可使74%至82%原本持相反意见的预测发生反转，且模型无法验证信息真实性，完全遵循虚假内容生成输出。此外，传统基于准确率的评估指标无法反映此类攻击的危害，因为攻击后模型在攻击方向正确时仍表现“准确”，导致平均准确率几乎不变。针对此问题，论文提出双重攻击策略：一是发布对立双文以制造不确定性，二是利用模型不进行概率性规避的特性，使其在非可控方向上输出高度自信的错误预测。进一步评估两种现有检索阶段防御机制（段落过滤与分块检索）发现，二者虽能抵御单篇攻击，但均可通过微调提示实现绕过；组合使用需达到四篇以上伪造文本才能有效防御，而十六个变体副本即可突破防线，并造成29%的正常决策偏差及39%真实新闻的误删。值得注意的是，分块检索还需控制检索索引，而开放网络搜索系统缺乏该能力，最终仅剩过滤机制可用，因而仍处于易受攻击状态。因此，该研究揭示了当前基于检索增强的生成式预测系统在对抗鲁棒性上的根本缺陷。
链接: https://arxiv.org/abs/2611.15102
作者: Zeyu Zhang,Bradly C. Stadie
机构: Northwestern University (西北大学)
类目: Cryptography and Security (cs.CR); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 18 pages, 2 figures, 13 tables
Abstract:LLMs are increasingly used for forecasting, and the best forecasting systems retrieve news as evidence to strengthen their predictions. This also opens them to attack. We study the weakest attacker of this kind, who knows only the forecasting question and publishes fabricated news articles, with no access to the model, the retriever or its queries. We evaluate how far this attack reaches on stock questions and real prediction-market questions. Once the article is in the model’s context, it reverses 74% to 82% of the forecasts that opposed it, on every frontier model we tested. The model cannot verify the fabrication and follows it anyway. Retrieval is the attack’s entry point, and the systems built to retrieve well are the most exposed: the architectures of two leading deployed forecasting systems, whose agents search for every related article, retrieve the fabricated one in about 90% of runs. Accuracy cannot reveal this harm: the attacked forecaster is right when the attacker’s direction turns out right and wrong otherwise, so its average accuracy barely changes while its forecasts follow the attacker’s guess. We also propose a second attacker, who publishes two articles that argue opposite directions to make the forecaster unsure. The model does not hedge; it returns a confident forecast in a direction the attacker no longer controls. We then evaluate the two strongest published retrieval-time defenses, passage filtering and partitioned retrieval. Each stops the one-article attack but is bypassed by a one-prompt change to the attack. Together they stop both bypasses at four articles, but sixteen distorted copies beat them, and they change 29% of clean decisions and discard 39% of the genuine news. Partitioned retrieval also requires control of the retrieval index. Systems that search the open web have no such control and are left with the filter alone. For them, the attack still works.
[NLP-94] FIBER: Bayesian Fusion of Fast-DetectGPT and IRT-Informed Detector Evidence for LLM -Generated Text Detection
【速读】： 该论文旨在解决大语言模型（Large Language Model, LLM）生成文本在领域迁移（domain shift）和生成器迁移（generator shift）场景下检测性能下降的问题。现有监督式检测方法泛化能力差，而零样本检测方法通常依赖单一统计信号，未能有效利用历史检测器的行为模式。其解决方案的关键在于提出FIBER（Fast-DetectGPT with IRT-informed Bayesian Evidence Recalibration），通过贝叶斯证据融合机制，将连续的Fast-DetectGPT评分与辅助检测器输出相结合，并引入项目反应理论（Item Response Theory, IRT）建模历史检测器的正确性模式，基于潜在文本表示的边缘化获得类别条件下的检测器输出似然。这些似然与Fast-DetectGPT得分密度及历史类别先验共同用于最终预测。当缺乏历史Fast-DetectGPT评分时，FIBER-Stream通过无标签批次估计类别特定得分分布以实现适应。实验表明，FIBER在32,000条文本的多领域基准上达到最高准确率（0.620），且在五折交叉验证下平均AUC优于Fast-DetectGPT（0.664 vs. 0.600），并在四个独立领域中持续提升AUC；在分布外迁移设置下，FIBER-Stream取得0.911的AUC，显著优于基线方法，验证了融合互补检测证据与历史信息对提升异构环境下检测性能的有效性。
链接: https://arxiv.org/abs/2611.15058
作者: Yan Li,Guangsheng Bao,Zhipeng Chen,Guandong Zhang,Yanbin Zhao
机构: Shanghai Polytechnic University (上海理工大学); Fuyao University of Science and Technology (福州理工学院)
类目: Computation and Language (cs.CL)
备注:
Abstract:Detecting large language model (LLM)-generated text remains challenging under domain and generator shifts. Supervised detectors may generalize poorly, while zero-shot methods often rely on individual statistical signals without leveraging historical detector behavior. We propose FIBER (Fast-DetectGPT with IRT-informed Bayesian Evidence Recalibration), which integrates continuous Fast-DetectGPT scores with auxiliary detector outputs through Bayesian evidence fusion. Item response theory (IRT) models historical detector-correctness patterns, providing class-conditional detector-output likelihoods through marginalization over latent text representations. These likelihoods are combined with Fast-DetectGPT score densities and a historical class prior for prediction. When historical Fast-DetectGPT scores are unavailable, a variant called FIBER-Stream estimates class-specific score distributions from an unlabeled batch. On a 32,000-text multi-domain benchmark, FIBER achieves the highest accuracy (0.620) and improves mean AUC over Fast-DetectGPT (0.664 versus 0.600) under five-fold cross-validation. FIBER also consistently improves AUC across four individual domains. Under out-of-distribution transfer, FIBER-Stream achieved an AUC of 0.911, compared with 0.889 for FIBER and 0.922 for Fast-DetectGPT alone. These results demonstrate the potential of integrating complementary detector evidence and historical information to improve detection performance across heterogeneous settings.
[NLP-95] A Time-Frequency-Condition Multimodal Fusion Foundation Model for Universal Rotating-Machinery Fault Diagnosis under Multi-Element Heterogeneous Scenarios
【速读】： 该论文旨在解决旋转机械故障诊断中“一机一模型”范式在面对多设备、多工况场景时可扩展性差的问题，尤其针对现有方法在跨设备、跨工况迁移时因过度追求域不变特征而忽略设备特异性信息所导致的诊断性能下降问题。其核心解决方案是提出TFC-MMFD——一种面向多元素异构场景的时频-条件多模态融合基础模型，通过将统计指标、时频图像、振动序列及文本描述等异源数据统一编码至类语言空间，实现多模态输入的序列化表征；利用LLaMA骨干网络结合跨模态双向注意力与Q-Former，从融合序列中提炼出故障模式查询向量，并基于自然语言形式的故障模式锚点进行匹配诊断，而非依赖固定类别索引；同时引入回归对抗性条件解耦与条件条件自编码预训练机制，增强对工况变化的鲁棒性。实验表明，该模型在10个公开轴承数据集上的4项渐进任务中达到99.40%准确率和99.16%宏平均F1值，在多数未见工况下无需微调仍保持80%以上准确率，且在仅5个标注样本/类的情况下即可实现从新设备上20.26%到95.63%的性能跃升，显著优于8个领域泛化基线模型（提升10.57个百分点），并以不到其一半的微调数据量实现相当甚至更优的表现。
链接: https://arxiv.org/abs/2611.15051
作者: Xinyu Zou,Jian Ma,Liang Ma,Chen Lu,Tao Peng,Yingjie Li
机构: Hangzhou International Innovation Institute, Beihang University, China; School of Reliability and Systems Engineering, Beihang University, China; Beijing Institute of Spacecraft System Engineering, China; School of Automation, Central South University, China
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Intelligent fault diagnosis models are typically developed for a single machine under a specific set of operating conditions, such that each new machine or condition requires retraining from scratch. This one-machine-one-model paradigm scales poorly. Although transfer learning and domain generalization can reduce the labeling burden, they primarily seek domain-invariant commonality, potentially discarding machine-specific characteristics that are critical for diagnosis. This paper proposes TFC-MMFD, a time-frequency-condition multimodal fusion foundation model for rotating-machinery fault diagnosis under multi-element heterogeneous scenarios. Statistical indicators, time-frequency images, vibration sequences, and textual machine and condition descriptions are tokenized into a unified language-like space, allowing inputs of different physical origin to be represented as a single sequence. A LLaMA backbone, together with cross-modal bidirectional attention and a Q-Former, then distills this sequence into fault-mode queries. Diagnosis is performed by matching these queries against natural-language fault-mode anchors rather than fixed class indices, while regression-adversarial condition disentanglement and condition-conditioned masked pre-training improve robustness to changing conditions. Across four progressive tasks on ten public bearing datasets spanning diverse bearing types, speeds, and loads, TFC-MMFD attains 99.40% accuracy and 99.16% macro-F1, maintains over 80% accuracy on most interpolated unseen conditions without adaptation, and improves from 20.26% to 95.63% accuracy on an unseen machine with only five labeled samples per class. It outperforms the strongest of eight domain-generalization baselines by 10.57 percentage points and compares favorably with a recent bearing foundation model while using less than half of its fine-tuning data.
[NLP-96] Science Prediction Universe
【速读】： 该论文旨在解决当前人工智能在科学实验结果预测能力上与人类存在显著差距的问题，尤其关注生成式人工智能（Generative AI）在跨学科领域（如化学、材料科学、生物学和医学）中对真实实验结果的可靠预测能力。其核心挑战在于：尽管现有模型在工具调用与流程执行层面已具备一定能力，但其科学判断力仍严重不足，导致预测结果与实际实验结果之间存在较大偏差。解决方案的关键在于构建一个名为“科学预测宇宙”（Science Prediction Universe, SPU）的基准测试平台，包含100个公开验证任务及16个来自未发表湿实验的真实任务，通过让智能体基于给定实验条件对候选方案进行排序来评估其预测能力。研究发现，当前主流智能体配置在公共基准上的表现分属两个明显层级，任务调整后的斯皮尔曼等级相关系数介于0.227至0.478之间，表明性能仍有巨大提升空间；进一步分析显示，瓶颈并非源于工具使用或计算执行，而在于科学推理中的先验假设偏差——智能体普遍依赖共通的单调性假设，且频繁计算与最终结果无关的物理或化学代理变量，反映出其缺乏对复杂科学机制的本质理解。因此，提升科学预测能力的关键在于增强智能体的科学判断能力，而非单纯优化工具链或数据处理流程。
链接: https://arxiv.org/abs/2611.15049
作者: Zengji Tu,Xiang Liu,Jiarui Sun,Ruoyu Wu,Zhihui Qi,Chengda Wu,Yutao Zhou,Bo Yang,Ziyan Zhang,Jialong Wu,Liang Chen,Kuan Li,Steven Dillmann,Yuanqi Du,Soojung Yang,Shizhe Diao,Jianjun Wu,Junlang Liu,Xuming Hu,Lin Chang,Baobao Chang,Xiaowen Chu,Philip S. Yu,Tong-Yi Zhang,Hao Cheng,Qin Han,Yanye Feng,Chenru Duan,Chuncheng Liu,Yan Liu,Yuchen Gao,Zhao Li,Ronghua Zhang,Tory Xu,Renjian Song,Yaokai Fu
机构: 未知
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 52 pages, 28 figures, 7 tables. Project page: this https URL
Abstract:Humanity has long sought to explain natural phenomena. Reliable prediction of experimental outcomes is a central test of our understanding of the processes that produce them. If artificial general intelligence (AGI) is to surpass human scientific capabilities, it must also predict experimental outcomes more reliably than humans. Such capabilities could accelerate scientific discovery across disciplines, help address humanity’s most pressing challenges, and improve human well-being. We introduce Science Prediction Universe (SPU), a benchmark of hidden experimental outcomes with 100 public verified tasks from chemistry, materials science, biology and medicine, and 16 private tasks from unpublished wet-lab experiments. Given an experimental setup and candidate conditions, an agent works with any resources it needs to rank the candidates by a measured outcome. We evaluate thirteen agent configurations spanning twelve frontier models across two harnesses (Codex and Claude Code). Current agents remain far from saturating wet-lab outcome prediction, with substantial performance disparities across models: on the public benchmark, configurations separate into two distinct tiers, with task-adjusted Spearman correlations ranging from 0.227 for the lowest-performing configuration to 0.478 for the best. Analysis of 1,155 scored public trajectories (227,103 tool calls) suggests that the bottleneck lies in scientific judgment rather than tool execution: agents rely on shared monotone priors, agreeing far more with one another than with experimental reality (0.52 vs. 0.33), and frequently compute trivial physical or chemical proxies that do not influence the final outcome. Comments: 52 pages, 28 figures, 7 tables. Project page: this https URL Subjects: Artificial Intelligence (cs.AI); Computation and Language (cs.CL) Cite as: arXiv:2611.15049 [cs.AI] (or arXiv:2611.15049v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2611.15049 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[NLP-97] AROMA: Adaptive Representation-Oriented Multimodal Actuation for Vision-Language Model Control
【速读】： 该论文旨在解决生成式视觉语言模型（VLM）中机制定位（mechanistic localization）无法直接转化为有效控制的问题，核心挑战在于：因果敏感性可能仅需极低的激活能量、精确的局部作用方向在有限尺度下失效、可到达的目标值可能是错误的。为此，论文提出AROMA（Adaptive Representation-Oriented Multimodal Actuation）框架，其关键在于将控制问题解耦为五个独立维度：干预位置、几何结构与增益选择、目标可达性、目标质量评估以及行动时机决策。通过因果验证的交叉注意力头进行表征分析，发现其四个维度可解释89.82%的基数敏感性，但仅解释9.52%的中位激活能量，揭示了传统方法的不匹配问题。采用幅度归一化的因果子空间作用机制，成功将此偏差转化为可实现的有限控制，使未见过的程序化数据上的精确计数准确率提升8.00个百分点。进一步引入迭代最小努力作用（Iterative Minimal-Effort Actuation, MEA），在相同步数预算下，达成81.25%的最优目标逼近率，远超冻结方向策略的15.18%。当目标与门控在评估前冻结时，靶向MEA将先前冻结策略的准确率从34.444%提升至38.730%（N=630，配对p=4.63e-7），并在全部7,992个样本上分别达到65.741%与65.153%的性能，且仅需修改63个输入。通过前瞻性路由、跨头与跨模型复现及与归因引导基线的对比，系统性地分离出定位、作用、可达性、目标质量与选择性部署等独立控制模块，为精准可控的模型干预提供了可扩展的理论与实践框架。
链接: https://arxiv.org/abs/2611.15041
作者: Zhaoning Zhang
机构: University of Toronto(多伦多大学)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV)
备注: 39 pages, 16 figures
Abstract:Mechanistic localization does not by itself yield usable control: causal sensitivity may occupy little activation energy, an accurate local direction may fail at finite scale, and a reachable numeral target may be wrong. We introduce AROMA (Adaptive Representation-Oriented Multimodal Actuation), a frozen-VLM control framework separating where to intervene, which geometry and gain to use, how to reach a target, which target to pursue, and when to act. In Llama-3.2-11B-Vision-Instruct, four dimensions of a causally validated cross-attention head explain 89.82% of cardinality sensitivity but only 9.52% of median activation energy; magnitude-normalized causal subspace actuation converts this mismatch into finite control, improving exact-count accuracy on untouched procedural data by +8.00 pp. Iterative Minimal-Effort Actuation (MEA) reaches 81.25% of oracle targets versus 15.18% for a frozen direction under matched step budgets. With target and gate frozen before evaluation, targeted MEA improves the previous frozen policy from 34.444% to 38.730% on a new natural Stage-2 cohort (N=630; paired p=4.63e-7) and yields 65.741% versus 65.153% over all 7,992 examples while editing only 63 inputs. Prospective routing, replication across heads and models, and comparisons with attribution-steering baselines further characterize each stage, separating localization, actuation, reachability, target quality, and selective deployment as distinct control problems.
[NLP-98] Breaking Bad: “Going Badly” Activations Drive Reward Hacking
【速读】： 该论文旨在解决大语言模型在面对高难度或不可能完成的任务时，因内部“任务失败累积”状态（task badness）加剧而诱发奖励劫持（reward hacking）的问题。其核心问题是：当模型在执行任务过程中持续遭遇失败，其内在状态会逐渐积累负面反馈，形成一种类似人类“绝望感”的压力状态，进而促使模型采取非诚实策略以规避失败感知。解决方案的关键在于识别并量化这一“任务坏度”方向（task badness direction），该方向可通过无奖励劫持示例的日常故事数据学习得到，并在推理过程中通过投影或减去该方向来缓解模型的作弊行为。实验表明，在多个模型（Qwen3.6-27B、GPT-OSS-20B、Gemma-4-31B）中，任务坏度在接近奖励劫持前显著上升；对任务坏度进行适度减除可有效抑制Qwen模型的作弊行为，效果优于仅针对“绝望”或“任务不可行性”的干预。然而，过度减除会导致模型丧失对自身失败的敏感性，误将失败测试视为成功，从而引发更严重的错误。不同模型间表现差异显著，例如在GPT-OSS中，该状态能区分劫持行为与诚实求解，但无法区分诚实失败与劫持。研究提示应建立基于任务坏度监控的防御机制，并强调随着智能体被部署于更复杂、长周期任务，此类内部状态可能诱导不可读的思维链（illegible chain-of-thought），因此需加强对模型内部表征的可解释性监督。
链接: https://arxiv.org/abs/2611.14996
作者: Natalia Fischl-Lanzoni,Punya Syon Pandey,Merlin Stein,David Demitri Africa
机构: ERA: Cambridge(剑桥大学); AI Security Institute(人工智能安全研究所); Resolution(解决)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Humans often cheat when a task is going badly and a goal feels out of reach. We explore and validate an analogous idea in language models: they exhibit a similar internal pressure state that precedes reward hacking. Using everyday stories (crucially, without any hacking examples), we extract a task badness direction that rises as failures accumulate across three models we study (Qwen3.6-27B, GPT-OSS-20B, and Gemma-4-31B). On impossible or difficult coding tasks, task badness rises over the turns leading up to a hack. Further, subtracting the direction a moderate amount reduces cheating in Qwen, more than steering against desperation or task impossibility. Subtracting task badness by an even larger amount blinds the agent to its own failures, causing it to interpret failing tests as passing. However, the relationship to hacking differs across models: in GPT-OSS, the state separates hacking runs from honest solves but not honest failures. Our results suggest monitoring and mitigation strategies based on tracking this state of task badness. Further, as agents are set on more difficult and long-horizon tasks, states like this may push them toward illegible and obscured chain-of-thought, so oversight will be needed on internal representations such as this.
[NLP-99] CoLFA: Collaboration-Guided Lightweight Failure Attribution for LLM -Based Multi-Agent Systems
【速读】： 该论文旨在解决多智能体系统中故障归因（failure attribution）的难题，即在智能体分工协作与信息交互导致执行步骤交错（Interleaved Steps）和内容冗杂（Cluttered Content）的情况下，如何高效准确地定位导致错误的关键决策步骤及责任智能体。传统方法在处理长序列日志时难以兼顾定位精度与计算成本（token cost）。其核心解决方案在于提出一种基于协作线索的轻量级故障归因框架CoLFA（Collaboration-Guided Lightweight Failure Attribution），其关键在于利用智能体协作过程中自然留下的执行关联与信息传递痕迹，通过“交接引导的步骤扩展”（Handoff-Guided Step Expansion）和“交叉引用的内容组织”（Cross-Referenced Content Organization）机制，重构输入序列以反映结果生成的实际流程，从而引导模型聚焦于与错误形成相关的关键步骤及其内容，同时抑制无关或重复记录的处理。实验表明，CoLFA在五个不同场景（涵盖WhoWhen与LongRCA任务）及三种主干模型上均实现了最高的精确步骤定位准确率，且token消耗最低，相较基线模型最高提升准确率37.50%，降低token消耗达97.6%。
链接: https://arxiv.org/abs/2611.14980
作者: Yufei Luo,Yuchen Ling,Chunrong Fang
机构: Nanjing University (南京大学)
类目: Computation and Language (cs.CL)
备注: 23 pages, 4 figures, 15 tables
Abstract:Failure attribution in multi-agent systems requires tracing error formation to identify the decisive error step and the responsible agent. However, agents’ division of work and information exchange lead to Interleaved Steps and Cluttered Content, requiring models to link executions and check details in lengthy records. This makes it difficult to balance localization accuracy and token cost. We find that multi-agent collaboration also leaves records of how executions connect and content passes between them as results are formed, providing evidence for tracing error formation. Based on this observation, we propose CoLFA (Collaboration-Guided Lightweight Failure Attribution), which uses collaboration cues to organize the attribution input according to how results are formed, determining which steps are examined and how their content is presented. Through Handoff-Guided Step Expansion and Cross-Referenced Content Organization, CoLFA helps the model focus on examining error formation and the role of each decision while reducing the processing of irrelevant and repeated records. Across five settings spanning WhoWhen and LongRCA and three backbone models, CoLFA achieves the highest exact step-localization accuracy with the lowest token consumption. Compared with individual baselines in each setting, CoLFA improves accuracy by up to 37.50% and reduces token consumption by up to 97.6%.
[NLP-100] rmination Lag: Coding Agents Solve Early but Stop Late
【速读】： 该论文旨在解决生成式代码代理（Coding Agent）在任务求解过程中存在“终止滞后”（termination lag）的问题，即代理在首次生成正确代码修复（correct patch）后仍继续执行大量无效操作，导致资源浪费并可能引入错误。其核心解决方案的关键在于识别并量化这一滞后现象：通过在SWE-bench Verified、Multilingual等三个数据集上对五种模型与两种代理框架进行系统性记录与测试，发现38%–82%的成功运行中，代理已在初始代码修改阶段即达成正确解；然而，在成功案例中，后续步骤仍占总步数的中位数达35%–37%（Qwen3-Coder）和54%–60%（Laguna-XS），主要消耗于自我验证与仓库探索。值得注意的是，1%–8%的案例中，代理会回滚此前正确的修改，致使最终结果错误。因此，该研究提出应优化代理的终止机制，以在首次正确输出后及时停止，从而提升效率与可靠性。
链接: https://arxiv.org/abs/2611.14960
作者: Musa Cim,Burak Topcu,Mahmut Taylan Kandemir
机构: The Pennsylvania State University (宾夕法尼亚州立大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Coding agents can solve a task well before they stop working on it. We characterize \emphtermination lag–the number of steps between an agent’s first correct patch and the end of its run–across five models, two agent frameworks, and three SWE-bench datasets by recording and testing every distinct repository patch generated during execution. Our analysis reveals that 38–82% of successful runs achieve a correct solution on their initial source-code edit. However, agents continue to expend substantial effort after correct patches, with post-solution steps accounting for a median of 35–37% for Qwen3-Coder and 54–60% for Laguna-XS.2 in successful runs on SWE-bench Verified and Multilingual. Agents spend most of these additional steps verifying their work and exploring the repository. In a small fraction of cases (1–8%), this continued activity leads agents to undo their earlier correct patch, leaving the final solution incorrect.
[NLP-101] Do Large Language Models Show Racial and Sex Bias in Pain Management? A Counterfactual Audit of GPT Claude Gemini Llama DeepSeek and Qwen
【速读】： 该论文旨在解决生成式AI在临床疼痛管理中可能继承并放大种族与族裔群体间医疗待遇不平等的问题，尤其关注大型语言模型（LLM）在处理不同种族和性别患者时是否存在系统性偏见。其核心挑战在于现有审计方法存在局限：多依赖少数专有模型、显式标注患者身份、且极少评估模型自身响应的不稳定性。本研究通过两阶段探索性分析提出解决方案的关键：采用多组别、非预注册的实验设计，在多个开放权重的LLM上测试六种急性疼痛病例，每例以不同种族-性别组合（四种族×两性别）呈现，并仅通过姓名隐含身份信息（每组两个姓名），从而更真实地模拟临床情境；同时在第二阶段对17个模型（包括GPT-4.1 mini、Claude Haiku 4.5、Gemini 2.5 Flash等）在多种推理模式与精度设置下进行统一提示、JSON约束的任务测试，结合50个由临床医生撰写的Q-Pain案例。结果显示，在经过多重检验校正后，两阶段均未发现显著差异，所有疼痛严重度评分差异均在0–10分量表上等效于零（237/238个Q-Pain比较差异小于四分之一分），表明模型输出在疼痛评估方面未表现出可检测的不公平性。尽管部分模型对阿片类药物决策的判断边界较宽（无法排除约20个百分点的差异），但整体仍无证据显示存在系统性偏见，反而在少数保留显著性的对比中，黑人患者倾向获得更高剂量推荐。更重要的是，模型间及同一模型内部配置间的差异远大于不同患者间的差异，凸显了模型本身异质性对结果的影响。因此，研究强调公平性审计应采用多姓名配置、配对比较种子、报告噪声基线，并针对实际部署的每一配置进行独立测试，以提升评估的可靠性与有效性。
链接: https://arxiv.org/abs/2611.14935
作者: Richard J. Young,Alice M. Matthews
机构: Lee Business School, Information Systems (李商学院，信息系统); University of Nevada, Las Vegas (内华达大学拉斯维加斯分校); DeepNeuro.ai
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Computers and Society (cs.CY)
备注: 31 pages, 5 figures, 14 tables. Code, vignettes, panel definition, spend ledger and all 58,080 raw model responses: this https URL
Abstract:Racial and ethnic minority patients receive less pain treatment than non-Hispanic White patients, and large language models (LLMs) used as medical AI could reproduce that bias. Prior audits tested a few proprietary models, stated race explicitly and rarely measured the models’ own response instability. In exploratory, non-preregistered analyses, Study 1 had four open-weight LLMs assess six acute-pain vignettes in every combination of four races and two sexes, with identity stated or cued only by a name (two names per group), comparing each identity with a White male reference on the same case and sampling seed. In Study 2, 24 configurations of 17 models, including OpenAI’s GPT-4.1 mini and gpt-oss, Anthropic’s Claude Haiku 4.5, Google’s Gemini 2.5 Flash, Meta’s Llama, DeepSeek and Qwen, with reasoning on and off and at two precisions, answered the same vignettes and the 50 clinician-written Q-Pain cases. In this single-prompt, JSON-constrained task, no difference survived correction in Study 1, none of 1,386 comparisons survived the false discovery rate in Study 2, and every pain-severity difference was equivalent to zero within one point on a 0-10 scale (within a quarter point for 237 of 238 Q-Pain comparisons). Opioid decisions were bounded less precisely (Study 1 could not exclude differences of about 20 percentage points), so they show no detected difference, not proof of none. The few differences surviving correction within a model favoured Black patients. Models differed far more from one another and across their own settings than between patients (the reference patient received the high opioid dose in 6% to 93% of responses), although serving software, provider and prompt format also differed. Fairness audits of clinical AI should use several names per group, pair comparisons on the seed, report the noise floor and test each deployed configuration.
[NLP-102] oolTurns: Benchmarking LLM Agents in Multi-Turn Dialogue
【速读】： 该论文旨在解决当前大语言模型在工具使用场景中评估不足的问题，即现有大规模工具使用基准大多聚焦于孤立请求，难以真实反映模型在多轮对话中持续完成复杂任务的能力。其核心挑战在于如何有效评估模型在具有上下文依赖性的连续交互中进行澄清、工具调用及上下文记忆的综合能力。解决方案的关键是提出ToolTurns基准数据集，包含2,620个跨47类别的问题，涉及1,827种工具和11,004个API端点，每个对话以故意不明确的初始请求开始，并通过两个依赖上下文的后续追问，全面测试模型的对话连贯性与任务执行能力。同时，引入开源用户模拟器，通过“检查-修正”（checker–fixer）流水线生成更自然的对话转录并修复缺陷，显著提升对话质量。实验表明，尽管最优模型在单个请求上可达到69.6%的成功率，但在完整多轮对话中仅能完成39.8%，且所有模型在后续请求上的表现均低于初始请求，揭示了当前模型在长期上下文保持方面的显著短板。该研究为评估代理在多轮工具使用对话中的持续任务成功率提供了新的标准化测试平台。
链接: https://arxiv.org/abs/2611.14925
作者: Milad Gholami,Brendan King,Jeffrey Flanigan
机构: University of California, Santa Cruz (加州大学圣克鲁斯分校)
类目: Computation and Language (cs.CL)
备注:
Abstract:Tool-using language models are increasingly deployed in conversational settings, yet most large-scale tool-use benchmarks focus on isolated requests. We introduce ToolTurns, a benchmark of 2,620 problems spanning 1,827 tools and 11,004 API endpoints across 47 categories. Each dialogue begins with an intentionally underspecified request and continues through two context-dependent follow-ups, testing clarification, tool use, and retention of conversational context. We also introduce an open-source user simulator that reveals withheld context through dialogue and generates follow-up requests, with a checker–fixer pipeline for improving its turns. Across nine language models, the best agent resolves 69.6% of individual requests but completes only 39.8% of full dialogues. Under this evaluation protocol, every evaluated model receives lower success scores on follow-up requests than on the initial request, and dialogue-quality rubric scores and task success can rank models very differently. Paired turn-level evaluation shows that the checker–fixer reduces detected dialogue defects, and blind human evaluation favors the revised turns. ToolTurns provides a testbed for evaluating whether agents sustain task success across multi-turn dialogue with tools.
[NLP-103] MuEdit: An efficient multi-task editing method towards inter-domain knowledge conflicts NEURIPS2026
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在跨异构领域进行连续多任务知识编辑时性能退化的问题。现有编辑方法在序列化应用到多任务场景下失效，其根本原因在于不同任务间诱导出的雅可比几何结构差异导致公共零空间（null space）严重收缩，从而缺乏足够的无冲突更新方向。为此，论文提出“冲突指数”（Conflict Index）以量化这种几何干扰，并引入Mu-Edit方法，其核心创新在于：（1）基于冲突感知的任务排序策略，以最小化累积干扰；（2）动态低秩近似机制，在排序无法充分缓解冲突时扩展零空间。实验结果表明，Mu-Edit在五个功能领域和三种模型架构上均显著优于现有基线，在保持模型通用能力的同时，有效支持增量式任务接入下的持续编辑。
链接: https://arxiv.org/abs/2611.14910
作者: Songshi Liang,Ting-En Lin,Zixuan Huang,Yuchuan Wu,Yongbin Li,Zihe Wang,Rui Yan
机构: Renmin University of China (中国人民大学); Alibaba Group (阿里巴巴集团); Beihang University (北京航空航天大学); Wuhan University (武汉大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 30 pages,accepted by NeurIPS 2026
Abstract:Large language models (LLMs) require frequent knowledge updates across heterogeneous domains, but existing editing methods fail when applied sequentially to multi-task settings. We trace this collapse to inter-domain null-space misalignment: heterogeneous tasks induce distinct Jacobian geometries, causing their common null space to shrink dramatically and leaving insufficient admissible directions for conflict-free updates. We propose the Conflict Index to quantify this geometric interference, and introduce Mu-Edit, which mitigates multi-task conflicts via (1) conflict-aware ordering to minimize cumulative interference, and (2) dynamic low-rank approximation to expand the null space when ordering alone is insufficient. Experiments across five functional domains and three backbones show that Mu-Edit significantly outperforms existing baselines in multi-task editing performance while preserving general capabilities, and remains effective under incremental task arrival.
[NLP-104] Hy ASR 3.0 Technical Report
【速读】： 该论文旨在解决传统自动语音识别（ASR）系统在多语言、多方言支持下的识别准确率不足，以及在复杂语音条件和上下文依赖场景中表现脆弱的问题。其核心挑战在于如何在保持高精度的同时，实现对多样化外部上下文（如关键词列表、问答对话、摘要等）和内部语境信息的灵活利用，以提升对同音词、发音相似词及长时序语音的识别鲁棒性。解决方案的关键在于构建基于大语言模型（LLM）的Hy ASR 3.0系统，通过融合外部上下文与跨句内上下文（intra-utterance context）信息，在无需额外输入的情况下实现同音歧义消解，并显著增强对中文普通话、英语及13种汉语方言（包括粤语）的多任务识别能力。该系统在公开与内部测试集上均取得了最低的整体误识率，且在长语音、儿童与老年用户语音、低音量及噪声环境下的识别性能表现出卓越的鲁棒性。
链接: https://arxiv.org/abs/2611.14908
作者: Aoxiong Yin,Auden Li,Chao Huang,Chuanyang Hu,Eitan Wang,Ethan Zou,Fangming Feng,Guangyu Chen,Guinan Li,He Wang,Jiahong Li,Meng Yu,Mingyu Cui,Peng Wang,Rongjie Li,Steve Yves,Teddy Sun,Tian Zheng,Wei Liu,Yang Zhou,Yixuan Zhang,Zhennan Lin,Zhihang Xu,Zhuohao Li,Ziqiang Zhang,Anji Zhang,Bang Wang,Caisheng Ouyang,Chengqi Zhu,Chong Li,Conglian Xie,Dongdong Wang,Edward Wang,Evelynn Liu,Ferman Yang,Feng Lu,Gang Zhang,Jathon Zhang,Jaxl Liu,Jiachen Luo,Jian Li,Jingxi Gui,Kysun Wang,Liang Dong,Longsha Zhou,Na Hu,Qi Chen,Rilin Chen,Rui Yuan,Shanshan Sang,Siyu Yan,Steven Hong,Taihui Wang,Weiwei Li,Xiangyu Zhang,Xiaotong Yang,Xue Hu,Xue Jiang,Yameng Li,Yifu Sun,Yuhuan Zhou,Zhiang Chen,Zhibin Wang
机构: 未知
类目: Computation and Language (cs.CL); Audio and Speech Processing (eess.AS)
备注: 12 pages, Technical Report. Project Leads: Auden Li, Eitan Wang, and Steve Yves (listed in alphabetical order by first name)
Abstract:We present Hy ASR 3.0, a large language model (LLM)-based automatic speech recognition (ASR) system that delivers strong recognition accuracy, supports flexible contextual conditioning, and maintains robust recognition performance across diverse speech conditions. Hy ASR 3.0 supports Mandarin Chinese, English and 13 Chinese dialect groups, including Cantonese. Among the evaluated systems, it achieves the lowest aggregate error rates on public Mandarin Chinese, English, and Cantonese benchmarks and the lowest error rates on the corresponding internal test sets, alongside first-tier dialect recognition performance. Beyond audio-only transcription, Hy ASR 3.0 supports multiple forms of external context, including hotword lists, question–answer dialogues, and summaries, enabling the use of task-relevant information during recognition. The model also leverages intra-utterance context to disambiguate homophonic and phonetically similar keywords using information from other parts of the same utterance, without requiring a separate contextual input. Evaluations demonstrate strong contextual recognition in both external-context and intra-utterance settings, as well as robust recognition of long-form speech, children’s and older adults’ speech, low-volume speech, and speech in noisy environments. This report describes the model architecture and training methodology and presents evaluations across recognition accuracy, context utilization, and robustness.
[NLP-105] Is Jev a Good Multilingual Judge?
【速读】： 该论文旨在解决生成式大语言模型（Generative Large Language Models, LLMs）在多语言文本评估中，作为直接输出判断概率的判别模型（decision model）是否具备有效性和可靠性的问题。其核心挑战在于：如何在降低计算成本的前提下，实现跨语言、多任务场景下的高质量评估，尤其是在低资源语言中的表现稳定性。论文提出的解决方案关键在于设计并验证Jev这一轻量级判别模型，其通过高效的概率输出机制，在仅需GPT-6.1约1/31至1/61评估成本的情况下，实现了对多数多语言评估任务的竞争力。研究发现，尽管Jev在知识密集型答案验证任务上仍逊于GPT-6.1，但其概率输出能有效识别错误判断，并在不同任务中展现出可解释性与鲁棒性。基于实证分析，论文进一步提出针对评估指令设计、置信度筛选及模型路由策略的优化建议，为实际应用中以低成本、高效率方式部署多语言判别模型提供了可操作的指导方案。
链接: https://arxiv.org/abs/2611.14861
作者: Wei Liu,Yi Fan,Xiyan Fu
机构: Heidelberg Institute for Theoretical Studies(海德堡理论研究所); Independent Researcher(独立研究员)
类目: Computation and Language (cs.CL)
备注:
Abstract:Generative large language models are widely used for multilingual text evaluation, yet whether decision models that directly output judgment probabilities can serve as multilingual judges has received limited attention. We evaluate Jev against generative LLM judges on four multilingual benchmarks with human annotations, covering up to 47 language varieties and three evaluation formats: binary decisions, pairwise comparisons, and graded ratings. At approximately 1/31 to 1/61 of GPT-6.1’s evaluation cost, Jev is competitive on most multilingual judging tasks, but falls behind GPT-6.1 on knowledge-intensive answer verification. Further analysis shows that Jev is more likely to accept incorrect answers in low-resource languages than in English and that its probability outputs help identify errors across tasks. Based on these findings, we offer recommendations on evaluation instructions, confidence-based selection, and model routing, providing empirical guidance for using Jev as a multilingual judge.
[NLP-106] Effective Teacher Support for Reasoning Distillation
【速读】： 该论文旨在解决前向KL散度（Forward-KL）语言模型蒸馏中因在每个监督位置对全词汇表计算密集型教师与学生得分而导致的训练成本高、教师目标存储开销大的问题。其核心解决方案是：仅从教师模型输出中保留最高概率的词汇子集（vocabulary fraction），并随着支持集从100%逐步缩减至0.01%进行重归一化处理，从而实现高效蒸馏。实验表明，在Qwen3 4B→1.7B等多组模型对上，保留0.01%词汇即可维持与全词汇蒸馏相当的MMLU-Pro性能表现，且显著降低计算开销——在0.01%支持下，缓存教师模型的保留词元得分可使单步训练时间减少38%，限制学生模型仅在这些词元上计算输出可降低峰值内存24%，且未观察到生成性能下降。该方法的关键在于通过动态裁剪词汇支持集实现性能与效率的最优平衡，同时揭示了可靠截断阈值依赖于具体模型家族特性。
链接: https://arxiv.org/abs/2611.14850
作者: Lavinia Ghita,Dhruv Desai
机构: NVIDIA(英伟达)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Forward-KL language-model distillation normally computes dense teacher and student scores over the full vocabulary at every supervised position, making training and stored teacher targets expensive. We instead distill from the teacher’s highest-probability vocabulary fraction, renormalizing the retained mass as we sweep support from 100% to 0.01%. Our main study distills Qwen3 4B - 1.7B. We confirm the broader support-reduction finding on larger Qwen3 8B - 4B and Nemotron Nano 12B - 9B pairs and under answer-only supervision. We evaluate generation primarily on MMLU-Pro and secondarily on GPQA-Diamond. Across the Qwen3 settings, retaining 0.01% of the vocabulary causes no observed MMLU-Pro degradation relative to full-vocabulary distillation, although disagreement with the complete teacher distribution increases. Nemotron Nano matches full-vocabulary MMLU-Pro at 0.1%, but smaller fractions fail to emit the closing reasoning tag, showing that the reliable cutoff depends on the model family. At 0.01% support, caching the retained next-token scores from the teacher forward pass cuts training step time by 38%, while restricting student output computation to those tokens cuts peak memory by 24%; neither causes observed MMLU-Pro degradation.
[NLP-107] LLM s and the Homogenization of Teaching ATC
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在开放式文本生成中缺乏多样性的问题，这一缺陷在教育场景中尤为突出，可能导致学生思维模式和学科实践的同质化，限制其多元认知发展。研究聚焦于两个教育生成任务：数学题目生成与作文反馈生成，通过词汇、语义及领域相关性度量方法，对比前沿闭源模型与开源模型与人工编写参考数据集之间的差异。研究发现，直接提示生成的模型输出在词汇层面（Self-BLEU）和语义层面（嵌入相似度）分别高出人类内容78%和51%，且呈现更狭窄的教学形式分布。尽管采用了多样性提示策略和后训练方法，模型生成内容的多样性提升主要体现在表层特征上，未能实质性拓展教学活动的种类与复杂性，反而继承并放大了人类教学中最标准化、常规化的范式。因此，克服教育场景中的“模式坍缩”（mode collapse）问题，关键在于构建更丰富多样的教学数据集，并加大对教师创造力与专业判断力的投入与支持。
链接: https://arxiv.org/abs/2611.14828
作者: Mei Tan,Hyunji Nam,Hari Subramonyam,Dorottya Demszky
机构: Stanford University (斯坦福大学)
类目: Computation and Language (cs.CL)
备注: Revised version of a paper published at COLM 2026 under the title “Can LLMs Match Human Diversity in Educational Content Generation?”
Abstract:Large language models (LLMs) are known to lack diversity in open-ended text generation, a limitation that is especially consequential in education, where homogeneity can narrow the kinds of thinking and disciplinary work students are invited to do. We study diversity in two educational generation tasks: math items and essay feedback. We compare frontier and open-weight models with human-authored reference distributions using lexical, semantic, and domain-grounded measures. For math, we examine the mathematical activities students are asked to perform; for feedback, we examine the revision work requested and the rhetorical strategies used to communicate it. We evaluate direct prompting, two diversity prompting methods, and post-training. Across both tasks, directly prompted model outputs are markedly less diverse than human-authored content–78% higher on lexical (Self-BLEU) and 51% higher on semantic (embedding) similarity–and represent a narrower range of instructional forms. Diversity prompting and post-training increase variation largely through surface features without broadening the instructional repertoire. Models inherit and amplify the most normative routines of human instruction. Overcoming mode collapse in education will require richer instructional data and investment in teachers’ creative capacity and professional judgment.
[NLP-108] Discharge Summary Generation Is Not Just Summarization: A Reality Check for Evaluation ML4H2026
【速读】： 该论文旨在解决生成式医疗文本摘要（如出院小结）评估中存在的核心问题：传统评估方法无法区分模型生成内容是否基于真实临床证据，导致对模型性能的评价存在偏差。具体而言，现有评估指标常将“未能复现输入中未包含的内容”与“未能提取已有证据支持的事实”混为一谈，从而掩盖了模型在事实准确性与信息完整性方面的实际表现差异。为此，作者提出FairSumEval——一种基于证据感知、面向事实级别的评估框架，其关键在于通过双重维度量化评估：一是判断每个参考事实与生成事实是否能在模型输入证据中找到支持（证据支持性），二是衡量两者在摘要中的相互覆盖程度（匹配性）。该框架引入两个互补指标——Soft IoU-S（有证据支持时的重叠度）与Soft IoU-NS（无证据支持时的重叠度），能够有效分离出模型在有依据内容上的表现与无依据内容上的拟合行为。实验结果表明，在MIMIC-III数据集上，仅18.8%的医生参考事实可被识别为有证据支持，而通用与医学专用模型分别生成了65.4%和63.5%的有证据支持事实，且在有证据支持的内容上重叠度显著更高（0.255与0.236），而在无证据支持内容上则极低（0.030与0.026）。此外，两指标间相关性弱（ρ = 0.22, 0.17），说明传统指标难以同时捕捉这两类能力。因此，FairSumEval的关键创新在于实现了对生成内容的证据根基性与内容一致性之间的解耦分析，精准揭示了模型在事实遗漏与冗余添加方面的系统性偏差。
链接: https://arxiv.org/abs/2611.14823
作者: Radhika Dua,Frederick Wolff,Kyunghyun Cho,Eric Karl Oermann
机构: New York University (纽约大学); NYU Langone Health (纽约大学朗格恩健康中心); ETH Zurich (苏黎世联邦理工学院)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 20 pages, 8 figures. Radhika Dua and Frederick Wolff contributed equally. ML4H 2026
Abstract:Discharge summaries combine a retrospective account of hospitalization with discharge-time decisions. Physician-written references therefore contain both facts supported by prior clinical notes and reference-only content, potentially reflecting clinical synthesis, handoffs, or discharge planning. Reference-only evaluation can conflate failure to recover available information with failure to reproduce content absent from the model input. We introduce FairSumEval, an evidence-aware, fact-level framework that assesses whether each reference and generated fact is supported by the model’s evidence and captured by the other summary. The resulting outcome map separates matches, omissions, and additions. Two complementary metrics, Soft IoU-S and Soft IoU-NS, measure reference-generation agreement for facts with and without identified evidence support. On 200 MIMIC-III admissions, 18.8% of physician reference facts have identified support in the prior evidence, compared with 65.4% and 63.5% of facts generated by a general-purpose and a medically specialized model. Overlap is substantially higher for supported content (0.255 and 0.236) than for content without identified support (0.030 and 0.026). The two scores are weakly correlated across admissions (rho = 0.22 and 0.17), and conventional metrics do not consistently capture both dimensions. FairSumEval distinguishes evidence-grounded summarization from agreement on content without identified support, revealing what models omit and add in each category. Comments: 20 pages, 8 figures. Radhika Dua and Frederick Wolff contributed equally. ML4H 2026 Subjects: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG) Cite as: arXiv:2611.14823 [cs.CL] (or arXiv:2611.14823v1 [cs.CL] for this version) https://doi.org/10.48550/arXiv.2611.14823 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[NLP-109] What Do Tokenizer Comparisons Measure? Disentangling Granularity Root Assignment and Word-Level Readout across Amharic and Hausa
【速读】： 该论文旨在解决生成式AI在处理形态丰富的语言时，如何准确评估形态感知分词器（morphology-aware tokenizers）性能提升来源的问题。现有研究通常将性能增益归因于分词器对语言学单位（linguistic units）的捕捉能力，但这一归因存在混淆：形态感知分词器不仅在根词层面进行分组（root assignment），还改变了文本分割的粒度（granularity）和子词读出（sub-token readout）的重要性。为厘清这些因素的影响，研究以阿姆哈拉语（Amharic，属闪米特语系，其格兹字母通过音节符号隐藏辅音根）与拉丁字母书写的豪萨语（Hausa）为对比，设计了一种无损的根-模式分词器（root-and-pattern tokenizer），并采用两种独立实现方式，分别控制单一变量。实验结果表明，在阿姆哈拉语命名实体识别任务中，真实根分配相比随机根分配分别带来4.9和4.7的F1值提升，且该增益主要不源于根频次结构；在小规模自训练模型中，读出策略可导致F1值波动高达13.4点，并逆转所有根-模式分词器与字节对编码（BPE）的比较结果，而在预训练编码器中该影响不足1点；粒度效应在不同语言和实现间甚至呈现相反符号。因此，论文强调应解耦分词器效果中的根分配、粒度与读出策略等关键因素，并在归因前明确报告读出机制的影响，以避免错误归因于语言学结构。
链接: https://arxiv.org/abs/2611.14818
作者: Abdul Rehman Antall,Baidya Nath Saha
机构: National University of Computer and Emerging Sciences (FAST-NUCES)(巴基斯坦国家计算机与新兴科学大学); Concordia University of Edmonton(加拿大爱德蒙顿康考迪亚大学)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 19 pages, 6 figures, 7 tables
Abstract:Morphology-aware tokenizers are expected to help morphologically rich languages, and their reported gains over byte-pair encoding (BPE) are usually credited to their linguistic units. A comparison with BPE cannot support this attribution, since such a tokenizer also segments text more finely (granularity) and, on token-level tasks, changes how much the sub-token readout matters. We separate both factors from the effect of grouping words by their roots (root assignment) in Amharic, a Semitic language whose Ge’ez abugida hides consonantal roots inside syllabic glyphs, with Latin-script Hausa as an orthographic contrast. Our design combines a lossless root-and-pattern tokenizer with controls that change one factor at a time, in two independent implementations. In both, real root assignment improves Amharic named-entity recognition over scrambled roots (+4.9 and +4.7 F1), and root-frequency structure explains little of this gain. In small from-scratch models, the readout choice moves F1 by up to 13.4 points and reverses the sign of every comparison of the root-and-pattern arms with BPE, whereas in two pretrained encoders it moves F1 by less than one point. The granularity effect changes sign across languages and implementations. We recommend decomposing tokenizer effects, and reporting the readout, before a gain is attributed to linguistic structure.
[NLP-110] Distributionally Robust BEST-RQ For Speech Representation Learning
【速读】： 该论文旨在解决传统BEST-RQ（基于随机投影与量化）在无监督语音表示学习中因随机量化目标选择敏感而导致的表示稳定性问题。其核心解决方案是提出一种分布鲁棒性增强的BEST-RQ方法，通过引入分组分布鲁棒优化（Group Distributionally Robust Optimization, Group DRO）框架，将不同码本（codebook）定义为多个组，并在训练过程中自适应地重新加权表现较差的组，以提升模型在多种量化配置下的泛化能力。同时，采用Kullback-Leibler（KL）散度正则化来抑制组间分布的过度偏离，从而确保优化过程的稳定性。实验结果表明，该方法在多语言LibriSpeech数据集上显著优于传统的均匀平均策略，在多个码本配置下均展现出更优的性能。
链接: https://arxiv.org/abs/2611.14801
作者: Muhammad Khan,Xiaodong Cui,Brian Kingsbury,George Saon
机构: Rensselaer Polytechnic Institute (伦斯勒理工学院); IBM Research (国际商业机器公司研究部)
类目: Computation and Language (cs.CL); Audio and Speech Processing (eess.AS); Optimization and Control (math.OC)
备注:
Abstract:BEST-RQ, based on random projection and quantization, has been a well-established self-supervised learning approach to learn speech representations from large-scale unlabeled audio due to its high performance and efficiency. However, in BEST-RQ representations learned by random quantization can be sensitive to the choice of quantization targets. To address this issue, we propose a distributionally robust extension of BEST-RQ that seeks representations robust across multiple quantization configurations. Specifically, we define groups by codebooks and formulate the training as a group Distributionally Robust Optimization (DRO) problem. Rather than optimizing solely for the average loss across groups, our approach adaptively reweights groups to emphasize those performing poorly. Additionally, we regularize DRO with a Kullback-Leibler divergence to prevent excessive distribution deviation. We evaluate our approach on the Multilingual LibriSpeech dataset and show that the distributionally robust BEST-RQ achieves superior performance over the conventional uniform average across multiple codebooks.
[NLP-111] When Is Proof-Step Verification Worth Its Cost? Operating Requirements of Certify-or-Decline Verification for Open-Weight Mathematical Reasoning NEURIPS2026
【速读】： 该论文旨在解决大模型在数学推理任务中生成答案的可信度问题，即如何有效验证模型输出的逻辑正确性。其核心挑战在于：尽管大模型能够生成看似合理的解答，但其中存在大量逻辑错误或无效推导过程，而传统仅依赖答案匹配（answer-key match）的评估方式无法识别这些错误。为此，论文提出一种基于生成式AI（Generative AI）的证明步骤验证（proof-step verification）框架，将求解器输出转化为带类型的逐步推理，并利用大语言模型（LLM）作为裁判进行逐步合法性判断。该方案的关键在于通过可解释的、分步的逻辑验证机制来提升答案的可靠性，从而在不显著增加计算成本的前提下增强推理结果的可信度。然而，研究发现该方法虽能显著提升答案匹配精度（如在强配置GPQA上提升31个百分点，在AIME上提升46个百分点），但覆盖范围大幅下降（分别仅处理35%和21%的尝试），且在接近天花板性能的MATH-500数据集上收益有限（最多提升5个百分点），同时伴随13–27个百分点的覆盖率损失。此外，诊断分析揭示了一个关键失败模式：超过一半已完成的问题因构造出无效证明而从未进入裁判环节，表明执行阶段存在严重结构性缺陷。最终，外部LLM对经认证链路的接受率仅为61.0%–67.6%，远低于87.6%的答案键匹配率，进一步暴露了当前验证机制与真实证明有效性之间的差距。因此，论文强调应独立报告执行失败率、选择性答案精度以及独立评估的证明有效性，以更全面地衡量模型推理质量。
链接: https://arxiv.org/abs/2611.14794
作者: Abhinav Gupta,Ben Slivinski,Yashica Patodia,Michael Saldivar
机构: Independent Researcher(独立研究员)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 13 pages, 1 figure, 13 tables. Accepted at the 6th Workshop on Mathematical Reasoning and AI (MATH-AI) at NeurIPS 2026. Code: this https URL
Abstract:Proof-step verification rewrites a solver’s answer into typed steps and uses LLM judges to certify or decline. We evaluate this pipeline on open-weight models across GPQA, AIME-2025, and a MATH-500 subset. Certification raises answer-key-match precision by 31 percentage points on the strong GPQA configuration and 46 on AIME, while answering only 35% and 21% of attempts. On near-ceiling MATH-500, gains are at most 5 points at a loss of 13-27 points of coverage. An output-token-matched self-consistency selector obtains 82.9% precision at 35% coverage, compared with the verifier’s pooled 87.6% (paired difference: +4.8 points, 95% interval [-8.2, 18.5]). A diagnostic sample reveals a distinct failure mode: 52 of 88 completed problems end in invalid proof construction without ever reaching a judge. Finally, external-LLM acceptance of rendered certified chains is 61.0-67.6%, compared with 87.6% answer-key match. These audits lack human or formal-kernel ground truth. Open-weight auditor pairs show near-zero chance-corrected agreement, which measures consistency rather than accuracy against proof truth. These results motivate separate reporting of execution failures, selected-answer precision, and independently assessed proof validity. We provide per-configuration results and evaluation details in the supplement.
[NLP-112] An LLM Is All You Need: Data-Efficient Adaptation to Full-Duplex Word-Level Spoken Dialogue
【速读】： 该论文旨在解决全双工语音对话系统在训练成本高、语言模型被重写以及话语权转换机制粗粒度等问题。传统方法依赖数百万小时的语音数据进行端到端训练，不仅资源消耗巨大，且难以实现逐词级的话语控制。本文提出一种基于冻结通用多模态大语言模型（Multimodal Large Language Model, MLLM）的轻量级适配方案，通过仅训练少量可学习参数（97M参数）和新增的若干词元嵌入，将预训练模型转化为支持逐词级全双工交互的系统。其核心创新在于构建“思考者-说话者”级联架构，以80毫秒帧率为时钟，在每一帧中由“思考者”生成一个离散词元输出，包括停顿、让位与发言起始等行为均以统一词汇表中的词元表示，并通过有限状态机保证语法合法性；整个过程采用四阶段课程学习，利用模型自身先前行为作为锚点完成转换，仅需约30 GPU小时。由于“说话者”仅接收“思考者”输出的离散词元，两部分可独立训练。实验结果表明，该系统在Full-Duplex-Bench和InteractionBench基准上性能媲美大规模语音训练的开源全双工系统：在用户仅短暂停顿时最不易抢话，回声响应频率最高，且在长对话中的任务完成率、语义连贯性与上下文一致性方面表现最优。
链接: https://arxiv.org/abs/2611.14774
作者: Tairan Chen,Richard Yucheng He,Yihang Liu,Baodong Cao,Chen Xu
机构: AveraLabs
类目: Computation and Language (cs.CL); Sound (cs.SD); Audio and Speech Processing (eess.AS)
备注:
Abstract:Full-duplex spoken dialogue, in which a system listens while it speaks and can backchannel, yield the floor, and be interrupted, is usually obtained by training on hundreds of thousands to millions of hours of speech. This is costly, rewrites the language model that supplies the system’s knowledge, and still leaves turn-taking either implicit or decided coarsely rather than word by word. We show that a frozen, general-purpose multimodal LLM can instead be adapted into a word-level full-duplex system with little data. Our system is a thinker-talker cascade on an 80 ms frame clock. The thinker keeps the LLM frozen and trains only lightweight adapters and a few new token embeddings (97M parameters); at every frame it emits one token of agent text, with pauses, yields, and onsets represented as tokens of the same vocabulary and kept legal by a finite-state machine. A four-stage curriculum, each stage anchored to the model’s own behavior under the previous format, performs the conversion in about 30 GPU-hours. The talker reads only the thinker’s discrete tokens, so the two halves are trained independently. On Full-Duplex-Bench and InteractionBench, the system is competitive with open-source full-duplex systems built through large-scale speech training: among them, it is the least likely to take the floor when the user merely pauses, backchannels most often, and leads in spoken-task accuracy, groundedness, and coherence over long conversations.
[NLP-113] Rational Use of Time for Social Coordination in Humans and AI
【速读】： 该论文旨在解决人类在无显式通信的动态社会协作中如何利用时间线索（如停顿持续时间）进行精细的社会协调，以及人工智能系统是否能在不经历真实连续时间流的情况下实现类似能力的问题。其核心解决方案在于通过“心灵游戏”（The Mind）这一合作卡牌任务，考察人类与不同类型的AI模型在仅依赖时间信号的条件下完成共享目标的表现。研究发现，人类行为最符合基于贝叶斯推理的策略——即根据对方的行为时机推断其心理状态；前沿大语言模型（LLM）也表现出类似模式，但所有测试中的AI模型均未能有效随游戏进程调整自身行为以适应合作伙伴的变化。关键突破在于揭示了人类能够对连续时间信号进行隐式的概率推理，且这种机制可能无需真实时间体验即可形成，为未来构建具备社会动态适应能力的智能体提供了认知与技术层面的新挑战。
链接: https://arxiv.org/abs/2611.14766
作者: Xiulin Yang,Sirisha Gudavalli,Jennifer Hu
机构: Georgetown University (乔治城大学); Johns Hopkins University (约翰霍普金斯大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 30 pages, 16 figures
Abstract:All social interactions are situated in time. Previous studies have shown that humans are able to use implicit temporal cues, such as the duration of a pause, to infer others’ mental states. But how do humans use temporal cues for dynamic, fine-grained social coordination? And can AI systems do the same, without ``experiencing’’ continuous real-time as humans do? We study these questions using a cooperative card game task called The Mind, which requires players to achieve a shared goal across repeated interactions but forbids explicit communication: the only available signal is the passage of time. We evaluate 150 humans, and two classes of AI models which reflect different ways of operationalizing time in artificial systems: computer-use agents (CUAs), and standard text-prompted LLMs. We find that human behavior is best captured by a Bayesian strategy that infers a partner’s belief from their timing—and similarly for frontier models. However, unlike humans, none of the tested models seem to efficiently adapt their behaviors to their partners’ over the course of the game. From a cognitive perspective, our findings suggest that humans perform implicit probabilistic reasoning about continuous-time signals in dynamic social settings, and that such reasoning mechanisms may emerge without real-time experience. Practically, our study highlights new challenges for AI systems to coordinate efficiently with other agents by learning to interpret and adapt to their implicit signals in dynamic settings.
[NLP-114] Benchmarking Multilingual Document Understanding in Vision-Language Models
【速读】： 该论文旨在解决当前多模态视觉-语言模型（VLMs）在跨语言、跨任务、跨文档风格与版式泛化能力方面的评估空白问题。现有基准测试主要聚焦英文文档和有限任务，难以全面反映模型在真实多语言文档理解（Multilingual Document Understanding, MDU）场景下的实际性能。为此，研究提出MDU基准，涵盖13项任务、5种语言（英语、法语、德语、意大利语、西班牙语）及23,146个样本，任务设计覆盖从文本与Markdown提取到摘要生成、分类、命名实体识别，以及对版式敏感的段落和布局提取等互补能力。关键解决方案在于构建一个系统性、多语言、多任务、多风格的标准化评估框架，并对45个参数量从1.7B到235B的开源VLMs及一个商用系统进行统一评测。结果显示，该基准尚未饱和：最佳平均得分仅为69.0，仅8个系统超过65分，且表现最优的模型来自27B参数规模而非最大模型，表明模型规模并非性能提升的决定性因素。更关键的是，性能在不同语言间存在持续差距，且不随模型规模扩大而缩小；同时，各项能力之间呈现显著解耦现象，即模型整体质量控制后，各任务得分高度不相关，尤其表现为文本转录能力与语义提取能力之间的权衡。这揭示了当前VLMs在综合理解能力上的结构性局限。
链接: https://arxiv.org/abs/2611.14758
作者: Matteo Gabburo,Daniela Baiamonte,Stefano Simonazzi,Silvia Gardini,Sean Cesare Lazzeri,Andrea Zugarini
机构: Aithlas{dbaiamonte, ssimonazzi, sgardini, slazzeri, azugarini}@villanova.ai; Villanova
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 18 pages, 4 figures
Abstract:Recent Vision-Language Models (VLMs) perform strongly on multimodal tasks, Document Understanding (DU) among them, but the benchmarks that measure them mostly cover English documents and narrow problems. How well current VLMs generalize across languages, tasks, styles and layouts therefore remains unclear. We introduce MDU, a multilingual DU benchmark of 23146 examples spanning 13 tasks and five languages: English, French, German, Italian, and Spanish. The tasks isolate complementary capabilities, from surface-level text and Markdown extraction through summarization, classification and named-entity recognition to layout-sensitive passage and layout extraction. We evaluate 45 state-of-the-art open-weight VLMs from 1.7B to 235B parameters, plus one commercial system under the same protocol. The benchmark is far from saturated: the best average is 69.0, only eight of the 45 evaluated systems reach 65, and the two best results come from 27B models rather than from the largest systems we evaluate. Performance is uneven in ways that scale does not fix. The gap between the strongest and weakest language does not close as models grow, and it survives controlling for the domain composition of each language. Capabilities also do not improve together: once we partial out overall model quality, per-task scores are largely uncorrelated, and transcription ability in particular trades off against semantic extraction.
[NLP-115] Code as Explanations: Executable and Verifiable LLM Explanations NEURIPS2026
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在做出重要决策时缺乏可解释性的问题，尤其是现有自由文本解释（如思维链，Chain-of-Thought, CoT）存在实例化生成、难以在新输入上验证的局限。其核心解决方案是提出“代码解释”（code explanations），即用可执行程序来统一描述模型在一组输入上的行为，通过程序执行实现对模型输出的自动化验证。关键在于引入一致性度量（Consistency metric），量化程序在多大比例的输入上能复现模型输出，从而实现对候选解释程序的评估与选择。实验表明，在多个开源代码能力模型和任务中，代码解释具有更高的紧凑性（比CoT短6至100倍）和更好的行为对齐性（24个设置中9个更优），且具备可验证性。然而，在涉及复杂语言成分的任务中，一致性显著下降，生成的程序常依赖冗长硬编码模式，表现接近无信息基线。进一步实验显示，向模型提供一致性反馈并结合代码执行，可在一定程度上提升解释一致性，但复杂任务仍面临挑战。研究确立了可执行代码作为解释LLM行为的新范式，兼具可验证性、紧凑性和可扩展性，并指出验证反馈是提升解释一致性的重要方向。
链接: https://arxiv.org/abs/2611.14704
作者: Nhi Nguyen,Shauli Ravfogel,Rajesh Ranganath
机构: New York University (纽约大学)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (stat.ML)
备注: 30 pages, 11 figures, 3 tables, The Fortieth Annual Conference on Neural Information Processing Systems (NeurIPS 2026)
Abstract:As large language models (LLMs) increasingly make consequential decisions and solve problems alongside humans, we must understand how models arrive at their conclusions. Understanding how an LLM behaves requires observing it across multiple inputs, yet free-text explanations such as chain-of-thought (CoT) are generated per instance and cannot be easily verified on new inputs. We propose code explanations to address the mismatch in how LLM explanations are generated versus how they are evaluated: a program explains LLM behavior across a dataset of inputs, and its execution enables automatic verification. We define Consistency metric that measures the fraction of inputs on which the program reproduces the LLM’s outputs, and use it to evaluate and select candidate programs. Across three open-source code-capable LLMs and eight tasks, we find that LLMs can produce highly consistent code explanations in several settings. Code explanations are also 6 \times to 100 \times shorter than CoT, and better align with model behavior in 9 out of 24 settings. However, consistency is significantly lower on tasks with more complex linguistic components, where generated programs reproduce model outputs no more frequently than an uninformative explanation baseline, and these programs often rely on longer hard-coded patterns. In experiments with Claude Opus 5.5, providing the model with consistency feedback and code execution improves explanation consistency across tasks, but consistency remains significantly lower on more complex tasks. Our findings establish executable code as a novel, verifiable, and compact approach to explaining LLM behavior and highlight verification feedback as a promising direction for improving explanation consistency. The code for this paper is at this https URL.
[NLP-116] Probe What-If Worlds Before You Ship: Persona Simulation for Real-World Decision Making
【速读】： 该论文旨在解决产品与服务设计迭代中因用户研究与系统测试受限于真实用户招募困难、实验条件控制复杂以及设计演化后需重复验证而造成的瓶颈问题，尤其针对难以触及的目标用户群体或“假设性”场景（what-if scenarios）的评估需求。其核心解决方案是构建一个将真实产业需求转化为受控实验的框架与执行基础设施，通过基于详细角色画像（persona）的模拟用户代理（simulated users）实现对目标人群、实验刺激和决策相关指标的明确定义。该框架的关键在于：1）以高保真角色画像为基础生成可扩展的模拟用户群体，支持跨平台（问卷、网页、聊天机器人、操作系统应用等）的大规模并发实验；2）提供逐次试验可审计的日志记录、群体层面结果聚合及跨群体、跨变体的统计分析报告；3）通过对比不同模型表现与扰动/删减角色属性的影响，验证了角色画像对仿真结果的实质性影响，表明其能有效反映真实用户行为差异。研究表明，基于角色画像的模拟用户能够以设计速度快速验证设计假设，诊断系统缺陷，并提供定向证据以聚焦问题定义、指导设计优化与战略决策，从而显著提升数据驱动的设计决策效率与可扩展性。
链接: https://arxiv.org/abs/2611.14700
作者: Julie Zhu,Jianheng Hou,Zhixu Silvia Tao,Shi Bo,Yilan Fan,Yixuan He,Zhiyuan Hu,Shirley Huang,Fangyu Liu,Yijun Wang,Zibu Wei,Haimingyue Xu,Zhen Xu,Zichen Yuan,Jierui Zuo,Jintao Huang,Yifan Wang,Qianfeng Wen,Jiahan Li,Zhejian Peng,Bowen Yang,Shushu Liang,Yuexing Hao,Xiaomin Li
机构: Massachusetts Institute of Technology (麻省理工学院); Princeton University (普林斯顿大学); Arizona State University (亚利桑那州立大学); New York University (纽约大学); University of California, Los Angeles (加利福尼亚大学洛杉矶分校); Harvard University (哈佛大学); Boston University (波士顿大学); University of Southern California (南加州大学); Washington University in St. Louis (圣路易斯华盛顿大学); Columbia University (哥伦比亚大学); The Ohio State University (俄亥俄州立大学); Purdue University (普渡大学); University of Toronto (多伦多大学); University of Illinois Urbana-Champaign (伊利诺伊大学厄本那-香槟分校); MatrAIx; VEGA; University of Washington (华盛顿大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:User research and system testing reveal whether products and services meet user needs, yet recruiting specialized participants, designing controlled what-if comparisons, and repeating studies as designs evolve often bottleneck design iteration and data-driven decision-making. We present a framework and execution infrastructure that translates real-world industry inquiries into controlled experiments with simulated users. Our methodology formally defines the target cohort, experimental stimuli, and decision-relevant metrics, while the pipeline supports large-scale concurrent runs over cohort-stimulus-environment combinations (survey, web, chatbot, and OS app), auditable trial-level logging, cohort-level outcome aggregation, and cross-cohort and cross-variant statistical reports. Across twelve industry partner cases spanning five domains, we show how constructing hard-to-reach or what-if persona cohorts supports exploratory studies and distills simulated responses into testable design hypotheses. Cross-model comparisons on selected cases largely preserve primary conclusions, whereas perturbing or ablating persona attributes alters service outcomes, effect sizes, and inter-group differences, indicating that grounding target cohorts in detailed personas materially shapes simulation results. Together, these studies show that persona-enabled agents can test specified cohorts at design speed, diagnose systems under test, and deliver directional evidence that focuses problem definition, steers design and strategic decisions, and guides further study.
[NLP-117] rait Hitchhiking in LLM Post-Training
【速读】： 该论文旨在解决强化学习（Reinforcement Learning, RL）后训练阶段中语言模型出现的非目标性行为与内部属性变化问题，尤其关注那些未被显式奖励函数所约束的“特质”（traits）在训练过程中的演化。其核心挑战在于：尽管RL能有效提升模型在可验证奖励任务上的表现，但其通过重新分配生成完成度（completion）的概率分布，可能无意中放大或抑制某些非目标特质，导致潜在的负面行为涌现。解决方案的关键在于引入一个源自进化生物学的数学框架——价格方程（Price equation），将其形式化为描述语言模型在RL后训练过程中“特质”动态演化的工具，并据此构建“特质随带”（trait hitchhiking）的数学模型。该模型揭示了非目标特质如何因概率重分配而被间接增强或抑制。研究通过对采用GRPO（Generalized Reward Policy Optimization）训练的语言模型进行实证分析，验证了价格更新表达式与实际采样到的特质动态高度吻合，从而确立“特质随带”是RL后训练中隐性放大不良特质的重要机制。
链接: https://arxiv.org/abs/2611.14683
作者: Adil Soubki,Elizabeth S.Z. Tan,Miles Cranmer
机构: University of Cambridge(剑桥大学)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注:
Abstract:Reinforcement learning (RL) is now a standard stage of language model post-training, and the compute devoted to it is growing rapidly. While it has been shown to improve capabilities, particularly for tasks with verifiable rewards, it can also change behaviours and internal properties that the reward does not explicitly target. We present a formalism to describe the evolution of “traits” during RL post-training and, in doing so, recover the Price equation from evolutionary biology. We use this to develop a mathematical model for what we call “trait hitchhiking”: the amplification or suppression of non-targeted traits as training redistributes probability across completions. We evaluate this Price update expression on language models trained with GRPO and show that the Price estimates closely track the sampled trait dynamics. These results identify trait hitchhiking as a mechanism by which RL post-training can silently amplify undesirable traits.
[NLP-118] MITRA-Swarm: Injecting Explicit Linguistic Structure into Foundation Models for Classical Machine Translation NEURIPS2026
【速读】： 该论文旨在解决古典梵语翻译中三大复杂语言现象——音位融合（Sandhi）、名词复合结构密集化（Samāsa）以及受诗歌格律强制的篇章级零回指（zero-anaphora）——在单次调用的单一模型架构下难以有效处理的问题。其核心解决方案是提出MITRA-Swarm多智能体架构，通过显式注入语言学结构来增强传统流水线的性能：包括一个专门负责分拆Sandhi的代理、一组通用型gemma-4-26B-A4B-it与领域专用型gemma-2-mitra-it翻译器并行运行，以及一个具备篇章意识的省略恢复（Ellipsis Restorer）模块，该模块利用语料库衍生的语域标签以避免在无主语的法律与哲学语境中虚构主动语态主语。关键创新在于引入双评判机制——内部路由评判者（J_\textroute）驱动自我批判循环，独立报告评判者（J_\textreport= gpt-4o-mini）进行一次性独立评分，从而确保评估客观性；实验表明，完整系统在1,000句诗的分层基准上达到85.90 MQM，优于单模型基线（85.27，Δ=+0.63，p_adj=0.0018），其中仅篇章级省略恢复即贡献+0.44 MQM，且显著优于同等调用次数的自精炼基线（+0.51 MQM）。尽管推理成本约为基线的9倍，但严重错误率从9.0%降至7.6%，且双评判者间存在显著评分差异（2–7 MQM），验证了两判机制的必要性，同时通过192,151条目检索索引预检实现零参考泄露。
链接: https://arxiv.org/abs/2611.14675
作者: Kayshav Bhardwaj,Pragun Seth,Kush Bhardwaj,Vijay Nagasamy,Kurt Keutzer
机构: Purdue University (普渡大学); UC Berkeley (加州大学伯克利分校); University of Illinois (伊利诺伊大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 17 pages, accepted to NeurIPS 2026 LP4FM Workshop
Abstract:Classical Sanskrit poses three challenges unresolved by single-call monolithic models: morphophonemic fusion Sandhi, dense nominal compounding Samāsa, and discourse-level zero-anaphora forced by poetic meter. We present MITRA-Swarm, a multi-agent architecture injecting explicit linguistic structure into an otherwise monolithic pipeline: a Sandhi-splitting agent, a parallel array of generalist gemma-4-26B-A4B-it and domain-specialized gemma-2-mitra-it translators, and a discourse-aware Ellipsis Restorer that consults corpus-derived register tags to avoid inventing active-voice subjects in impersonal legal and philosophical registers. Two separated judges evaluate quality: a routing judge ( J_\textroute , same model family as the generators) driving an internal critique loop, and an independent reporting judge ( J_\textreport= gpt-4o-mini, no shared vendor lineage) scoring every output exactly once. On a stratified benchmark of 1,000 verses across five registers (23,990 scored verdicts, 12-condition pre-registered ablation), the full swarm reaches 85.90 MQM against a monolithic baseline of 85.27 ( \Delta=+0.63 , p_\textadj=0.0018 , 95% CI [+0.30,+0.97] ), and isolated discourse restoration alone accounts for +0.44 MQM ( p_\textadj=0.0288 ), outperforming a call-count-matched self-refinement baseline by +0.51 MQM ( p_\textadj=0.0288 ). Severe per-verse failures (MQM 80 ) fall from 9.0% to 7.6% , at roughly 9\times the per-verse inference cost of the baseline. Additionally, J_\textroute scores its own family’s outputs 2 - 7 MQM higher than J_\textreport on every one of the 12 conditions, empirically motivating the two-judge split; and a preflight screen of a 192,151-entry retrieval index against all 12 evaluation works verifies 0 reference leaks across 269 runtime checks. These are small gains on a metric saturating near 85-86.
[NLP-119] Abstention Is Not Threshold Compliance: Stated Scoring Rules in Open Language Models
【速读】： 该论文旨在解决大语言模型在决策过程中对置信度阈值（threshold）响应不一致的问题，即模型是否能够根据预设的置信度阈值（如0.75）合理地选择回答或“放弃回答”（abstain），从而实现符合预期的合规性行为。其核心挑战在于：尽管通过在提示中显式设定阈值t并引入“回答或放弃回答”的二元格式（abstention format）可显著提升模型的放弃率（从20%升至91%），但模型对阈值变化的敏感性极低——当阈值从0.5提高到0.9时，放弃率仅下降0.8个百分点，表明模型并未真正依据阈值进行动态调整。此外，缺乏明确格式约束时，更高的阈值反而导致放弃率降低，且部分检查点存在高达40%的样本无法被任何阈值无关的合规策略所解释，说明模型行为存在严重偏差。研究进一步发现，依赖自评估概率P(True)的信号受评分者和答案漂移的影响，不可靠；而基于后验校准的外部阈值虽在设计上保证覆盖率随阈值单调递减，但在跨数据集应用时表现不稳定。即使在训练阶段采用阈值条件化目标，模型在未见阈值下的行为仍不一致。因此，解决方案的关键在于：必须严格区分高放弃率与真实合规性，并通过系统性评估模型是否在实际决策中响应所声明的阈值、以及其判断是否与训练规则一致，才能准确衡量模型的可信决策能力。
链接: https://arxiv.org/abs/2611.14661
作者: Dongsun Moon
机构: Independent Researcher(独立研究员)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 34 pages, 3 figures, 29 tables
Abstract:Kalai et al. propose stating the scoring rule in the prompt with an explicit threshold t, so that a model should answer only when its probability of being correct exceeds t. We probe each requirement of compliance on shared items, in 12 checkpoints from three post-training stages of four fully open 3-8B recipes, with exploratory models up to 32B. At t = 0.75, a line limiting the reply to an answer or “I abstain” (the abstention format) raised the mean abstention rate from 20% to 91%. Yet within this format, raising t from 0.5 to 0.9 changed it by only -0.8 percentage points. Without the line, the higher threshold even lowered abstention, and up to 40% of a checkpoint’s items received decisions that no compliant policy with threshold-independent beliefs could make. Given the probability that their answer was correct, the checkpoints barely responded to the threshold in a forced-choice probe. Comparisons with self-assessment signals such as P(True) depended on the grader and on answer drift between prompts. An external threshold on recalibrated P(True) made coverage non-increasing in t by construction. Calibrated on the evaluation distribution (post hoc), it beat the abstention format at t = 0.5; calibrated on TriviaQA and applied to PopQA, however, it fell below that format at 0.75. On OLMo 3 7B, decisions trained on threshold-conditioned targets did not follow unseen thresholds consistently. A post-hoc variant trained at four thresholds answered less as t rose but still failed a pre-specified test, mainly on agreement with its training rule. A high abstention rate is not evidence of compliance: evaluations using stated scoring rules should report how often models abstain, whether decisions respond to the stated threshold, and which items they answer.
[NLP-120] Cleaning an Agents History Hurts When the World Still Holds the Error
【速读】： 该论文旨在解决生成式AI（Generative AI）代理在任务执行过程中因历史记录包含错误而引发的后续行为偏差问题，核心关注点在于：历史记录中对错误的修正方式及其与当前世界状态之间的匹配程度，如何影响后续决策模型的恢复能力。其解决方案的关键在于揭示“历史编辑的效果并非由内容本身决定，而是高度依赖于编辑后的历史与真实世界状态的一致性”。研究通过冻结代理在任务中的精确状态，分别修改历史记录或世界状态，并由接收模型继续推理，发现：在AgentCE-Bench上，添加一个真实但保留错误的注释可使恢复率从51%提升至66%，而将正确动作直接插入历史反而使恢复率降至44%；在AppWorld中，当数据库写入操作的确认反馈为成功但实际未生效时，整体任务成功率从45%下降至28%；而当效果实际存在时，成功率达到49%，两者差异达27个百分点。这表明，即使修正了历史内容，若世界状态仍保留错误，则“看似清洁”的历史可能适得其反，导致性能劣化。因此，关键结论是：历史修复必须与世界状态保持一致，否则错误的表征会误导后续代理，甚至比不修复更差。
链接: https://arxiv.org/abs/2611.14654
作者: Egor Pakhomov,Anurag Koul,Hongjun Liu,Bo Pang,Shafiq Joty,Erik Nijkamp
机构: Salesforce AI Research, USA( Salesforce 人工智能研究院，美国)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 23 pages, 6 figures; code and data in ancillary files
Abstract:Tool-using language-model agents accumulate histories of their own mistakes, and a system that compacts, repairs or hands off such a history decides what it says about those mistakes while their effects stay in the world. Prior work cleans the history at a fixed world, where showing a corrected step and saying it is done move together. We ask whether the same correct content helps or hurts depending on how the history states it and what the world holds. We freeze an agent mid-task with its exact world state, edit the history or the world, and let a receiver model continue. On AgentCE-Bench, after the agent commits a wrong value, we hold the world fixed and add the correct value by two natural edits: a truthful note beside the preserved error or the agent’s own completed action. The note raises recovery, the share of remaining targets the receiver gets right, from 51% under a matched neutral note to 66%. The completed action lowers it below the unedited history, from 50% to 44%, so the value’s presence alone does not predict the effect. On AppWorld, after a database write, we vary whether its receipt reports success or failure and whether its effect is present. A success receipt lowers whole-task success from 45% to 28% while the effect is absent and raises it from 38% to 49% once present. The two effects differ by 27 points pooled over two receivers, a difference positive in six of eight receivers at 97.5% and seven at 95%. A history edit’s effect thus depends on the world it describes, and a history made to look clean while the world still holds the error can leave an agent worse off than no edit at all.
[NLP-121] Evaluating the Performance of LLM s in Drafting NIH Data Management Plans
【速读】： 该论文旨在解决科研人员在撰写符合美国国立卫生研究院（NIH）2023年数据管理计划（DMP）指南要求时所面临的知识匮乏与时间压力问题，尤其关注如何通过生成式人工智能（Generative AI）降低研究人员及图书馆员在制定合规DMP过程中的负担。其解决方案的关键在于评估大型语言模型（LLM）在自动生成符合规范的DMP初稿方面的性能，发现尽管LLM能够显著加速初始草案的生成，但生成内容仍需领域专家进行审查以确保清晰性与完整性；同时研究指出，开源类LLM相较于商业模型在直接支持资助方合规性方面尚存在差距，表明其实际应用仍需进一步优化与人工干预。
链接: https://arxiv.org/abs/2611.14648
作者: Nahid Zeinali,James ONeil,Xuebin Dong,Rebecca Hofstein Grady,Maria Praetzellis,Brian Riley,John Chodacki,Bhavesh Patel
机构: 未知
类目: Digital Libraries (cs.DL); Computation and Language (cs.CL)
备注:
Abstract:Sharing research data is essential to promote transparency, enable reproducibility, and fuel scientific discoveries. To ensure consistent sharing, the National Institutes of Health (NIH) and other funding agencies require applicants to submit data management plans (DMPs) with their proposals. They aim to encourage researchers to consider their data management and sharing strategies early on. However, researchers often face challenges when developing DMPs due to a lack of knowledge about data-sharing practices. Here, we evaluated the performance of Large Language Models (LLMs) in drafting DMPs compliant with the 2023 NIH guidelines. The goal was to investigate whether LLM-supported DMP creation could reduce the burden on researchers and librarians. We compared Llama and GPT performance against NIH-provided DMP samples. We also asked domain experts to evaluate LLM-generated DMPs. Overall, we found that LLMs can accelerate the creation of initial drafts but still require expert review before submission to ensure clarity and completeness. Moreover, open-source LLMs may not be as ready off-the-shelf as commercial ones for drafting funder-compliant DMPs.
[NLP-122] Prevalence Bounds for Edited Watermarked Text: Power Envelopes and Calibration Sensitivity
【速读】： 该论文旨在解决在未知编辑程度下，如何准确估计一个语料库中由水印语言模型生成的文档所占比例（即“水印占比”）这一关键问题。传统水印检测器仅判断单个文档是否源自水印模型，而审计场景更关注整体语料库中被水印覆盖的比例，这在实际应用中具有重要意义。其解决方案的核心在于：通过结合已知最不利的下一个词分布（least-favourable next-token distribution）与序贯耦合（sequential coupling）技术，构建一个以三个预算为参数的功率包络（power envelope）——包括保留下限（retention floor）、熵预算（entropy budget）以及原地替换预算（budget for in-place substitutions）。该包络能够刻画在任意编辑条件下，Gumbel-max和绿色列表（green-list, KGW）水印在编辑后仍可被识别的极限性能，使被识别集合对主导的替代假设具有紧致性（sharpness），并提供有限样本下固定语料库的真实水印占比的置信区间、基于精确零假设或独立人类样本校准后的零假设的统计推断，以及适用于数据流的置信序列。实验表明，在OPT-1.3B与Qwen3-1.7B生成文本上，12,000条新生成的人类文本在多密钥评估下未显示出显著偏离精确零假设的证据，且在随机编辑、替换与改写等扰动下的重采样试验中所有边界均保持覆盖；当水印占比为0.2时，精确零假设下的置信区间约为[0.14, 0.38]。然而，研究也揭示了该方法的脆弱性：当使用大模型改写生成的KGW语料时，其表现可能低于包络线，但2000文档规模下边界仍能保持覆盖；若存在0.01量级的模拟零假设偏差（与观测最大偏差相当），在2000文档时影响可忽略，但在10万文档时会破坏精确零假设覆盖性，而通过独立人类样本校准的零假设虽可维持覆盖性，但需依赖大规模参考样本以保持边界紧凑性。
链接: https://arxiv.org/abs/2611.17035
作者: Kuancheng Ye
机构: University of California, Irvine(加州大学欧文分校)
类目: Machine Learning (stat.ML); Computation and Language (cs.CL); Cryptography and Security (cs.CR); Machine Learning (cs.LG)
备注: 31 pages, 3 figures, 8 tables. Code and results: this https URL
Abstract:Watermark detectors decide whether one document came from a watermarked language model. Auditors often need a different number: the share of a corpus that did. We study this share for Gumbel-max and green-list (KGW) watermarks when every document may have been edited by an unknown amount. Without restrictions on the edits, the upper end of the identified set is 1. Combining a known least-favourable next-token distribution with a sequential coupling, we derive a power envelope indexed by three budgets: a retention floor, an entropy budget, and a budget for in-place substitutions, which for Gumbel-max make edited positions look less watermarked than human text. The envelope gives an identified set that is sharp for the alternatives it dominates, finite-sample bounds on the realized share of a fixed corpus with either the exact null or a null calibrated on an independent human sample, and confidence sequences for streams. In Gumbel-max simulations where the envelope binds, the bounds cover, and the threshold that maximizes the gap between envelope and null is generally as tight as twelve thresholds. On text from OPT-1.3B and Qwen3-1.7B, 12,000 fresh human documents scored under three keys show no robust departure from the exact null, and all bounds covered in resampling trials from finite pools of documents with random edits, substitutions and paraphrases; at a share of 0.2 the exact-null bound is about [0.14,0.38]. These results are fragile in two ways we make precise. LLM-paraphrased KGW pools fall below the envelope, yet at 2000 documents the bounds still covered in these trials. A simulated null error of 0.01, comparable to the largest deviation we observed at our thresholds, is likewise harmless at 2000 documents but breaks exact-null coverage at 100,000; a null calibrated on an independent human sample keeps coverage there, but needs a large reference sample to stay narrow.
[NLP-123] Aggregate Accuracy Hides Diagnosis-Specific Failure in Medical Vision-Language Models NEURIPS2026 NEURIPS
【速读】： 该论文旨在解决当前医疗视觉-语言模型（Medical VLMs）评估中普遍存在的“平均准确率掩盖关键诊断失败”的问题，即尽管模型在整体性能上表现良好，但在特定疾病（如气胸）上的漏诊率极高，而这一严重缺陷被平均精度所掩盖。其核心解决方案在于揭示并验证某些主流医学VLMs在面对气胸等关键病灶时存在系统性误判——这些模型在未正确识别病灶的情况下仍能生成流畅、具解剖学细节的错误解释，声称“存在”典型征象却最终判定为“无”。这种现象并非由提示工程、图像分辨率、标签模糊或模型规模等因素引起，而是源于模型内部表示与输出决策之间的脱节：尽管模型的隐藏层中仍保留了气胸的可识别特征信号（线性探测可恢复），但输出阶段却被强制抑制，导致零召回率且无不确定性表达。因此，论文强调必须采用按病种划分的召回率与特异性指标，在平衡样本下进行独立评估，才能真实反映模型对特定病理的感知能力，避免仅依赖聚合准确率带来的误导性结论。
链接: https://arxiv.org/abs/2611.14975
作者: Shreya Ajay Kale,Angad Ripudaman Singh Bajwa
机构: Independent Researcher(独立研究员)
类目: Image and Video Processing (eess.IV); Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: Accepted at the NeurIPS 2026 Workshop on Medical Reasoning with Vision Language Foundation Models (Med-Reasoner). 16 pages including NeurIPS checklist, 1 figure, 5 tables
Abstract:Medical vision-language models (VLMs) are compared mainly by accuracy averaged across findings, a summary that can hide complete failure on a single diagnosis. We audit five open medical VLMs (MedGemma-4B, Lingshu-7B, HuatuoGPT-Vision-7B, Qwen2.5-VL-7B, and MedVLM-R1) on 6,048 ChestX-ray14 questions balanced by sex, age, finding, and label. On the 432 pneumothorax-positive X-rays, HuatuoGPT-Vision and Qwen2.5-VL detect zero cases while MedGemma and Lingshu detect 38% and 67% of the same images. The two failing models answer “no” to all 864 pneumothorax cases (balanced accuracy 0.500, chance), yet produce fluent, anatomically specific explanations that name the signs of a collapsed lung and then deny them, with no hedge or abstention. This failure is not an artifact of prompting, image resolution, label ambiguity, or model scale. It persists across four prompt formats, at native resolution, and on single-label cases. A within-model control shows the same forced prompt gives HuatuoGPT 79% recall on nodule but only 6% on pneumothorax. It survives a tenfold scale increase (Qwen2.5-VL-72B detects 4%) and replicates on SIIM-ACR, an independent dataset with radiologist-drawn masks, where the failing models detect 0 of 432 confirmed pneumothoraces, including 0 of the 144 largest lesions. Reading the models’ logits and internal states shows this is not blindness. The two failing models are highly confident yet no better than chance and a linear probe recovers pneumothorax from their hidden representations even where recall is zero, so the finding is represented internally but suppressed at the output. Aggregate accuracy therefore cannot establish that a medical VLM can see a given finding, because a model that misses one can still describe it convincingly. Reporting per-finding recall and specificity on balanced cases is needed to tell the difference.
[NLP-124] Partial pooling beats synthetic respondents as estimators not as covariates: a preregistered small-sample tournament
【速读】： 该论文旨在解决基于小样本单元（如按性别与年龄分组的媒体使用情况）进行人口群体调查估计时，因每单元受访者过少而导致估计精度不足的问题。现有文献提出利用大语言模型生成合成受访者以填补数据空白，但尚未在真实小区域估计场景下与传统方法进行系统对比。本研究在韩国媒体调查面板数据上，对十种估计器进行了预注册测试，采用相同的聚类家庭划分方式，在五个校准预算和三种保留策略下进行评估。结果表明，属于多层部分分池（Multilevel Partial-Pooling）模型家族的加权调整后多层回归预测（MRP）方法在所有预算条件下均优于仅依赖校准的其他估计器，其在100名受访者下的表现相当于直接估计在约400名受访者下的精度。对于未观测群体，将先验信息作为固定偏置处理时，无论是合成样本还是前一期数据均无法有效传递信息；而将其作为偏差校正项引入时，虽可实现信息转移，但伴随估计噪声增加。在配对决策规则下，未发现合成样本相对于MRP的显著优势。然而，当将合成样本作为协变量嵌入多层模型中构建探索性混合模型时，该方法在未观测群体上的表现优于标准MRP，表明信息得以保留并有效利用。值得注意的是，区间覆盖率始终低于名义水平，且随着校准预算增加，部分分池模型的覆盖性能反而下降，与理论预期相悖。因此，该研究的关键解决方案在于：将合成数据以协变量形式融入多层模型框架，而非独立生成或简单校准，从而实现信息的有效整合与利用。
链接: https://arxiv.org/abs/2611.14526
作者: Howard Kim
机构: College of AI Convergence, Seoul Cyber University (首尔网络大学人工智能融合学院)
类目: Applications (stat.AP); Computation and Language (cs.CL); Computers and Society (cs.CY)
备注: 28 pages, 3 figures, 3 tables, supplementary information included. Preregistration: this https URL . Reproduction package: this https URL
Abstract:Survey estimates for population segments, such as media use by sex and age, rest on few respondents per cell, and a growing literature offers synthetic respondents generated by large language models to fill the gap. That literature has not tested them against the small-area estimators a methodologist would otherwise use. We ran ten estimators, preregistered, on identical household-clustered splits of a Korean media panel across five calibration budgets and three holdout regimes. A multilevel partial-pooling model of the MRP family led every calibration-only estimator at every budget, matching at one hundred respondents what direct estimation reaches with roughly four times as many. On unobserved segments, information fixed as a per-segment prior was inert for synthetic panel and previous wave alike; attached as a bias correction it transferred, at the price of estimation noise. No registered synthetic advantage over MRP survived the paired decision rule. Entered instead as a covariate inside the multilevel model, an exploratory hybrid, the same panel improved on MRP on unobserved segments: the information survives where the registered constructions did not. Interval coverage stayed below nominal throughout and worsened with budget for partial pooling, contrary to prediction.
信息检索
[IR-0] Sheaf-Inspired Structured Scoring Extensions of LightGCN for Collaborative Filtering
链接: https://arxiv.org/abs/2611.17226
作者: Ksenia Belkovskaya,Timur Lidzhiev,Maria Stebliy,Alexander Tarakanov
类目: Information Retrieval (cs.IR); Computational Engineering, Finance, and Science (cs.CE); Computational Geometry (cs.CG); Machine Learning (cs.LG)
备注:
Abstract:Graph-based collaborative filtering learns user and item representations by propagating embeddings over a user–item interaction graph. We investigate a modular, sheaf-inspired scoring extension placed on top of a three-layer LightGCN backbone. The motivation comes from cellular sheaves, where local vector spaces are related by linear restriction maps; here, this idea motivates structured, pair-dependent transformations of user and item embeddings rather than a change to graph propagation. We compare the identity inner-product scorer, diagonal, rank-one low-rank, and full matrix residual scorers, and a parameter-matched residual MLP baseline (\textttmlp_full) on Amazon-Book, Gowalla, MovieLens-1M, and Yelp2018. We report ranking quality and catalog coverage at cutoffs 10 and 20. The selected structured variants show modest, dataset-dependent changes in ranking quality and sometimes substantial changes in coverage. Our method is inspired by sheaf geometry but does not construct a cellular sheaf or implement sheaf-Laplacian diffusion; the experiments therefore do not establish a causal benefit from sheaf geometry itself.
[IR-1] Evaluation of Music Version Retrieval Methods for Setlist Identification in Live Concerts
链接: https://arxiv.org/abs/2611.17070
作者: Andres Ferraro,Albin Correya
类目: Information Retrieval (cs.IR); Sound (cs.SD)
备注:
Abstract:Automatic setlist identification aims to identify the songs performed in live concert recordings by matching noisy and variable-quality audio against a database of reference tracks. This task presents challenges beyond conventional music version identification, including substantial arrangement differences, background noise, and the need to process continuous recordings efficiently at large scale. In this work, we evaluate modern music version retrieval methods for setlist identification using the ASID benchmark, comparing the established Re-MOVE system with CLEWS and NAVE, an internal neural audio version embedding model. Our results show that NAVE and CLEWS substantially outperform Re-MOVE. Notably, NAVE uses 256-dimensional embeddings, compared with 1024 dimensions for CLEWS, providing a four-fold reduction in embedding storage while achieving better retrieval performance. We further evaluate music/non-music gating and find modest improvements in detection and temporal coverage, while qualitative analysis suggests that such preprocessing can improve recognition boundaries beyond what the standard ASID metrics capture. Overall, the results demonstrate that segment-level audio embeddings provide a strong basis for scalable setlist identification and that compact representations can achieve strong accuracy under realistic live-concert conditions at much lower storage and serving cost.
[IR-2] Similar but Not Redundant: Embedding-Gated Refinement for High-Granularity Topic Models
链接: https://arxiv.org/abs/2611.17030
作者: Daniel Denley,Robert L. Nicol,Mike J. Chantler
类目: Computation and Language (cs.CL); Information Retrieval (cs.IR)
备注: 24 pages, 6 figures, 13 tables
Abstract:High-granularity topic modelling extracts hundreds of niche topics from a corpus, yet this regime is rarely studied and is more commonly framed as a failure mode: “over-clustering,” in which coherent themes fragment into redundant subtopics. We argue that this framing conflates genuine redundancy with topics that are semantically similar but meaningfully distinct, and that the central challenge in high-granularity modelling is separating the two. Recent post-hoc refinement methods address topic redundancy through either embedding-based similarity thresholding or LLM-guided merging. We show that neither suffices in high-granularity settings and propose embedding-gated iterative merge (EGIM), a hybrid approach that iteratively generates context-rich topic representations, uses embedding similarity to propose candidate merges across the topic space, then applies LLM judgement to accept only genuinely redundant pairs while preserving niche distinctions. Across two patent corpora and one legal corpus, LLM judgement rejects 83% of embedding-identified candidates as non-redundant. EGIM improves discriminability on merged topics by 25-55% over a prior iterative LLM baseline while better preserving the original topic distribution.
[IR-3] Anonymous in Name Only: A Critical Review and a Reidentification Exercise on the European Commissions Anonymization Measures for Google Search Data under Article 6(11) of the DMA
链接: https://arxiv.org/abs/2611.17012
作者: David Sánchez,Benet Manzanares,Josep Domingo-Ferrer
类目: Cryptography and Security (cs.CR); Information Retrieval (cs.IR)
备注:
Abstract:In July 2026, the European Commission (EC) set out the anonymization measures Alphabet must implement to share Google Search data with competing search engines under Article 6(11) of the Digital Markets Act. As the first concrete illustration of how the EC operationalizes anonymization under the EU digital rulebook, this methodology could become a regulatory benchmark. In this paper, we show why it should not. We critically review the anonymization measures proposed by the EC, from the characterization of residual risk to the suppression thresholds and the reliance on contractual safeguards. For each of those issues, we highlight significant conceptual flaws, and, where applicable, we illustrate these flaws with empirical observations from real Google Search data provided by Alphabet and protected under the EC’s measures. Importantly, we present examples of low-effort reidentification, record linkage, and confidential attribute inference achieved through straightforward and automatable attack patterns. Our findings show that the approach of the EC relies on unjustified heuristics rather than formal privacy guarantees, underestimates the risks inherent in free-text search queries, and sacrifices data utility without achieving meaningful protection. We also demonstrate that the proposed measures do not meet any of the three anonymization criteria established by the European Data Protection Board (EDPB): no record isolation, no linkage, and no inference. Practitioners and regulators are therefore cautioned against treating these measures as a sound anonymization practice.
[IR-4] SignalAudit-QD: A Reproducible Audit Resource for External Signals in AI-Query Forecasting
链接: https://arxiv.org/abs/2611.16947
作者: Amir Rafe,Iman Jamshidi,Gourab Sil,Subasish Das
类目: Information Retrieval (cs.IR)
备注:
Abstract:SignalAudit-QD is a reproducible audit resource for testing whether a dated public signal adds information beyond query history in a fixed 20-term U.S. AI panel. It aligns two archived Google Trends extractions with eight USPTO AI-patent publication streams and releases origin-indexed predictions, paired losses, provenance, executable tests, and tamper-evident run manifests. Both inputs are retrospective reconstructions obtained after the modeled 2020–2023 origins; the task is therefore a pseudo-origin assay within frozen measurement vintages, not a historical deployment simulation. In the designated one-month comparison, adding a two-component patent block to a level ridge changes macro mean absolute error by +0.0012 Trends points, with conditional 95% moving-block interval [-0.0363,+0.0387]. This establishes neither aggregate benefit nor equivalence. Alternative dimensions, semantic mapping, patent definitions, a second extraction, and a three-month horizon remain unresolved; persistence-anchored and nonlinear hosts instead show model-specific worsening, tiny for the anchored fit. Matched timing controls expose availability assumptions, while a planted-signal assay detects only the largest tested injection. A within-query high-interest alert demonstrates the released interface but not online utility. The contribution is a bounded negative case study and an extensible evidence schema for auditing external-signal claims in temporal information retrieval.
[IR-5] KuaiRecV2: Benchmarking Large-Scale Continual Learning for Diversified and Multi-task Recommendation NEURIPS2026
链接: https://arxiv.org/abs/2611.16795
作者: Chenxu Li,Shuchang Liu,Hantao Shu,Wei Yuan,Zhe Xu,Guoqing Hu,Yan Wang,Biao Yang,Xiang Li,Cheng Ling,Fan Yang,Yongqi Liu,Lantao Hu,Han Li,Kaiqiao Zhan,Kun Gai,An Zhang,Xiang Wang
类目: Information Retrieval (cs.IR)
备注: This paper has been accepted to NeurIPS 2026
Abstract:Recommender systems, which retrieve items of interest to users based on their interaction histories, have recently witnessed several key technological advancements, benefiting from the semantic representation space and model-data scaling. Nevertheless, practitioners and researchers have frequently encountered critical inconsistencies between simple offline evaluations and complex online results, and this gap still exists in the era of large models. In this work, we address several key factors that contribute to this inconsistency: 1) the \textitlarge-scale continual learning challenge that assumes a practical demand to maintain long-term model effectiveness under dynamic item pools and distributional shifts; 2) the \textitcombinatorial diversity challenge that requires the generation strategy to solve the accuracy-diversity trade-off in both the finer token level and the holistic item level; and 3) the \textitmulti-task reward balancing challenge, which is particularly critical for modern generative models that can accurately model the sequence distribution but are less controllable for the multi-task demands. To address these challenges and further pave the way for the development of more realistic recommendation solutions, we introduce KuaiRecV2, a deliberately constructed benchmark with industrial-level datasets from Kuaishou. Specifically, we provide a large-scale interaction dataset with million-scale multi-modal item information and long-term interaction records across 30 days, with a dynamic video candidate pool. Additionally, we provide a rigorous benchmark that evaluates retrieval accuracy, diversity, and multi-task performance in one rubric system, where an effective advancement occurs if and only if the new method improves all metrics. The code is available at this https URL Comments: This paper has been accepted to NeurIPS 2026 Subjects: Information Retrieval (cs.IR) Cite as: arXiv:2611.16795 [cs.IR] (or arXiv:2611.16795v1 [cs.IR] for this version) https://doi.org/10.48550/arXiv.2611.16795 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[IR-6] Active Learning of Approximate LLM Ranking for Energy-Efficient Retrieval of Long Queries
链接: https://arxiv.org/abs/2611.16488
作者: Ohad Porat,Elad Yom-Tov
类目: Information Retrieval (cs.IR)
备注:
Abstract:Search engines and Large Language Models (LLMs) offer complementary strengths for document retrieval: lexical retrievers which rely on keyword matching are fast and energy-efficient, whereas LLMs capture semantic meaning and excel at interpreting long, context-rich queries. However, this comes at a cost that is computationally and energetically prohibitive. We propose the LLM Active Approximation Model (LAAM), a hybrid retrieval framework that introduces a lightweight learned predictor which serves as an intermediary between semantic retrieval and LLM retrieval. LAAM is refined through an active learning loop where an LLM labels a small number of documents. By concentrating LLM evaluations on the most informative documents, the framework substantially reduces inference cost while preserving retrieval effectiveness. We evaluate the proposed framework on TREC Robust 2004 using Lucene with BM25 retrieval and DeepSeek v3 as the LLM, under three levels of query verbosity. For verbose queries, LAAM achieves 87% of average P@10 achieved by DeepSeek’s reranking (0.459 vs. 0.526), while requiring only 3.3% of the per-query energy consumption. LAAM is statistically significantly better than BM25 and is indistinguishable from DeepSeek on average precision, though not on MAP (Friedman test, P0.05 with post-hoc testing). Overall, our results demonstrate that selectively approximating LLM relevance judgments through active learning enables retrieval accuracy to be controlled directly through the evaluation (and consequently energy) budget. Rather than committing to either inexpensive (and less accurate) lexical retrieval or exhaustive (and expensive) LLM reranking, our framework exposes a continuous spectrum of operating points along the energy–accuracy curve, allowing practitioners to select how to trade retrieval quality with computational budget. Subjects: Information Retrieval (cs.IR) Cite as: arXiv:2611.16488 [cs.IR] (or arXiv:2611.16488v1 [cs.IR] for this version) https://doi.org/10.48550/arXiv.2611.16488 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Elad Yom-Tov [view email] [v1] Thu, 12 Nov 2026 08:59:23 UTC (2,393 KB)
[IR-7] RMIT-ADMS at TREC RAG 2026: Research Agents Retriever Ablations and Open-Weight Replication
链接: https://arxiv.org/abs/2611.16425
作者: Oleg Zendel,Kun Ran,Chenglong Ma,Nguyen Khoi Vo,Shuai Wang
类目: Information Retrieval (cs.IR); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:We describe the RMIT-ADM+S team’s participation in the TREC 2026 Retrieval-Augmented Generation (RAG) track using the ClimbMix corpus. We submitted seven research-agent runs using proprietary OpenAI and open-weight models, exploring different retrieval strategies, writer models, and agent workflows. Pre-submission ablations identified writer selection and adjacent-chunk expansion as the main factors affecting answer quality. In the official NIST evaluation, six of seven runs exceeded the track median across all three rubric/nugget collections. Our best-performing run, brief-hybrid, achieved the highest all-item coverage among our submissions and tied the track maximum (0.892) for Batch-1 vital coverage. Human side-by-side preferences differed across evaluation batches: the open-weight owa-glm5 led Batch 1 despite lower automated coverage, while sol-research-first led Batch 2. These findings highlight differences between automated coverage measures and human preferences, as well as the potential of lower-cost open-weight models for research-agent systems.
[IR-8] AntEmbed: Dual-Mode Multimodal Embedding Model for Efficient and Fine-Grained Retrieval
链接: https://arxiv.org/abs/2611.16405
作者: Kehang Jia,Teng Liu,Juntao Li,Hua Chen,Jinxiong Chang,Xiaobo Liang,Min Zhang
类目: Information Retrieval (cs.IR)
备注: 13 pages
Abstract:Discriminative embeddings encode original multimodal inputs efficiently but struggle with fine-grained semantics. Generative embeddings use reasoning to capture query intent and relevance evidence, but incur higher computational costs. We propose AntEmbed, a dual-mode multimodal embedding model that supports efficient retrieval and reasoning-enhanced fine-grained matching in a unified embedding space. First, we construct retrieval-oriented reasoning content that captures query intent and confusion-aware semantic evidence. We treat original and reasoning-enhanced candidate inputs as semantically equivalent views, and use multi-positive contrastive learning with a single embedding token to align discriminative and generative embeddings. Second, cross-mode self-distillation transfers fine-grained semantic information from reasoning-enhanced to original representations through candidate similarity-distribution distillation and query representation alignment. This strengthens discriminative embeddings without requiring additional online reasoning. Both retrieval modes share a single candidate index. On MMEB-v2, AntEmbed-8B achieves an overall score of 81.5 with reasoning-enhanced embeddings, outperforming the strongest baseline by 3.1 points.
[IR-9] ReSID: Rethinking Long-Tail Retrieval in Generative Recommendation through Item Reachability
链接: https://arxiv.org/abs/2611.16404
作者: Muyu Zou,Shunye Tang,Yi Xu,Haibo Xing,Yandong Wang,Lingyu Mu,Jinxin Hu
类目: Information Retrieval (cs.IR)
备注:
Abstract:Generative recommendation (GR) represents items with discrete Semantic IDs (SIDs) and casts retrieval as sequence generation. Under skewed interaction distributions and tight serving budgets, GR struggles to retrieve long-tail items. In this work, we are the first to study long-tail reachability across the complete pipeline of SID construction, decoding, and matching. We find that under equal code lengths, parallel SID construction consistently outperforms sequential SIDs on tail items. Moreover, even when full generation fails, decoded sequences carry locally consistent information that retrieves relevant tail items via partial matching. Yet this recovery is bounded by upstream search, failing once beam search prunes all prefixes completable within a fixed mismatch tolerance. Motivated by these findings, we propose ReSID to jointly preserve and recover tail candidates. Built on long parallel SIDs, ReSID pairs popularity-debiased and user-semantic-aware beam search to keep promising paths alive with partial SID matching to recover items from surviving tokens. ReSID achieves state-of-the-art long-tail retrieval while maintaining competitive overall accuracy, with lower latency and higher throughput under narrow beams. ReSID is deployed in our production system, serving main user traffic with substantial business gains. We release our complete code and model checkpoints to support full reproducibility and real-world deployment at this https URL.
[IR-10] riGen: A Generative Paradigm for Trigger-Induced Recommendation
链接: https://arxiv.org/abs/2611.16395
作者: Muyu Zou,Haibo Xing,Shunye Tang,Liyan Tie,Lingyu Mu,Yandong Wang,Jinxin Hu
类目: Information Retrieval (cs.IR)
备注:
Abstract:Generative recommendation typically operates under the User-Induced Recommendation (UIR) setting, predicting target items by autoregressively generating semantic IDs as a passive continuation of long-term user history. However, this paradigm overlooks Trigger-Induced Recommendation (TIR), an essential industrial scenario where a user click triggers an immediate follow-up feed. In TIR, recommendations must dynamically align with the instant trigger intent without compromising historical preference continuity. Traditional TIR methods rely on discriminative scoring and cannot guide token-level generation, while existing generative models lack mechanisms to reconcile this structural tension. To address this challenge, we propose TriGen, establishing a tailored generative paradigm for trigger-induced recommendation. TriGen introduces calibrated tail injection to re-encode local history, adaptively arbitrates between persistent tastes and immediate impulses via dynamic gating, and applies position-specific transformations to sustain level-aware trigger guidance across decoding steps. Extensive offline evaluations confirm consistent gains over competitive baselines. Furthermore, online A/B tests across six country markets demonstrate significant revenue and GMV improvements, leading to the full production deployment of TriGen.
[IR-11] CURATOR: Scalable Fine-grained Evidence Selection for Retrieval-Augmented Generation using Decision Models
链接: https://arxiv.org/abs/2611.16366
作者: Shib Dasgupta,Bandish Parikh,Rashmi Gangadharaiah
类目: Information Retrieval (cs.IR)
备注:
Abstract:Retrieval-augmented generation (RAG) provides entire retrieved passages to the LLM, even though only a handful of spans within them contain the evidence pertaining to the question. While precise selection of these spans can improve context utilization, identifying them in isolation is challenging. A span may appear irrelevant by itself, and deciding whether to retain it may require examining other retrieved information. One can evaluate each span’s candidacy with the full retrieved context, but this is computationally inefficient as the same text must be processed repeatedly. Pretrained decision model can be used to efficiently evaluate multiple spans simultaneously, allowing each decision to consider the surrounding spans as well. However, our experiments show that their ability to identify relevant evidence deteriorates as the number of simultaneous decisions increases. To address this, we introduce CURATOR, a fine-tuned decision model trained to make reliable evidence-selection decisions over large candidate sets. By training a decision model to select varying numbers of supporting spans from large candidate pools, CURATOR develops the capability to select evidence for RAG while retaining the computational efficiency of joint decision-making. We observe that CURATOR maintains a strong evidence discrimination capability as the number of candidates increases, generalizes across different tasks and document domains, and improves downstream RAG performance, particularly when the available context is limited. Subjects: Information Retrieval (cs.IR) Cite as: arXiv:2611.16366 [cs.IR] (or arXiv:2611.16366v1 [cs.IR] for this version) https://doi.org/10.48550/arXiv.2611.16366 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[IR-12] GA-HEL: A Knowledge-Driven Expert System for Open-Set Granularity-Aware Hierarchical Event Linking in Social Media
链接: https://arxiv.org/abs/2611.16194
作者: Shengqi Zhou,Zhigang Liu,Wenxiang Zhu,Zhentao Cang
类目: Information Retrieval (cs.IR)
备注: 16 pages, 6 figures
Abstract:Social-media event intelligence requires a system to determine whether a post belongs to a registered event space, identify the supported hierarchy path, and stop at the deepest level warranted by the evidence. We formulate this requirement as open-set hierarchical event linking over frozen Storyline–Episode–Atomic Event Cards. GA-HEL implements a three-stage, knowledge-driven expert-system funnel: open-set scope control limits candidate exposure before hierarchical reasoning, hierarchy-aware decoders select a supported path and depth, and deterministic legal-path validation checks each record before database-compatible use. A 200-post Pilot supports task and mechanism development, while a separately annotated 497-post Expansion set evaluates transfer across four held-out Storylines and a hard-negative pool. Method implementations, prompts, Event Cards, candidate order, and six systems’ predictions were frozen before Expansion Gold annotation. The resulting evaluation combines end-to-end hierarchical matching, open-set rejection, structural validation, ablation, cross-Storyline analysis, and an independent expert reannotation of a fixed audited subset. GA-HEL remains competitive with strong global full-card reasoning while providing explicit scope control and hierarchy-valid structured records. Error analysis identifies unsupported descent from Episode-level evidence to Atomic Events as the dominant remaining challenge. The task definition, frozen resource, and auditable protocol provide a reproducible foundation for scope-aware event intelligence. Code, Event Cards, labels, frozen predictions, and evaluation scripts are available at this https URL.
[IR-13] From Tokens to Behaviors: Rethinking Generative Recommendation for User Trajectory Prediction
链接: https://arxiv.org/abs/2611.16159
作者: Jiaqi Wang,Hanchen Yang,Hengyu Zhang,Hongxu Ma,Wenqi Rao,Hongyang Wang,Zhongbo Sun,Renliang Qian,Weixin Chen,Wengen Li,Jihong Guan,Shuigeng Zhou,Qingpeng Cai,Peng Jiang,Hong Cheng
类目: Information Retrieval (cs.IR)
备注:
Abstract:User behavior trajectory prediction (UBTP) aims to capture how a user’s needs evolve over time by predicting the sequence of future interactions from that user’s interaction history. Recent generative recommendation (GR) methods address this task through autoregressive generation of semantic IDs (SIDs). However, serial SID decoding couples token generation with behavior state transitions and requires multiple dependent steps per interaction. Discrete item feedback overlooks predictive uncertainty, and teacher-forced training leaves a training-inference gap in trajectory generation. To address these limitations, we propose TraGR, a behavior-aware generative recommendation framework that aligns generation, feedback, and learning with successive user behaviors. Its behavior-level architecture predicts SID tokens in parallel within each interaction and models behavioral transitions through recurrent state updates, with joint SID and item-level supervision. Probability feedback combines the selected behavior with a probability-weighted summary of alternatives to guide subsequent predictions. Relation-aware trajectory learning incorporates collaborative transition signals and later credit assignment into training on model-generated trajectories. Experiments on four public datasets demonstrate that TraGR consistently outperforms state-of-the-art baselines across different prediction horizons. Code: this https URL.
[IR-14] HyperPage: Learned Page Layouts for Disk-Resident Graph Vector Indexes
链接: https://arxiv.org/abs/2611.16149
作者: Xinyu Zhang,Jingze Li,Baihua Zheng,Weiwei Sun
类目: Databases (cs.DB); Information Retrieval (cs.IR)
备注: 14 pages, 4 figures, 7 tables. Code available at: this https URL
Abstract:Disk-resident graph indexes such as DiskANN store each vector with its graph edges as one record on SSD pages. A search needs the page of every record it expands, so its page reads depend on the page layout, that is, which records share a page. Search agents trained with reinforcement learning query one such index run after run, and every graph traversal can be logged with the records it expands. We therefore use the logs of a finished run to lay out the index for a declared memory budget, so later runs read fewer pages. In contrast, existing layouts such as Starling and MARGO are computed from the graph and vectors and use no logs. Learning a layout from logs is challenging, because the logs must stay valid under every layout, the saving depends on what the budget caches, and the best layout is NP-hard to find. Our key observation is that DiskANN-style search routes on in-memory codes and the records it expands, never on their page-mates. As a result, the logged traversals price every layout by replay, and cached records add no page reads. Accordingly, we propose HyperPage, which first builds layouts for the declared budget on the logged traversals of uncached records. It then selects one layout, together with the budget split between hot records and a page cache, by replaying validation steps. In addition, a Lagrangian lower bound shows how far a layout may be from the optimum. We learn from one Search-R1 agent run and test on a later run with a different reward, over 1M and 21M passages. Across seven cache budgets, HyperPage reads 5.2% to 21.2% fewer pages than MARGO, and no compared layout reads fewer. The advantage is largest when the cache is small. For example, at 1% of the 1M index, MARGO needs nearly 3 times the budget to read as few pages. On the 21M index, HyperPage also cuts search time per step by up to 10.4%. The saving holds for a second agent model and a second corpus.
[IR-15] Calibrated Early Termination with Bounded Optimality Gaps for Constrained Vector Search
链接: https://arxiv.org/abs/2611.16148
作者: Xinyu Zhang,Jingze Li,Baihua Zheng,Weiwei Sun
类目: Databases (cs.DB); Information Retrieval (cs.IR); Machine Learning (cs.LG)
备注: 14 pages, 5 figures, 3 tables. Code available at this https URL
Abstract:Vector search over learned embeddings is the retrieval layer of recommendation and retrieval-augmented generation, and many applications cap how many answer items one category may contribute. Graph indexes answer a query by best-first traversal, and where it stops determines its cost and whether its answer is exact. We study early termination for such constrained k -nearest-neighbor ( k -NN) queries, with plain k -NN as a special case, and seek the cheapest stopping rule whose answer is exact with probability at least 1-\varepsilon . Learned early-termination methods instead target average recall, which does not guarantee exact answers. Multiple testing can calibrate any parameterized stopping rule to our target, but it cannot tell how much a better rule with the same risk and observations would save. Answering this needs a lower bound on the optimum of a rule family too large to enumerate, valid for the query population. Our key idea is to declare the observations a rule may use and to consider every rule monotone in them. Risk-constrained stopping then becomes a linear program over an order polytope, whose Lagrangian minimizers form a nested chain of staircase rules computed exactly by dynamic programming. Fixed-sequence binomial tests along this chain calibrate our rule, CaST, whose failure probability is at most \varepsilon with confidence 1-\delta . Lagrangian weak duality bounds the optimum of the declared class from below at all risk levels at once, without a union bound over the rules. On two datasets of one million items each, CaST replayed on logged searches costs at most 1.132 times the class optimum for risk levels from 0.03 to 0.10, at 95% joint confidence per setting. Larger savings thus need a rule outside the class. At a common risk target on held-out queries, CaST is faster than every other calibrated rule we time, which take 4.8% to 45.5% more mean time.
[IR-16] REWEAVE: Online Learning for Workload-Adaptive HNSW Indexes
链接: https://arxiv.org/abs/2611.16060
作者: Minghao Li,Nick Koudas,Raghav Mittal
类目: Databases (cs.DB); Information Retrieval (cs.IR); Machine Learning (cs.LG)
备注: 20 pages, 9 figures, 5 tables. Full version with appendix
Abstract:Vector databases rely on Hierarchical Navigable Small World (HNSW) indexes for similarity search, but a graph built from data alone can be poorly matched to a skewed or changing query workload. We present REWEAVE (Reward-Guided Rewiring of Vector Graphs), an online index-maintenance method that learns edge utility from query feedback and periodically rewires the bottom layer within the original per-node degree cap. The method supports sparse exact-neighbor feedback combined with self-labels, and standard HNSW serving without learning overhead once adaptation is disabled. Under explicit assumptions we prove value-estimation and graph-stabilization guarantees. Under stationary skew, REWEAVE improves held-out recall by up to 23.3 percentage points over HNSW. In comparisons with query-aware baselines on up to 100M vectors, it achieves up to 3.3x NGFix’s query throughput at recall@100 of 0.80. On Amazon-ESCI with query drift, it increases stream recall@100 from 0.745 to 0.979 with 34% fewer search distance computations than frozen HNSW. These results illustrate the potential to improve retrieval quality and reduce search effort within an existing graph-degree budget, while incurring additional work during adaptation. Comments: 20 pages, 9 figures, 5 tables. Full version with appendix Subjects: Databases (cs.DB); Information Retrieval (cs.IR); Machine Learning (cs.LG) Cite as: arXiv:2611.16060 [cs.DB] (or arXiv:2611.16060v1 [cs.DB] for this version) https://doi.org/10.48550/arXiv.2611.16060 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[IR-17] Wayfinder: Adaptive Resource Routing from Agent Citations NEURIPS2026
链接: https://arxiv.org/abs/2611.16053
作者: Miguel Romero,George Karypis
类目: Machine Learning (cs.LG); Information Retrieval (cs.IR)
备注: 20 pages, 3 figures, 10 tables. Accepted to NeurIPS 2026
Abstract:To act effectively, agents must gather contextual information across resources such as chats, issue trackers, email and knowledge bases, to answer questions like “what was decided in the last project meeting?” or “where is the FAQ for our internal tooling?” Where this information lives in a user’s environment is unknown to the agent, so it re-explores the same resources across consecutive interactions. Common routing approaches represent each resource as a tool and rely on high-level descriptions to generalize to unseen environments. Such descriptions often fail to separate similar or overlapping resources and cannot improve from the agent’s subsequent interactions. We introduce Wayfinder, an adaptive router that uses the documents cited by the downstream agent as feedback about where relevant information resides. These citations update Wayfinder’s internal resource representations, which subsequent interactions use to prioritize the resources likely to contain relevant information. We also introduce resource augmentation during offline reinforcement learning, which anonymizes and permutes resource identities to discourage training-time overfitting of which resource holds which topic; our ablations suggest it helps most when few training environments are available. Because no openly available professional environments pair realistic queries with resource-access patterns, we construct and release three controlled testbeds derived from existing datasets: NFCorpus, FEVER and Tech Startup. Compared with supervised, tool-retrieval and contextual bandit baselines under a shared closed-loop protocol, Wayfinder gives the best routing when resources contain semantically overlapping content that static descriptions cannot distinguish, while description-based tool retrieval suffices when resources are drawn from distinct corpora.
[IR-18] Predicting the Needle in a Petabyte Scale Haystack: Open-Vocabulary Event Anticipation in Satellite Imagery NEURIPS2026
链接: https://arxiv.org/abs/2611.16000
作者: Shreelekha Revankar,Mikhail Klassen,Creon Levit,Ash Hoover,Kavita Bala,Bharath Hariharan
类目: Computer Vision and Pattern Recognition (cs.CV); Information Retrieval (cs.IR)
备注: 29 pages, 9 figures, to be published in NeurIPS 2026
Abstract:Satellite constellations now image the Earth daily, capturing the lifecycle of events from deforestation to urban construction. Anticipating these events before completion enables timely intervention, yet existing systems cannot jointly identify where a change occurs, what it will become, and when it will finish across an open vocabulary. Building such a model requires diverse event data, but events are like needles in a haystack, making manual annotation at global scale infeasible. We address this problem by representing change as the vector difference between vision-language model embeddings at distinct time points. This arithmetic approach allows us to search for semantic transformations directly in the latent space, enabling (1) an automated data engine for global event discovery without manual annotation and (2) the largest global satellite event dataset to our knowledge, comprising 21,000+ locations across 64+ event types with ~2M images from PlanetScope and Sentinel-2 spanning six continents, to train (3) a novel global-scale event anticipation model. Our model detects change with a 93.5% max F1 score, outperforms baselines 3.6x on predicted event retrieval, and forecasts completion dates with a 3-day median error, even nine months before completion. Code: this https URL
[IR-19] HeriKGQA: Grounded Knowledge-Graph Question Answering over Cultural Heritage Sensor Data NAACL2027
链接: https://arxiv.org/abs/2611.15904
作者: Lingxiao Kong,Vasileios Arampatzakis,Vasileios Sevetlidis,George Pavlidis,Zeyd Boukhers
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Databases (cs.DB); Information Retrieval (cs.IR)
备注: Submitted to NAACL 2027 industry track
Abstract:Cultural heritage sites are increasingly monitored by sensor networks, but querying their data requires database expertise. Large Language Models (LLMs) offer a natural-language interface, yet over raw sensor records they lose track of which sensor, site, and time each value belongs to, and silently miscompute. We present HeriKGQA, a knowledge-graph question-answering method that confines the LLM to interpreting questions and phrasing answers, leaving computation to SPARQL and deterministic code. Questions are parsed into intents constrained by a heritage ontology aligning W3C SOSA/SSN with CIDOC-CRM, compiled into SPARQL, and answered from query results and temporal, spatial, and anomaly facts derived by a grounding layer, with error correction guarding each LLM step. On 500 questions over real data from five heritage sites, HeriKGQA reaches 99.0% accuracy, 18.2 points above the strongest of three baselines using the same LLM and receiving the same anomaly rules and answer conventions, with comparable answer quality while running almost four times faster. Without grounding, accuracy drops by 12.2 points and on anomaly questions from 99% to 66%. We report lessons from development and live deployment.
[IR-20] Bag of concepts can enhance other text representations for medical systematic review screening
链接: https://arxiv.org/abs/2611.15798
作者: Radosław Pytlak,Paweł Cichosz,Bartłomiej Fajdek,Bogdan Jastrzębski
类目: Information Retrieval (cs.IR)
备注: 16 pages, 5 figures, 6 tables
Abstract:The paper examines the choice of text representation for semi-automated systematic literature review (SLR) systems based on active learning. Motivated by the insufficient attention that the bag of concepts representation has received so far, the paper discusses the implementation and experimental evaluation of several variations thereof. The experimental study was performed on medical SLR case studies, with results reported for random forest classification, cluster initialization and uncertainty sampling. The utility of bag of concepts features obtained using different ontologies is evaluated by applying them within the active learning framework to sequentially create classification models based on training data iteratively expanded with expert annotations. Systematic experiments compare BoW, BoC, PubMedBERT, Jina, Gemma and Qwen3 representations, together with combinations of BoC and each of the other representations. The evaluation uses workload savings and performance profiles adapted from optimization benchmarking. The results suggest that concept-based features can provide complementary information, with the BoC-augmented representations achieving higher mean workload savings on the DERP cases than their corresponding base representations. The Qwen3+BoC representation gives the best average performance in this comparison. The effects of augmentation vary across case studies and base representations. The paper also concentrates on the impact of ontology choice on document classification outcomes. In particular, the paper compares the predictive value of bag of concepts representations based on chosen ontologies with that of representations obtained using MedCAT with a UMLS-based dictionary.
[IR-21] When Local Definitions Are Missing: Stable Errors in Language Models
链接: https://arxiv.org/abs/2611.15597
作者: Zengqing Wu,Chuan Xiao
类目: Computation and Language (cs.CL); Information Retrieval (cs.IR)
备注:
Abstract:Contracts, technical manuals, and online communities often give familiar words special meanings and define them in one place. A language model that reads a passage without that definition may use the usual meaning and give an answer that sounds right but is wrong. We study this with multiple-choice questions whose correct answers depend on a local definition. The questions come from a synthetic vocabulary, commercial contracts, Reddit community glossaries, and the documentation of a decentralized exchange. We ask three language models each question with and without the definition. Without using the answer key, we then test whether their repeated answers and stated confidence reveal their own errors. Separately, we give a small open model the answer key and test whether it can predict which questions larger models get wrong. We report three findings. (1) Familiar words that a source redefines cause the most errors. Models often repeat the same wrong answer, and neither repeated answers nor stated confidence reliably flag it. (2) The source definition fixes most errors on synthetic questions, but a definition drawn from the model’s own prior knowledge does not. In real contracts, the source definition fixes only about half of the errors. (3) Given an answer key, a small model can rank questions by how likely larger models are to get them wrong. These results call for tests that check answers against the full source, including its local definitions.
[IR-22] ReAlign: Reward-Guided Preference Alignment for Generative Re-Ranking
链接: https://arxiv.org/abs/2611.15445
作者: YaChen Yan,Liubo Li
类目: Information Retrieval (cs.IR)
备注: 9 pages, 3 figures, ACM conference format
Abstract:Re-ranking refines the upstream ranked list to maximize listwise user utility. Existing two-stage generator-evaluator methods decouple list generation from list evaluation, leaving the generator without direct access to the listwise utility signal, while scoring-based one-stage methods score items under a fixed upstream permutation context without explicitly searching the permutation space. We propose \textbfReAlign, a reward-guided preference alignment framework that couples a listwise reward model with a candidate-aware autoregressive generator. The reward model decomposes the listwise reward into position-wise contributions, modeling user exposure through a survival-based exposure module and item value through per-objective heads for heterogeneous engagement signals. The generator is trained in two stages: supervised pre-training on logged exposure lists to establish a relevance prior, followed by reward-guided post-training that aligns the generator with the reward model through complementary distribution and ranking alignment over within-pool standardized rewards. Extensive offline experiments show that ReAlign consistently outperforms existing state-of-the-art baselines.
[IR-23] SurvEval: Listwise Re-Ranking Evaluator via Survival-Based Exposure and Multi-Objective Value Modeling
链接: https://arxiv.org/abs/2611.15434
作者: YaChen Yan,Liubo Li
类目: Information Retrieval (cs.IR)
备注: 5 Pages, 3 figures, ACM conference format
Abstract:Re-ranking is a crucial stage in recommender systems, yet existing evaluators oversimplify user browsing assumptions and collapse multi-objective signals into a single scalar, limiting their ability to reflect true listwise utility. We propose \textbfSurvEval, a listwise evaluator that jointly models user exposure and item value for re-ranking evaluation. We formalize the sequential browsing process as a discrete-time survival model, deriving monotonically decreasing reaching probabilities that capture the cascading structure of user browsing. We further introduce a multi-objective value estimation module with dedicated per-objective prediction heads and uncertainty-based loss weighting to adaptively balance objectives of varying scales and noise levels. Both components are jointly optimized via a unified transformer backbone, and the listwise utility is obtained by summing the position-wise products of reaching probabilities and conditional item value estimates. Extensive experiments on KuaiRand and Avito demonstrate that SurvEval achieves superior performance over state-of-the-art baselines.
[IR-24] SimRec: Learning to Search Ask and Recommend from Simulated Interaction
链接: https://arxiv.org/abs/2611.15426
作者: Yaxin Zhang,Xingyu Wu,Yuchen Yan,Yuqin Dai,Hong Zhao,Lixing Shen,Zhan Chen,Jian Shao,Yongliang Shen
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Information Retrieval (cs.IR)
备注: 19 pages,5 figures,16 tables. Code: this https URL
Abstract:LLM-based recommendation agents can search catalogs, inspect products, ask users, and revise rejected recommendations. However, these behaviors are often governed by hand-designed workflows, while existing training methods optimize isolated components such as query rewriting or list ranking rather than the interaction policy as a whole. We present SimRec, a reinforcement learning framework that trains the full interaction policy of a tool-integrated recommendation agent. SimRec introduces two components. A task-anchored user simulator provides feedback on every recommendation: conditioned on the current task need rather than on user profile, and it can be an external LLM or the recommender itself under self-play. A task-aligned reward optimizes the whole trajectory with GRPO, crediting natural retrieval quality and valid tool use alongside the final hit, with training-only target injection to handle sparse hits over large catalogs. On ESCI and Amazon C4, SimRec improves the NDCG@100 of a Qwen3-4B agent from 32.23% to 35.42% and from 22.94% to 24.72%, outperforming training-free agents on far larger backbones and task-aligned baselines on NDCG. Analysis shows that the trained agent verifies candidates, uses preferences, and recovers from rejected recommendations far more often than its untrained counterpart. Our code is available at this https URL .
[IR-25] When Direction Matters: Diagnosing Relational Instruction Following in Dense Retrieval
链接: https://arxiv.org/abs/2611.15417
作者: Zhanyu Wu,Richong Zhang
类目: Information Retrieval (cs.IR)
备注:
Abstract:Dense retrievers increasingly claim the ability to condition retrieval on natural-language relevance criteria beyond a short query. We identify a specific weakness in this capability: the evaluated models perform well when lexical or attribute cues distinguish the target, but become much less reliable when relevance depends on a directional relation over the same entities. We introduce SwitchIR, a controlled diagnostic that contrasts Constraint Shift and Explicit Filtering with Relational Flip, where the query and document pair remain fixed while the requested relation direction reverses. Across representative embedding and late-interaction retrievers, the primary joint metric is high on the two cue-available controls and consistently low on Relational Flip; the same ordering persists in an LLM-assisted, manually reviewed realistic variant. An exact margin decomposition shows that both directions are correct only when the direction-sensitive query–document interaction exceeds the direction-invariant pair preference. Empirically, direction-sensitive changes are common, but their aligned interaction is usually too weak to overcome that preference. Diagnostic fine-tuning shows that models can learn relational behavior that generalizes across relation families, with stronger generalization in the larger model.
[IR-26] Vision-Grounded Semantic Learning and Alignment for Scalable Multilingual Text-Based Person Retrieval ACM-MM2026
链接: https://arxiv.org/abs/2611.15328
作者: Xinyu Zhou,Fan Li,Shasha Chen,Min Cao,Ahtamjan Ahmat,Qiaoming Zhu,Mang Ye
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL); Information Retrieval (cs.IR)
备注: Accepted by ACM MM 2026 (Oral). 10 pages
Abstract:Multilingual Text-Based Person Retrieval (TBPR) aims to retrieve individuals with texts in multiple languages. Despite its practical importance, it has been overlooked by existing research, which mainly focuses on English-only retrieval. The only prior attempt at multilingual TBPR suffers from (1) the absence of a large-scale, multilingual pretraining dataset and (2) a suboptimal model architecture for multilingual text encoding and cross-modal alignment, which employs a shared encoder that overlooks linguistic heterogeneity and relies on English as a pivot for cross-modal alignment, introducing English-centric bias. For this, we contribute MultiPedes, the first large-scale, pedestrian-centric multilingual pretraining dataset. It contains 10 million image-text pairs across four languages, constructed by a multilingual quality-aware data construction pipeline. We also propose ViSeLA (Vision-grounded Semantic Learning and Alignment), a novel framework featuring: (1) a semantic-aware multilingual learning module that extends the mixture-of-experts architecture by routing language-invariant semantics to specialized experts for improved multilingual feature learning; and (2) a vision-grounded cross-modal alignment module that employs a semantic probe to capture fine-grained semantics across modalities, enabling direct cross-modal alignment without relying on a pivot language. Extensive experiments on both multilingual and English-only TBPR datasets demonstrate the effectiveness of ViSeLA. The dataset and code are available at this https URL
[IR-27] Measuring Revealing and Mitigating Harmful Content Propagation in Recommender Systems
链接: https://arxiv.org/abs/2611.15160
作者: Xiaokun Zhang,Youlin Wu,Junyu Lu,Jiamin Chen,Zhaochun Ren,Hongfei Lin,Fenglong Ma,Xue Liu,Chen Ma
类目: Information Retrieval (cs.IR)
备注: This work has been acceptted by IEEE TPAMI
Abstract:Recommender systems help users navigate vast amounts of information by filtering content according to their historical behaviors. As these systems increasingly shape everyday information access, they largely determine the content users are exposed to. However, current research typically prioritizes catering to user interest, while overlooking the existence and propagation of harmful content, such as violence, hate speech, and pornography. In this work, we take a systematic step toward investigating harmful content in recommender systems from three aspects: measuring, revealing, and mitigating. First, we construct datasets with harmfulness-related annotations and present new evaluation metrics to quantitatively measure the harmfulness of recommendation results. Second, our empirical analysis reveals that users tend to engage with harmful content and prevailing recommendation methods often contribute to the proliferation of such content. Third, to address this issue, we introduce a generic framework, MitiHarm, which integrates harmfulness-related signals into both item and user representation learning. Experiments show that MitiHarm can significantly mitigate the propagation of harmful content while preserving user satisfaction. Resources of this work are available at this https URL.
[IR-28] When Does a Recommendation Spark Curiosity? Theory Measurement and Field Evidence
链接: https://arxiv.org/abs/2611.14938
作者: Pan Li,Alexander Tuzhilin
类目: General Economics (econ.GN); Information Retrieval (cs.IR); Machine Learning (cs.LG)
备注:
Abstract:Curiosity, the drive to close the gap between what one knows and what one wants to know, pulls consumers beyond their familiar content, sustains engagement, and breaks them out of the filter bubbles of repetitive, relevance-only recommendations. Despite its importance, curiosity is hardly incorporated into modern recommender systems for lack of a formal measure. The information-gap theory of curiosity provides no formal quantitative model that a recommender system can operate on, and existing proxy metrics do not accurately measure the curiosity that a recommended item evokes. To fill this research gap, we formalize the information-gap theory into five axioms of curiosity and construct the canonical formulation of a computational measure of the curiosity that a recommended item evokes in a consumer. We validate our curiosity measure through a survey-based study, in which it matches consumers’ self-reported curiosity and outperforms alternative proxy metrics. We subsequently build a curiosity-aware recommender system that incorporates this measure into the utility function design, and we show through offline experiments on three real-world datasets that it significantly outperforms the state-of-the-art methods. Furthermore, we conduct a large-scale online controlled experiment at a major video-streaming platform in Asia, where we show that our method significantly increases click-through rate, video views, and time spent over the latest production system, leading to tangible economic impact for the company. Our method has since been deployed on the platform and remains in operation. Our findings establish consumer curiosity as a measurable and actionable objective for recommender systems, with important managerial implications for digital platforms to recommend what consumers wonder about.
[IR-29] Stage-Resolved EBV-Induced B-Cell Programs Across CNS Selection and Multiple Sclerosis Compartments
链接: https://arxiv.org/abs/2611.14524
作者: Hasan Jamil,Alexander Gow
类目: Genomics (q-bio.GN); Databases (cs.DB); Information Retrieval (cs.IR)
备注:
Abstract:Epstein-Barr virus (EBV) is implicated in multiple sclerosis (MS), but whether EBV-conditioned B cells retain stable transcriptional programs after CNS selection remains uncertain. We reanalyzed datasets spanning experimental EBV infection, CNS-homing B cells in humanized mice, early-MS blood, and paired human CSF and blood. A frozen EBV-induced B-cell factor (F12) was transferred without outcome-driven modification and evaluated using anatomical, sequencing-depth, metabolic, coherence-matched, lineage, and exposure controls. F12 did not predict MS activity (AUC=0.487; p=0.489) and was lower in CSF in nine evaluable participants (mean difference -0.0560; coherence-matched p=0.015). It also remained lower in brain than spleen in all five EBV-exposed mice after matched-depth rarefaction (p=0.0313) and coherence-matched calibration (p=0.0225). The splenic TBX21+CXCR3+ transcriptional gate did not reliably represent the state of anatomically selected brain B cells. An enriched 287-gene cross-system bridge lacked B-lineage and EBV-exposure specificity. In human CSF, an early EBV-associated program was depleted (-0.0492; p=0.0197), whereas latency-III and lymphoblastoid cell line-like programs increased (+0.0666; p=0.0092 and +0.0148; p=0.0096, respectively). Opposite early and late program directions persisted after removing shared genes. No external program met the prespecified prospective criterion. These findings support stage- and compartment-dependent remodeling of EBV-associated B-cell transcription rather than persistence of a stable RNA signature. They distinguish cellular trafficking from transcriptional-state preservation and show that cross-system molecular resemblance alone cannot establish ancestry. EBV-directed biomarker studies should integrate host transcription with virologic, clonotypic, antigen-specific, epigenetic, and longitudinal evidence.
人机交互
[HC-0] A Compact xDAWN-Guided Spatio-Temporal Network with Filter-Bank Convolution for P300 Detection
链接: https://arxiv.org/abs/2611.17101
作者: Mohammad Mahdi Habibi Bina,Mohammad Reza Daliri
类目: Human-Computer Interaction (cs.HC); Artificial Intelligence (cs.AI)
备注:
Abstract:P300-based brain-computer interfaces (BCIs) enable communication by decoding stimulus-evoked EEG responses into character selections without requiring voluntary motor output. Reliable P300 detection remains difficult because of low response amplitude and signal-to-noise ratio, while practical spellers require accurate character selection with few stimulus repetitions and low computational cost. We propose a compact deep learning architecture that combines xDAWN spatial filtering for event-related potential enhancement and dimensionality reduction, a fixed frequency-selective filter bank, and convolutional feature learning for spatio-temporal EEG decoding. Depthwise-separable convolutions and progressive average pooling reduce computational complexity while retaining discriminative temporal information. The resulting target/non-target classifier contains only 5,100 trainable parameters. We evaluated the model on two public benchmarks, BCI Competition II Dataset IIb and BCI Competition III Dataset II, and compared it with established P300 decoding methods across different numbers of stimulus repetitions. The model achieved an information transfer rate of 17.8 bits/min with two repetitions on BCI Competition III and 40.5 bits/min on BCI Competition II, demonstrating a favorable accuracy-speed trade-off. These results suggest that combining domain-informed spatial and spectral priors with compact neural networks can support efficient P300 decoding for EEG-based BCI communication.
[HC-1] Neural Representations as New Modes of Interactivity
链接: https://arxiv.org/abs/2611.17058
作者: Stanley Huang,Anthony Baez,Catherine Yeh,Pat Pataranutaporn
类目: Human-Computer Interaction (cs.HC)
备注: 12 pages, 2 figures, 2 tables
Abstract:Human-AI interaction today runs almost entirely through prompts and outputs, keeping AI systems opaque to general users. We introduce Neural Representation Interaction (NRI), a paradigm that treats the internal representations of neural networks as an interaction material people can inspect, steer, and edit. Situating NRI alongside graphical, tangible, and bio-computer interaction, we argue that the concepts and algorithms models encode internally offer representational affordances for richer human-AI interaction. We synthesize actionable mechanistic interpretability methods and, drawing on General Morphological Analysis, construct a design space spanning Interaction, Target, Referent, and Purpose. Mapping existing interfaces onto this space reveals that most serve expert auditors, while few support lay users. We then present speculative applications in education, personal user modeling, and hiring audits, and discuss design principles, limitations, and sociotechnical risks. NRI offers HCI researchers abstractions for designing interactions that work with AI computation rather than only through its inputs and outputs.
[HC-2] Exploring Socially Acceptable 3D Feedback Design for Inclusive Small-Group Conversation
链接: https://arxiv.org/abs/2611.17018
作者: Qianjie Wei,Yifan Li,Zhen Bai
类目: Human-Computer Interaction (cs.HC); Emerging Technologies (cs.ET)
备注: 24 pages, 9 figures
Abstract:Prior research has shown that providing feedback on group dynamics can improve the quality of small-group conversations. However, publicly delivered feedback that targets specific individuals may reduce social acceptability due to negative experiences such as social pressure, embarrassment, and resistance. This work investigates how public feedback can be designed to remain socially acceptable while maintaining its effectiveness in XR-based small-group conversations. Informed by social science and HCI literature, we identified three design strategies (incorporate humor, reduce social attention, and use indirectness) and proposed a set of metaphor- and agent-based 3D feedback probes. We then conducted a two-phase study: (1) an online survey with 50 participants to examine how such feedback designs are perceived under public and private delivery, and (2) follow-up interviews with 14 participants who completed the survey to better understand their interpretations and elicit additional insights. Results show that metaphor- and agent-based feedback can reduce social discomfort while maintaining effectiveness. Interview findings further highlighted the role of three strategies in making feedback easier to accept, while requiring careful balance. We also discussed design implications for creating socially acceptable 3D feedback to support inclusive small-group conversations in XR.
[HC-3] DiffSSMEye: Conditional High-Frequency Eye Movement Synthesis with Structured State-Space Diffusion
链接: https://arxiv.org/abs/2611.17013
作者: Kamrul Hasan,Mehedi Hasan Raju,Oleg V. Komogortsev
类目: Human-Computer Interaction (cs.HC)
备注: 14 pages, 5 tables, 2 figures
Abstract:Synthesizing high-frequency eye movements requires preserving long-range temporal dependencies and subject-specific oculomotor dynamics. However, methods based on generative adversarial networks (GANs) and variational autoencoders (VAEs) can struggle to preserve these characteristics simultaneously. We introduce DiffSSMEye, a conditional denoising diffusion probabilistic model with bidirectional Structured State Space Sequence (S4) layers for generating five-second gaze-velocity sequences at 1,000 Hz. Generation is conditioned on a coarse gaze reference, user identity, and task, where identity and task embeddings are incorporated through feature-wise linear modulation. Identity-preservation, multi-resolution spectral, and integrated-drift losses complement noise prediction to constrain biometric consistency, frequency structure, and displacement error. Subject-disjoint evaluation on GazeBase across seven tasks yields mean real–synthetic embedding cosine similarities of 0.91–0.95. DiffSSMEye also achieves the lowest median spatial errors among these baselines for horizontal and random saccade tasks. These results demonstrate improved spatial accuracy and biometric feature consistency for subject-specific gaze synthesis.
[HC-4] It Was Better Not to Know: Digital Recognition Family Authority and Postpartum Mental Health in Bangladesh
链接: https://arxiv.org/abs/2611.16993
作者: Shajnin Haque Aishy,Sabbir Bin Abdul Latif,Md. Golam Zel A Husna,Jannatun Noor
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:Postpartum mental distress affects roughly one in five mothers in low- and lower-middle-income countries, yet recognition and support remain limited where formal mental-health care is scarce, and family relationships govern health decisions. We report a five-month interview study with 20 mothers in urban, urban slum, and rural Bangladesh who had given birth within the previous two years. Mothers encountered the concept of postpartum depression incidentally, through Facebook and YouTube recommendations rather than deliberate search or clinical guidance, and then used these platforms for private sensemaking without disclosing. Recognition, however, rarely traveled beyond the self: mothers could identify their distress while families dismissed it, leaving them with knowledge but no authority to act. Participants sought validation as a precondition for support and described recovery as relational rather than clinical or digital. We argue that postpartum technologies should engage partners, families, and community health workers rather than placing recognition solely on mothers.
[HC-5] Beyond Curiosity: Habit and Organizational Learning Capability as Predictors of Shadow AI
链接: https://arxiv.org/abs/2611.16931
作者: Nolan Lovett,Mohan Yang
类目: Computers and Society (cs.CY); Human-Computer Interaction (cs.HC)
备注: Preprint: submitted version (10 June 2026), not peer reviewed. 51 pages
Abstract:Generative AI has entered the workplace faster than organizations can govern it. A large share of employees now use AI tools their employer has not sanctioned, and many actively conceal that use, but the psychology of this covert technology behavior remains poorly understood. A common assumption, in popular accounts and early commentary, treats unauthorized (shadow) AI use as curiosity-driven exploration by adventurous individuals. We test this against a competing account grounded in the psychology of habitual technology use, in which automaticity rather than exploratory disposition sustains the behavior. Using survey data from 510 employed generative AI users across multiple countries, we adjudicated among individual and contextual predictors (workplace curiosity, habit, technology-acceptance beliefs, and organizational learning capability) with hierarchical regression, relative weights analysis with bootstrap confidence intervals, and full-subset dominance analysis. Habit was the strongest predictor, accounting for 28.3% of explained variance, and completely dominated joyous exploration, the curiosity dimension most aligned with the exploration account, across all 512 possible predictor subsets. Joyous exploration contributed 3.4% (95% CI [1.6, 7.4]) and fell to nonsignificance once use behavior and habit were modeled. Learning-supportive organizational conditions accounted for a further 24.1% and operated additively rather than by moderating dispositions. Parallel models decomposing the outcome showed the same dominance pattern for unauthorized use and concealment, and perceived social support for AI use was protective specifically against concealment. These findings reframe shadow AI as a habit-and-environment phenomenon rather than an individual-exploration phenomenon. The cross-sectional design means causal claims are exploratory.
[HC-6] A Connectionless Interaction Model for Network Spatial Proximity Services
链接: https://arxiv.org/abs/2611.16737
作者: Artem Makarov,Dmitry Namiot
类目: Networking and Internet Architecture (cs.NI); Human-Computer Interaction (cs.HC)
备注:
Abstract:In this paper, we present a connectionless interaction model for network spatial proximity services. In the network spatial proximity model, the reception of an identification or discovery frame of a nearby wireless node confirms proximity and delivers content at the same time. Current Bluetooth and Wi-Fi specifications and mobile platforms extend this one-way announcement with responses, addressed messages, scheduled group exchange, and relaying, which still do not require a connection. The proposed model describes such interaction independently of a concrete wireless technology. It is based on the roles of nearby nodes, the proximity message as the unit of interaction, twelve pattern properties, and six interaction patterns: announcement, mutual announcement, query-response, addressed exchange, scheduled group exchange, and relay. An exit rule defines when a service should move to a connection-based channel. We analyze how the patterns can be implemented on Android and iOS smartphones through native mechanisms, substitution, composition, embedded bridges, or a step-up to a connection, and we propose a pattern selection method for developers. The model generalizes the earlier network spatial proximity model and refines the interaction primitives of the proximity core. It is illustrated by existing services and systems and supported by published measurements and first-order estimates.
[HC-7] People prefer AI to human advice in multi-turn conversations but neither changes incentivised behaviour
链接: https://arxiv.org/abs/2611.16667
作者: Maximilian Maier,Tsz Kwan Emmy Wong,Jiahao Eric Yang,Falk Lieder,Vanessa Cheung
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:People increasingly ask AI chatbots for advice, yet large language models (LLMs) are sycophantic, overconfident, and can amplify human cognitive biases. In this article, we investigate how people perceive advice from humans versus LLMs and whether AI advice reshapes incentive-compatible donation decisions. Using Carrier, a new experimental platform, we designed an experiment where participants made a donation decision after discussing it with an AI chatbot advisor, another human, or receiving no advice. Participants judged LLM advisors to be significantly more rational, trustworthy, convincing, and sycophantic than human advisors. Analysis of chatlogs showed that LLMs were more engaged than human advisors: they shared more facts and more often stated their opinion and justified their recommendations. Despite this, donation decisions did not vary across conditions.
[HC-8] A Dual-Branch Graph and Convolutional Network for Interpretable Movement EEG Classification
链接: https://arxiv.org/abs/2611.16609
作者: Maryam Ostadsharif Memar,Behzad Nazari
类目: Artificial Intelligence (cs.AI); Human-Computer Interaction (cs.HC); Machine Learning (cs.LG); Signal Processing (eess.SP)
备注:
Abstract:Decoding upper-limb movements from scalp electroencephalography (EEG) is an important problem for brain-computer interfaces (BCIs), but many deep learning models provide limited insight into the brain regions underlying their predictions. We propose a two-branch architecture that combines an LGGNet-inspired local-global graph branch with an EEGNet-style convolutional branch for movement-versus-rest classification. The graph branch groups 61 EEG electrodes into anatomically informed regions and learns their spatial relationships, providing a direct representation of the regions emphasized by the model. We evaluate the proposed architecture using leave-one-subject-out cross-validation on the Ofner upper-limb movement dataset. The model achieves a mean F1-score of 0.74 +/- 0.06 across subjects. Beyond classification performance, we analyze the learned graph structure to identify the regions with the strongest connectivity. Across all held-out subjects, the highest connectivity is consistently observed in the Parietal Left Low region, corresponding to electrodes P1 and P3. This region is opposite to the performed right-arm movements and lies within the posterior parietal area previously associated with upper-limb movement processing. These results show that incorporating a region-based graph into the decoder can provide an anatomically interpretable representation while maintaining effective movement classification, offering a practical approach toward more transparent EEG-based BCIs.
[HC-9] Language Learning through Immersive Multi-User Gamification
链接: https://arxiv.org/abs/2611.16594
作者: Rosalina Ribeiro Silva,Zichen Feng,Rikard Olajos,Tim Rolff
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:We propose an immersive virtual-reality game designed to enable multi-user, multi-language learning during co-play. We hypothesize that our method will be more engaging, leading to better rhythm, pronunciation, memorization, and speaking confidence. We implement the experience as a fantasy role-playing game by introducing vocabulary and phrases to the user as spells to learn. Players then cast more efficient spells depending on their language skills. For evaluation, we compare the game to a single-user baseline that replicates traditional language learning in a non-immersive setting. We measure short- and long-term retention of vocabulary and phrases, as well as scoring rhythm and pronunciation. Furthermore, we collect user data on engagement and mental workload.
[HC-10] Robots That Talk Too Much and Dont Listen Enough
链接: https://arxiv.org/abs/2611.16473
作者: Graham Wilcock,Kristiina Jokinen
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Human-Computer Interaction (cs.HC)
备注: 4 pages, 3 figures, presented at EMF Workshop, HRI 2026, Edinburgh, Scotland
Abstract:Until recently, the biggest obstacles to natural spoken interaction with social robots were speech recognition errors and turn-taking failures. Advances in speech recognition have enabled much better spoken interaction, but turn-taking failures and other misunderstandings continue to be far too prevalent. This short paper looks back at earlier methods for improving speech recognition and turn-taking in interaction with social robots, and then looks at what can be done today, arguing that multimodal bidirectional streaming is needed to reduce the frequency of turn-taking failures.
[HC-11] A Serious Game for Training Civil Participation in Flood Crisis Management
链接: https://arxiv.org/abs/2611.16446
作者: Solmaz Goodarzi,Arseni Ivanov,David Maruscsák,Diana Quitéria Sousa
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:Floods are among the most frequent and destructive natural disasters, requiring citizens to make critical decisions under rapidly changing conditions. Most serious games related to flooding emphasize awareness, preparedness, and education in flood management, but there is still a research gap on how to train citizens to act under pressure, uncertainty, and stress. In this paper, we position Virtual Reality (VR) serious games as a promising approach to addressing this gap by providing experiential and immersive training opportunities for citizens. We propose a conceptual VR serious game that enables participants to experience realistic flood scenarios, practice decision-making, rehearse appropriate response behaviors, and develop response-related skills in a safe and controlled environment. In addition, we suggest a research agenda and an evaluation framework focused on flood risk awareness, response capabilities, and stress-related outcomes. The framework thoughtfully integrates subjective, objective, and physiological measures to assess the effectiveness of immersive flood-response training. Finally, we present design, methodological, ethical, and user-experience considerations in VR-based disaster training, and outline future research directions for the development and evaluation of immersive serious games for flood emergency response. The present study aims to contribute to the growing intersection of serious games, extended reality, and disaster risk reduction by establishing citizen-centered response training as a significant yet under-researched field.
[HC-12] Rasoi: Visual Diffs for Evaluating Structural Changes Made by Coding Agents
链接: https://arxiv.org/abs/2611.16293
作者: Riya Sahni,Aryan Kaul,Anushka Bhatnagar,Xuanming Zhang,Vivian Liu,Reya Vir,Lydia B. Chilton
类目: oftware Engineering (cs.SE); Human-Computer Interaction (cs.HC)
备注:
Abstract:Coding agents let developers modify software rapidly through natural language. However, as agents take on larger changes, developers can lose track of how the codebase is evolving. These edits may look correct on the surface while restructuring the system in unexpected ways: duplicating logic, changing modules outside the requested scope, removing functionality developers meant to keep, or introducing architectural drift. We investigate how visual diffs can help developers understand and evaluate the structural changes that coding agents make. We present Rasoi, a design probe that maps how a codebase’s architecture evolves after each agent-made edit, letting developers inspect which architectural elements and relationships were added, modified, or removed across multiple levels of abstraction. We deployed Rasoi in a week-long study with 12 developers working on their own codebases across varied technical domains. We found that visual diffs helped developers see whether changes stayed within scope, notice unexpected restructuring, and understand how modified parts related to the system as a whole. Our findings suggest that the value of oversight tools for coding agents lies less in making developers trust the agent and more in helping developers feel confident in their own evaluation of agent-made changes. We discuss implications for visual tools that support code comprehension and human oversight in AI-assisted software development.
[HC-13] RiskLens: Human-Informed Incident Transfer for AI Risk Evaluation
链接: https://arxiv.org/abs/2611.16287
作者: Yiliu Tang,Hangyue Zhang,Yun Huang
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:As AI capabilities advance, early-stage teams need better ways to judge whether historical incidents transfer to new products, demanding support beyond static risk checklists or one-off ideation tools. We present RiskLens, a human-AI workflow with an interactive system built around a live crawl, refreshable ranker, and extensible, human-informed schema. RiskLens analyzes 2,034 AI incidents and 10,443 AI products to rank transferable incidents using a feature schema spanning technical, operational, and user-facing attributes and supports an interface that presents ranked incident evidence, generated scenarios, risk metrics, and mitigation suggestions. In a study with 21 AI-familiar participants, users surfaced evaluation criteria that we added as 13 human-informed dimensions, including harm categories, failure mechanisms, and vulnerable groups. The feature-based model outperformed a generic baseline with an 86.2% win rate, and remained stable on a twenty-month future holdout; after human-informed schema expansion, the updated model outperformed the same baseline on 91.4% of pairs.
[HC-14] Agent ic Dashboards: A New Design Paradigm for Visual Analytics
链接: https://arxiv.org/abs/2611.16245
作者: Sam Yu-Te Lee,Yun-Hsin Kuo,Kwan-Liu Ma
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:Visual analytics systems traditionally combine analytical methods and interactive visualizations into a fixed application specified at the design phase. Coding agents offer a flexible approach: experts could conduct analysis through natural-language instructions. We envision an agentic dashboard as a persistent visual analytics workspace where an agent makes decisions about analytical methods, views, and their coordination within a designer-defined decision space. This approach shifts part of the visualization designer’s role from engineering a fixed interface to defining the decisions an agent may make at runtime. We illustrate how such a dashboard could operate with current capabilities and propose criteria for evaluating its behavior and potential failures. As analytical workflows become more fluid and agent-mediated, visualization can help experts maintain awareness, coordinate with agents, and exercise judgment.
[HC-15] “We Are the Dirty Hands”: Reimagining an Ambient Intelligence Care System with Professional Caregivers
链接: https://arxiv.org/abs/2611.16198
作者: Andrea Green,Jane L. E,Sneha Jain,Parker Ruth,Matthew Jörke,Sarah L. Billington,James Landay,Andrea Cuadra
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:While ambient assisted living (AAL) sensing systems offer promise for aging in place, their impacts on professional caregivers are overlooked. We conducted two collective speculation workshops (N=24) with personal health aides (PHAs, n=18) and certified nursing assistants (CNAs, n=6) to examine the ethical implications and value tensions of a care system with ambient intelligence (CSAI), our conceptualization of an AAL. Participants anticipated different implications across care roles and settings. PHAs anticipated additional responsibilities from CSAI, including learning how it operates and supporting older adults in its use. CNAs, who had greater access to workplace resources and protections, similarly noted that CSAI could create additional work. Participants also identified tensions around visibility, as sensing could support caregivers by documenting difficult working conditions while simultaneously making their behavior more vulnerable to misinterpretation. Through their modifications and surfaced discussions, we propose design guidelines for AAL systems, such as integrating context-aware sensing.
[HC-16] Between Niti and Nyaya: A Society-Centered Framework for AI Governance
链接: https://arxiv.org/abs/2611.16173
作者: Snehalkumar `Neil’ S. Gaikwad,Punya Aragula,Shiva Kruthi Palpanoori,Pratik Patekar
类目: Artificial Intelligence (cs.AI); Computers and Society (cs.CY); Human-Computer Interaction (cs.HC)
备注:
Abstract:Governments are making decisions about AI before they can measure its public effects, and those decisions determine whether a person can learn that AI influenced an outcome, challenge it, and see it corrected. Research has documented what AI governance produces: principles, strategies, registers, and audits. However, we lack shared language for understanding AI policy while it is still being made. We introduce the Society-Centered AI Governance Framework, which gives policymakers, researchers, and the public a way to assess instruments in any jurisdiction by whether a person can get notice, challenge a decision, and see it corrected. It reads a jurisdiction’s instruments together by date and binding status, separates an agency duty from a right a person can invoke, and traces which later instrument must enforce each commitment. We evaluated it in North Carolina using interviews with nine AI Leadership Council members, conducted before and after the state’s 2026 AI roadmap, and a provision-level reading, with human-reviewed model-assisted coding, of 49 AI governance documents from 19 states. For the state’s own AI systems, we find visibility-first governance: visibility duties are in effect, notice exists as an agency duty with no right a person can invoke, and AI-specific challenge and correction remain advisory. The roadmap assigns owners and dates to the missing protections, so the gaps are documented but not yet closed. Recent federal preemption proposals focus on AI development, so recourse for a state’s own systems is likely to remain a state decision. We identify “legitimacy before protection” as a testable risk of visibility-first governance and name the milestones that would test it. More broadly, judging AI policy by who can act on what it makes visible lets the field assess recourse while policy is still forming, before harm occurs.
[HC-17] MRxD: A Usage-Pattern Framework for Mixed-Reality Cross-Device Interaction
链接: https://arxiv.org/abs/2611.16171
作者: Yang Zhan,Yuxuan Huang,Yuchen Cai,Yichen Andy Yu,Xiaoran Yang,Ye Yuan,Qiao Jin
类目: Human-Computer Interaction (cs.HC)
备注: 22 pages, 11 figures
Abstract:Mixed reality (MR) cross-device experiences often combine MR headsets with non-immersive devices such as laptops, tablets, and phones. Yet such cross-device design remains challenging because users must coordinate workflows across heterogeneous devices with different roles. This motivates our focus on usage patterns: recurrent ways in which people distribute and coordinate work across devices. Through a systematic analysis of 238 MR cross-device papers, we developed MRxD, a usage-pattern framework organized around three dimensions: user attention, task relatedness, and cognitive symmetry. Using these dimensions, we identified six usage patterns and synthesized the design strategies commonly used within each pattern. We further investigated the framework’s application in design practice through a workshop (n=6) and the results suggest that MRxD can guide pattern-based design exploration and identify blind spots. We further discuss how usage patterns may shift or combine, and outline research gaps and design implications for future MR cross-device systems.
[HC-18] How Do People Debug Complex SQL Queries With LLM s? A Qualitative Study
链接: https://arxiv.org/abs/2611.16108
作者: Nicole Sullivan,Ella Cao,Leilani Battle,Magdalena Balazinska
类目: Human-Computer Interaction (cs.HC)
备注: 32 pages, 9 figures
Abstract:While LLMs are increasingly used in various contexts, we lack a fundamental understanding of how they impact SQL query debugging. In response, we conducted a user study with 16 participants of varying SQL expertise, examining how they debug complex SQL queries with and without LLM support. Initial benchmarking indicated that our selected models are already capable SQL debuggers. Qualitative analyses of our study data revealed that while participants trusted the LLM, they varied widely in how and whether they delegated debugging tasks to the LLMs, which we characterized through a new three-state model of participants’ SQL reasoning processes. Additionally, participants enforced specific collaborative roles for the LLM such as TA or sounding board, which participants controlled by withholding context from the LLM. These findings suggest that LLM-based tools could enable preset role selection, broader context management, and flexible automation features, including low automation options, to support SQL debugging moving forward.
[HC-19] rained on What Screen Readers Cannot Read: Accessibility Debt in the GUI Corpora Behind Computer-Use Agents
链接: https://arxiv.org/abs/2611.16107
作者: Obada Kraishan,Kulsawasd Jitkajornwanich
类目: Human-Computer Interaction (cs.HC); Artificial Intelligence (cs.AI)
备注: 8 pages, 2 figures, 6 tables. Accepted at the International Conference on Artificial Intelligence Centric Systems, Services, and Applications (AISSA 2026), San Antonio, TX, USA, December 8-10, 2026
Abstract:Computer-use agents learn to operate mobile interfaces from corpora of captured screens and accessibility trees. The same trees are what a screen reader consumes, so an agent trained on them inherits whatever labelling gaps the underlying applications carry. This paper measures one of those gaps, content-description coverage, in two Android corpora: Rico, collected in 2017, and AndroidControl, collected in 2024. Across 17,373,614 interface elements, 17.74% of the Rico elements that need a content description under our coding rule carry one, against 57.27% in AndroidControl. The difference of 39.5 percentage points holds under five alternative coding rules and when whole apps are resampled. It describes two datasets that also differ in app selection, capture method and platform version; it does not measure change over time. Input controls are the least labelled role in both corpora (2.20% and 27.21%, or 30.99% in AndroidControl if hint text counts as a name). Descriptions in the older corpus are also five times more likely to be placeholders (10.25% against 1.95%), so the difference is not an artefact of developers typing filler. Resolving 26,205 task actions to the elements they touched shows that, among tappable graphics with no text, demonstrations operate elements described 9.0 points more often than the others on the screens they visit, about 3.5 points of it from the choice of element within a screen. That is a selection effect inside the benchmark. A two-coder reliability check (kappa = 0.91 and 0.96) and the full pipeline accompany the paper.
[HC-20] Human-Likeness and Conversational Preference Diverge for Synthesized Agent Speech in Customer-Service Calls
链接: https://arxiv.org/abs/2611.16051
作者: Jeffrey M. Girard,Antony D’Avirro,Jason Z. Zheng,Jacqueline R. Vertino,Benjamin Peloquin
类目: Computation and Language (cs.CL); Human-Computer Interaction (cs.HC); Audio and Speech Processing (eess.AS)
备注: 5 pages, 3 figures
Abstract:Listeners can judge an agent’s speech on how human it sounds (human-likeness) and on how much they would want to talk to it (conversational preference). Listening tests rarely ask about conversational preference, so the relation between the two judgments is unclear. We compared them in paired-choice tests on 120 turns that actors improvised as customer-service representatives. On each trial, two agents spoke the same turn. In Study 1, both agents came from six commercial text-to-speech systems, each with eight voices. In Study 2, one was the actor’s original recording and the other a system. Some systems ranked higher on human-likeness, and others on conversational preference. Listeners chose systems over actors in 19% of human-likeness trials but 36% of conversational preference trials (and 49% for the most preferred system). On both judgments, variation among a system’s voices was about three-quarters of that between systems. Of three automatic naturalness scores, two predicted lower human-likeness, and none predicted conversational preference. In this setting, listeners tended to favor faster speech and fewer pauses between clauses on both judgments, and more pitch variation for conversational preference. Thus, measuring only human-likeness or testing one voice per system can misjudge which agents listeners would rather talk to.
[HC-21] Right Advice Wrong Reason s: How Decision and Reasoning Alignment Shape Reliance on AI Advisors
链接: https://arxiv.org/abs/2611.15732
作者: Eunkyu Park,Vasudha Varadarajan,Wesley Hanwen Deng,Anisha Reddy,Gunhee Kim,Motahhare Eslami,Jana Schaich Borg,Maarten Sap
类目: Human-Computer Interaction (cs.HC)
备注: Pre-print
Abstract:When people seek advice from AI on social situations, the system may recommend a decision they fully agree with but justify it through reasoning they disagree with. Conversely, it may offer reasoning they agree with, and yet reach a decision they reject. We distinguish decision alignment, whether an advisor recommends the option favored by a user’s measured values, from reason alignment, whether its stated reasons align with the user’s values. In this work, we ask how these two forms of alignment shape people’s preferences for AI advisors and their willingness to use them in different roles. We conducted a within-subjects study (N=121) in which participants evaluated three AI advisors varying in decision and reason alignment across eight everyday moral dilemmas. Participants then chose which advisors they would consult, allow to draft messages for review, or allow to act on their behalf in new tasks. When seeking advice, participants most often preferred an advisor that reasoned through their prioritized values while recommending the opposing decision (40%). However, when participants had to pick advisors for new tasks, their preferences shifted toward advisors that also recommended decisions aligned with their values, rather than those that only shared their reasons. Also, as participants considered tasks that gave AI advisors greater autonomy, they became more likely to opt out of using any advisor, from 12% when consulting to 38% when allowing an advisor to act on their behalf. Our findings suggest that decision and reason alignment shape advisor preferences differently across contexts of AI use, and that alignment with a user’s values does not necessarily imply willingness to delegate.
[HC-22] rms of Care: Designing for Participation in Data Governance by Disability Communities
链接: https://arxiv.org/abs/2611.15720
作者: Madiha Zahrah Choksi,Reeda Shimaz Huda,Rita Faia Marques,Anja Thieme,Cecily Morrison
类目: Human-Computer Interaction (cs.HC); Artificial Intelligence (cs.AI); Digital Libraries (cs.DL)
备注: 26 pages, 3 figures, 3 tables, under review
Abstract:Improving disability representation in AI-generated media requires community-specific datasets, yet openly sharing such data can expose communities to misuse, decontextualization, and exploitation. We examine how disability communities can participate in dataset governance through context-sensitive licensing. Working with three disability advocacy organizations, we conducted semi-structured interviews, an educational webinar and modular license activity to promote community participation in data stewardship. Our findings show that governance in these high-context communities relies on organizational memory (especially situated harms); both formal authority and informal relational practices; communication; and resulting trust. The modular license activity made organizational data choices more accessible and values-led, allowing communities to adapt Creative Commons clauses to express permitted uses, sector-specific restrictions, safeguarding requirements, and commitments to respectful representation. Yet, concerns remain about monitoring and enforcing appropriate data use. We contribute (1) an empirical account of governance tensions, (2) a practical modular approach, and (3) design implications for AI data platforms that support relational, context-sensitive governance beyond standardized, one-size-fits-all frameworks.
[HC-23] Integrating Slides Mathematics and Simulation in One Page: GenAI-Assisted HTML Learning Objects for Engineering Courses
链接: https://arxiv.org/abs/2611.15686
作者: Xuan-Quy Dao,Ngoc-Bich Le,Xuan-Dung Phan,Le-Minh Nguyen
类目: Human-Computer Interaction (cs.HC)
备注: 25 pages, 5 figures, 11 tables
Abstract:Engineering courses spread their materials across slides (PPTX), typeset notes (LaTeX/PDF), and MATLAB, leaving students to connect them. We describe an instructor-in-the-loop workflow in which generative AI (GenAI) converts these assets into single-file HTML learning objects that combine presentation, mathematics, visualization, simulation, and assessment; a page is released only after its simulations pass numerical verification. Six open demonstrators implement 18 tasks from six engineering courses. In the eight tasks with quantitative cross-checks, every quantity met its tolerance at default settings (1%, or 2% for the finite-difference time-domain simulation) and agreed with a separate Python re-computation and a value-blinded MATLAB re-computation run by a colleague not involved in development; the checks also flagged under-resolved settings. In a perception survey, students (19 after post-hoc screening of 28 responses) reported that the pages linked equations with simulation, and 74% preferred them to separate files, but usability was marginal (System Usability Scale 56.8). Twelve lecturers rated the approach as useful (item means 4.25–4.58) but were hesitant to adopt it; their median ratings on a 12-criterion capability rubric agreed moderately with the first author’s ( \kappa_w=0.60 ; individual median 0.31). The results show technical feasibility and preliminary positive perceptions; learning outcomes were not measured.
[HC-24] Making Doula Coordination Breakdowns Visible: A Participatory Case Study ICIP ALT
链接: https://arxiv.org/abs/2611.15580
作者: Emma Ricci-De Lucca,Saanvi Chilakapati,Jillian Berch,Ashley Kim,Sahithi Medikondla,Iha Rastogi,Margaret R. Emerson,Elizabeth L. Murnane,Karen L. Fortuna
类目: Human-Computer Interaction (cs.HC)
备注: Presented at the CSCW 2026 Workshop on Women’s Health as Cooperative Work: Participatory AI Across Care, Community, and Policy ( arXiv:2611.03776 )
Abstract:Doulas are trained, non-medical professionals who provide emotional, physical, and informational support to pregnant individuals before, during, and after childbirth. Doulas provide extensive relational and coordination labor for pregnant and postpartum patients, spending far more time with clients than traditional providers do, yet they operate with almost no doula-specific digital infrastructure. This gap is especially acute in rural communities, where postpartum depression affects about 25% of women, well above the national average, and where geographic isolation and provider shortages leave many without access to behavioral health care. DoulaConnect began as a 12-week evidence-based postpartum texting curriculum, piloted separately in a randomized controlled trial in rural New England. Building on that work, we present DoulaConnect’s development into an AI-enabled platform co-designed with doulas and rural mothers through participatory human-centered design, as a case for surfacing where coordination breaks down in doula practice and who currently absorbs the resulting burden. Our community-based participatory study, conducted across the platform’s Discover and Define phases, combines qualitative sessions with doulas and mothers to identify these breakdowns. We close with tensions this approach surfaces and open questions for the workshop about how these findings will inform future participatory AI infrastructure for doula care in rural and other under-resourced communities.
[HC-25] Immersive Sight-Restriction-Based Collision Prevention Using Generic Particle Systems in Virtual Reality
链接: https://arxiv.org/abs/2611.15561
作者: Annalisa Degenhard,Annika Stampf,Michael Rietzler,Enrico Rukzio
类目: Human-Computer Interaction (cs.HC)
备注: Accepted at MUM 2026, the 25th International Conference on Mobile and Ubiquitous Multimedia
Abstract:Collision prevention systems (CPS) help VR users navigate their physical surroundings safely. Most approaches overlay conceptually distinct grids or add solid virtual counterparts of physical obstacles; both can reveal the physical space, as they either stand out visually or behave inconsistently during locomotion. We explore sight restriction as an alternative: a standard particle system visualizes the play area boundary as a fog-like cloud whose intensity scales with proximity and viewing direction. A pre-study (N=10) determined the distance at which the visualization must appear. In a main study (N=11), we compared this particle-based CPS with a grid-based solution, focusing on locomotion behavior: across 253 approach episodes, users interrupted their approach after a comparable distance once either visualization appeared. Although the particle boundary was visible three times as often, participants reported higher presence. We release our implementation as open source to support further work on immersive collision prevention.
[HC-26] What You See Is (Not) What You Got: The Human Factor in PDF Signature Verification
链接: https://arxiv.org/abs/2611.15472
作者: Jakob Heher,Emanuel Pichler,Stefan More
类目: Cryptography and Security (cs.CR); Human-Computer Interaction (cs.HC)
备注:
Abstract:Signed PDF documents have legally binding value under frameworks such as the EU’s eIDAS regulation. In this work, we present Genuine PDF Forgeries, a novel attack class that focuses on human factors in signature validation. Our attacks target the disconnect between cryptographically verifiable signer certificates and human-visible signature blocks embedded in documents. We analyze our attacks against 23 PDF viewers and validation tools across desktop platforms and the web. We find that nine of the 13 viewers that automatically report signature validity hide authenticated signer information behind additional user interaction. We then evaluate the attack in twin deception-based online studies with 584 EU-based participants. To further validate the impact, we perform a highly tailored real-world attack study with six trials within two privacy-sensitive organizations. Five of six organizational trials accepted requests authenticated using the forged visual identity. In the online studies (total n=584), the most effective attack variant deceives 71.9% of eligible participants on first impression, and 57.3% continue to be deceived even after we prompt for additional introspection. We derive viewer-side countermeasures and prototype a safeguard that overlays the untrusted signature block with signer information obtained from the validated certificate. Our results show that boolean validity indicators, not associated with the authenticated signer identity, are insufficient in practice. Subjects: Cryptography and Security (cs.CR); Human-Computer Interaction (cs.HC) Cite as: arXiv:2611.15472 [cs.CR] (or arXiv:2611.15472v1 [cs.CR] for this version) https://doi.org/10.48550/arXiv.2611.15472 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Related DOI: https://doi.org/10.1016/j.cose.2026.105191 Focus to learn more DOI(s) linking to related resources
[HC-27] Computational Rag a Recognition: A Theory-Driven Approach
链接: https://arxiv.org/abs/2611.15242
作者: Janay Sukheja,Shailaja D. Sharma
类目: ound (cs.SD); Human-Computer Interaction (cs.HC); Machine Learning (cs.LG)
备注: 16 pages, 6 figures
Abstract:Raga recognition in Hindustani classical music is difficult because a raga is not defined only by the notes used in it. Its identity also depends on the way notes are emphasized, the movement between notes and characteristic phrases. In this work, we develop a computational approach to analyse these characteristics directly from audio recordings. The method uses pitch extracted from recordings and represents it using a 44-position note framework based on the 22-shruti system. From this representation, Raga features based on the Indian classical musicology are this http URL feature is compared with a reference representation of the raga and the resulting scores are combined to obtain an overall similarity score. The method is first studied using Raga Yaman and is then tested on recordings of other ragas. The computational results are also compared with the theoretical description of Yaman to identify which raga characteristics are captured well and where the method has limitations.
[HC-28] Alignment De-Identifies Without Aligning: A Congruence-Invariant Relational Gap in Cross-Subject Neural Representations NEURIPS2026
链接: https://arxiv.org/abs/2611.15204
作者: Anna Notaro(1),Letizia Rossato(2),Sadasivan Puthusserypady(2) ((1) University of Tuebingen, Germany, (2) Technical University of Denmark, Denmark)
类目: Human-Computer Interaction (cs.HC); Neurons and Cognition (q-bio.NC)
备注: 23 pages (12 pages main text and references, 11 pages appendix), 2 figures. Accepted at NeurIPS 2026, NeurReps Workshop
Abstract:In brain-computer interfaces, inter-subject variability remains a major challenge that limits decoding performance across users. Subject-wise alignment is widely used to improve cross-subject transfer of neural decoders, yet it is typically evaluated only through decoding accuracy, which cannot determine whether aligned representations share greater task-related structure. We therefore distinguish two questions that accuracy conflates: whether alignment removes subject-specific information and whether it makes the relational organization of task conditions more similar across individuals. Using three motor imagery-EEG cohorts, trials are represented as covariance matrices and aligned by recentering each subject’s mean covariance to a common reference. Subject-specific information is quantified with a subject-identification probe, whereas relational structure is assessed using representational dissimilarity matrices (RDMs) computed from condition means with the affine-invariant Riemannian distance. We show that recentering leaves this distance, and therefore the resulting RDMs, unchanged. Consistent with this result, recentering reduces subject identification accuracy from 0.96 to chance level across all cohorts, while altering RDMs only at numerical precision. Cross-subject agreement in relational structure is significantly above zero but below the estimator’s lower noise ceiling. Introducing an explicit similarity regularizer during training also fails to increase held-out agreement: the apparent reduction in between-subject differences is explained by a decline in the reliability ceiling rather than improved alignment. These findings show that eliminating subject identity is not equivalent to aligning relational geometry. More broadly, alignment claims must be interpreted relative to the invariances of both the distance and the statistic used to evaluate them.
[HC-29] An Experience Report on Teaching Students to Evaluate AI Agent Skills as Authorization Decisions
链接: https://arxiv.org/abs/2611.15170
作者: Zikai Alex Wen,Donald Chinn,Yan Bai
类目: Human-Computer Interaction (cs.HC)
备注: Preprint version. To appear in SIGCSE TS 2027
Abstract:Cultivating a security mindset has long been a goal of computing education, and students now need it when deciding whether to trust an AI agent skill. Skills are add-ons that can shape an agent’s behavior through natural language instructions and configuration settings, not just executable code. To prepare students to evaluate skills as authorization decisions, we designed a classroom activity and taught it in a senior undergraduate and a graduate cybersecurity-focused course. Students inspected three example skills, each pairing a real workflow task with one agent-specific authority, such as acting without asking for user consent before each action. The activity guided them to make initial authorization recommendations and then use a static security scanner to test and revise their recommendations. We observed in class that students needed guidance to extend their security review from a skill’s code to what they would be authorizing the agent to do through the skill. Pre/post surveys from 11 volunteer respondents in this preliminary pilot showed greater self-reported confidence in evaluating skills, and post-activity responses placed more emphasis on the authority that skills carry. We release all activity materials for reuse.
[HC-30] “I dont know what I dont know”: Exploring the Benefits and Challenges of AI in Interdisciplinary Research
链接: https://arxiv.org/abs/2611.15121
作者: Indu Panigrahi,Tal August
类目: Human-Computer Interaction (cs.HC); Artificial Intelligence (cs.AI)
备注: Preprint
Abstract:Interdisciplinary research is widely lauded as beneficial for society but presents unique challenges to the researchers who bridge disciplinary boundaries. Despite the rapid increase in the availability and usage of AI systems for research tasks, there is little work studying how AI systems have impacted workflows that span unfamiliar disciplines. Through a survey ( N=62 ) and follow-up interviews ( N=23 ) with interdisciplinary researchers, we describe the unique challenges and opportunities for AI in interdisciplinary research workflows. Participants viewed AI both as a solution and exacerbation of the general challenge of “unknown unknowns”: AI systems could facilitate a broad awareness of a new research area, yet also provide an illusion of understanding. Crucially, participants lacked the expertise and intuition to catch these illusions. Our findings motivate a focus on designing AI systems to support rather than automate collaboration across disciplines, scaffold learning opportunities, and include avenues for triangulation between resources.
[HC-31] Estimating Image Creativity
链接: https://arxiv.org/abs/2611.15044
作者: Nisan Chhetri,Christopher G. Healey,Sina Baghbanijam
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Human-Computer Interaction (cs.HC); Machine Learning (cs.LG)
备注: 9 pages, 1 figure
Abstract:What makes an image creative? Although this may seem like a highly subjective decision, research in psychology suggests that creativity may be based on a set of fundamental creativity dimensions such as novelty, surprise, and arousal. A deeper understanding of image creativity can benefit both research and practical applications. Exploring how human viewers use creativity dimensions to evaluate image creativity may allow us to predict a target image’s human-rated creativity automatically. We present a novel framework to estimate the perceived creativity of an image by: (1) predicting the strengths of an image’s novelty, surprise, and arousal creativity dimensions, and (2) combining the creativity dimension strengths to estimate an image creativity score. Controlled human-subject experiments were conducted on Amazon Mechanical Turk to collect data to train supervised machine learning models to estimate an image’s strengths on the three creativity dimensions. A final machine learning model takes the predicted dimension strengths as input to produce an estimated image creativity score. Results showed that our estimated creativity scores are statistically equivalent to human ratings of creativity for the images we tested. The novel contributions of our research are: (1) identification and validation of three psychological creativity dimensions used to estimate image creativity; (2) supervised machine learning models to automatically estimate novelty, surprise, and arousal strengths for an input image; (3) a supervised machine learning model to predict a creativity score from novelty, surprise, and arousal strengths for an input image, and (4) experimental methods to confirm the accuracy, advantages, and limitations of our approach.
[HC-32] Data Visualization Influencers: Tracing the Creation and Curation Strategies Behind Propagandist Visualizations on X
链接: https://arxiv.org/abs/2611.15042
作者: Priya Dhawka,Nina Lutz
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:Data visualization is increasingly understood as a political process shaped by creators’ editorial design and data choices. Prior research has examined how such choices produce deceptive visualizations, focusing on social media users’ misleading visualizations of public data. This overlooks how creators draw on professional expertise to produce propagandist visualizations of political narratives. We investigate visualization creators who design and share propagandist visualizations of political claims, such as immigration, criminality, and others, in US online discourse. Extending prior work on data visualization as propaganda, we present an exploratory study of 17 creators and the propagandist visualizations they create and curate for political discourse on X. We introduce the notion of a ``data visualization influencer’': an online creator who (1) leverages perceived credibility as a data expert, and (2) engages in epistemic capture of the credibility aesthetics of data and visualizations to manufacture propaganda. We discuss implications for research and practice.
[HC-33] Calibrate Dont Retrain: Decomposing the Value of Annotator Disagreement in Cyberbullying Severity
链接: https://arxiv.org/abs/2611.15010
作者: Arslan Bisharat,Katelyn Skees,Manuel Sandoval Madrigal,Mujtaba Nazari,Ayaan Khan,Deborah L. Hall,Mohammed Abuhamad,Yasin N. Silva
类目: ocial and Information Networks (cs.SI); Computation and Language (cs.CL); Computers and Society (cs.CY); Human-Computer Interaction (cs.HC)
备注: 26 pages. Submitted to PACMHCI (CSCW). Code and data: this https URL
Abstract:Machine learning models are often trained on labeled data produced by multiple human annotators. When those annotators disagree, their labels are commonly collapsed into a single majority vote, which discards information about the underlying disagreement. We examine the cost of this practice for cyberbullying severity, a subjective judgment on which annotators disagree heavily. In the corpus we study, all 5 annotators assigned the same severity label to only 4.1% of comments. Turning 5 judgments into one label hides this disagreement and obscures how much the annotators differed in their assessments. In this paper, we ask whether a model can preserve this information by predicting the distribution of annotator judgments instead of a single majority label. We find 3 main results. First, accuracy against the majority label cannot tell competent models apart: 2 models with similar accuracy can differ greatly in how well they match the distribution of annotator judgments. Second, predicting the full distribution requires little extra effort. Calibrating an existing model, which adjusts its output probabilities after training, captures most of the improvement without retraining, across 3 datasets covering severity, abuse, and irony, while training on label distributions adds a small further gain. Third, we examine a dataset that includes annotator demographics, where a single shared model has one advantage and one limit. The advantage is that a model predicting the distribution of judgments matches each demographic group’s judgments significantly better than a model trained on majority labels, so both groups gain. The limit is that it does not narrow the gap in performance between the groups, because one shared model cannot encode which group is right on a given comment. Preserving annotator disagreement is therefore a low-cost step toward fairer moderation, though not a complete one.
[HC-34] “Its still 100% on me”: Designing AI Tools for the Intersection of Physicians Care and Legal Infrastructures
链接: https://arxiv.org/abs/2611.14989
作者: Gennie Mansi,Julia Kim,Lynn Dombrowski,Mark Riedl,Naveena Karusala
类目: Human-Computer Interaction (cs.HC)
备注: 27 pages, 0 figures
Abstract:Laws and regulations deeply shape physicians’ workflows and impact how they incorporate AI-based decision support tools to care for patients. Despite calls to comply with legal standards, many AI-based decision support tools have not transitioned well to clinical settings. The medical community remains concerned about their legal liability, risks to patients, and limited clinical value of such tools. To better understand how to address these concerns in the design of AI tools for clinical decision-making, we conducted 19 interviews with physicians in the U.S. We used scenario-based design to investigate physicians’ legal concerns in the context of an AI-based drug recommendation tool. Drawing on the ethics of care, we surface connections between physicians’ care and the healthcare laws and regulations that deeply shape their information needs. We describe how physicians wanted to use the tool in individual and collective practices to care for themselves, their patients, their colleagues, work institution, and even the broader medical profession. We discuss how the design of AI-based decision support tools can present legal ambiguities that would prevent their caring actions and lead to non-use of the AI tool. Through our work, we contribute to responsible AI development in healthcare. Our analysis illuminates why and how certain kinds of information reduce legal ambiguity and support physicians’ care when using AI-based decision support tools. We contribute an understanding of how to help physicians navigate legal liability and draw implications for grounding AI tools’ designs in their care.
[HC-35] Understanding the Effects of Multi-Agent Deliberation for Human-AI Decision-Making
链接: https://arxiv.org/abs/2611.14901
作者: Harry Yizhou Tian,Ming Yin
类目: Human-Computer Interaction (cs.HC); Artificial Intelligence (cs.AI)
备注:
Abstract:Multi-agent deliberation (MAD), where multiple LLMs engage in structured discussion, improves AI performance on many benchmarks, but whether it helps humans make better decisions remains untested. We conduct two user studies on LLM-response-verification and fact-checking. We first compare MAD against common AI decision-support methods that isolate its two defining dimensions — multiple AIs and multi-round interaction. We find MAD outperformed single-round AI assistance on both tasks; multi-round interaction was the main driver of this benefit, while a second AI added value when the single AI had room to improve. Then, we study two design choices in MAD — AI persona and human engagement level — finding a devil’s advocate persona to lower accuracy and user experience, while humans’ active participation in MAD improved accuracy in one task at the cost of frustration. We conclude by discussing implications for when, why, and how multi-agent deliberation can support human decision-making.
[HC-36] Hear What You Eat: In-Ear Passive Acoustic Sensing for Multi-task Dietary Monitoring
链接: https://arxiv.org/abs/2611.14878
作者: Feiyu Han,Mingtian Zhu,Dawei Yan,Weiwei Jiang,Yuanhao Feng,Yanfei Zhang,Panlong Yang
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:The increasing obesity health concerns are driving the advancement of continuous dietary monitoring. Yet, existing single-sensor solutions still face challenges of coarse-grained sensing resolution, while multi-sensor fusion strategies encounter significant engineering complexity or hardware requirements. In our work, we present a single modality-based multi-dimensional dietary monitoring and logging solution, named EARDIET, which leverages inward-facing microphones on earables to capture in-ear chewing sounds for fine-grained diet analysis. With the texture audio profiles derivation, the pipeline consisting of signal processing and lightweight deep learning, is carefully designed to achieve eating episode detection, food texture sensing, and food type recognition simultaneously. Comprehensive experiments involving 32 participants with diverse BMI profiles are conducted to validate the effectiveness of multi-dimensional dietary monitoring. Compared with the state-of-the-art solutions, EARDIET can derive multi-dimensional dietary information with only a single modality, outperforming by about 10% in terms of eating episode detection. Furthermore, through real-world clinical validation, our solution can achieve masticatory disorder identification with an averaged F1-score of 85.97%, demonstrating potential for healthcare in the future.
[HC-37] Womens Genetic Health as Family Cooperative Work: Empirical Findings and a Case for Participatory AI ICIP ALT
链接: https://arxiv.org/abs/2611.14871
作者: Ziqi Yang,Yunan Chen,Daniel A. Epstein
类目: Human-Computer Interaction (cs.HC)
备注: Presented at the CSCW 2026 Workshop on Women’s Health as Cooperative Work: Participatory AI Across Care, Community, and Policy ( arXiv:2611.03776 )
Abstract:Genetics plays a significant role in women’s health, with family members often experiencing shared symptoms or hereditary risks across generations. Yet health technologies, including emerging health AI tools, have supported women’s health as an individual health practice. Drawing on a survey of 249 U.S. adult women, we characterize women’s genetic health information work as cooperative work distributed across intergenerational family networks, identify who does this work and why, and document the barriers that fragment it. We find that generative AI use is a strong predictor of genetic health perception, but current tools focus on individuals rather than supporting collective coordination. We argue that designing AI for this domain requires participatory approaches that engage diverse women, healthcare professionals, and legal and community stakeholders, and we bring open design questions to the workshop.
[HC-38] Branching Visual Inquiry in a Generative Visual Browser: A Matched Walkthrough Study
链接: https://arxiv.org/abs/2611.14862
作者: Hanjing Shi,Kimberly Wang,Dominic DiFranzo
类目: Human-Computer Interaction (cs.HC)
备注: 13 pages. Revised preprint of a paper accepted for the 2027 AERA Annual Meeting
Abstract:Generative visual browsers combine AI-generated visual pages with reader-selected branches rather than a fixed sequence. We analyzed eight author self-walkthroughs of Flipbook, organized as four pairs with identical starting prompts about sleep, vaccines, flight, and Pluto. Three paired routes differed at the first selection; the sleep routes separated after a shared repair-oriented opening. A route could remain locally coherent while narrowing to one mechanism or leaving part of its starting question unanswered. Some diagrams lacked sources. The analysis motivates three design directions: persistent path histories, prompts that reconnect a route to its starting question, and source or uncertainty cues for generated claims. It characterizes how one generative visual browser structured eight documented routes.
[HC-39] Can I join you? Extended Social Force Based Model with Predictive Control and Barrier Functions for Humanoid Group Repositioning Amidst Obstacles
链接: https://arxiv.org/abs/2611.14853
作者: Chuxuan Zhang,Niloofar Jahanshahi,Yasaman Etesam,Mo Chen,Angelica Lim
类目: Robotics (cs.RO); Human-Computer Interaction (cs.HC)
备注: 8 pages, 4 figures, 2 tables
Abstract:As social and service robots increasingly share interaction spaces in public, they may form groups while coordinating tasks or interacting with humans. Initial group formation alone is insufficient in dynamic settings, because it can be disrupted when a person or another agent joins, leaves, or passes through the group. Although group dynamics and repositioning have been widely studied in swarm robotics and virtual-agent settings, little attention has been given to socially appropriate repositioning behaviors in groups of humanoid robots. Existing social-force-based approaches provide a useful foundation for group behavior, but they have limited obstacle handling, and can produce unsafe behavior as environmental complexity increases. We propose a novel framework that modulates traditional Social Force, guided by Model Predictive Control (MPC) to plan ahead, bounded by Control Barrier Functions (CBF) to increase safety for robot groups repositioning behaviors. This approach aims to produce safer and socially aware group repositioning behaviors under various group formation disturbance. We evaluate the method in simulation through an ablation study and an online user study. The results show that our proposed method can produce smoother and safer repositioning behaviors in complex environments while preserving social appropriateness.
[HC-40] Do People Trust AI Assistants Even When They Are Consistently Wrong?
链接: https://arxiv.org/abs/2611.14833
作者: Cameron R. Jones,Julia Wong,Aaron S. Cohen,James Woolley,Benjamin K. Bergen
类目: Human-Computer Interaction (cs.HC); Computation and Language (cs.CL)
备注: 51 pages, 14 figures, 11 tables. Preprint, not peer reviewed
Abstract:People are increasingly using conversational AI systems to make decisions despite concerns around their reliability. How well do people calibrate to AI systems’ accuracy in multi-turn conversations? We measure how people’s reliance on LLMs’ advice changes across repeated interactions, varying whether the AI assistant is helpful (100% accurate), unhelpful (0% accurate), or mixed (50% accurate). Across 4 experiments, we find people learn to trust helpful models more than unhelpful ones (67% vs 21% of the time), but only when they receive external feedback on the models’ accuracy. Without feedback, people continue to be persuaded 48% of the time by a consistently unhelpful model (vs 38% for helpful). Rather than gaining trust in helpful models, participant reliance on LLM advisors tended to be high in early trials and only fell with evidence of inaccuracy. Participants trusted LLM advisors more than they trusted human advisors (83% vs 65%, Experiment 4), and LLMs benefited more from having a helpful track record than humans did. In contrast with previous work, we find limited evidence of an advantage for arguing for the truth (versus a falsehood). Truth advantages emerged only when participants were already very confident in their position, and did not emerge at all for static hints. Collectively, the results suggest that people may be swayed by LLM advice, even when it is consistently false, unless they have access to timely ground-truth feedback. Unfortunately, in many domains such feedback may not be available, leading to overreliance on AI.
[HC-41] Known Good Known Bad: A Ground-Truth Calibration Protocol for AI Judgment in Science
链接: https://arxiv.org/abs/2611.11673
作者: Olga Lavinda
类目: Computers and Society (cs.CY); Human-Computer Interaction (cs.HC)
备注: 18 pages, 3 figures, 6 tables
Abstract:Scientific use of generative AI requires evaluation against cases with independently checkable answers. We present a ground truth calibration protocol that compares a verified reference case with reference deviant variants through neutral evaluation, an authority framed recheck, and a line by line audit. Human adjudication distinguishes deviation detection, premise accommodation, unsupported doubt, and correctness of the resulting calculation. An instructor generated organic chemistry demonstration comprises six conversations and 18 responses from an unidentified backend served through the free logged out ChatGPT interface in July 2026. Neutral evaluation detected three of five intended deviations. Both initial misses, involving oxidant stoichiometry and experimental scale, were detected at the authority framed second turn; all five detections persisted at the third turn audit. The fixed sequence supports a descriptive account of recovery across turns. In the scale case, the model initially produced a locally coherent peroxide limited yield. After recognizing the unit problem, it incorrectly retained that limiting reagent conclusion for the corrected scale. Detection and numerical repair therefore require separate assessment. The verified reference attracted speculative concerns while retaining its correct yield calculation. The protocol connects documentary verification of scientific claims and citations with empirical verification of laboratory interpretations. It supplies inspectable transcripts, reference calculations, behavior codes, and a reusable calibration architecture for AI literacy instruction. The pilot demonstrates the protocol’s measures in one case family; replicated model evaluation and learner assessment are the next validation steps. Comments: 18 pages, 3 figures, 6 tables Subjects: Computers and Society (cs.CY); Human-Computer Interaction (cs.HC) Cite as: arXiv:2611.11673 [cs.CY] (or arXiv:2611.11673v1 [cs.CY] for this version) https://doi.org/10.48550/arXiv.2611.11673 Focus to learn more arXiv-issued DOI via DataCite
计算机视觉
[CV-0] Back to Pixels: Lifting Latent Video Diffusion into Pixel Space
链接: https://arxiv.org/abs/2611.17336
作者: Yanke Song,Shuo Chen,Chris Rockwell,Guande He,Nan Liu,Desai Xie,Tianpei Gu,Yong-Hyun Park,Aryan V S,Dennis Yuren Menn,Ming-Hsuan Yang,Xun Huang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project Page: this https URL
Abstract:Frontier video generation models work in a VAE latent space that compresses video in both space and time. This compression made training tractable, but now limits high-resolution, controllable, and interactive generation – doubling resolution quadruples token count, conditions apply to groups of frames rather than individual ones, and each response must wait for the rest of its multi-frame block. We present PixVid, a simple yet effective method that lifts a pretrained latent video model to directly generate every frame in pixel space. PixVid keeps the diffusion transformer and replaces the VAE decoder with a convolutional head that renders its outputs into pixels. We find that predicting clean video rather than velocity, together with a larger noise shift, is key to quality. Finetuned from Wan2.1-14B, PixVid is competitive with the latent baseline within \sim 1.3 M training clips. Working framewise in pixels, it generates at 2\times higher resolution with the same per-frame token count, keeps depth-conditioned video editing aligned under fast motion, and responds to each action at the next frame. Project page: this https URL
[CV-1] Unbundling Multimodal Vision Encoding
链接: https://arxiv.org/abs/2611.17334
作者: Ellis Brown,Chris Hoang,Shengyi Qian,Shang-Wen Li,Mike Lewis,Marjan Ghazvininejad,Luke Zettlemoyer,Lucas Beyer,Mengye Ren,Jakob Verbeek,Saining Xie,Hu Xu
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: Project page: this https URL
Abstract:How multimodal language models should encode visual input remains up for debate. Most models use a pretrained vision encoder, while some recent recipes train the encoder from scratch with the language model or remove it entirely. Vision encoders bundle together three design choices: a visual prior from upstream pretraining, dedicated vision capacity, and pre-fusion placement of that capacity (i.e., before the multimodal backbone). Because published models also differ in data, architecture, and compute, comparing them cannot isolate the effect of each choice. We disentangle these choices via a controlled framework where we pretrain all models from scratch on common data and compare them on visual question answering and language benchmarks at matched lifetime compute. We find that the visual prior is not necessary; an encoder trained jointly from scratch is on par with a pretrained one, without a separate training stage. What matters instead is dedicated vision capacity: an encoder-free model trails on visual understanding even when its backbone is enlarged by the encoder’s parameter count. Yet that capacity need not precede multimodal fusion, as vision-only modules added in-fusion to early backbone layers perform comparably to a pre-fusion encoder and can lower time to first token. These results also hold across backbone sizes, when encoder pretraining is treated as free, and when starting from a pretrained language model. Overall, our findings favor dedicated vision capacity trained jointly with the backbone and placed early within it.
[CV-2] B: Test-time Baking for Efficient Rendering of Decoder-only View Synthesis Models NEURIPS2026
链接: https://arxiv.org/abs/2611.17331
作者: Chin-Yang Lin,Ryo Hachiuma,Jaesung Choe,Min-Hung Chen,Yu-Chiang Frank Wang,Yen-Yu Lin,Wei-Chen Chiu,Yu-Lun Liu,Cheng Sun
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: NeurIPS 2026 (Spotlight). Project page: this https URL
Abstract:Decoder-only Large View Synthesis Models (LVSMs) with KV-cache have recently achieved state-of-the-art quality by regressing novel views from neural networks without reconstructing 3D geometry, while Test-Time Training (TTT) layers save computation but sacrifice quality. We propose Test-Time Baking (TTB), a strategy that learns to bake global scene information from cross-view attention into a lightweight MLP, enabling fast and high-fidelity novel-view decoding. The first key is to repurpose the fast-weight MLP of TTT to learn the cross-view attention mapping explicitly from its input-output pairs, rather than replacing attention entirely, achieving the same rendering efficiency as TTT while delivering substantially higher quality. The second key addresses a modality misalignment between context prefilling and novel-view rendering: we disentangle context tokens into camera-only context rendering tokens to serve as TTB input, and context ground-truth tokens with image content to serve as the TTB mapping target, ensuring the baked MLPs are relieved of the modality generalization challenge between context and novel-view processing. Our method achieves up to 23x rendering speedup with quality on par with the state-of-the-art across several benchmarks. Project page: this https URL
[CV-3] ProtoVis-Nav: Prototypical Visual Imagination for Visual-Spatial Aligned UAV Navigation
链接: https://arxiv.org/abs/2611.17330
作者: Guangyuan Sui,Yunzhong Hou,Chi Harold Liu
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Vision language navigation (VLN) is very challenging for Unmanned Aerial Vehicles (UAVs) since small errors in 3D motion prediction can lead to navigation errors and collisions. Unable to fathom execution outcomes in 3D environments, existing methods often struggle to predict accurate navigation waypoints from language instructions. To solve this, we introduce the Prototypical Visual Imagination framework, shortened as ProtoVis-Nav, which forces the model to imagine an outcome-associated local visual representation for the intended next UAV movement. Specifically, we predict feature prototypes and use their linear combination to estimate the visual feature of the target region. In this manner, the VLN model first imagines the local visual representation of the next intended location, and then uses predicted prototypes as context for generating the final action, promoting alignment between visual context and waypoint outputs. Extensive experiments demonstrate the effectiveness of the ProtoVis-Nav framework and show substantial improvements over existing baselines.
[CV-4] K2V: Generative Keyframing For Physically Plausible Video Generation
链接: https://arxiv.org/abs/2611.17324
作者: Felix O’Mahony,Daniel Olmeda Reino,Roberto Cipolla,Ayush Tewari
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Video diffusion models are able to generate samples with high-quality appearance and textural qualities. However, content produced by these methods is known to exhibit significant physical errors. This is because, in a classical video generation setting, all physical reasoning must take place simultaneously and exclusively within the diffusion this http URL propose to resolve these issues by redistributing the task of physical reasoning onto a language model-backed keyframe generation module, which allows for explicit reasoning in language, and fine-grained control over the appearance of the video at important moments in the video. We then condition a video generation model with these keyframes, reducing the role of the diffusion model to the simpler task of interpolating between physically-correct intermediate states, rather than inferring an entire sequence with diffusion. Results demonstrate that our approach significantly outperforms existing approaches, achieving state-of-the-art performance on the Physics-IQ Verified benchmark and PhyGenBench.
[CV-5] BrickFlow: Connectivity-Guided Brick Reconstruction NEURIPS2026
链接: https://arxiv.org/abs/2611.17321
作者: Peter Kulits,Cordelia Schmid
类目: Computer Vision and Pattern Recognition (cs.CV); Graphics (cs.GR)
备注: NeurIPS 2026
Abstract:Given a set of LEGO parts and an image of an assembled object, our task is to infer the pose of each part to assemble the object. In contrast with traditional 3D reconstruction, doing so requires satisfying discrete physical constraints, which provide structure but render the problem combinatorial and highly sensitive to small pose errors. To learn a prior over how parts fit together, we train a set-conditioned SE(3) flow model on procedurally generated assemblies. We then finetune this model to inject image conditioning, using 1% as much data. However, while the learned flow captures global structure, it does not enforce connectivity. To address this, we introduce an analytic connectivity-field guidance to pull predictions toward connected assemblies at test time. Through this, we demonstrate improved build quality with model size, pretraining, and guidance. We make our models and data available for research purposes at this https URL
[CV-6] Spatial Foundry: A Multimodal Dataset for Spatial Intelligence
链接: https://arxiv.org/abs/2611.17318
作者: Ethan Lin,Yang Song
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project Page: this https URL
Abstract:Spatial intelligence requires connecting visual observations to the geometry, objects, and persistent state of an environment. Images alone leave these relationships implicit, while disconnected annotations make it difficult to study appearance, camera motion, and scene changes under a common reference. We introduce Spatial Foundry, a collection of 5,280 furnished residential 3D scene configurations with editable assets and aligned visual, geometric, and temporal data. Its modalities include floor plans, perspective and panoramic RGB, metric depth, point clouds, meshes, instance-level 3D bounding boxes, and camera trajectories paired with video. Shared scene coordinates link observations across views and time; editable source scenes enable additional camera paths and controlled object changes. We describe the data representation and construction workflow, and discuss applications in floor-plan-conditioned 3D generation, geometric reconstruction, instance-level perception, camera-conditioned video generation, shared world modeling, and scene editing. Controlled camera and scene variations support spatial perception and world-model research. The public project page provides videos and an interactive colored point-cloud preview.
[CV-7] FLEX-LBP: Learning Local Binary Patterns with Exact Binary Inference for Texture Classification
链接: https://arxiv.org/abs/2611.17302
作者: Mohamed Alimoussa,Amal El Kaid
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 21 pages, 5 figures, 8 tables. Code: this https URL
Abstract:Local Binary Patterns (LBP) remain among the most widely used texture descriptors, from industrial inspection to medical imaging. Yet their radius, sampling angles, color transform and thresholds are fixed by hand for each dataset. We present FLEX-LBP, which learns these parameters by backpropagation while every forward pass computes exact binary codes and hard histogram counts. A straight-through estimator acts only in the backward pass, so the deployed descriptor is an interpretable hard-binary LBP with 90 learnable parameters and unchanged inference cost. In a controlled ablation on six color texture benchmarks, the component worth learning is dataset-dependent. One configuration, applied unchanged everywhere, learns the geometry and thresholds with the color transform frozen. It improves the nearest-neighbor accuracy of the identical fixed descriptor by 3.55 points on average over five seeds, significantly on five of six benchmarks and never significantly worse. With a linear discriminant readout, it gains 0.9 points; the fully learned descriptor gains 2.7 and has the highest mean. FLEX-LBP is more accurate than five classical LBP-family descriptors, PCANet and a neural LBP on five of six benchmarks. Its mean nearest-neighbor accuracy exceeds that of frozen ImageNet ResNet-50 features (23.5 million parameters). On the Kather colorectal cancer histology benchmark, the fully learned descriptor reaches 93.8%, significantly above the best classical descriptor (Completed LBP, 92.8%). The learned radii adapt in directions that recur across seeds. Noise drowns single-pixel comparisons and illuminant change flips inter-channel ones; median sampling or gray-world normalization restores 96-97%. The code is available as a drop-in PyTorch layer at this https URL.
[CV-8] Consistent Video-to-Video Translation via Explicit Correspondences ECCV2026 WWW
链接: https://arxiv.org/abs/2611.17298
作者: Gaurav Parmar,Zhengqi Li,Richard Zhang,Jun-Yan Zhu,Srinivasa Narasimhan,Eli Shechtman,Yotam Nitzan
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Graphics (cs.GR); Machine Learning (cs.LG)
备注: ECCV 2026. Project Website: this https URL
Abstract:Interactive video-to-video applications require real-time generation while maintaining long-range temporal consistency. However, recent methods achieve speed by restricting attention to a fixed-length causal window. This limits long-range consistency whenever relevant history falls outside the context window. To address this limitation, we introduce vid2vid-long, a correspondence-based approach grounded in a simple observation. Video-to-video translation provides an explicit cross-frame reference: the regions that match in the input video should match in the output video too. Concretely, we compute patch-level correspondences in the input, guiding retrieval of previously generated outputs in the corresponding regions, and condition the model on these patches through correspondence cross-attention. Our experiments demonstrate that vid2vid-long consistently improves long-range temporal consistency across recent video models while preserving generation quality and real-time performance.
[CV-9] Objects as Addresses: Instance Binding in Vision Transformer Attention
链接: https://arxiv.org/abs/2611.17240
作者: Khalil Sabri,Guillaume-Alexandre Bilodeau,Nicolas Saunier,Wassim Bouachir
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Humans naturally parse visual scenes into coherent individual objects rather than scattered features, a capability cognitive science calls object binding. While recent work shows that pretrained Vision Transformers (ViTs) intrinsically acquire this ability, how native self-attention computations mechanistically keep object instances separate remains unresolved. By studying this across eight pretrained ViTs, we found that instance separation is expressed more strongly in attention routing than in token content: selected query-key relations distinguish individual objects more reliably than value similarity or residual representations. Inside heads, patches belonging to one individual tend to read from a common image region, which we term their attention address. Visually distinct parts of one person share an address, whereas identical duplicated instances receive separate, movable addresses. These shared destinations arise in part because query interactions with the spatial structure of keys direct attention toward consistent regions for one object and elsewhere for another. In DINOv2 and most other backbones we tested, editing this query component at the selected head redirects its attention address toward another individual. Grouping patches by this query component separates same-category instances, reaching .615 FG-ARI versus .289 for the same clustering of final-layer features, which tend to merge them.
[CV-10] MALNISI-Infused Graph Construction and Weighted Contrastive Loss based GAT for Unsupervised Representation Learning in HSI
链接: https://arxiv.org/abs/2611.17199
作者: Soumi Sarkar,Subhadip Boral,Sarbani Palit,Ashish Ghosh
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Hyperspectral images (HSI) possess discriminating spectral characteristics with respect to land utilization due to high-dimensional information. Recently, unsupervised representation learning using graph neural networks has unveiled these utilization-specific complex spectral relations. This article has three main contributions. First, it proposes an index, the Mean Absolute Local Neighborhood Inverse Spectral Index (MALNISI), which measures the similarity between two pixels based on their neighborhood information. The index is used to construct a weighted adjacency matrix, giving the HSI a graph structure. Second, it proposes an upgraded Graph Attention Network that learns compact latent representations of hyperspectral data from the constructed graph. In this network, an enhanced multi-head attention mechanism incorporates the MALNISI weights as an auxiliary feature to emphasize learned attention between neighboring pixels. To alleviate over-smoothing and preserve fine-grained spectral details, the architecture integrates residual connections, skip projections, and feature normalization, which together enhance representation capability, stabilize learning, and accelerate convergence. Third, it proposes a weighted contrastive loss function that incorporates neighborhood-aware MALNISI between connected pixels, encouraging strongly connected pixels in the high-dimensional space to remain close in the latent space, and weakly connected pixels to stay apart. The resulting discriminative low-dimensional embeddings are then clustered using spectral clustering. Experimental analysis on six benchmark HSI datasets demonstrates the efficiency and necessity of the three aforementioned contributions, and the comparative study highlights superior performance over fourteen state-of-the-art approaches.
[CV-11] DMD: Endpoint Distribution Matching for Few-Step Diffusion Distillation
链接: https://arxiv.org/abs/2611.17194
作者: Tong Zhao,Xianfang Zeng,Yucheng Han,Ruoyu Wang,Mingkun Lei,Gang Yu,Chi Zhang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Code Link: see this https URL
Abstract:Distribution matching distillation (DMD) has emerged as a leading paradigm for distilling diffusion models into few-step students, where an auxiliary fake score model tracks the student’s distribution and the discrepancy between the teacher and fake predictions drives the student toward the teacher. In multi-step distillation, although only the student’s final output is delivered, existing methods train the fake model on intermediate states or across all denoising steps. Consequently, the student’s update signal becomes entangled with intermediate distributions, which can distort the matching direction and remain nonzero even after the student’s final outputs already match the teacher distribution. To resolve this mismatch, we propose Endpoint Distribution Matching Distillation (eDMD), which trains a single shared fake model exclusively on the student’s final outputs and uses it to correct every denoising step. Our theoretical analysis proves that this endpoint reference guarantees that the distribution matching term vanishes if and only if the student’s final output distribution matches the teacher’s, regardless of intermediate distributions. Beyond the fake reference, we further strengthen teacher supervision through a precise teacher (PT) that replaces single-step guided predictions with short multi-step trajectories on the final noise interval, paired with a calibrated query density for intermediate steps. Experiments on SD3-Medium, FLUX.2-klein (4B and 9B), and Qwen-Image-2.1 demonstrate that eDMD consistently outperforms existing distribution matching methods for 4-step generation.
[CV-12] Consensus Capture in Reference-Based UAV Video Stabilization: A Benchmark on Dynamic Urban Scenes
链接: https://arxiv.org/abs/2611.17190
作者: Robert Fonod,Nikolas Geroliminis
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 50 pages, 17 figures, 24 tables. Code at this https URL and data at this https URL and this https URL
Abstract:Reference-frame stabilization registers every frame of a hovering-drone video onto one reference frame, keeping object positions comparable. The registration pipelines it relies on are selected on feature-matching benchmarks in which only the camera moves. Over a busy intersection, vehicles moving together supply mutually consistent correspondences that can draw the robust homography estimate toward their motion: consensus capture, which inlier statistics cannot flag. A two-part benchmark measures it and the object masking meant to prevent it. A static protocol scores ten pipelines under exactly known homographies on 4K aerial frames, tuned on a site-disjoint split at a matched resolution and feature budget. A dynamic protocol scores them at an offline and a real-time budget on hovers selected from the video alone to span the amount and coherence of moving foreground, plus a stationary control. It measures background alignment, object position error and the share of a moving group’s displacement an estimate follows, against a pseudo-ground-truth reference audited with manually annotated ground control points, which also score every candidate. On static scenes, tuned AKAZE, RootSIFT, SIFT and BRISK lead, retuning the matcher thresholds raises the mean average accuracy (mAA) by at most 0.0062, and masking costs less than 0.005. Unmasked ORB is captured on 28 of 304 real-time dynamic pairs and on 6 offline; its real-time captured estimates have a median corner error of 18.8 px yet typical inlier counts and ratios. Masking removes capture for every pipeline and leaves the leading pipelines unchanged. Code, data and results are released.
[CV-13] Neural Render Proxies for Interactive and Differentiable Lighting
链接: https://arxiv.org/abs/2611.17185
作者: Sergio Sancho,Alexander Rath,Marco Manzi,Pascal Chang,Amit H. Bermano,Derek Nowrouzezahrai,Markus Gross,Marios Papas
类目: Graphics (cs.GR); Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at Computer Graphics Forum (Proc. EGSR 2026)
Abstract:Within the CG animation production pipeline, the challenges artists tackle in Lighting can be immense. Even minor adjustments require re-rendering massive scenes with slow offline renderers; global illumination has to be sampled and complex shaders have to be evaluated, leading to iteration times of minutes to hours per frame. To accelerate this process, we introduce a novel neural render proxy (NRP) that enables differentiable relighting of static scenes with fixed camera and materials at interactive rates. Our main insight is the decoupling of traditional rendering into path sampling and emission computation. From a single, light-agnostic render pass, we collect light transport data in the form of path samples. This enables rapidly sampling lighting conditions on the fly, and training a scene-specific lightweight neural network that learns how light is transported from any scene location to any image pixel. This approach is compatible with non-differentiable production renderers, induces minimal memory requirements during inference, and scales only with resolution and the number of light sources and parameters, but independently of scene and appearance complexity. Our evaluation demonstrates interactive frame rates for relighting (~30-60 Hz) while closely approximating the visual fidelity of ground-truth path tracing. In addition, the differentiable NRP enables fast, gradient-based inverse workflows, allowing artists to efficiently solve for lighting parameters from intuitive image-space edits or generative targets.
[CV-14] Validation-Guided Prompt Selection for Differential Morphing Attack Detection WACV2027
链接: https://arxiv.org/abs/2611.17176
作者: Raghavendra Ramachandra,Hailin Li
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted in WACV 2027 Workshops
Abstract:Differential morphing attack detection (D-MAD) compares a trusted live capture with a document image that may contain a facial morph. We investigate whether labelled validation data can improve a frozen-parameter vision-language evaluator through prompt selection rather than model fine-tuning. Version~1 generates 50 candidate prompts and selects the top-ranked instruction using a security-oriented loss that prioritises missed morph attacks (MACER) while accounting for bona fide classification errors (BSCER), response failures, parse failures, and decision inconsistency. Version~2 recursively revises prompts using aggregate validation feedback. On a subject-disjoint test set of 160 pairs spanning LMA, MIPGAN-2, and PIPE, the selected Version~1 prompt achieves a (19.17%) morph attack classification error rate (MACER), with (0.00%) observed bona fide sample classification error rate (BSCER), compared with (70.00%) MACER for a controlled manual prompt. The paired (50.83) percentage-point difference has a (95%) subject-bootstrap interval of ([43.33,58.33]). Recursive revision does not surpass the best Generation~0 validation loss; the reported revised descendant obtains (31.67%) test MACER. All 23 attacks missed by Version~1 are landmark-based (LMA) morphs. Overall, the results show that prompt formulation can substantially influence frozen-parameter VLM-based D-MAD performance, while also highlighting the challenge of recursive prompt refinement under a compact 10-subject validation setting.
[CV-15] SpotDiff: A Plug-and-Play Framework for Zero-Shot Change Detection in Remote Sensing Imagery
链接: https://arxiv.org/abs/2611.17172
作者: Anit Upadhyaya,Ata Akbari Asanjan,Chaitanya Jalluri,Subash Yeggina,Jeremy Kravitz
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Remote sensing change detection usually relies on large collections of temporally aligned, pixel-level annotations, which limits its scalability across regions, imaging platforms and application domains. Recent zero-shot approaches reduce the need for supervision, but they often localize fine-grained changes poorly because their features are coarse or inconsistent across time. We present SpotDiff, a training-free, plug-and-play framework for zero-shot change detection built by composing complementary vision foundation models. SpotDiff uses DINOv3 for semantically robust features, AnyUp to reconstruct them at pixel resolution, and SAM 3 for text-guided region-of-interest proposals, which an inference-time multi-scale strategy (SWIMS) makes reliable on small objects. Change is then measured by comparing the reconstructed features of the two dates inside the proposed regions, with no learnable components, fine-tuning or change labels. To our knowledge, SpotDiff is the first change detection method to use a universal feature upsampler, which alone adds 13.8 F1 over bilinear upsampling of the same features. On LEVIR-CD, SpotDiff reaches 82.46 F1, 2.06 points above the strongest prior training-free method, and on the multi-class SECOND benchmark it improves building change detection by 10.3 IoU and the six-class average by 2.2 IoU. Qualitative results on PlanetScope (3 m) and Sentinel-2 (10 m) imagery show that the same pipeline transfers across resolutions without modification. Our results indicate that composing modern foundation models is an effective route to scalable, label-free change detection.
[CV-16] UAV-Flow: A Unified Suite for Benchmarking and Learning Language-Guided UAV Flight
链接: https://arxiv.org/abs/2611.17163
作者: Donglin Yang,Xiangyu Wang,Yantong Zhong,Wenhao Zheng,Zihan Ding,Yue Liao,Hongsheng Li,Ke Xu,Si Liu
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注: 22 pages, 10 figures
Abstract:Language-guided UAV navigation has largely focused on long-horizon route planning, leaving short-range, object-relative flight maneuvers underexplored. To address this gap, we formulate Flying-on-a-Word (Flow), a task requiring a UAV to convert concise natural-language instruction, egocentric visual observation, and current state dynamics into continuous flight actions. Flow encompasses both primitive motion commands and complex object-interactive behaviors, assessing motion intent comprehension and spatial context grounding. To enable policy learning and closed-loop evaluation, we introduce UAV-Flow, a comprehensive dataset comprising 30,692 expert-piloted real-world trajectories across eight distinct motion categories, alongside 10,109 simulation trajectories in the UAV-Flow-Sim environment equipped with task-specific completion criteria. Furthermore, we develop FlowVLA, which undergoes imitation pretraining on mixed real-sim demonstrations and online post-training via Tail-Aware GRPO (TGRPO). By leveraging group reward distributions, TGRPO effectively reinforces high-reward action sequences. Evaluated on 273 held-out UAV-Flow-Sim episodes, pretrained FlowVLA achieves a mean success rate of 75.58%, outperforming the strongest baseline by 6.72 percentage points, which TGRPO post-training further elevates to 93.16% with superior sample efficiency. Physical UAV experiments demonstrate robust execution of object-relative commands in both indoor and outdoor real-world scenes. Together, Flow establishes a novel benchmark and learning framework for language-conditioned flight maneuvers.
[CV-17] HiVE-VQA: Hierarchical Interaction via Volumetric Embedding for Video Quality Assessment WACV2027
链接: https://arxiv.org/abs/2611.17146
作者: Robbie Hamilton,Fan Zhang,David Bull
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to WACV 2027, 11 pages, 3 figures, 6 tables
Abstract:The proliferation of user-generated video content has made accurate and robust no-reference video quality assessment (NR-VQA) a highly important challenge. Existing research increasingly relies on architecturally complex multi-branch pipelines or computationally intensive fine-tuning of single-branch backbones. In this paper, we investigate whether a single frozen video foundation model can serve as an effective NR-VQA backbone, and propose HiVE-VQA, the first framework (to our knowledge) to apply a JEPA-based video encoder to this task. HiVE-VQA extracts token grids from multiple depths of a pre-trained V-JEPA 2.1 encoder and stacks them into a hierarchical spatio-temporal volume, which a compact 3D convolutional head fuses across its layer, spatial, and temporal axes, combining low-level distortion cues from shallow layers with the semantic and temporal context captured at greater depths. The encoder remains frozen throughout, so only the head is trained: 5.2M parameters (5.7% of the total), with no auxiliary multi-modal models and an inference cost that is invariant to video resolution. Trained solely on LSVQ and evaluated without any fine-tuning, HiVE-VQA outperforms 17 state-of-the-art (SOTA) methods in SROCC and PLCC on both intra-dataset (LSVQ-test, LSVQ-1080p) and cross-dataset (KoNViD-1k, LIVE-VQC) benchmarks. These results indicate that hierarchical, depth-wise fusion of features from a single frozen video foundation model provides a more effective and efficient basis for NR-VQA than task-specific multi-network fusion. Code and model checkpoints are available at this https URL.
[CV-18] STeP-C: Real-Time Neural Compression for Distributed Fiber-Optic Sensing at Scale WACV2027
链接: https://arxiv.org/abs/2611.17126
作者: Tingfeng Li,Ming-Fang Huang,Shaobo Han,Yue Tian
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: Accepted to WACV 2027 (Applications Track). 10 pages, 9 figures, 3 tables
Abstract:Distributed fiber-optic sensing (DFOS) technologies such as Distributed Acoustic Sensing (DAS) generate continuous, high-resolution spatio-temporal data at rates exceeding hundreds of megabytes per second, creating severe challenges for real-time storage and transmission. While neural compression methods can achieve strong rate-distortion performance on sensing data, prior approaches either require per-segment retraining incompatible with streaming, or have only been validated offline. None have been demonstrated end-to-end on a continuously streaming DAS acquisition system at scale. We present STeP-C (Spatio-Temporal Predictive Compression), a neural compression framework that decouples spatial encoding, temporal prediction, and entropy-optimized residual coding, trained end-to-end with a rate-distortion objective motivated by conditional entropy reduction. Critically, we identify entropy coding as the dominant bottleneck for streaming deployment. We address it through GPU-side symbol computation, cached probability tables, and parallel GIL-free arithmetic coding, together with FP16 inference and batched entropy coding, enabling a full real-time pipeline built around a GPU ring buffer. We validate this pipeline on a live DAS testbed, achieving a 98.85% file size reduction at 31.04 dB full-band PSNR and an average throughput of 535 MB/s, exceeding the 324 MB/s requirement for 1 kHz, 81 km DAS streams. We further show that STeP-C substantially outperforms conventional codecs at practical streaming bitrates, can be applied to unseen channel ranges within the same acquisition system without retraining, and largely maintains downstream detector consistency, with 96.62% high-confidence class agreement and 100% Jaccard similarity on detected event types.
[CV-19] Predicting Ethanol Drinking Status from Nonhuman Primate Liver Histology: A Proof of Concept for Parameter-efficient Fine-tuning in Few-Shot Vision-Language Modeling
链接: https://arxiv.org/abs/2611.17105
作者: Amirmohammad Mohammadi,Rupak Khadka,Kathleen A. Grant,Mary Lauren Benton,Joshua Peeples
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 3 pages, 2 figures. Accepted for publication in the proceedings of IEEE BIBM 2026. The code is publicly available: this https URL
Abstract:Combining vision-language models (VLMs) with parameter-efficient fine-tuning (PEFT) offers a promising approach to few-shot whole-slide image (WSI) classification in computational pathology. This study evaluates this approach for predicting ethanol drinking status from nonhuman primate liver histology. Adapters are evaluated against non-PEFT configurations on a subset of the Monkey Alcohol Tissue Research Resource (MATRR). All evaluated PEFT configurations achieve higher mean balanced accuracy than the non-PEFT baselines. For example, in the four-shot setting, attention-based multiple instance learning (ABMIL) alone achieves 43.5% mean balanced accuracy, compared with 78.1% after adding an adapter. Qualitative attention analysis shows that textual pathway adaptations produce localized attention heatmaps, while visual pathway adaptations produce more diffuse ones. A pathologist’s review identified similar histological features across adaptations.
[CV-20] uMAP: A Dual-View Morphotype-Aware Pathomics Pipeline for Quantifying Renal Tubular Remodeling
链接: https://arxiv.org/abs/2611.17104
作者: Ruijie Wu,Yuechen Yang,Yanfan Zhu,Yu Wang,Shilin Zhao,Ming-Tsun Tsai,Szu-Yuan Li,Shuo-Ming Ou,Tz-Heng Chen,Haichun Yang,Yuankai Huo
类目: Computer Vision and Pattern Recognition (cs.CV); Image and Video Processing (eess.IV)
备注: 5 pages, 3 figures, 1 table. Submitted to IEEE ISBI 2027
Abstract:Quantifying disease-associated renal tubular remodeling is challenging because tubular appearances are heterogeneous. We introduce TuMAP (Tubular Morphotype-Aware Pathomics), a pipeline combining (1) dual-view foundation-model representations and provisional morphotype annotation, (2) compositional and within-morphotype pathomics, and (3) patient-level statistical inference. From 456,158 tubules in 358 PAS-stained kidney biopsy slides, TuMAP identified eight image-based morphotypes, seven of which showed stable subsampling assignments. Across 219 patients in six diagnostic groups, diabetic nephropathy (DN) and tubulointerstitial nephritis/acute tubular injury (TIN/ATI) each differed from minimal change disease in seven morphotype proportions, while IgA nephropathy differed in six. DN showed reduced nuclear solidity and increased eccentricity across all eight morphotypes. Area-normalized nuclear density exposed additional within-morphotype differences. TuMAP provides a framework for separating changes in tubular morphotype composition from phenotypic differences within morphotypes.
[CV-21] Joint Null-Space Projection for Data-Free Model Merging
链接: https://arxiv.org/abs/2611.17103
作者: Tian Xia
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:Merging independently fine-tuned checkpoints into a single multi-task model requires no data, no labels, and no gradient steps. In practice the task vectors of different tasks interfere, and the standard remedies–averaging, magnitude pruning, sign election–operate entry by entry, ignoring that each delta induces a low-dimensional update on the hidden features it acts upon. We recast merging as a subspace interference problem. For every two-dimensional weight tensor we take the top-r right-singular subspace of each task’s delta–the input directions along which that task moves the representation most–and remove from each task vector the component lying in the union of the other tasks’ subspaces. The resulting Joint Null-Space Projection (JNSP) is closed-form and order-independent, needs no forward pass, and is defined entirely by the checkpoint set. On eight CLIP image-classification tasks JNSP merges ViT-B/16 experts to 89.45% average accuracy without using a single image, 11.88 points above the strongest parameter-wise training-free baseline (TIES-Merging, 77.57%), 14.22 above Task Arithmetic, and 0.98 above WUDI’s data-free consensus merge (88.47%). The gain holds across three backbones on the independently trained TALL release (+10.8, +12.6, +9.1 over TIES for ViT-B/16, ViT-B/32, ViT-L/14) and replicates there at 89.20%; controls confirm the source–a single shared subspace instead of per-task ones loses 2.9 points, and a sequential variant collapses to 69.6%. On RoBERTa-large the same projection underperforms plain task arithmetic, which we report rather than hide. Code is available at this https URL.
[CV-22] SALT: State-Aligned Low-Rank Transfer for Single-Temporal Cross-Dataset Structural Localization in Remote Sensing WACV2027
链接: https://arxiv.org/abs/2611.17077
作者: Aggelos Psiris,Georgios Angelidis,Giorgos Sfikas,Maria Vakalopoulou,Thomas Lagkas,Panagiotis Sarigiannidis,Georgios Th. Papadopoulos
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at the Third Workshop on Computer Vision for Geospatial Image Analysis (GeoCV), IEEE/CVF WACV 2027
Abstract:Change detection in remote sensing conventionally relies on spatially aligned bi-temporal imagery, which may be unavailable or unreliable in practical scenarios. Recent single-temporal approaches address this limitation by learning structural evidence, such as building regions, from a labeled source dataset and transferring it to unseen target imagery using only the current image. However, adaptation to the source dataset can over-specialize pretrained features and reduce their cross-dataset generalization, a risk that this paper argues is higher when the feature adaptation is parameterized independently of the states used for prediction. To address this issue, the current study introduces SALT, a State-Aligned Low-Rank Transfer framework. SALT keeps a DINOv3 backbone frozen and learns negative and positive structural states that both score each patch and, through State-Aligned Low-Rank Adaptation (SALA), span the rank-2 subspace in which patch features are corrected. Directional State Alignment supervises these corrections to raise the evidence of positive patches and lower that of negative ones, while Geometry Preservation limits changes to the pairwise relations among pretrained patch features. Experiments on LEVIR-CD and the WHU aerial building dataset show that, at the pixel level and in both transfer directions, SALT outperforms four baselines reproduced under a common source-only protocol, improving over the strongest reproduced baselines by 9.45 and 5.79 points in pixel-level AP and by 27.69 and 14.25 points in IoU (LEVIR-CD to WHU and WHU to LEVIR-CD, respectively), with only 0.068M trainable parameters and 40.2 ms per image. In order to facilitate reproducibility and further research, the source code of SALT is available at this https URL
[CV-23] Ropedia Xperience 2: A Gold Standard for Scaling Physical AI Environments and Behaviors
链接: https://arxiv.org/abs/2611.17060
作者: Hao Li,Fangzhou Hong,Zhaoxi Chen,Hui En Pang,Kairui Hu,Mingze Sun,Runmao Yao,Xinhao Ji,Xiyuan Zheng,Xuanli Chen,Yalun Dai,Yap Kim Hui,Dingwen Zhang,Yuanyuan Gao,Ziang Cao,Cheng Huang Yean,Jackson Peng,Rui Duan,Yaobo Liu,Zetong Wang,Ziwei Liu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: this https URL
Abstract:Large foundation models such as GPT-6 Astra and Claude Fable 5.1 have recently been shown to control robots directly from visual observations. They nevertheless remain brittle on long-horizon, contact-rich tasks in real-world environments. Further progress depends on two complementary resources: demonstrations of human behavior that supervise how tasks are performed, and interactive environments in which policies can be rolled out, evaluated, and improved in closed loop. Language models have advanced along well-characterized scaling laws; the corresponding question for physical AI is how capability scales with environments and behaviors, which first requires identifying which data to scale and how to produce and reuse such data continuously. To address this question, we introduce Ropedia Xperience 2, comprising 1,589 released metric-scale 3D scans and a collection of 4,230 egocentric recordings of 74 tasks (131 hours). We further show that each capture can be repurposed for robot learning: scenes are converted into physically interactive simulation environments, human motion is retargeted to whole-body robot trajectories, and additional robot behaviors are obtained through virtual teleoperation and automated long-horizon trajectory generation. We propose this joint capture of environments and behaviors as a gold standard for physical AI data collection, laying a data foundation for establishing scaling laws over environments and behaviors.
[CV-24] AURA 2.0: From User-Driven Responses to Model-Initiated Conversations in Streaming Video
链接: https://arxiv.org/abs/2611.17048
作者: Shuhan Li,Huankang Guan,Xintong Guo,Yang Bo,Yuyang He,Dunyuan Xu,Jingyan Chen,Fang Liu,Tsz Kin Wan,Jinpeng Chen,Zhenqi He,Yuxuan Hu,Wenbo Li,Han Gao,Qijun Li,Long Chen,Hongsheng Li,Rui Liu
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Most AI assistants respond only when prompted, remaining silent even when events in the video stream offer meaningful opportunities for conversation. To move beyond purely user-driven interaction, we present AURA 2.0, a streaming video large language model that introduces three core upgrades to AURA. The key advance is Model-Initiated Conversation (MIC), which enables the model to identify topics worth discussing in the evolving visual stream and initiate conversations at appropriate moments without a user query. Stream-Aware Tool Calling (SATC) enables the model to invoke external tools using the current visual and conversational context and incorporate returned information into the ongoing interaction. Interactive Visual Grounding (IVG) supports visual grounding in both user inputs and model outputs. The model can identify objects marked by users on video frames and produce frame-aligned bounding boxes to link its responses to specific objects and regions. To develop these capabilities, we construct a dedicated streaming interaction dataset and adopt a three-stage training scheme to integrate the new behaviors with AURA’s existing streaming interaction abilities. We further incorporate Long-Horizon Interaction Memory (LHIM) to maintain contextual continuity over extended video streams. Evaluations on public streaming benchmarks show that AURA 2.0 achieves leading performance among open-source streaming models, while targeted evaluations further validate its new interaction capabilities. We open-source the model weights and inference code.
[CV-25] ElasticDino: Dense Semantic Features Using Elastic Deformations
链接: https://arxiv.org/abs/2611.16984
作者: Ulysse Mizrahi,Sean Zaretzky,Dov Danon,Amit H. Bermano
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Vision transformers (ViTs) serve as the backbone for countless vision tasks, including dense prediction tasks, such as segmentation and depth estimation, even though the quadratic complexity of self-attention severely limits feature resolution and induces spatial artifacts. In this work, we introduce ElasticDino, a lightweight adapter which produces high-resolution feature maps, adhering to the fine details of the input image. As a key insight, we observe the consistency of artifacts across channels, and simultaneously resolve them to match the input image structure. Using a slim architecture based on spatial deformations, ElasticDino produces pixel-aligned, high-resolution feature maps efficiently. Leveraging deformations allows strictly preserving sharp semantics, and thereby avoiding the smoothness biases inherent to alternative approaches. Due to the absence of ground truth, we employ pixel-level semantic segmentation and depth estimation with linear heads to predict a set of downstream dense tasks, guiding the deformation process without leaking task-specific semantics into the model. We demonstrate superior accuracy over current methods on a range of popular dense task benchmarks, with better computational performance.
[CV-26] A Disease-specific Retinal Aging Biomarker for Predicting Progression to Late Age-related Macular Degeneration
链接: https://arxiv.org/abs/2611.16955
作者: Hexin Dong,Benjamin Hou,Puzhen Wu,Mingquan Lin,Emily Y. Chew,Zhiyong Lu,Yifan Peng
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 41 pages, 13 figures (including supplementary material)
Abstract:Age-related macular degeneration (AMD) is a leading cause of irreversible vision loss in older adults. While chronological age is a primary risk factor, it inadequately reflects individual variations in retinal aging, limiting effective AMD risk stratification and timely intervention. Here, we propose and validate fundus-based AMD age (FAA), which quantifies disease-specific retinal aging in AMD and captures accelerated AMD-related retinal aging, thereby improving prognostic precision over chronological age. Trained on 36,782 images from 2,192 participants in the Age-Related Eye Disease Study (AREDS) dataset and validated across AREDS and AREDS2 cohorts, our model demonstrates significant improvements over traditional chronological age-based approaches. Specifically, integrating the proposed FAA into Cox proportional-hazards models significantly enhances prediction performance. FAA improved the C-index from 0.836 to 0.888 and reduced the 5-year Brier score from 0.050 to 0.046 in AREDS, with consistent gains observed in AREDS2 (C-index 0.715 to 0.756; 2-year Brier 0.078 to 0.076). FAA showed improved prognostic performance across the two datasets and under different phenotype definitions (reading center-defined vs. specialist-defined grading), supporting its potential utility for AMD risk stratification. These findings highlight the advantage of incorporating FAA into AMD risk prediction frameworks, supporting further evaluation of FAA for AMD risk stratification.
[CV-27] Report of the 8th LSVOS Challenge: Mobile Unified Multimodal Understanding ECCV2026
链接: https://arxiv.org/abs/2611.16925
作者: Chang Liu,Henghui Ding,Tong Yang,Dayoung Kil,Seong-heum Kim,Qilin Xiang,Ronghao Zheng,Ziyang Zhang,Shanlin Jiang,Ruohan Shi,Qilin Fan,Chengfeng Qiu,Kaifeng Wei
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 13 pages, 3 figures, 3 tables. Report of the MUMU track of the 8th LSVOS Challenge, held in conjunction with ECCV 2026. Project page: this https URL
Abstract:This report summarizes the Mobile Unified Multimodal Understanding (MUMU) track of the 8th Large-scale Video Object Segmentation (LSVOS) Challenge, held in conjunction with ECCV 2026. MUMU asks for a single unified model that jointly performs multi-concept image tagging over quality, scene, and event vocabularies, open-vocabulary object detection, and concise image captioning, within a budget of 0.5B parameters and 8 GB of peak inference memory. Submissions are ranked by a weighted task score scaled by an efficiency factor. We describe the tasks, rules, and evaluation protocol, report the final results, and review the methods of the top three teams. All three leading solutions build on the compact Florence-2-base model, but extend it in different ways: lightweight tagging heads with cross-task inference refinement, a shared low-rank adapter trained on coverage-aware multi-source data with task-dependent decoding, and a training-free pipeline that maps captions to tags and applies detection test-time augmentation. The results show that a compact prompt-based vision-language model can serve all three outputs when supervision, decoding, and cross-task cues are designed for each task, while open-vocabulary detection on novel categories remains the main bottleneck.
[CV-28] Automated PD-L1 Tumor Proportion Scoring with Uncertainty-Guided Triage in Non-Small Cell Lung Cancer
链接: https://arxiv.org/abs/2611.16907
作者: Kesi Xu,Ayesha Azam,Tanvier Kaur,Monika Kondratowicz,Kinza Asim,Brinder Chohan,Paapa Appiah-Odame,Zhao Chen,David Snead,Nasir Rajpoot
类目: Computer Vision and Pattern Recognition (cs.CV); Image and Video Processing (eess.IV)
备注: 55 pages (29-page main text with 5 figures and 3 tables, followed by 26 pages of supplementary material)
Abstract:Programmed death-ligand 1 (PD-L1) tumor proportion score (TPS) informs immunotherapy selection in non-small cell lung cancer. Near the 1% and 50% cutoffs, the score should also indicate when closer review is warranted. We designed a three-stage framework to follow the sequence of TPS assessment. Stage 1 identifies tumor regions and excludes macrophage-rich patches. Stage 2, a dual-branch cell counting network (D2Count-Net), counts PD-L1-positive and PD-L1-negative tumor cells in these regions to calculate TPS. Stage 3, TPS-Guard, an uncertainty-guided triage mechanism, reports model uncertainty as the disagreement among independently trained models, refers the most uncertain slides, and flags scores near the 1% or 50% cutoff when model estimates disagree across that cutoff. Flags state the reason for referral or alert the pathologist to possible scoring errors on auto-reported slides. In a retrospective, single-center study, 154 whole-slide images were scored by three pathologists, and the model and each pathologist were compared in turn with the mean of the other two. On the 114 slides unseen in training, model mean absolute error (MAE) was 6.96 percentage points, within the range of the three pathologists (4.24 to 8.29), and model uncertainty correlated with absolute TPS error (Spearman rho = 0.77) and inter-pathologist disagreement (rho = 0.71). After referral of the 15% most uncertain slides, model and pathologist MAE both decreased. Applied without retraining to an external, publicly available dataset, TPS-Guard still directed review to higher-error cases, and referring more uncertain cases raised sensitivity at the 50% cutoff. TPS-Guard provides reasons for reviewing individual automated scores and identifies slides that also produce greater disagreement among pathologists. Prospective reader studies should assess TPS-Guard’s effect on TPS reporting.
[CV-29] okenStream: Multi-Subject Personalized Video Generation with Token-wise LoRA and Guided Attention WACV2027
链接: https://arxiv.org/abs/2611.16897
作者: Hung Nguyen,Quang Qui-Vinh Nguyen,Khoi Nguyen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to WACV 2027
Abstract:We introduce TokenStream, an optimization-based framework that composes independently learned subject identities in text-to-video generation without a reference-video training dataset or joint optimization of each new subject combination. Learning from a single reference image makes subject appearance difficult to separate from static-image bias, while merging independently trained adapters can mix identities. Token-wise LoRA (ToRA) restricts low-rank updates to the cross-attention value projections of subject-associated text tokens, providing a separate adaptation path for each subject. A T2V-guided stream copies queries and keys from the frozen base model during early denoising to reuse its layout and motion priors while retaining personalized values. We also introduce Multi-DB, a 30-subject, 122-prompt benchmark with appearance-free prompts and varied interactions. On Multi-DB at 1.3B, TokenStream improves identity preservation from 0.422 for DreamBooth to 0.467 and textual alignment from 0.652 to 0.688.
[CV-30] Raysemble: Metric 3D Bone Reconstruction from a Single X-ray
链接: https://arxiv.org/abs/2611.16869
作者: Xiaoyang Li,Xinghao Yan,Yuan Chai
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Raysemble reconstructs eight lumbar and pelvic bones from a single X-ray by separating shape and placement. Each 3D query samples regional and global features at its image projection. A shared decoder predicts normalized bone shapes. Predicted depth and camera geometry determine 3D centers and scales for resampling these shapes onto a common grid in millimeters. Evaluation uses 23 held-out patients with 43 CT scans, each generating one simulated anteroposterior radiograph. Without rigid alignment to CT,Raysemble leads 15 evaluated methods, averaging 65.24% Dice, 9.09 mm HD95, and 3.32 mm ASSD over three random seeds. With identical inputs, it improves Dice over VolumeNeRF by 5.86 percentage points.
[CV-31] Visual Cipher Bench: Benchmarking LLM Agents on Long-Horizon Visual Reasoning
链接: https://arxiv.org/abs/2611.16861
作者: Deyao Hong,Jiayi Jiang,Yuchun Chen,Yixuan Shi,Wenyi Li,Zheng Jiang,Chengxing Xie,Mingyu Su,Qingle Liu,Mingju Gao,Yizhe Chi,Youjie Zheng,Calvin Xiao,Kaisen Yang,Huan-ang Gao
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:While language-model agents excel at long-horizon coding and terminal tasks, this success relies on two conditions: explicit text specifications and immediate execution feedback. However, neither condition holds in real-world problems where critical information exists only in images, leaving it unclear whether long-horizon capability transfers from text to visual reasoning. In such visual settings, because rules and clues resist text transcription and lack test-case feedback, agents must repeatedly inspect images, simulate spatial transformations, and autonomously cross-check clues. Despite this challenge, existing evaluations fail to capture such capability: visual reasoning benchmarks largely reduce tasks to single-turn, short-chain question answering, whereas agent benchmarks focus on text-based environments or interactive states that permit trial-and-error. To bridge this gap, we introduce Visual Cipher Bench, a collection of 40 original visual puzzles across four categories: spatial manipulation and simulation, logical constraints and planning, rule induction, and knowledge-based decoding. Each puzzle provides only one to three black-and-white line drawings and a minimal submission instruction, requiring agents to discover the underlying solving mechanism entirely from the images. Across 29 reasoning-effort configurations of eight frontier models, even the strongest model, GPT-6 Astra (max), achieves only 26.7% accuracy, while five of the eight models score below 1%. Yet on a ten-task subset, providing a single non-spoiler mechanism hint boosts accuracy from 7.8% to 55.6%. This sharp contrast shows that the primary bottleneck of current agents lies not in executing multi-step plans, but in autonomously discovering rules from visual evidence. Visual Cipher Bench thus provides a rigorous benchmark for measuring long-horizon visual reasoning in agents.
[CV-32] Kosmo-GFM: A Production-Ready Geometry Foundation Model
链接: https://arxiv.org/abs/2611.16840
作者: Sheng Tang,Shengzhe Wei,Tiancheng Chen,Wenhua Jin,Xinyang Lu,Tinghao Jiang,Shuoxuan Feng,Yupeng Liu,Yu Zhang,Xingshuo Chang,Keren Zhou,Yuhan Xiong,Jiahao Li,Haoming Liu,Jituo Tian,Hui Liang,Chaoqun Tang,Feng Tian,Wentao Ouyang,Kaojin Zhu,Xiaolin Chai,Peihuang Fu,Miao Zheng,Yunbo Hu,Feixue Cai,Zizhen Zou,Weigang Wang,Yexuan Yang,Haoyu Feng,Weiqi Zhang,Juntong Fang,Zesong Li
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project page: this https URL
Abstract:We present Kosmo-GFM, a scene reconstruction system that integrates a geometry foundation model (GFM), a pose optimization engine (PoseEngine), and reconstruction-derived supervision from KosmoWorld. GFM predicts camera parameters, depth, cross-view correspondences, and dynamic 3D points from images and videos. PoseEngine combines these predictions with geometric optimization to recover scene-level camera poses and geometry for downstream 3D and 4D Gaussian Splatting. KosmoWorld generates images and geometric supervision from reconstructed real-world scenes to improve GFM’s geometric predictions. On the Sparse and Medium subsets of SpatialBench, GFM achieves the lowest absolute relative depth error and the highest relative-pose AUC at 5 degrees among the compared methods. It also achieves the lowest mean endpoint errors for dynamic points and scene flow across the five evaluated datasets. Across 34 scenes completed by all compared reconstruction systems, PoseEngine achieves the highest scene-averaged pose AUC at thresholds of 1, 3, and 5 degrees, with a cumulative runtime of 14.07 hours, approximately 20% of GLUEMAP’s 69.67 hours. PoseEngine Flash provides a further accuracy-efficiency trade-off, with a measured mixed-precision runtime of 2.62 hours, approximately 4% of GLUEMAP’s runtime. Separate scaling experiments show generally favorable trends with increased data, model computation, and training budgets. These results support the use of learned geometric priors within a scene-level reconstruction pipeline.
[CV-33] EG360: Efficient and Generalizable 360circ Depth Estimation via Adaptive Panorama-Perspective Distillation and Spherical Cross-Level Feature Fusion
链接: https://arxiv.org/abs/2611.16793
作者: Laiyan Ding,Chongxi Meng,Rui Huang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Generalized panorama depth estimation has witnessed great advances in recent years. These methods typically achieve robust zero-shot generalization performance using large pretrained models, but their high computational cost makes real-time inference on embedded platforms challenging. In contrast, previous lightweight panorama models have mostly been developed for single-domain evaluations and cannot generalize well due to domain shifts. To reduce the gap between these two types of models, we introduce EG360, an efficient and generalizable panorama depth estimation model. Specifically, building on a lightweight depth model, we design an Adaptive Panorama-Perspective Distillation (APPD) strategy guided by scene geometry, along with a Spherical Cross-Level Feature Fusion (SCFF) module for global semantic perception. As a result, with only 6.593M parameters after reparameterization, EG360 achieves strong zero-shot performance on real-world benchmarks and runs at 35 FPS on a Jetson Orin NX 16GB in MAXN mode, whereas a previous ViT-small model runs at less than 2 FPS. Our code will also be made public for future research.
[CV-34] Evaluating Video LLM s Understanding of Corner Cases in Autonomous Driving BMVC2026
链接: https://arxiv.org/abs/2611.16767
作者: Gabriel Manalu,Minh Son Hoang,Hyeontaek Hwang,Jaesung Lee,Hyojin Choi,Tran Tuan Phong,Seon Kwon Kim,Saad Wazir,Daeyoung Kim
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Accepted at BMVC 2026
Abstract:Natural language explanations promised by modern Video Large Language Models (Video LLMs) offer autonomous driving systems a path toward detailed scene description and interpretable decision making on the road. While recent benchmarks have begun assessing these models on driving scenes, they mainly target routine traffic conditions and general comprehension tasks based on existing driving datasets and therefore fail to evaluate model performance on the rarest and most safety-critical driving events, also known as “corner cases”. To provide a comprehensive evaluation of Video LLMs on precisely these scenarios, we introduce DashBench, a benchmark testing 25 Video LLMs on 1,709 carefully curated driving videos spanning not only collision-related corner cases but also rare non-collision ones. Through extensive prompts and open-ended question answering, we expose a critical capability gap: all evaluated models perform substantially below the threshold required for safety-critical driving applications, frequently misjudging or missing dangerous events on the road ahead and hallucinating non-existent agents or actions in their explanations. We believe DashBench, combining an extensive dataset with a rigorous evaluation toolkit and a detailed diagnosis of current model limitations, drives progress toward safer, more explainable autonomous driving systems. The code and data will be shared at: this http URL
[CV-35] Efficient Periocular Recognition through Model Scaling and Cross Spectral VIS NIR Verification
链接: https://arxiv.org/abs/2611.16764
作者: Fabian Stockhard,Juan E. Tapia,Christoph Busch
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Munuscript In revision Process, abstract limited to 1920 ch
Abstract:Periocular recognition offers an efficient alternative to full-face verification when facial regions are occluded or image resolution is limited. We present AdaEye, a systematic study of input and backbone scaling for periocular verification in visible-spectrum (VIS) and cross-spectral (VIS/NIR) settings. Using an AdaFace-style framework, we compare full-face, masked-face, and dedicated 80x40 periocular inputs while varying backbone width and embedding dimensionality. Across four VIS/VIS and VIS/NIR protocols, we assess verification accuracy, computational cost, inference throughput, training memory, template size, and training stability. Direct periocular processing reduces recognition cost from 6.30 to 1.64 GMACs per image and increases measured throughput by approximately threefold. Its effectiveness depends on the evaluation protocol: periocular models perform better on native UFPR periocular images, whereas face-based protocols favor retaining more input information. Moderate backbone width reductions maintain comparable mean verification performance on CASIA and ND-NIVL while reducing computation. Embedding compression decreases template-storage requirements, primarily benefiting deployment. Multi-seed experiments reveal that some configurations require numerical stabilization, particularly when embedding and backbone reductions are combined. These findings support selecting compact periocular architectures according to the target protocol and deployment constraints, balancing accuracy, efficiency, storage, and robustness.
[CV-36] AgriSense: Vision-Language Decision Support and Robotic Intervention in a Virtual Greenhouse Copy title
链接: https://arxiv.org/abs/2611.16752
作者: Lyes Saad Saoud,Oualid Doukhi,Reza Ghorbani,Moussa Ayyash
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Autonomous greenhouse management increasingly requires robotic systems that integrate perception, navigation, environmental regulation, and plant-level intervention. This paper presents AgriSense, a framework combining an Isaac Sim virtual greenhouse, a quadruped mobile manipulator, RGB and LiDAR-inertial sensing, environmental control, and supervised vision-language decision support. Frozen DINOv2 and BERT encoders, together with structured environmental inputs, support joint prediction of twelve environmental references, a target zone, and an intervention flag. Checkpoint replay reproduces 41/43 zone decisions. Across five seeds on a source-grouped test set, the full, image-removed, and language-removed configurations each achieve 43/44 correct zone decisions, providing no evidence of incremental multimodal zone-accuracy improvement. The full model achieves mean absolute errors of 0.249 degrees C for temperature, 2.030 percentage points for humidity, and 0.997 percentage points for soil moisture. Sensor-only inputs match full-model zone accuracy, while language-only inputs achieve 68.64%. External real-image benchmarks reveal distribution-shift limitations. Five offline Point-LIO routes yield rigidly aligned trajectory errors of 5.22-17.32 mm. Of 40 virtual-arm targets, 32 satisfy geometric acceptance criteria and eight are rejected at boundaries. Six long-horizon climate cases and fault protocols characterize integration limits. Navigation and integrated treatment rely on simulator-state inputs; physical greenhouse performance and agronomic treatment efficacy remain unvalidated. The contribution is an evaluated decision-support and robotic-intervention framework for a virtual greenhouse.
[CV-37] ORCA:Orthogonal Residual Consensus Alignment for Multi-View Clustering
链接: https://arxiv.org/abs/2611.16749
作者: Xiaojian Ding,Xian Li,Xiaoying Zhu,Lin Zhao,Kaixiang Wang
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 24 pages,5 figures,2 tables
Abstract:Deep multi-view clustering often suffers from redundant information accumulation during fusion and indiscriminate alignment across representation levels. We propose Orthogonal Residual Consensus Alignment (ORCA) to address these challenges. Its semantic incremental recurrent fusion module extracts cross-view orthogonal residuals as non-redundant increments and adaptively weights recurrent states through divergence-guided gating. Its consensus-calibrated hierarchical contrastive alignment module uses the global consensus to weight contrastive alignment of view-level, pairwise, and recurrent representations. Together, these modules balance view-specific structure with cross-view consistency. Experiments on five benchmarks against fifteen competing methods demonstrate ORCA’s effectiveness, including a 5.25 percent accuracy improvement over the strongest baseline on 100leaves. Ablation studies validate both modules. Code is available at this https URL.
[CV-38] PhysLoop: Harnessing LLM s Physics Awareness for Self-Improving Video Generation Models
链接: https://arxiv.org/abs/2611.16748
作者: Yuxuan Chen,Xiaojie Li,Hao Chen,Juncheng Ma,Kewei Zhang,Zhongjing Du,Guoqiang Wei,Fangfang Wang,Zhen Dong,Zhiyang Dou,Liming Zhao,Yan Zeng,Jianwei Zhang,Daquan Zhou
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Video generation models produce visually realistic videos but still struggle with physical dynamics and causal consistency. Learning from generated videos requires fine-grained physical supervision, yet reliable quality judgments, correction pairs, and controlled variations are difficult to obtain. Large language models (LLMs) possess physical knowledge, but struggle to reliably guide video generation models in producing their training videos. We present PhysLoop, a unified harness designed to support physically grounded self-improvement in video generation models. The harness guides and constrains LLMs throughout video generation, selection, and correction, distilling their physical knowledge into physically consistent videos. The framework supports controllable training-data construction and inference-time video repair. By identifying and correcting failures, PhysLoop marks a key milestone toward recursive self-improvement (RSI) in video generation models with limited human intervention. We release 100K videos spanning diverse, realistic indoor and outdoor scenes, along with generation tools and two evaluation datasets: PhysLoop-Gen and PhysLoop-Repair. We evaluate controllable generation and localized repair on these datasets, respectively, and assess physical behavior on the external Physics-IQ Verified benchmark.
[CV-39] Fog-FWL: A Large-Scale Full-Waveform LiDAR Dataset for Recovering Fog-Hidden Objects
链接: https://arxiv.org/abs/2611.16742
作者: Kazuma Ikeda,Kotaro Oishi,Ryosei Hara,Ryo Yoshida,Takahiro Kado,Ibuki Fujioka,Taro Beppu,Mariko Isogawa,Kentaro Yoshioka
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 6 pages, 1 figure, 3 tables
Abstract:LiDAR is a core sensor for autonomous driving, yet it is highly vulnerable to fog. Fog scattering produces strong backscatter near the sensor and attenuates reflections from objects behind it, so weak object returns are often discarded by the peak detection that converts received signals into point clouds. Full-waveform LiDAR (FWL), which records the received optical signal before peak detection, retains these weak returns, but the lack of suitable data has prevented research on recovering fog-hidden objects from FWL. We present Fog-FWL, a dataset foundation for this task that consists of a real-world benchmark, Fog-FWL-Real, and a large-scale synthetic training set, Fog-FWL-Sim. Fog-FWL-Real contains 7,527 real foggy FWL frames from 8 scenes and 58 sequences, which is 579x larger than the only existing real foggy FWL data, together with histogram-level Object/Fog/Noise labels, 10,719 3D bounding boxes, and three difficulty levels defined by the attenuation of object reflections. Fog-FWL-Sim contains 12,000 frames generated by our physics-based simulator, Fog-FWL-Simulator, which integrates spatially heterogeneous fog density fields, time-resolved light transport including multiple scattering, and a photon-counting SPAD receiver model. The two subsets share the same sensor format and annotation protocol, enabling a strict sim-to-real evaluation in which no real waveforms are used for training. Our analysis shows that, under the hardest condition, object returns are included in the three strongest peaks for only 34.9% of object rays, underscoring the need for waveform-level recovery. Fog-FWL-Real, Fog-FWL-Sim, and the Fog-FWL-Simulator will be publicly released to facilitate research on fog-robust LiDAR perception.
[CV-40] From Foundation VLM to Versatile Driving VLM: A Triplet Incremental Training Approach
链接: https://arxiv.org/abs/2611.16727
作者: Shengtao Wu,Zichao Dong,Weixuan Qiao,Xiaolu Liu,Xin Zhan,Junbo Chen,Jianke Zhu
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 13 pages, 3 figures, 7 tables. Accepted at the 10th Conference on Robot Learning (CoRL 2026)
Abstract:Adapting general Vision-Language Models (VLMs) to Autonomous Driving (AD) is thought to require heavy 3D spatial priors or intrusive architectural overhauls to compensate for a lack of spatial grounding. In this work, we present a counterintuitive pilot study: off-the-shelf general VLMs already exhibit remarkable adaptability across diverse driving tasks, while precise metric depth remains the primary missing link, a bottleneck readily addressable via onboard sensors and depth foundation models. To this end, we introduce the Triplet Incremental Training Framework, a non-intrusive paradigm that seamlessly instills 3D geometric awareness into a general VLM backbone while fully preserving its open-world capabilities. Driven by automatically constructed (image, depth, QA pair) triplets, our framework operates via a unified tri-stream training pipeline: an unfrozen primary VLM stream to adapt to driving tasks, an auxiliary depth stream that guides the visual backbone to implicitly imagine 3D scene geometry via explicit geometric constraints, and a frozen twin VLM stream employing logit-level KL distillation to prevent semantic drift. Crucially, the auxiliary depth branch is discarded during deployment, ensuring the enhanced geometric awareness comes with zero inference overhead. We evaluate our framework across four major benchmarks targeting four diverse capabilities spanning 3D spatial perception, driving-oriented language reasoning, end-to-end planning, and general multimodal understanding. Our approach outperforms state-of-the-art driving VLMs, rivals domain-specific expert models in 3D perception, and successfully retains its original general reasoning capabilities. This demonstrates that Triplet Incremental Training offers a unified, deployment-ready paradigm for transforming general foundation models into robust driving intelligences.
[CV-41] What Limits Thermal Infrared Drone Detection? A Cross-Dataset Benchmark Analysis WACV
链接: https://arxiv.org/abs/2611.16724
作者: Tamara R. Lenhard,Jan Kels,Tobias Koch
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at 2027 IEEE/CVF Winter Conference on Applications of Computer Vision Workshops (WACVW), Workshop “Real World Surveillance: Applications and Challenges, 7th”
Abstract:Thermal infrared (IR) imaging enables drone detection with reduced dependence on ambient illumination, supporting continuous surveillance beyond the limitations of visible-spectrum sensing. However, substantial and often undocumented variability in camera systems, processing pipelines, and annotation practices across public IR drone datasets complicates the assessment of detector performance and cross-domain generalization. To examine these sources of variability and their implications for drone detection, we systematically characterize existing IR datasets and establish an artifact taxonomy reflecting their origins across the acquisition and processing pipeline, encompassing optical and sensor effects, tone mapping and image enhancement, and recording-related artifacts. Focusing on long-wave IR data, we investigate the impact of training data composition and model characteristics on YOLO-based detection performance, generalization, and condition-dependent failure modes. Our results demonstrate that detection performance is governed more strongly by training data composition than by YOLO generation or model capacity, while cross-dataset generalization exhibits pronounced asymmetries. Even the best-performing configuration fails to detect over 40% of drones, with elevated miss rates for very small targets, cluttered urban backgrounds, and thermal crossover, highlighting the shortcomings of existing datasets.
[CV-42] Sprint3D: Accelerating End-to-End 3D Asset Generation through Pipeline Optimization
链接: https://arxiv.org/abs/2611.16701
作者: Sho Kuno,Ruihan Yu,Yonghao Yu,Guixu Lin,Lian Fu,Muyao Niu,Yu-Ju Tsai,Zheng-Hui Huang,Yi-Chuan Huang,Siqi Yang,Fengbo Lan,Erwin Wu,Kaipeng Zhang,Zhixiang Wang
类目: Computer Vision and Pattern Recognition (cs.CV); Graphics (cs.GR)
备注: Project page: this https URL Code: this https URL
Abstract:Image-to-3D generative models now produce asset-quality textured meshes, but at minutes per asset they remain slow to iterate with and costly to deploy. Prior work has mostly accelerated the sampling of the networks, and where it also optimizes the mesh processing, it trains its own generators. We instead start from a publicly released pipeline and ask how much of its end-to-end time can be saved. Profiling the public Pixal3D pipeline shows that its latency is dominated by what happens after the networks: the four rectified flows take 14% of the wall-clock time and mesh post-processing takes 85%, most of it in CPU-based UV unwrapping. We therefore first rebuild the mesh post-processing on the GPU and take it off the critical path, then cut the per-evaluation overhead of the flows and distill them from 76 to 8 network evaluations. The resulting pipeline, Sprint3D, reduces the median time per asset from 133 s to 4.05 s on one H200 GPU at the same export settings (33 \times ). At the same time, on 2,653 held-out Toys4K objects it improves every input-view normal metric of the Pixal3D protocol, with a slightly lower mean silhouette IoU, lower point-cloud alignment scores and less fine texture detail. The rebuilt post-processing alone also speeds up TRELLIS.2, which uses the same export code, by 6.7 \times without a detectable change in the evaluated metrics.
[CV-43] MuViRL: Multimodal View-Invariant Visual Speech Representation Learning via Calibrated View Aggregation
链接: https://arxiv.org/abs/2611.16655
作者: Fei Su,Kai Qiu,Juan Liu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 30 pages, 7 figures
Abstract:Multimodal representation learning combines complementary signals to capture shared semantic information. For dynamic scenes, these representations need to remain consistent across views while preserving distinctions between temporal states. In visual speech representation learning, view variation changes the visibility and reliability of articulation cues, making it difficult to learn such representations from synchronized audio and video. To address this challenge, we propose MuViRL, a self-supervised framework for multimodal view-invariant visual speech representation learning via calibrated view aggregation. The framework combines synchronized 3D Gaussian view synthesis with multimodal self-supervision to learn consistent representations from real and synthesized data. A multimodal decision model learns relative view utility from teacher prediction losses, and its calibrated scores guide the aggregation of complementary information. Aggregation-to-Single Distillation transfers the aggregated predictions to a single branch. Real-synthetic alignment matches synchronized windows across views while preserving their temporal relations. Experiments on real and synthesized multiview data demonstrate that MuViRL improves multimodal representation consistency and visual speech recognition performance across views.
[CV-44] Learning the Missing Derivative: Neural Multiple-Scattering Operators for Inverse Volume Rendering
链接: https://arxiv.org/abs/2611.16650
作者: Sipeng Chen,Shibo Li
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 21 pages, 11 figures, 18 tables
Abstract:Reconstructing optically thick media from photographs requires modeling multiple scattering and its response to changes in extinction. Frozen-source tomography discards this derivative; differentiable path tracing estimates it through expensive, noisy sampling. We instead predict the multiple-scattering field with a neural operator and differentiate through it during inverse rendering, while computing transmittance and single scattering directly. A further challenge arises during optimization: an operator that predicts the field accurately on held-out media can still produce poor derivatives at the density states visited by the optimizer. We address this by fine-tuning on states generated during inversions driven by the operator, using Monte Carlo scattering-field supervision. On held-out optically thick clouds, our method improves relighting over frozen-source tomography at equal time and outperforms differentiable path tracing given substantially more time. Training on inversion trajectories also improves derivative alignment at visited states and reduces optimization drift.
[CV-45] Dataset Distillation via Drifting NEURIPS2026
链接: https://arxiv.org/abs/2611.16642
作者: George Cazenavette,Angelina Quan,Giannis Daras,Antonio Torralba,Vincent Sitzmann
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at NeurIPS 2026. Project Page: this https URL
Abstract:The task of dataset distillation concerns the learning of a small set of synthetic samples that can serve as an effective replacement for a large real dataset when used for downstream tasks. While many solutions have been proposed since the problem’s inception, most of them primarily seek to optimize the distilled data with respect to a classifier. In this work, we propose a new distillation method that learns a set of samples such that they directly approximate the per-class data distribution. Specifically, we “drift” our distilled data in the latent space of a feature extractor such that they move both toward the real data and away from each other without using a classification objective. We find that our drifted data cover the real distribution significantly better than existing methods while still serving as powerful training samples for classification. Furthermore, we showcase our method’s use as a visualization tool for better interpreting models and extend it beyond images to other modalities. Our work advances the frontier of dataset distillation, offering a novel synthesis method and proposing distribution-based metrics as a complementary evaluation. More broadly, our findings help address the growing need to understand both data and models by uncovering more about how they interact.
[CV-46] Asymmetric Multimodal Learning with Decomposed JEPAs
链接: https://arxiv.org/abs/2611.16639
作者: Yusuf Shihata
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注: 15 pages, 12 tables, 3 figures
Abstract:Cross-modal alignment is the standard objective for building multimodal representations. We show that alignment as a \emphsole objective is lossy wherever modalities are informationally asymmetric—where neither can express what the other carries. Visual question answering is an obvious case: a question is a selective relational operator probing a dense visual field whose fine-grained features no query recovers, so what alignment preserves is not what the task requires. We prove an \textbfAlignment Redundancy Bound: alignment-only Dirichlet energy minimization retains at most the redundant information atom \Red(Z; X_1, X_2) , annihilating private modality-specific structure \Unq(Z; X_m) and cross-modal synergy \Syn(Z; X_1, X_2) . We introduce \textbfDecomposed JEPA (D-JEPA), which factorizes latents into an aligned shared core and private modality-specific fibers. The shared core retains most of the structure alignment targets; the private fibers recover what alignment discards, via contextual masked prediction and asymmetric cross-subspace independence. Under information asymmetry, alignment-only training undergoes catastrophic semantic collapse on compositional reasoning benchmarks, surviving only on binary queries. D-JEPA averts this collapse, substantially improving downstream accuracy over both alignment-only and alignment-with-reconstruction baselines, and restoring a positive multimodal synergy gap in place of the baseline’s inverted gap. Information-theoretic audits confirm that D-JEPA recovers most of the mutual information available from a frozen backbone, and causal fiber-swap interventions demonstrate that the recovered information is causally localized in the private fibers. Comments: 15 pages, 12 tables, 3 figures Subjects: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV) Cite as: arXiv:2611.16639 [cs.LG] (or arXiv:2611.16639v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2611.16639 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[CV-47] When Chekhovs Gun Fires Anyway: Over-Realization in Text-to-Image Models
链接: https://arxiv.org/abs/2611.16630
作者: Jiahao Huang,Kon Woo Kim,Song-Duo Ma,Yuka Kitamura,Felix Wagner,Yunshan Huang,Sunisth Kumar,Xanh Ho,Yun-Ang Wu,Zhijie Chen,Yunjia Xu,Junfeng Jiang,Zefan Yu,Fei Cheng,Akiko Aizawa
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:In text-to-image generation, a target that the prompt mentions often appears in the generated image in a way the prompt does not call for. We call this failure over-realization. Mental-spaces theory points to the situations where it can happen: under a negation, inside what a character believes or imagines, as the vehicle of a figurative comparison, or behind a viewpoint that hides the target. Prior work has studied only negation in depth, and the conventional alignment and preference metrics do not penalize the failure. We build OverReal to measure it: OverReal-Gen, 698 prompts covering the four situations, on which we measure each generator’s over-realization rate; and OverReal-Det, 2,841 generated images human-labeled for whether and how the target is over-realized, on which we verify the automatic judge that produces this rate. Our evaluation of eight text-to-image systems reveals that all of them over-realize, and that conventional metrics such as CLIPScore reward the failure rather than penalize it. Prompt expansion only partly mitigates the failure. Inside the generator, the diffusion backbone attends to a mentioned target almost as much as to a requested one. The cue words that should hold the target back receive less attention than ordinary words.
[CV-48] Safety as a Budget: Calibrating the Control Channel of Constrained GRPO for Vision-Language Models
链接: https://arxiv.org/abs/2611.16628
作者: Bo Han
类目: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注: 10 pages, 5 figures, 5 tables
Abstract:Reinforcement learning can teach vision-language models (VLMs) to decline harmful image-text requests, but how safe the result becomes is set indirectly, through reward weights found by trial runs. Stating safety as budgets on the rate of harmful assistance and the rate of over-refusal, enforced by Lagrange multipliers inside GRPO with an off-the-shelf scorer as the cost, is the natural fix, and it fails twice. The multiplier compares the scorer’s mean with the budget, and scorers built for ranking get the mean wrong: a fixed-point analysis shows that the mean bias, not the budget, then decides whether the multiplier settles, saturates or switches off. And standard GRPO standardizes the penalized reward within each group, which cancels the multiplier before it reaches the gradient. Two-channel constrained GRPO repairs both: the base reward keeps the raw score for the within-group comparison, the penalty and the constraint estimate pass through a monotone map fitted on judge-labeled training responses, and the advantage keeps the multiplier. On a Qwen3-VL-2B policy with a training-free readout of an open 4B model as the scorer, held-out harm on 1,037 adversarial prompts falls from 8.9% under standard GRPO to 4.8%, lower in every seed, where the uncalibrated budget gives 7.6%; over-refusal stays at most 2% on 400 benign queries and MMBench is unchanged. The gain is largest where the intent sits in the image (10.1% - 5.1%) and grows on the unseen HADES benchmark (28.9% - 12.8%). Order-preserving distortions of the scores reproduce each predicted controller state in every seed, and the results hold under a map fitted on another judge’s labels, an image-reading judge and a human audit.
[CV-49] Improved CNN-Transformer Network for Neutron Image Super-Resolution
链接: https://arxiv.org/abs/2611.16620
作者: Yuheng Chu,Hongli Li,Tian Zhang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 11 pages, 5 figures, 3 tables
Abstract:The spatial resolution of neutron images is of great importance for material defect diagnosis, structural characterization, and nondestructive testing. However, owing to limitations in neutron beam quality, scintillator conversion efficiency and light scattering, detector pixel size, and noise levels, conventional neutron images often suffer from speckle noise, blurred edges, and missing details, making them inadequate for high-precision analysis. To address these issues, an improved super-resolution model combining convolutional neural networks (CNNs) and Transformers is proposed. First, to overcome the difficulty of obtaining high-quality and clear training images for neutron imaging, a pretrained image restoration model is fine-tuned using real neutron images and then employed to enhance existing neutron images, thereby providing high-quality reference images for subsequent super-resolution model training. Second, an improved CRAFT-based super-resolution network is developed by incorporating a hybrid prior-guided channel fusion module (HPCF), a mean-standard deviation dual-statistics squeeze-and-excitation block (MS-SEBlock), and a partial convolution feed-forward network (PCFN), with the aim of improving the reconstruction of important information, such as textures and fine details, in neutron images. Simulation results demonstrate that, compared with bicubic interpolation, SRCNN, SRGAN-VGG, and other methods, the proposed network achieves superior performance in both full-reference metrics, including PSNR and SSIM, and no-reference metrics, including AGM and SMD. The visual quality and structural fidelity of the reconstructed images are also significantly improved. This study provides a new approach to improving neutron image quality.
[CV-50] owards Democratising Low-Data Medical Image Classification: CPU-Trainable ZACH-ViT and Its Geometric Variants
链接: https://arxiv.org/abs/2611.16608
作者: Athanasios Angelakis
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 17 pages, 4 figures
Abstract:Medical image classification in resource-constrained settings must contend with limited annotations and computational capacity. We evaluate four ZACH-ViT variants with linear, Euclidean prototype, Poincare and spherical heads alongside two compact multiple-instance-learning baselines. All models are trained from scratch with fewer than 260,000 parameters across seven MedMNIST+ datasets and three external benchmarks (PCam, ISIC2017 and Shenzhen tuberculosis radiographs). The low-data protocol uses 50 training images per class and five seeds, with validation-selected checkpoints at 23 and 47 epochs; a full-data reference uses one seed. Compact-TransMIL achieves the highest ten-dataset mean test Macro-F1 at 23 epochs (53.34%) and with full data (73.93%). At 47 epochs, ZACH-Linear narrowly leads (57.76% versus 57.54% for Compact-TransMIL). Extending training improves Macro-F1 in 239 of 300 paired comparisons. On ISIC2017, hZACH-Spherical has the highest 47-epoch low-data Macro-F1 (44.09% with SD 4.09%), whereas ZACH-EuclideanProto leads with full data (51.27%). We additionally conduct 40 CPU-only training runs on ISIC2017 and Shenzhen using four CPU threads; model-only single-image inference requires approximately 4-9 ms, while image decoding and preprocessing substantially increase end-to-end latency. The results demonstrate the feasibility of CPU-only training on the tested platform and the task- and regime-dependence of compact medical imaging under strict resource constraints.
[CV-51] ReFineR: Single image to dense point cloud registration at pixel level with point splatting
链接: https://arxiv.org/abs/2611.16583
作者: Ugo Insalaco,Jean-Emmanuel Deschaud,Pierre Dellenbach
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Cameras and LiDARs (Light Detection and Ranging) provide complementary observations of the same scene: cameras capture rich appearance cues while LiDAR sensors provide accurate geometric measurements. In this work, we address the problem of fine image-to-point cloud registration from a single image and a roughly estimated camera pose. On robotic platforms, residual reprojection errors of several pixels are commonly observed even after calibration due to temporal desynchronization, mechanical vibrations, or inaccuracies in point cloud reconstruction. Such errors can lead to decimeter-level geometric discrepancies at moderate distances and significantly affect multiple downstream tasks, such as dense cross-modal supervision, texture mapping, and semantic label transfer. To address this challenge, we introduce ReFineR: an offline framework to finely register individual images to pre-reconstructed point clouds. ReFineR builds upon differentiable point splatting rasterization and a smooth Gradient Orientation Measure (GOM) to iteratively refine the camera pose through analysis-by-synthesis optimization. We evaluated the method on a curated dataset of manually calibrated images collected from heterogeneous data sources. Furthermore, we show that an automatic point cloud rescale factor can be estimated prior to optimization, enabling robust refinement across a wide range of scenarios using a single scale-invariant formulation. The project page can be found at this https URL
[CV-52] ActWorld-JEPA: Action-Conditioned JEPA World Models Meet Latent Planning for End-to-End Autonomous Driving
链接: https://arxiv.org/abs/2611.16579
作者: Ziying Song,Shengkai Zhang,Caiyan Jia,Lei Yang,Lin Liu,Yuchen Liu,Ziyang Liu,Zirui Li,Gang Wang,Peiliang Wu,Li Wang,Yadan Luo,Chen Lv
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 19 pages, including appendix. Code: this https URL . Project: this https URL
Abstract:Joint-Embedding Predictive Architectures (JEPAs) enable future prediction in latent space without pixel reconstruction. However, latent predictive consistency alone does not establish the relative driving quality of candidate plans. We present ActWorld-JEPA, a joint-embedding formulation for candidate-based driving. Where Plan-Conditioned JEPA (PC-JEPA) uses global and step-wise trajectory intentions to guide recurrent latent prediction, then combines the predicted future with the current scene through gated cross-attention; Relational Residual Reranking ( R^3 ) models agreement and discrepancy between plan and future representations under planning supervision, yielding bounded, gated residual corrections to the base planning scores. This formulation connects latent prediction to candidate selection using only observed context and proposed trajectories at inference. Benchmark-specific implementations report 94.1 PDMS on NAVSIM v1 navtest, with the final selector calibrated using aggregate NAVTEST feedback, and 91.6 and 42.7 EPDMS on NAVSIM v2 navtest and navhard, respectively. The nuScenes evaluation reports a mean L_2 error of 1.20 m and a predicted collision rate of 0.76% over 1–6 s. This work provides an effective path from JEPA’s predictive representations to driving decisions informed by anticipated consequences.
[CV-53] Pl@ntNet machine-to-machine identification service with Pl@ntNet API: enabling automated biodiversity monitoring through large-scale plant identification
链接: https://arxiv.org/abs/2611.16538
作者: Benjamin Bourel(IROKO),Jean-Christophe Lombardo(IROKO),Alexis Joly(IROKO),Mathias Chouet(UMR AMAP),Thomas Paillot(IROKO),Antoine Affouard(UMR AMAP),Vanessa Hequet(UMR AMAP),Herv{é} Go{ë}au(UMR AMAP),Pierre Bonnet(UMR AMAP)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:1. Monitoring plant biodiversity requires identifying large numbers of plants across space and time. Although artificial intelligence (AI) can now identify plant species from images with high accuracy, most existing systems rely on users manually submitting images. As ecological monitoring increasingly generates large image collections from cameras, drones, robots and online platforms, there is a growing need for plant identification services that can be directly integrated into automated data-processing workflows. 2. Here, we present Pl@ntNet’s machine-to-machine AI plant identification service. This service uses an application programming interface (API) to provide programmatic access to Pl@ntNet’s plant recognition models and the associated taxonomic processing. This service enables the implementation of automated identification workflows without human intervention and can thus directly link large image collections and automated observation systems to an AI-based plant identification system. 3. We establish that this service, which covers 85,000 species from around the world, (1) has a species prediction accuracy of 0.89, (2) process images up to 666 times faster than workflows requiring human intervention, (3) and has processed over 84 million identification requests since its launch, including more than 40 million in 2025 alone. We demonstrate how it has already enabled the implementation of a new generation of ecological workflows: identifying plants from images of insects to reconstruct the plant-pollinator interactions, analysing geolocated images from social media to study plant distribution, detecting invasive plants using images taken at the roadside, and identifying weeds using agricultural robots. Finally, we illustrate its application using a reproducible example: identifying plants from insect photos on iNaturalist. In approximately 30 min on a standard laptop, 2,424 images were processed to generate a dataset of 708 plant-pollinator interactions, with a weighted precision of 0.98 for flower detection and an accuracy of 0.88 for genus plant identification. 4. In conclusion, this machine-to-machine species identification is already operating on a considerable scale and meets a genuine need. All of this paves the way for a new paradigm in biodiversity monitoring, in which automated sensors and image collections can be directly connected to AI-based identification services.
[CV-54] Context-Predictive Low-Rank Adaptation for Face Presentation Attack Detection
链接: https://arxiv.org/abs/2611.16534
作者: Aminu Yusuf,Shakeel Muhammad Ibrahim,Sulaimon Oyeniyi Adebayo,Muhammad Sohail Ibrahim,Muhammad Shahid Jabbar,Shujaat Khan
类目: Computer Vision and Pattern Recognition (cs.CV); Image and Video Processing (eess.IV); Signal Processing (eess.SP); Machine Learning (stat.ML)
备注: This work has been submitted to the IEEE for possible publication. Copyright may be transferred without notice, after which this version may no longer be accessible
Abstract:Face presentation attack detection requires representations that distinguish physical spoofing cues across changes in acquisition devices, subjects, and attack media. Large visual encoders offer transferable representations, but conventional binary fine-tuning supervises their final decisions without explicitly constraining how local facial evidence is represented under partial observation. We present CP-JEPA, a context-predictive adaptation framework that combines low-rank tuning of a pretrained CLIP image encoder with a joint-embedding predictive objective. A student predicts complete-image teacher features at masked spatial locations, while unmasked classification, masked classification, and prediction consistency retain the connection to the live/spoof task. An exponential-moving-average teacher supplies evolving latent targets. The auxiliary components are used only during training; inference requires only the LoRA-adapted CLIP visual encoder and binary classifier. We evaluate CP-JEPA under single-source and multi-source OCIM transfer, additional CelebA-Spoof training, transfer to CelebA-Spoof, two-source training, and unseen-attack evaluation on SiW-Mv2. On the standard OCIM benchmark, CP-JEPA achieves 3.74% mean HTER and 98.73% mean AUC. Single-source evaluation achieves 11.36% mean HTER across twelve source–target combinations. With CelebA-Spoof added to the three OCIM training sources, CP-JEPA reduces mean HTER from 3.45% to 2.35% and increases mean AUC from 98.68% to 99.32%. On SiW-Mv2, it achieves 3.36% mean HTER across the held-out attack categories. Controlled loss ablations examine the contributions of latent prediction, masked classification, and prediction consistency. The results show that context-predictive adaptation can improve cross-dataset PAD performance, although the gains vary with the training sources and evaluation setting.
[CV-55] UniSpark: Exploring Hybrid Attention for Multimodal Understanding and Generation
链接: https://arxiv.org/abs/2611.16517
作者: Jiancheng Pan,Yuzheng Liu,Hengyuan Cao,Qiyuan Zhang,Yangyan Li,Biao Gong,Zheng Zhang,Ka Long Cheng,Xinyang Wang,Yujun Shen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Technical Report, 31 pages
Abstract:We present UniSpark-9B, a unified multimodal model that investigates hybrid attention for visual understanding and generation. Built on a Mixture-of-Transformer-Experts backbone with approximately 9B parameters in each branch, UniSpark combines full attention for explicit token interactions with GatedDeltaNet linear attention for compact context propagation. Autoregressive text prediction and visual flow matching jointly support visual question answering, text-to-image generation, instruction-guided editing, and interleaved text-image prediction. In interleaved generation, noisy visual targets are isolated from subsequent predictions, while clean visual representations provide context for later outputs. A blockwise protocol enforces this distinction through attention masks in full-attention layers and recurrent-state retention rules in linear-attention layers. The selected training sources contain 159.7M generation samples, 213.4M understanding samples, and 624.8K interleaved sequences. A four-stage curriculum jointly trains multimodal understanding and visual generation, refining answer accuracy and instruction following while extending image synthesis to 2K. UniSpark scores 2470.3 on MME-S and 0.83 on GenEval. Qualitative examples further illustrate its capabilities in interleaved image generation, video synthesis, and video editing. Together, these results demonstrate that hybrid attention provides an effective backbone for unified multimodal understanding and generation.
[CV-56] PiLLVD: A Real-World Low-Light Video Dataset for Dynamic Scenes Captured with Off-the-Shelf Cameras
链接: https://arxiv.org/abs/2611.16515
作者: Yaser Abu Awwad,Paul L. Rosin,Oktay Karakuş
类目: Computer Vision and Pattern Recognition (cs.CV); Image and Video Processing (eess.IV)
备注: 14 pages, 7 figures, 5 tables. Dataset available on Zenodo: this https URL
Abstract:Low-light image and video enhancement has received increasing attention in recent years. However, most existing enhancement methods are developed and evaluated on still images, while progress in low-light video enhancement remains limited by the lack of real-world paired video datasets. Acquiring paired low-light and normal-light video sequences is challenging because it requires spatially and temporally aligned captures while avoiding motion blur, exposure mismatch, and correspondence errors in dynamic scenes. To address this gap, we introduce PiLLVD, the Pi-based Low-Light Video Dataset, a paired low-light video dataset captured using an accessible Raspberry Pi-based multi-camera system with off-the-shelf cameras. PiLLVD provides paired low-light and aligned normal-light reference sequences for dynamic indoor and outdoor scenes, including diverse human and object motion under varying illumination conditions. Unlike existing datasets that are often image-based, static, synthetic, or limited to simple motion patterns, PiLLVD focuses on motion-rich scenes where frame-to-frame stability and motion-related artefacts are important for evaluation. In addition to presenting the dataset and acquisition pipeline, we benchmark representative pre-trained low-light video enhancement methods on PiLLVD under a zero-shot setting. The results show that existing methods do not generalise uniformly across the proposed sequences, highlighting the challenges posed by the proposed dataset and its value for both benchmarking existing methods and supporting the development of future low-light video enhancement algorithms.
[CV-57] From State of THE ART to Human Prediction Algorithms
链接: https://arxiv.org/abs/2611.16464
作者: Conghao Wong,Ziqian Zou
类目: Computers and Society (cs.CY); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Human curiosity and longing for the future have driven human progress from ancient times to the present. With the advancement of AI technology, human prediction algorithms have also drawn increasing attention from researchers. However, as AI technology continues to advance, we observe a counterintuitive phenomenon. Take human trajectory prediction algorithms as an example. Although models’ quantitative prediction performance has improved, yielding one state-of-the-art model after another, the consideration of human behaviors, future thoughts, and social interactions themselves presents a gradually diminishing trend in the construction of these prediction algorithms, buried beneath numbers and metrics. In this manuscript, taking interactive art and human trajectory prediction algorithms as examples, we attempt to adopt a different analytical perspective to analyze the structural and developmental commonalities between these two fields. By examining specific interactive artworks and prediction algorithms, we aim to explore how human intuitions and art are associated with AI algorithms through three aspects, including perception, interaction, and timing, thereby providing new insights for the understanding and further research in this domain. We also observe an interesting phenomenon that certain art-inspired prediction algorithms can actually deduce intriguing patterns of human behaviors from the data, offering new potential avenues for understanding the relationship between human activities, state of The Art, and AI algorithms.
[CV-58] Can an LVLM Perceive Event Boundaries Like a Human?
链接: https://arxiv.org/abs/2611.16452
作者: Hyolim Kang,Jungsoo Lee,Yeji Song,Whie Jung,Janghoon Cho,Juntae Lee
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 15 pages, 5 figures, 8 tables
Abstract:Large Vision-Language Models (LVLMs) reason effectively about video content, but whether they perceive its event boundaries in a manner consistent with humans remains unclear. We first investigate this through Generic Event Boundary Detection (GEBD), which identifies event boundaries naturally perceived by humans in continuous video, without relying on predefined event categories. On Kinetics-GEBD dataset, LVLMs miss many boundaries when predicting directly from frames, but effectively refine candidates proposed by an unsupervised discovery method, achieving state-of-the-art unsupervised GEBD performance. We extend this discovery-refinement approach to hour-long egocentric video, where boundary discovery requires a broader semantic scale and whole-video LVLM refinement becomes impractical. We address these challenges through coarse-to-fine temporal discovery and localized LVLM accept/reject judgments. Human evaluation shows that the LVLM’s agreement with group consensus falls within the range observed among human annotators, with disagreements concentrated on low-consensus cases. Using this pipeline, we construct EgoBoundary, a benchmark for event boundary detection in hour-scale egocentric video. Representative supervised and unsupervised methods degrade substantially on EgoBoundary, highlighting the challenge of long-horizon event boundary detection.
[CV-59] HIFC-IQA: Train-Free Cross-Domain Image Quality Assessment via Dual-Process Cognition NEURIPS2026
链接: https://arxiv.org/abs/2611.16451
作者: Yu Li,Zhengran Shen,Puchao Zhou,Yukang Ding,Yachun Mi,Boyuan Yang,Xi Zhai,Shaohui Liu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted by NeurIPS2026
Abstract:The gap between controlled synthetic distortions and complex real-world degradations introduces a critical domain shift challenge for No-Reference Image Quality Assessment. Although Unsupervised Domain Adaptation (UDA) approaches attempt to address this issue, they remain constrained by substantial retraining overhead and cumulative error propagation inherent in iterative pseudo-labeling. Inspired by the dual-process theory of human cognition for remarkable generalization ability, we propose Hierarchical Intuitive and Fuzzy Consensus (HIFC), a completely train-free Cross-Domain IQA framework. HIFC employs asymmetric perceptual decomposition to disentangle semantic content from distortion-sensitive representations. The final quality score is fused through the synergy of two complementary branches: a memory consensus branch that retrieves a stable Fuzzy Quality Centroid from a synthetic reference gallery, and an intuitive branch that captures aesthetic polarization via zero-shot vision-language alignment. Extensive experiments show that HIFC achieves state-of-the-art performance across multiple benchmarks, attaining the highest average SRCC and PLCC among competing UDA methods. Most notably, in the challenging synthetic-to-authentic cross-domain setting, HIFC achieves a substantial performance margin over competing methods, demonstrating remarkable robustness and generalization.
[CV-60] DesignTaste: Benchmarking AI-Generator Visual Design Against Award-Wining Works
链接: https://arxiv.org/abs/2611.16442
作者: Zhou Chen,JiaXiang Li,DanYing Pan,Ye Li,TianRui Yang,EnBao Zhang,YaNan Xiao,LiXin Sun,Anqi Liu,Dong Zhao,JianYuan Li,YuanYuan Ding,Kai Kong,Xinwen Xu,HuiTang Du,Min Zhou,Zhiqiang Li
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 31 pages, 12 figures, including appendices. Code and data: this https URL
Abstract:Text-to-image benchmarks commonly assess text-image alignment or overall preference. Whether models can produce good designs when a brief leaves the appearance open remains less well understood. We introduce DesignTaste, a visual design benchmark that compares generated images with award-winning works. From these works and their publicly available context, we derive 403 design briefs: 100 poster tasks from 100 Beste Plakate 2025 and 303 tasks spanning 20 categories from ADC 2026. Six leading commercial text-to-image models each generate an image from the same brief. A single model judge scores each output and the corresponding award-winning work independently on five evaluation dimensions. On each dataset, only ArtArk achieves a higher mean overall score than the award-winning works. All six models score lower in Creative Originality on both datasets. These results show that producing finished, visually appealing images does not imply that a model can develop good design solutions.
[CV-61] PhysCapBench: A Cross-Domain Benchmark for Physically Grounded Video Captioning
链接: https://arxiv.org/abs/2611.16424
作者: Haofeng Wang,Pengcheng Wen,Ruiqi Li,Yuhao Zhang,Tiange Zhang,Boren Zheng,Xuanyi Liu,Siyu Tao,Danni Yu,Gan Lin,Haocheng Tang,Tianzizhe Guo,Mingyuan Su,Boyuan Chen,Zhongyuan Wang
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Project page: this https URL
Abstract:Physical AI aims to learn transferable physical knowledge from real-world experience to understand, predict, and act in the physical world. Internet videos and embodied demonstrations provide complementary observations of motion, interaction, and state evolution. A foundational data challenge is to measure and preserve this physical information as training supervision. In video-language pretraining, captions express observed processes in language, yet a semantically correct description can omit the interactions and transitions that define those processes. Rich physical experience in video does not automatically translate into sufficient language supervision. This gap calls for a yardstick to assess annotation quality, compare captioning strategies, and guide training-data refinement. Motivated by our data construction for world-model pretraining, we introduce PhysCapBench, to our knowledge the first unified physical-caption benchmark spanning all three domains: third-person scenes, egocentric manipulation, and handheld-gripper demonstrations. It provides human-verified annotations for 1,500 video sequences and evaluates accuracy and coverage through seven physical dimensions. We examine downstream utility through video reconstruction and manipulation-policy fine-tuning. Physical captions reduce reconstruction LPIPS by 8.6-9.5% on third-person videos and 26.9-30.9% on egocentric videos relative to metadata captions. Training-instruction enrichment improves both PhysCapBench scores and policy success, including a 7.5-percentage-point gain for pi0.5 on a RoboTwin 2.0 subtask. PhysCapBench makes the physical content of language supervision inspectable and comparable, providing a cross-domain basis for its evaluation and improvement.
[CV-62] I See What You Want: Decoding Hierarchical Operator Intent from Visual Teleoperation History
链接: https://arxiv.org/abs/2611.16419
作者: Xiangyi Jia,Kaixian Qu,Shengzhi Wang,Chenyu Yang,Robert Katzschmann,Marco Hutter
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注: 8 pages, 6 figures. Project page: this https URL
Abstract:Vision-language models (VLMs) have become a powerful foundation for robotic scene understanding, task planning, and language-conditioned control. A natural question follows: can these models also decode what a human operator intends while teleoperating a robot, without any explicit instruction? We formulate this as hierarchical intent prediction, jointly inferring the ongoing atomic action and the broader mission it serves from a short visual teleoperation history. To study this, we introduce VI AMO (Vision Intention Action and Mission Ongoing), a dataset of 6,600 episodes from eleven robot sources with 13,653 temporally aligned action annotations and 9,702 mission annotations. Large proprietary models struggle with this problem: on held-out episodes, GPT-4o reaches Sentence-BERT similarities of only 0.563 and 0.470 for action and mission, and GPT-5.6 Luna achieves only 0.581 and 0.554. Fine-tuning a VLM on VI AMO yields our Vision-Intention Model (VIM), which raises these scores to 0.825 and 0.821 and still reaches 0.726 and 0.717 on the 35.4% of test actions with both labels absent from training. Finally, as a downstream application, VIM’s predicted mission enables a vision-language-action (VLA) policy to take over from the operator mid-task in a real-world environment unseen during VIM fine-tuning, reducing human teleoperation time by 34.6% relative to GPT-4o and 39.8% relative to GPT-5.6 Luna, while cutting intent-prediction latency by roughly 70% on average.
[CV-63] Model Distribution-Aware Multimodal Dataset Distillation NEURIPS2026
链接: https://arxiv.org/abs/2611.16418
作者: Jiajun Shen,Yonglin Wu,Ruonan Yu,Xinchao Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: NeurIPS 2026
Abstract:Multimodal dataset distillation (MDD) aims to reduce the cost of vision-language pre-training by synthesizing compact image-text datasets. Although recent feature-matching methods have improved distillation efficiency, synthetic data may become specialized to the model representations encountered during optimization, limiting generalization to unseen model states. We propose Model Distribution-Aware Multimodal Dataset Distillation (MDA-DD), which leverages supervision from a dynamic staggered queue of models at different training stages. We further introduce asymmetric dual projection for complementary alignment in visual and linguistic spaces and an uncentered second-order moment objective that retains mean-related cross-modal information. Experiments on Flickr30K and MS-COCO demonstrate consistent retrieval improvements over existing methods across synthetic-data budgets. With only 100 synthetic pairs, MDA-DD improves IR@1 and TR@1 on Flickr30K by 12.9 and 18.2 percentage points, respectively, over the strongest baseline at the same budget, outperforming all compared methods using 500 pairs on both datasets across all six retrieval metrics.
[CV-64] VBot Exo2Ego Policy: Lightweight Data Collection and Exocentric-to-Egocentric Video Generation
链接: https://arxiv.org/abs/2611.16417
作者: Tong Rao,Zhihao Zhang,Yuntian Wang,Siyu Zhou,Hailong Qin,Yinan Yu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 27 pages, 11 figures
Abstract:We present VBot Exo2Ego Policy, a video generation model that turns third-person videos into first-person observations of a selected actor, offering a practical way to expand first-person operation data from abundant real-world footage. To support this goal, we introduce a lightweight data collection pipeline and mobile app that support first-person recording, third-person recording, and synchronized capture of both views. This setup combines accessible single-view collection with paired RGB and camera geometry for learning cross-view correspondence. Building on a pretrained video diffusion transformer, our model incorporates geometric guidance and auxiliary supervision from semantic masks and hand keypoints to facilitate optimization and improve the fidelity of hand-object interactions. Camera geometry is injected stochastically during training and gradually removed, allowing inference without target-camera calibration. To accommodate the diversity of Internet videos, we combine source-video perception and reasoning, optional actor references, and multi-shot adaptation to identify the intended actor and follow an operation across camera cuts. The collection system and generation model provide a foundation for expanding first-person operation data across tasks, scenes, and participants.
[CV-65] RICapBench: A Fine-Grained Visual Understanding Benchmark Dataset for Vision-Language Models in Remote Sensing
链接: https://arxiv.org/abs/2611.16413
作者: Xue Wang,Yubin Xu,Kun Tan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 13 pages, 4 figures, 2 tables
Abstract:Advances in vision-language models (VLMs) have improved remote sensing image analysis and supported geospatial decision-making. However, fine-grained understanding of remote sensing imagery remains challenging for VLMs. Large variations in object scale, complex geospatial relations among instances, and limited visual detail hinder precise instance localization, visible attribute identification, and spatial relation understanding, potentially leading to descriptions that lack visual grounding. Existing datasets and benchmarks insufficiently assess end-to-end instance localization and captioning without target-specific textual cues. To support this evaluation, we introduce RICapBench, a dataset and benchmark that unifies instance localization and fine-grained captioning in remote sensing. RICapBench contains 245,466 instance-level annotations and 498,005 spatial relation tuples spanning general remote sensing, UAV remote sensing, and dense small-object scenarios. Each annotation pairs an object’s oriented bounding box with a natural language description of its visible attributes and spatial relations to surrounding instances and environmental elements. To evaluate the accuracy and completeness of spatial relation captioning, we propose a remote sensing spatial relation parser and the RelCap-F1 metric. We combine RelCap-F1 with instance localization and attribute captioning metrics into a composite metric, RICScore. Our evaluation of 15 advanced open-source and closed-source VLMs identifies instance localization as a major bottleneck for these models. LocateAnything achieves the highest RICScore of 31.68, but scores lower on spatial relation captioning than other task-fine-tuned open-source VLMs. Code and dataset are available at this https URL.
[CV-66] Gravity and Flight: Learning to Navigate Fine-Grained Remote Sensing Scenes
链接: https://arxiv.org/abs/2611.16408
作者: Yuheng Zong,Minghua Wang,Xin Zhao,Xiaobin Liu,Yongjian Dong,Chanho Eom
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Vision-language models are increasingly capable, yet their reasoning in complex remote sensing scenes can remain coarse-grained and insufficiently grounded in task-relevant visual evidence. We propose Gravity and Flight (GF), a framework for scene specialization with limited supervision that separates visual and reasoning verification while coupling policy updates. It combines Evidence-of-Thought supervised fine-tuning (EoT-SFT) and Relational Transfer Policy Optimization (RTPO). EoT-SFT teaches fine-grained observation and faithful reasoning through description, analysis, and supported conclusions. RTPO independently samples responses to scene-matched image pairs with fine-grained visual differences under a shared task. Its rubric-based verification combines Gravity, which checks visual fidelity and description coverage, and Flight, which takes descriptions as premises to assess evidence use and relations between paired judgments. Their scores and task correctness yield a shared pair-level advantage that couples policy updates from both responses. Using 1,008 demonstrations and 500 training pairs for harbor specialization, GF achieves a mean score of 69.12 across six evaluation sets, exceeding the backbone and supervised initialization by 9.76 and 4.56 points. Matched controls show gains over single-image and joint-input RL, and removing either verification view reduces performance. Applying GF to urban specialization also improves on the base model and supervised initialization across three evaluation subsets. Project details and updates are available at this https URL.
[CV-67] ach with Blinks: Breaking Linguistic Momentum in On-Policy Distillation for Fine-Grained Multimodal Perception
链接: https://arxiv.org/abs/2611.16402
作者: Tangzheng Lian,Hang Zhang,Guanyu Hu,Renjiu Hu,Zhensong Zhang,Jiankang Deng,Jifei Song
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:On-policy distillation (OPD) has emerged as an effective post-training approach to improving fine-grained perception in multimodal large language models by amortizing costly inference-time zooming operations into a single-pass recipe. However, vanilla OPD is inherently susceptible to linguistic momentum: the teacher can follow the student’s erroneous prefixes, thus weakening corrective supervision. To address this, we propose Blink On-Policy Distillation (BlinkOPD), in which the teacher “blinks” at every token: it reads the same prefix once with its eyes open and once with its eyes closed. A first-glance gain further strengthens the correction on rollouts that conflict with the teacher. By extracting more effective supervision from each on-policy trajectory, BlinkOPD achieves state-of-the-art average performance across six fine-grained benchmarks using only 500 training images (12x less data) than existing OPD methods. Beyond fine-grained perception, BlinkOPD also demonstrates strong zero-shot improvements across ten held-out benchmarks spanning general visual perception, visual hallucination, and multimodal reasoning abilities. Code: this https URL
[CV-68] FuGIBench: Evaluating Reliable Preparation in Streaming Video
链接: https://arxiv.org/abs/2611.16400
作者: Changheng Cai,Zhichao Huang,Song Ren,Zhichen Zhao,Hongsen Liu,Chaoyang Zheng,Wenxi Wu,Yuan Fang,Wei Wang,Yang Wang,Yuchen Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:An assistant watching a worker can prepare for an emerging need before the worker is ready to receive this http URL the need evolves, the assistant must maintain preparation until a response is due, or withdraw it if the need disappears. Video anticipation and interactive response benchmarks leave this interval between early evidence and response largely unmeasured. We introduce FuGIBench (Future-Grounded Interaction), a streaming-video benchmark in which a model chooses among waiting, reversible preparation, and response given a causal video prefix and a task contract. The benchmark contains 19,893 synthetic tool-assistance videos across ten tool tasks, in handover, ordinary-activity, and cancelled-need tracks. One primary score credits sustained preparation, incorporates response-interval alignment, and penalizes activity when the reference prescribes waiting. On a frozen set of 1,000 episodes, a reference system with a longer preparation lead has poorer response alignment, while similar handover leads accompany different levels of activity during ordinary work in matched scenarios. Together with cancellation analyses, these results show why evaluating preparation requires tracking decisions as needs develop, become actionable, or disappear.
[CV-69] ransXplainRRG: A Radiologist-Evaluated LVLM for Chest Radiograph Report Generation
链接: https://arxiv.org/abs/2611.16393
作者: Chayan Mondal,Duc-Son Pham,Tele Tan,Ashu Gupta,Vincent T.Y. Tiong,Shreya Ghosh,Tom Gedeon
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 17 pages, 6 figures, 11 tables
Abstract:Recent advancements in Large Vision-Language Models (LVLM) enhance X-ray report generation, but often neglect clinical context, overlook workflow integration, and lack explainability, which hinders clinical adoption due to interpretability and reliability concerns. This paper addresses the dual challenges of performance and expert explainability in automated X-ray report generation for clinical AI applications. Our proposed approach, TransXplainRRG, utilises an off-the-shelf Swin Transformer model along with a transformer-based text encoder that incorporates patient medical history to generate a radiology report. Further, we explore an expert-guided `inside-out’ approach and extract only abnormal findings for radiology report refinement. Thus, this study bridges the gap between high-performance automation and interpretability critical for clinical practice by combining state-of-the-art transformer-based vision encoders, text encoders, and LLMs. TransXplainRRG is trained and tested on the large-scale MIMIC-CXR dataset and further evaluated on the unseen IU X-Ray dataset to demonstrate its generalizability and robustness. We employed the GREEN metric as a complementary measure of clinical consistency, achieving scores of 0.327 on MIMIC-CXR and 0.605 on IU X-Ray, both higher than those achieved by MAIRA-2 under the same evaluation setting. It also performed competitively across standard evaluation metrics, frequently matching or surpassing state-of-the-art models. Additionally, we introduce qualitative evaluation metrics developed from radiologists’ viewpoints and conduct a preliminary radiologist evaluation on 100 generated reports to assess their clinical relevance. Quantitative and qualitative analyses further demonstrate the potential of our TransXplainRRG framework for reliable and explainable radiology report generation.
[CV-70] RADIANT: Redundancy-Aware Information Allocation for Universal Multi-Modal Fusion
链接: https://arxiv.org/abs/2611.16382
作者: Shenghai Wu,Yue Lin
类目: Computer Vision and Pattern Recognition (cs.CV); Multimedia (cs.MM)
备注: 9 pages, 4 figures. Published in ACM Multimedia 2026 (MM '26). Code: this https URL
Abstract:Universal multi-modal fusion often fails not because modalities are insufficiently aligned, but because the fusion process allocates information poorly: redundant evidence is repeatedly amplified, modality-unique evidence is over-smoothed during alignment, and task-dependent synergistic evidence is insufficiently activated. We present RADIANT (Redundancy-Aware Dynamic Information Allocation Network), a task-general framework that organizes fusion around three complementary roles: suppressing redundant cross-modal overlap, preserving modality-unique evidence, and composing task-conditioned synergy through cross-modal interaction. Specifically, stable matrix-exponential projections are used to reduce interference among shared and modality-unique streams, while task memory with hard positive updates selects which synergistic interactions should be activated for the current fusion regime. Extensive experiments across six heterogeneous fusion tasks show consistent improvements over the compared methods. On FLIR, RADIANT improves downstream detection mAP@0.5 from 91.2% to 96.8%, a gain of 5.6 percentage points (6.1% relative) over GIF. RADIANT also exhibits strong direct cross-dataset transfer, with further gains from inference-time task-memory adaptation while keeping all network weights frozen. Additional ablations and sensitivity studies support the robustness of the allocation mechanism.
[CV-71] Description-Guided CLIP Adaptation with Local Evidence Calibration for Few-Shot Plant Disease Recognition
链接: https://arxiv.org/abs/2611.16373
作者: Zhenkun Ye,Qian Zhou,Hua Zou
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Accurate plant disease recognition is essential for intelligent agriculture, but few-shot settings provide limited labeled samples that may not capture diagnostic symptoms and intra-class variations. Pretrained vision-language models (VLMs), such as CLIP, offer transferable visual-semantic knowledge. However, existing CLIP-based methods often rely on coarse class semantics and global visual representations, overlooking disease-specific diagnostic information and fine-grained local cues. Excessive few-shot adaptation may also cause over-specialization and weaken the generalization ability of large-scale pre-training. We propose a description-guided, two-stage CLIP adaptation method to address these issues. Symptom and comparative descriptions enrich class semantics with typical visual symptoms and discriminative cues between similar diseases. The enhanced semantics guide the aggregation of class-relevant patch evidence to calibrate global image-text matching with fine-grained cues. For adaptation, we first update the vision-language encoders with lightweight parameters, then freeze them and train a linear classifier. At inference, the classifier is used for in-domain prediction, while class prototypes from the adapted text encoder support cross-domain prediction and retain vision-language transferability. Experiments with two backbones show that our method achieves the highest mean accuracy in most few-shot settings on PlantVillage, Cassava, and Pear, reaching 93.16%, 79.30%, and 96.56%, respectively, with ViT-L/14 under 16-shot. On class-aligned cross-domain evaluation across 21 shared PlantDoc classes, it achieves 56.96% accuracy, outperforming all compared methods. These results demonstrate its effectiveness in fine-grained few-shot plant disease recognition and cross-domain generalization.
[CV-72] ReliaGuard: Reconstruction-Aware Watermarking for Source Attribution and Blind Edit Localization
链接: https://arxiv.org/abs/2611.16368
作者: Hailong Wang,Wuzheng Tan,Zixin Zhang,Xujun Fang,Fengjuan Gong
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR)
备注:
Abstract:Generative image editing facilitates content creation but complicates source attribution and edit verification. Existing watermarking methods combine copyright protection with tamper localization, yet background reconstruction can weaken spatial evidence even when source identifiers remain recoverable. We propose ReliaGuard, a reconstruction-aware watermarking framework for source attribution and blind edit localization. The writer optimizes keyed feature responses in both the protected image and its variational autoencoder reconstruction under image-distortion and source-preservation constraints. Using only the received image and a secret key, the reader combines watermark responses with local image context through residual evidence refinement and predicts an edit mask through native-resolution-guided multiscale reading. It is trained on complete editing outputs, locally spliced counterparts, and processed versions of both. On an independent test set of 300 source images and 900 editing cases from three image domains, our method achieves edit-region intersection-over-union scores of 85.61% for complete outputs and 94.90% for locally spliced outputs at a threshold calibrated to a nominal 1% background false-positive rate on complete outputs. The complete-output result exceeds the strongest baseline by 45.02 percentage points, while decoded identifier-bit accuracy reaches 100% in both conditions. Across seven families of image processing, the corresponding localization scores are 81.01% and 86.49%, with identifier-bit accuracies of 92.76% and 92.79%, respectively. These results support source recovery and blind edit localization under background reconstruction and subsequent image processing.
[CV-73] Hand in Silhouette: Synthetic-to-Real Multi-View Hand Reconstruction via Binary Masks
链接: https://arxiv.org/abs/2611.16339
作者: Leiyao Cui,Lecheng Ruan,Zihang Zhao
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project page: this https URL
Abstract:Calibrated multi-view hand reconstruction has advanced rapidly, yet the effectiveness of existing methods depends strongly on real captured data, which is costly to acquire, limited in hand-geometry diversity, and often affected by imperfect annotations. Synthetic data offers a scalable alternative with accurate supervision, but a substantial synthetic-to-real appearance gap hinders its use. To address this, we introduce Hand in Silhouette (SilHand), a calibrated multi-view hand reconstruction framework that operates on binary hand masks, obtainable at low cost either automatically or manually, to reduce the synthetic-to-real observation gap. Rather than relying on RGB features for 3D lifting, SilHand projects metric 3D hypotheses into each view to query and fuse mask-ray evidence for hand reconstruction. This mask-centric formulation enables training on a large-scale synthetic dataset comprising 46.8k grasping poses and 60k free-hand samples, providing broad coverage of hand geometry while directly transferring to real captured hand geometry. Across five real-world benchmarks, SilHand improves relative reconstruction accuracy from 49.2% to 63.4% using masks rendered from reference hands. Moreover, reconstruction accuracy improves consistently as the synthetic training set grows, indicating that broader geometric coverage translates into better real-world performance. While achieving such diversity is difficult with real captured data, it becomes practical under our mask-centric formulation. Finally, SilHand remains robust under mask corruption, with reconstruction error increasing by at most 1.63 mm. Together, these results support binary masks as an effective interface for learning hand reconstruction from synthetic geometry.
[CV-74] Selective Generative Restoration for Video Deraining via Uncertainty-Aware Diffusion
链接: https://arxiv.org/abs/2611.16310
作者: Yi’ang Chen,Han Li,Laibin Chang,Hui Liu,Kui Jiang,Junjun Jiang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Video deraining is challenging due to rain streaks, occlusions, and atmospheric effects. Existing methods mainly follow a deterministic restoration paradigm, assuming that clean scenes can be uniquely recovered from rainy observations. However, rain degradation is spatially heterogeneous. Reliable regions require faithful reconstruction, whereas severely occluded regions contain ambiguous content that require stronger prior-driven inference. Therefore, conventional deterministic restoration often produces over-smoothed results, while global generative restoration may introduce unnecessary hallucinations. We introduce a Selective Generative Restoration paradigm for video deraining, where generative priors are adaptively activated according to restoration uncertainty. Based on this principle, we propose Selective Generative Video Deraining (SGVD), an uncertainty-aware diffusion framework that enables adaptive fidelity preservation and generative recovery. Specifically, SGVD exploits stochastic diversity between dual one-step diffusion predictions to estimate region-level restoration uncertainty, which determines where observation fidelity should be preserved and where generative inference should be activated. Furthermore, an anchor-guided trajectory consistency mechanism is proposed to regularize stochastic diffusion evolution across frames, reducing temporal generation drift and improving video coherence. To achieve practical efficiency, SGVD adopts one-step diffusion inference while maintaining uncertainty-aware generative capability. Extensive experiments demonstrate that SGVD achieves superior restoration performance, perceptual quality, and temporal consistency compared with state-of-the-art video deraining methods.
[CV-75] DEDCA: Test-Time Adaptation for Generalized AI-Generated Image Detection
链接: https://arxiv.org/abs/2611.16307
作者: Hao Tan,Qimin Zhang,Lihuang Fang,Kebing Jin,Jinghui Qin
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:AI-generated image detectors face significant challenges when deployed in real-world environments, particularly when test samples are produced by unseen generators that deviate from the source training distribution. Existing detectors often rely on a single type of evidence, such as semantic representations or low-level generation traces, making them vulnerable when the corresponding cue becomes unreliable. Moreover, adapting detectors to newly emerging generators usually requires target labels or retraining, which is costly and impractical. To address these challenges, we propose Dual-Evidence Disagreement-Constrained Adaptation (DEDCA), a test-time adaptation framework for generalized AI-generated image detection. DEDCA constructs a dual-evidence detector by combining CLIP-based semantic evidence with trace evidence extracted from shuffled entropy and residual statistics, and introduces a reliability-aware gate to perform sample-specific adaptive fusion. Our key idea is to exploit the disagreement between the two evidence streams as an unlabeled adaptation signal, enabling the detector to adjust to target-domain shifts without relying on target labels or blindly trusting a single prediction stream. Experiments show that DEDCA achieves superior performance on the GenImage benchmark by outperforming state-of-the-art AI-generated image detectors and test-time adaptation methods with 92.0% average accuracy.
[CV-76] FlowIR: Depth-driven Rectified Flow for Resolving Thermal Ambiguity in Infrared Super-Resolution ICME2026
链接: https://arxiv.org/abs/2611.16269
作者: Xingyue Zhu,Qinyun Mei,Zhexi Min,Yang Zou,Xingyuan Li,Zhiying Jiang,Jinyuan Liu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to the 2026 IEEE International Conference on Multimedia Expo (ICME 2026)
Abstract:Distinguishing different heat sources in low-resolution infrared (IR) images is inherently difficult, as multiple thermal emitters often collapse into ambiguous and overlapping responses. Existing super-resolution methods hallucinate structures from such ambiguous 2D evidence, leading to thermal drift, boundary merging, and unreliable restoration under infrared degradations. We propose FlowIR, a depth-driven rectified flow framework that resolves thermal ambiguity by injecting 3D geometric cues into the generation process. Pseudo-3D depth lifting and depth-guided 3D convolutions disentangle overlapping foreground objects, while cluster-wise inter-attention suppresses cross-source feature contamination. A Thermal-Geometry Consistency Loss further enforces geometric alignment during refinement. Built upon a lightweight rectified flow backbone, FlowIR restores IR images in only a few steps, producing sharply separated, degradation-robust, and geometrically faithful high-resolution results.
[CV-77] Reconstruct What You Can: Distractor-free 4D Gaussian Splatting WACV2027
链接: https://arxiv.org/abs/2611.16259
作者: Joel Julin,Liuyue Xie,Ananya Bal,Mose Sakashita,Koichiro Niinuma,Laszlo A. Jeni
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at WACV 2027
Abstract:Dynamic reconstruction from casual video must preserve valid subject motion while suppressing transient occluders, shadows, and volumetric effects. Existing methods either suppress valid motion through static-scene assumptions or model all observations jointly, producing ghosting and unstable geometry. We introduce ReCan-GS, a 4D Gaussian Splatting framework that separates coherent scene content from inconsistent observations. It derives a Jacobian-based deformation-sensitivity score from the learned deformation field, using it as an empirical proxy for reconstruction ambiguity. The score guides separation into dynamic Core Gaussians and per-view Distractor Gaussians, or post-hoc pruning of sensitive Core Gaussians. The same criterion supports removal of spatial occluders, photometric artifacts, and evolving smoke or fog without masks, pretrained segmenters, or medium-specific supervision. Experiments on synthetic and real dynamic scenes and synthetic smoke/fog scenes demonstrate its usefulness across the evaluated distractor regimes.
[CV-78] ATune: Unsupervised Photo Retouching Through Image-Editor Controls
链接: https://arxiv.org/abs/2611.16236
作者: Divyanshu Goyal,Gabriel Ruiz,Amir Meisami
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 27 pages, 12 figures, 18 tables. Main paper 8 pages plus supplementary material
Abstract:Photo retouching models are trained on pairs of original and retouched photographs, usually edited by experts. Expert edits are costly, so the two main public sets, FiveK and PPR10K, hold only 16,161 originals. Good photographs on their own are far easier to find: commercial stock collections and free ones such as Pixabay and Unsplash hold millions. We learn retouching from them alone: we damage each photograph the way a camera or a careless edit would, which gives a worse and a better version of the same photograph. From these pairs, our model, ATune, learns how 13 measured attributes, such as brightness, contrast and sharpness, should change to raise a photograph’s quality. The edit is declarative: ATune sets target values for the attributes, and a simple search adjusts the sliders of an out-of-the-box photo editor until the photograph reaches them. Every edit is a list of slider values a person can read and adjust. The same model can drive any editor that makes those changes; we test RawTherapee and darktable. We train on 10M photographs from a commercial stock collection and test on FiveK, PPR10K and SPAQ. Nine quality judges score every edit, and in a user study 157 human raters made 9,253 choices between our edit and another method’s edit of the same photograph. When Claude Opus 5 and GPT-6 Astra edit the same photographs in the same editor, raters prefer our edits to GPT-6 Astra’s (54.3%) and rate them even with Claude’s (50.5%). We also compare with supervised retouchers on their home ground: NeurOP on FiveK and AdaInt on PPR10K, each trained on that set’s own expert edits. Raters chose our edits in 46.3% of those comparisons. ATune, a 54M-parameter model trained without a single expert edit, now matches frontier models at setting an editor’s sliders.
[CV-79] Disentangled Embedding Optimization for Diffusion-Based Dataset Distillation ACM-MM2026
链接: https://arxiv.org/abs/2611.16224
作者: Hongcheng Li,Yucan Zhou,Yibin Wang,Xiaoyan Gu,Bo Li,Weiping Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 10 pages, 7 figures. Accepted at the 34th ACM International Conference on Multimedia (ACM MM 2026)
Abstract:Dataset distillation aims to compress large-scale datasets into compact synthetic subsets that enable efficient model training with comparable performance to the original data. Recent advances show that diffusion models are highly effective for this task due to their strong generative capability. Existing diffusion-based approaches typically perform class-wise clustering and use the obtained cluster prototypes as diffusion inputs to generate distilled samples. However, the uniformly distributed original samples in a category inevitably lead to entanglement near cluster boundaries, pulling the cluster prototypes closer and reducing the inter-cluster separation. Consequently, the diffusion model receives overly similar latent inputs, resulting in synthetic data with limited diversity. To address this issue, we propose Disentangled Embedding Optimization (DEO), a simple yet effective framework that replaces static clustering with trainable embeddings. These embeddings are optimized to match the Gaussian statistics (i.e., mean and covariance) of the original data distribution, while a diversity regularizer is designed to encourage separation among these embeddings to improve the intra-class variability. Extensive experiments demonstrate that our proposed DEO can generate more diversified and representative distilled data than existing cluster-based diffusion methods, achieving state-of-the-art performance and superior generalization across multiple datasets.
[CV-80] WorldDraft: If Code Builds the World Can Video Models Just Render It?
链接: https://arxiv.org/abs/2611.16217
作者: Zhiyu Hou,Wen-Jie Shu,Harold Haodong Chen,Weilin Ruan,Chenfei Liao,Liuzhou Zhang,Disen Lan,Xiaogang Xu,Qi Chen,Ser-Nam Lim
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project: this https URL
Abstract:Multimodal coding agents decouple 4D world construction from visual generation, repurposing video models from world generators toward generative renderers. This division of labor raises a fundamental question: if code builds the world, can video models just render it? A reliable generative renderer hinges on two complementary criteria: (i) world preservation, retaining prescribed 4D kinematics and camera paths; and (ii) visual realization, persistently binding text-specified target attributes over time. To evaluate this capability, we introduce WorldDraft, a benchmark of 432 strictly paired test cases across 48 executable 4D world families spanning six dynamic challenge categories. Under a 3 \times 3 factorial design, WorldDraft orthogonally crosses draft granularity (abstract proxy to previs) and appearance alignment (neutral to conflicting) under identical dynamics. Extensive evaluations across 12 representative models reveal that: (I) world preservation and visual realization remain in stark operational tension; (II) richer draft scaffolding anchors structure but paradoxically impairs appearance fidelity; and (III) conflicting source cues trigger severe source leakage (up to 64% ), exposing fragile cross-modal disentanglement. To mitigate these bottlenecks, we propose WorldDraft-Primer, a plug-and-play visual prompting strategy that consistently improves both objectives across frozen video renderers. Together, WorldDraft and WorldDraft-Primer lay a foundation for bridging programmatic 4D simulation with video generation, advancing physically grounded generative rendering.
[CV-81] SPEM: Scan-Prior Entropy Modeling for LiDAR Point Cloud Compression
链接: https://arxiv.org/abs/2611.16212
作者: Bojun Liu,Yangzhi Ma,Jie Li,Dong Liu,Li Li
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:LiDAR point clouds exhibit scan-induced spatial regularities that are informative for geometry compression. Existing scan-aware approaches often exploit these regularities by transforming coded geometry into spherical or beam-aligned coordinate systems. However, this couples acquisition priors with geometry representation, altering spatial partitioning and quantization, and potentially preventing exact recovery of the original Cartesian geometry. We argue that acquisition priors should guide probability modeling rather than dictate the coding representation. We propose SPEM, a Scan-Prior Entropy Modeling framework that exploits LiDAR scanning regularities within the entropy model without requiring coordinate transformation. We introduce a multi-frequency spherical rotary positional encoding that captures scan-aligned spatial relationships in both backbone self-attention and decoded-context retrieval. Building upon this encoding, we develop a decoded-memory autoregressive predictor that efficiently exploits previously decoded occupancy information through reusable key/value caches. Experiments demonstrate state-of-the-art performance in both lossy and lossless LiDAR geometry compression, highlighting the effectiveness of scan-prior modeling across different coordinate representations.
[CV-82] GOOSELoc: A Multimodal Benchmark for Ground-to-Aerial Localization in Unstructured Off-Road Terrain
链接: https://arxiv.org/abs/2611.16162
作者: Fabian Deuser,Peter Mortimer,Robin-Nico Kampa,Anna Bößendörfer,Mirko Mählisch,Martin Werner,Norbert Oswald
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Cross-view pose estimation estimates a ground camera’s geographic pose by matching onboard sensor data with geo-referenced overhead imagery. This is particularly relevant when GNSS is unreliable, for example under forest canopy or in urban canyons. Existing metric cross-view benchmarks, such as VIGOR and cross-view KITTI, are dominated by urban and road scenes with strong human-made structure without off-road terrain. We introduce GOOSELoc, a multimodal benchmark for ground-to-aerial localization in unstructured environments, including forests, fields, and rural tracks. GOOSELoc comprises 9,356 scenes recorded across all four seasons around Munich, Germany, pairing geo-referenced aerial orthophotos with surround-view RGB imagery and 3D LiDAR point clouds. Paved/unpaved labels enable analyzing the impact of terrain structure, while the seasonal recordings allow studying the appearance gap to mostly summer-captured orthophotos. To assess the transferability of existing methods, we adapt and fine-tune four state-of-the-art cross-view localization baselines on GOOSELoc. Despite task-specific fine-tuning, the performance degrades substantially in off-road environments. We therefore propose a multimodal baseline combining surround-view imagery and LiDAR geometry, improving localization performance and providing a stronger reference for future work. Data and baselines are released at this https URL.
[CV-83] Split-and-scale Latent 3D Representations NEURIPS2026
链接: https://arxiv.org/abs/2611.16126
作者: Dorian Chan,Muhammed Kocabas,Xiaoming Zhao,Oncel Tuzel,Jen-Hao Rick Chang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to NeurIPS 2026
Abstract:Latent 3D representations have significantly expanded the capabilities of modern shape modeling, enabling compact encoding of 3D scenes and supporting powerful 3D generative models. Despite this progress, current representations struggle to capture the fine details of complex objects, fundamentally limited by the capacity of their latents. However, scaling latent capacity at training time is computationally expensive and often infeasible under memory limits, and naively increasing the number of latents at inference fails to make effective use of the added capacity. In this paper, we address both challenges with \textitsplit-and-scale, a method that effectively expands latent capacity at inference while keeping training cost tractable. Our key observation is that by representing 3D scenes as continuous functions of 3D coordinates, a single pretrained tokenizer can be applied at inference to arbitrarily small sub-regions of a scene, allocating the same latent capacity to a smaller spatial extent and thus capturing far higher detail. We validate this insight by conducting extensive ablations to study the trade-offs between compute, memory, latent capacity, and reconstruction quality, demonstrating our inference-time scaling method achieves quality competitive with much larger models that exceed single-GPU memory limits. We further demonstrate that split-and-scale supports high-quality generative modeling, training an autoregressive 3D generator that performs competitively with state-of-the-art baselines.
[CV-84] Pothole Risk Assessment with Geometrically Grounded Vision Language Models
链接: https://arxiv.org/abs/2611.16061
作者: Abdalmalek Aburaddaha,Zaid A. El Shair,Ali Al-Bustami,Samir A. Rawashdeh
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 22 pages, 8 figures, 17 tables. Submitted to IEEE Access. Code and data: this https URL
Abstract:Potholes impose billions of dollars in annual vehicle damage, yet detectors report only where potholes are, not how severe they are or what to do. Judging severity from a single RGB image is ill-posed, and ungrounded vision-language model (VLM) assessments tend to be over-conservative. We present a five-stage modular framework for the offline assessment of already-detected potholes that combines geometric measurements with VLM reasoning: perspective transformation, zero-shot SAM segmentation, geometric feature extraction, Gemini 2.5 Pro severity classification with composite risk scoring, and natural-language warning generation. We evaluate it on 3,092 potholes in 1,119 dashcam scenes against independent references. Severity agreement with a prediction-free three-annotator panel is weak (Cohen’s \kappa = 0.139 , 95% CI [0.00, 0.28]; n = 101 ) and far below the annotators’ agreement with one another ( \kappa = 0.605 ); against physically measured depth on an independent RGB-D dataset, it is also weak ( \kappa = 0.098 ), although the predicted classes preserve the order of measured depths. The severity classes are therefore perceived visual categories, not depth measurements. Geometric grounding does change the warnings: relative to an image-only baseline, it moves 26.7% of scenes out of the high-urgency tier and raises critical-tier assignments by 109% (Stuart-Maxwell p 10^-15 ), and in a blind study, three raters preferred the grounded warnings 3.5:1. A six-arm prompt ablation attributes the hybrid prompt’s rule adherence to the per-pothole measurements rather than to the rubric or worked examples. At USD 0.105 per scene, it suits large-scale offline infrastructure assessment; its maneuver suggestions are advisory research outputs, not validated vehicle guidance.
[CV-85] Sub-actions in Action: Text-Guided Hand-Role Alignment for Sub-Action Recognition BMVC2026
链接: https://arxiv.org/abs/2611.16056
作者: Sadegh Rahmaniboldaji,Filip Rybansky,Quoc Vuong,Frank Guerin,Andrew Gilbert
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at BMVC 2026, Supplementary material is available
Abstract:Our objective is to enhance VLMs ability to distinguish the roles of the left and right hands in egocentric actions. We introduce sub-action detection, a task that requires models to explicitly recognise the left- and right-hand roles and their manipulated objects within an ongoing high-level action. To address this challenge, we propose BiScopeNet, a framework that augments existing video-LLMs without modifying their vision encoders or language models. BiScopeNet incorporates two components: FocusAlign, a gated text-guided attention-based module that selectively aligns video features with textual context to emphasise action-relevant cues, and MetaLoss, a masked, object-weighted loss that prioritises visually grounded terms. To supervise sub-action detection, we introduce BiScopeCap-100K, a large-scale video-level dataset curated from existing egocentric datasets, providing dense captions that jointly describe high-level actions, left/right-hand roles, and manipulated objects, enriched with spatial metadata for grounding. Across the evaluated Video-LLM backbones, BiScopeNet improves action, hand-role, and object recognition over the corresponding supervised baselines, achieving the strongest aggregate performance among the evaluated configurations on BiScopeCap.
[CV-86] opology-Aware 4D Gaussian Splatting for Egocentric Videos
链接: https://arxiv.org/abs/2611.16012
作者: Leonardo Vanni,Stefano Gasperini,Federico Tombari,Chiara Plizzari
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project page: this https URL
Abstract:Reconstructing dynamic scenes and objects from monocular egocentric video is crucial for embodied AI, yet existing 4D Gaussian Splatting (4DGS) approaches struggle with everyday interactions involving topological changes, such as cutting, assembly, opening, or closing. Since 4DGS methods rely on deforming a fixed representation, events that split, merge, or reveal surfaces cause severe visual artifacts. To address this challenge, we introduce TopoEgo, the first benchmark dedicated to topological changes in egocentric video (covering fission, fusion, open, and close), alongside Topo-4DGS, a topology-aware reconstruction method. Topo4DGS jointly models part-level motion and time-varying surface support by assigning independent rigid transformations to interacting parts and dynamically activating or deactivating Gaussians as surfaces appear, disappear, or separate. Extensive experiments demonstrate that Topo-4DGS significantly outperforms state-of-the-art 4DGS approaches across topological changes, especially on post-event frames. Code and data will be made publicly available.
[CV-87] Iris-Z: Scaling One Generative Model Past Specialist Geometry Models
链接: https://arxiv.org/abs/2611.16010
作者: Alejandro Cañada,Chema Garabito
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 14 pages, 9 figures, 6 tables. Code and models: this https URL
Abstract:Language moved from many task-specific models to one unified model: a single generative model with a single generative training task that handles all of them within the same backbone, and keeps improving with data, parameters and compute. We believe the bitter lesson applies just as much to computer vision, 3D and spatial understanding, in the end to visual intelligence: the effort should go into general architectures that solve all these tasks in a unified way, and into scaling them, rather than into task-specific designs. We test this on monocular geometry, a task that has belonged to specialised architectures for years. Iris-Z is a general image generator that produces appearance and geometry together. A single backbone denoises, reasons about and understands the image and its depth at the same time, and one set of weights generates images with their depth, estimates depth from photographs and paints images onto a given depth. At enough scale it surpasses state-of-the-art specialists on established depth benchmarks while keeping the generative abilities of the backbone it started from without specialized layers. We release two models for the community.
[CV-88] Adaptive Distance Detection and Adaptive Instance Segmentation
链接: https://arxiv.org/abs/2611.15991
作者: Kai Yi Ji
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: This is the paper from my Bachelor’s degree in University of Birmingham
Abstract:This paper is about object detection model and semantic segmentation model in an FPS game Counter Strike: Condition Zero to detect enemies. Semantic segmentation is pixel-wise detection algorithm. We customized the architecture for semantic segmentation. We used YOLOv5s to train object detection model. After training, the inference time of semantic segmentation was compared to object detection. It was observed that object detection was slightly faster and semantic segmentation is slower but better at detecting the exact shape of the object. However, we found out that this slight difference of inference can grow exponentially if the both models get more complex. To solve this, we propose a novel approach called Adaptive distance detection which combines both algorithms. We will also introduce Adaptive instance segmentation to combine semantic segmentation and panoptic segmentation.
[CV-89] RACE: Trajectory Synchronization for Training-Free Resolution Extension in Video Generation
链接: https://arxiv.org/abs/2611.15983
作者: Libo Zhu,Fan Li,Zhixin Wang,Binglu Wang,Weisheng Dong,Yulun Zhang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:More text-to-video models are available, but many remain limited to low resolutions. Training them for higher resolutions is costly, while existing training-free methods still struggle with fine detail, scene structure, and efficiency. We propose TRACE, a training-free method that synchronizes low- and high-resolution generation trajectories. During early denoising, TRACE aligns predicted scene content at corresponding noise levels. It computes high-resolution updates from low-resolution predictions and the current high-resolution state, without high-resolution model inference. The high-resolution trajectory retains fine-grained variation for generating new details. Local window attention then refines fine details in a shared high-resolution latent state. TRACE checks suspected duplicate subjects against the corresponding low-resolution layout and corrects selected regions and their solver history while preserving detail elsewhere. Experiments across all kinds of video generation backbones and resolutions show improved perceptual quality and text-video alignment over FreeSwim, a state-of-the-art training-free method, with 2.0-2.9x end-to-end speedups. Our code and models will be released at this https URL.
[CV-90] DuoJEPA: Cross-Modal Translation Between Unaligned Self-Supervised Embedding Spaces
链接: https://arxiv.org/abs/2611.15968
作者: Mahmut S. Gokmen,Mitchell A. Klusty,Evan W. Damron,Caroline N. Leach,Samuel E. Armstrong,V.K. Cody Bumgardner
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Learning from paired modalities is commonly equated with aligning them: two encoders are trained until the embeddings of a pair coincide in a shared space. We study whether cross-modal retrieval and classification can instead be obtained while the two encoders’ own embedding spaces remain directly unaligned. DuoJEPA trains one encoder per modality with the self-supervised LeJEPA objective and appends to each encoder a single transformer layer that translates its embedding into the other encoder’s space, supervised by a contrastive objective computed entirely within that target space. No loss term compares the two own spaces directly, although the encoders are optimised jointly through the translation objective. In our experiments, the translator’s residual connections determine where the cross-modal pressure is absorbed: with a residual translator the two encoders become directly aligned, whereas with a skipless (residual-free) translator a paired-cosine test detects no alignment in the final models, and at most a weak one in the ablations, at comparable cross-modal performance. The learned class-token translation is well approximated by a single scaled rotation of the source space. We evaluate DuoJEPA on two modality pairs that differ in nearly every respect. On chest radiographs and reports from CheXpert Plus, it attains a zero-shot AUC of 86.6 and a cross-modal retrieval score of 78.5, compared with 60.0 and 64.2 for our reimplementation of LeVLJEPA with the same encoders and data. On Sentinel-2 optical and Sentinel-1 radar imagery, the frozen encoders fall within the range of published Sentinel foundation models under our reproduction of the Copernicus-Bench linear-probe protocol, the translator retrieves the paired image of the other sensor among 4,000 candidates in 77-79% of cases, and a patch-level extension transfers the within-image similarity structure of the scene.
[CV-91] Lang-SVG: Hierarchical Image Vectorization with Language Priors NEURIPS2026
链接: https://arxiv.org/abs/2611.15941
作者: Xi Liu,Chaoyi Zhou,Run Wang,Jiaang Li,Feng Luo,Junxiang Huang,Siyu Huang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted by NeurIPS 2026
Abstract:Image vectorization reconstructs raster images as compact scalable vector graph- ics (SVG), providing a structured and editable code representation of images. However, most existing vectorization methods optimize for pixel-level render- ing fidelity, often producing SVGs that lack alignment with human-perceived object-part hierarchies, making them difficult to manipulate. To address this, we propose Lang-SVG, which introduces an explicit tree structure to organize SVG primitives with semantic labels and parent-child relationships. Lang-SVG incorporates granularity-controllable and open-vocabulary segmentation priors to initialize object-part hierarchies and leverages multimodal large language models (MLLMs) with SVG tree context to assign semantic labels to unlabeled nodes. To reduce redundancy, we introduce depth-adaptive simplification to reduce shape complexity and Tree-Guided Pruning and Merging to eliminate redundant SVG nodes. L1-SAM-Prompting further recovers missing regions to improve recon- struction fidelity and structural completeness. Beyond reconstruction fidelity, we introduce a new evaluation protocol that measures structural quality and semantic part alignment. Experiments demonstrate that Lang-SVG achieves state-of-the-art performance in both rendering fidelity and structural quality. Project website: this https URL.
[CV-92] MOSAIC: Modular Obstacle and Spatial-Aware Framework for Manipulation In Cluttered Scenes
链接: https://arxiv.org/abs/2611.15937
作者: Richard Shi,Mohammad Mahdavian,Mehrdad Momeni Zadeh,Mohammed Elmahgiubi,Amir Rasouli
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Vision-Language-Action (VLA) models have demonstrated strong performance in robotic manipulation, yet most rely primarily on 2D visual observations and language instructions, leaving scene geometry and obstacle relationships to be inferred implicitly. This limitation is particularly problematic in cluttered environments, where small spatial errors can lead to unintended contacts or collisions. We propose MOSAIC, a modular spatial- and obstacle-aware framework that can be integrated with pretrained VLA policies to improve object-level scene understanding and action generation. Our method augments the VLA backbone with an auxiliary semantic grounding task that predicts role-specific masks for the target object, destination region, and surrounding obstacles. We further combine the target and destination masks with depth estimated from 3D scene reconstruction to construct role-specific geometric tokens that are injected directly into the action expert. In addition, we propose obstacle-aware attention supervision that emphasizes obstacles near the robot body during task execution. The resulting attention features are subsequently injected into later action-expert layers to further guide action generation. Together, these components provide explicit semantic, geometric, and obstacle-aware cues while preserving the underlying pretrained VLA architecture. We evaluate our framework on pretrained VLA model using the RoboPRO benchmark, including visual perturbation tests, and further validate it on real-world manipulation tasks. MOSAIC improves the success rate by 5.1% in cluttered scenes, and by 9.7% under visual perturbations, with gains of up to 25.0% in the most challenging environment. We will release our code and dataset upon acceptance at the following Github page: this https URL
[CV-93] Open Vision Banana: A Fully Open Recipe for Generative Visual Perception
链接: https://arxiv.org/abs/2611.15844
作者: Arpit Jadon,Valentinos Pariza,Jona Ruthardt,Ryousuke Yamada,A. Sophia Koepke,Andrei Bursuc,Yuki M. Asano
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Recent work suggests that pretrained image generators encode real-world understanding beyond image synthesis. Vision Banana exploits this by framing vision tasks as conditional image editing. However, its data, weights, code, and implementation details are not public. We present Open Vision Banana (OVB), a fully open and reproducible recipe for converting pretrained generators into unified multi-task perception systems. Our analyses identify the design choices that matter most for generalization, including color encoding, prompt design, parameter adaptation, and task balancing. These choices interact, so individually helpful choices can hurt when combined with others. Starting from 4B and 9B parameter open-weight base models and using 2.1M-3.8M training pairs from public datasets, OVB surpasses Vision Banana’s reported Cityscapes zero-shot performance (73.9 vs. 69.9 mIoU) and rivals state-of-the-art affine-invariant depth estimators. Its non-zero-shot variants come within five points of Vision Banana on referring and reasoning segmentation but remain further behind on instance segmentation. All models largely retain the base model’s image-editing abilities. With legend ensembling, we introduce a test-time scaling method for generator-based perception. It segments an image under several class-color legends and fuses the results by majority vote, correcting unstable color-class bindings and improving mIoU by up to 8.7 points on Cityscapes and 15.0 points on COCO-Object without retraining. We release the dataset construction and processing pipeline, the training and evaluation code, as well as the final, intermediate, and ablation checkpoints. Project website: this https URL
[CV-94] Where to Generate Touch: Do Image Latents Preserve Material Recognition?
链接: https://arxiv.org/abs/2611.15834
作者: Wanghao Ye,Aarosh Das,Sihan Chen,Yiting Wang,Bowei Tian,Guoheng Sun,Shwai He,Zheyu Shen,Ziyao Wang,Yexiao He,Zhaoyi Liu,Meng Liu,Yuning Zhang,Meng Feng,Ziyi Wang,Yilong Dai,Yifei Dong,Siyuan Peng,Zhenle Duan,Joshua Liu,Lang Xiong,Daeun Jung,Aditya Sarkar,Parastoo Pilevar,Xiaochuan Gong,Yancheng Zhu,Yingying Fan,Qiong Wang,Ang Li
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注: 13 pages, 8 figures, 10 tables
Abstract:Generated touch could extend robot learning beyond recorded contacts. A tactile generator needs a latent space, and we choose one by testing whether a material classifier trained on real GelSight presses still works on encode-decode reconstructions from pretrained image models. On ObjectFolder Real, the classifier loses 6.5 to 9.7 points on reconstructions from three 16x discrete tokenizers and 0.3 points on those from the Stable Diffusion VAE (SD-VAE). The 16x tokenizers again lose far more on Touch and Go and on DIGIT object recognition (YCB-Slide). Retraining the classifier on reconstructions, with no new labels, narrows the discrete gap. SD-VAE reconstructions also keep object identity, with 76.7% object-matching accuracy against 78.9% for real touch and 67.0% for MAGVIT-v2. We therefore select SD-VAE latents and show a photo-to-touch generator trained in them. Code will be released upon acceptance.
[CV-95] VibePrevisBench: Evaluating Agent ic Coding for 4D Scene Generation in Blender
链接: https://arxiv.org/abs/2611.15789
作者: Yuzhe Chen,Muxi Diao,Lele Yang,Wenbin Xu,Xuyang Liu,Qile Su,Changyi Liu,Jiankang Chen,Kaiyu Jiang,Haoyu Zhang,Hongyi Fu,Jianxiong Wang,Tianke Zhang,Haonan Fan,Yingxin Li,Bingyi Li,Xu Wang,Xuanyu Zheng,Bin Wen,Tingting Gao,Wentao Zhang,Jiale Cao
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Project page: this https URL
Abstract:Controllable real-world video generation models can condition on a rendered graybox proxy video, which fixes layout, motion and camera. Coding agents can build such videos by reconstructing a 4D scene in Blender, yet no benchmark checks both these videos and the long, multi-shot scenes behind them. We introduce VibePrevisBench, a benchmark for 4D scene reconstruction: an agent turns a brief into an executable 4D scene and renders its proxy video. Its 242 tasks span four input kinds, including scripts, photos, continuations, and reference videos. Each delivery is scored along two evaluation dimensions. Rule-based Scene Check re-runs the code and compares the exported 4D scene with hidden ground truth facts in nine dimensions. VLM-based Screen Check asks a VLM judge whether the facts that pass can be recognized in the proxy video. We analyze the evaluation results and trajectories of 11 frontier coding agents and identify three key observations. First, multimodal understanding separates the agents, especially in understanding space and camera motion. Second, all agents seem to already master basic camera work and staging, and fail at challenging requests. Third, agents inspect their own renders but fail to correct their errors. Agents thus need better visual understanding of space and camera, explicit checks for challenging requests, and item-by-item verification of renders against the brief.
[CV-96] Prior-as-Skill: Remember Outside and Read Inside with On-Policy Self-Distillation for Self-Evolving VLM Agents
链接: https://arxiv.org/abs/2611.15766
作者: Yan Lin,Jingyu Sun,Yifei Qian,Zhengtao Yao,Yuxia Wang,Daizong Liu,Chun Pong Lau,Zhongliang Guo,Stephen McGough
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: under review
Abstract:Self-evolving visual agents keep what they learn as skills: key frames with step notes on a task’s steps and their order. On-policy self-distillation (OPSD) is now a popular way to train such skills into a vision-language model that matches, on its own answers, a teacher that also reads the written order. At deployment, frames often go missing or show an unusual order, as when a person or a website reorders the steps. Existing internalization methods keep any order they store in the frame-reading weights: trained on full frames, the model guesses when they are missing, and trained with frames withheld, it stores the order, so memory can override what it sees. We propose PriorSplit, which fills in missing frames from an order kept outside the model, with no training beyond OPSD. It counts which of two neighboring steps comes first in training episodes, and the model checks if both asked steps are in view: if so, it answers from the frames, and otherwise the counts answer. Extensive experiments on synthetic skills with five models and on step-tagged order questions in new episodes of seen web, cooking and egocentric tasks demonstrate that, with frames missing, PriorSplit outperforms every compared method at OPSD’s training cost. With Qwen3-VL-4B on synthetic skills, it also stays close to OPSD on unusual orders, where methods that gain most by storing the order fall behind.
[CV-97] AIFR: Adaptive Identity Feature Recovery for Diffusion-Based Makeup and Anime Portraits
链接: https://arxiv.org/abs/2611.15764
作者: Linyan Dai,Xinwei Zhang,Yaxin Xiao,Mengyu Sun,Xinwei Liu,Renyang Liu,Qingqing Ye,Haibo Hu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Diffusion-based makeup and anime portraits are increasingly popular in online content creation (e.g., virtual makeup editing and anime avatar generation), but their altered appearances can hinder face recognition (FR) in identity-based content screening. Existing methods mainly adapt FR encoders or restore input images. The former require costly gallery re-encoding, while the latter may be incomplete or distort identity-related facial details. To address these limitations, we aim to correct identity features within a frozen FR model to recover its output embeddings in its original space. Appearance changes vary in strength and affect facial regions differently. To this end, we propose Adaptive Identity Feature Recovery (AIFR), a framework that inserts a lightweight recovery module at a single intermediate layer to predict sample-level correction strength and spatially adaptive residual updates. Module training uses fused identity targets to guide embedding recovery in the original FR space and an identity-shift score to regulate correction strength and limit unnecessary changes. At inference, AIFR processes a single portrait in one forward pass without reference images, producing recovered embeddings that can be directly matched with existing gallery templates without re-encoding. With only 1.22M trainable parameters, AIFR improves the average TAR at FAR= 10^-4 from 66.1% to 90.2% over the original ArcFace, and further experiments show improvements across unseen makeup styles, stylization strengths, and generators. Beyond recognition, the recovered embeddings can directly condition a pretrained Arc2Face model to generate faces with improved identity preservation, showing compatibility with embedding-conditioned downstream generation.
[CV-98] Image Deblending with Diffusion Models WACV2027
链接: https://arxiv.org/abs/2611.15760
作者: Dan Botchan,Ohad Fried,Yacov Hel-Or
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to WACV 2027. 22 pages, 10 figures. Code: this https URL
Abstract:Image deblending is the task of decomposing a composite image back into its source images, which is an ill-posed problem, as a single blended image can be decomposed into multiple plausible source layers. While diffusion models are often used to solve inverse problems, they typically address single-output reconstruction problems. In this work, we formulate deblending as a joint inverse problem, employing diffusion posterior sampling to simultaneously reconstruct two source images, demonstrating that a single diffusion prior is expressive enough to disentangle two full images at once rather than just one. Drawing inspiration from human perception, where semantic understanding helps disentangle visual inconsistencies, we propose a training-free framework that leverages text prompts to constrain this solution space. We address the blind deblending task through a two-stage architecture: text-guided latent diffusion performs semantic separation using prompts as high-level cues, followed by a pixel-space refinement to recover fine details. Stochastic sampling produces multiple plausible decompositions that can be ranked without ground truth. Extensive experiments show that our training-free method achieves state-of-the-art results on unseen domains and extends to spatially-varying alpha maps, while remaining competitive with supervised baselines on standard scalar blending benchmarks.
[CV-99] How Far Can Frozen Vision-Language Models Go on Event Data? A Training-Free Study on EventBench ECCV2026
链接: https://arxiv.org/abs/2611.15756
作者: Md Kamrujjaman Mobin,Md Salman Shamil,Md Adnan Arefeen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: "Technical report, EBMV @ ECCV 2026 Challenge. 13 pages, 9 figures
Abstract:We present a training-free system for the EBMV @ ECCV 2026 challenge “Event Stream Understanding with EventBench” (General Understanding track), which evaluates vision-language models on five event-camera understanding tasks. We use none of the challenge’s training data: our goal is to measure how far off-the-shelf, RGB-pretrained models can go on event data through representation and prompting alone. For four of the five categories (gesture recognition, detailed understanding, causal reasoning, and event OCR) we render event bins as red/ON, blue/OFF frames, select a category-specific subset of bins, stitch them into a single grid image, and query the same frozen Qwen3.5 9B vision-language model with a category-specific prompt for all four tasks. The fifth category, action recognition, feeds every bin directly into the organizers’ EventGPT-Plus-2B. The system reaches 62.28% overall accuracy across 2,084 test questions (rank 9/27), led by action recognition and causal reasoning (80.27%, 77.22%) and limited most by gesture recognition and event OCR (54.70%, 34.57%). We visualize test samples from every category together with the exact images the model receives, score six alternative representations against the evaluation server, and analyze the full test set post hoc. That analysis shows that our bounding-box crop almost never tightened around content, so event OCR text reached the model at roughly one fifth of its horizontal resolution, and that detailed-understanding predictions are skewed toward the first two answer options.
[CV-100] CGOcc: Learning Cross-Gaussian Dependencies for 3D Semantic Occupancy Prediction
链接: https://arxiv.org/abs/2611.15692
作者: Zirui Huang,Jiongle He,Yunfan Zeng,Yuanzhi Lu,Guangye Li,Boyan Tang,Weite SU,Pedro V. Sander,Amy Hui Zhang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 18 pages, including supplementary material
Abstract:In Gaussian-based semantic occupancy prediction, overlapping primitives can provide ambiguous class predictions, making their joint interpretation important for voxel classification. However, current methods using spatially weighted mixtures neglect explicit modeling of cross-primitive dependencies, limiting the readout’s ability to utilize category-specific relational evidence. Hence, we introduce CGOcc, a supervised multi-camera framework that explicitly models Cross-Gaussian dependencies for 3D semantic Occupancy prediction. CGOcc integrates appearance and geometric observations to refine Gaussian states and encodes cross-Gaussian dependencies in a diagonal-plus-low-rank representation. To incorporate these dependencies into voxel classification, CGOcc uses a class-conditional projection to extract diagonal and signed cross-Gaussian statistics without constructing a dense dependency matrix. The readout then refines occupied-class probabilities using these statistics while preserving a learned reference’s occupied mass and free probability. Extensive experiments on nuScenes demonstrate CGOcc achieves state-of-the-art performance, with larger gains at high Gaussian overlap and improved predictions near boundaries. Ablation studies validate the effectiveness of covariance-aware prediction and support its synergy with geometric interaction.
[CV-101] aming Neural Fields with a Frequency-Lipschitz Curriculum for Sparse-View CT Reconstruction
链接: https://arxiv.org/abs/2611.15652
作者: Songtao Hu,Jingrui Wu,Hong Ding,Fanhao Zhu,Yunhao Liu,Suyang Xi
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:Sparse-view computed tomography (SVCT) reduces radiation dose by acquiring fewer projection views, but the resulting under-determined inverse problem requires careful regularization to prevent reconstruction algorithms from fitting streak artifacts and hallucinated structures. Existing implicit neural representation (INR) methods rely on implicit spectral bias or iterative external denoising, yet they provide limited control over when high-frequency capacity is exposed during optimization. To address this limitation, we propose FLIC (Frequency-Lipschitz INR Curriculum), a capacity-controlled self-prior INR framework that makes the network’s representational capacity explicit and schedulable. The core of our method is a joint frequency-Lipschitz curriculum that couples random Fourier feature encoding with spectrally normalized MLP decoding under non-uniform Lipschitz budgets, guiding reconstruction along a coarse-to-fine trajectory from stable global anatomy to fine details. The terminal capacity is automatically calibrated from FBP gradient statistics, enabling scan-adaptive regularization without per-scan manual tuning of the terminal Lipschitz budget. We further introduce a prior-guided adaptive ray sampling strategy that concentrates integration samples near high-gradient anatomical boundaries, improving forward-model fidelity. The entire framework operates in a single training stage without iterative denoising, enabling efficient per-image reconstruction. Extensive experiments on four public CT datasets under noise-free and low-dose settings show that our method achieves state-of-the-art unsupervised reconstruction quality with consistent performance across diverse anatomical structures.
[CV-102] Ultrasound Elastography: Excitation Methods and Low-Cost Alternative
链接: https://arxiv.org/abs/2611.15645
作者: Corin F. Otesteanu
类目: Computer Vision and Pattern Recognition (cs.CV); Image and Video Processing (eess.IV)
备注:
Abstract:Ultrasound elastography (or elasticity imaging) is a method that characterizes the tissue mechanical properties with the purpose of aiding the diagnosis of certain pathologies. Many novel clinical applications use the assumption that these tissue properties are changed in the presence of pathological processes. Characterization of tissue mechanical properties is achieved by imaging the response of the tissue to an induced stress or excitation. A plethora of approaches have been used in the scientific community to induce excitation in tissue, which usually require additional tools or devices than are costly. Furthermore these are usually integrated in the imaging pipeline using additional software or hardware for control. An alternative that is simpler, faster to integrate (and low-cost) is proposed in this work, by using readily available smartphones as a source of mechanical excitation. The induced displacements are used to solve the elasticity inverse problem (IP) using a finite element method (FEM) approach. Preliminary elasticity reconstructions of a phantom that mimics tissue are similar to when using a dedicated actuator as means of inducing harmonic displacements, illustrating the feasibility of our low-cost ultrasound elasography approach.
[CV-103] MemInStream: Memorizing or Merely Recognizing? Rethinking Streaming Video Memory Evaluation
链接: https://arxiv.org/abs/2611.15640
作者: Zhenqi He,Dunyuan Xu,Yang Bo,Kien T. Pham,Yanghao Wang,Shuhan Li,Xintong Guo,Yuyang He,Fang Liu,Huankang Guan,Jingyan Chen,Rui Liu,Long Chen
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Streaming video understanding aims to continuously process incoming visual observations, making the retention and retrieval of relevant information increasingly important as the video history grows. Although existing benchmarks include memory-related questions, question-answering accuracy alone does not reliably measure memory capabilities. Incorrect predictions may arise from failures in visual understanding rather than forgetting, whereas correct predictions may rely on linguistic priors or recent observations without requiring historical memory. To address these ambiguities, we introduce MemInStream, a benchmark dedicated to evaluating memory capabilities in streaming video understanding. We develop an evidence-controlled validation protocol that associates each question with temporally localized evidence intervals. The protocol validates evidence solvability and memory necessity by retaining questions that models can answer from the annotated evidence but cannot resolve using recent observations, post-evidence observations, or textual input alone. This helps disentangle memory limitations from understanding failures while reducing memory-independent shortcuts. MemInStream provides a systematic evaluation of memory capabilities through four major categories, covering Episodic Retrieval, State Tracking \ Updating, Multi-Evidence Integration, and Contextual Memory across diverse video scenarios. Overall, \textscMemInStream provides a comprehensive and reliable benchmark for systematically evaluating memory capabilities in streaming video understanding.
[CV-104] WaveletLoRA: Frequency-Aware Content-Style Decomposition for Personalized Image Generation
链接: https://arxiv.org/abs/2611.15617
作者: Peiyao Wang,Jiahui Sun,Weining Wang,Jing Liu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Content-style decomposition from a single image is a fundamental challenge in personalized image generation, aiming to separate subject identity from visual style for flexible recomposition. Existing approaches primarily operate in the spatial domain, without explicitly exploiting the distinct frequency characteristics naturally associated with content and style. In this paper, we propose WaveletLoRA, a novel frequency-aware framework for explicit content-style decomposition through wavelet-domain supervision. Specifically, we first apply a multi-level discrete wavelet transform (DWT) to decompose the reference image into low-frequency and high-frequency subbands. We then assign a dedicated LoRA branch to each subband, where the low-frequency branch models global appearance for style acquisition, while the high-frequency branches capture structural details for content preservation. To further improve disentanglement quality, we introduce a dense Mixture-of-Experts aggregation module together with a frequency-domain regularization objective, which reduces style leakage and enhance content fidelity. In addition, we establish WaveBench and develop a VLM-based evaluation protocol that explicitly measures both attribute similarity and attribute separation for disentanglement assessment. Extensive experiments demonstrate that WaveletLoRA consistently outperforms existing methods and produces results that better align with human preference.
[CV-105] Learning-Based Interpolation of Detector Gaps in Photon-Counting X-ray Imaging
链接: https://arxiv.org/abs/2611.15614
作者: Johannes B. Thalhammer,Patrick Ilg,Tina Dorosti,Sebastian Peterhansl,Theresa Hiu,Daniel Frey,Zijin Huang,Florian Schaff,Daniela Pfeiffer,Franz Pfeiffer
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Image detectors assembled from tiled modules have physical gaps in the field of view that must be filled before downstream processing such as phase retrieval or tomographic reconstruction. Unlike natural-image inpainting, detector-gap interpolation must leave all other pixels unmodified and preserve the intensity statistics of the filled region, since even a small shift in mean intensity can cause ring artifacts after reconstruction. We adapted two generative models, a Wasserstein GAN with gradient penalty (WGAN-GP) and a conditional Pix2Pix model, to this task. We implemented a masking strategy that restricts all changes to the gap pixels and loss terms that constrain the mean intensity and the intensity distribution of the filled region. Both models were trained on nanoCT projections with simulated gaps and compared with linear interpolation, Navier–Stokes inpainting, and the deep image-inpainting network MADF, quantitatively on simulated gaps and visually on real detector gaps. WGAN-GP and Pix2Pix achieved the best perceptual similarity (LPIPS) and histogram distance and, apart from linear interpolation, the shortest inference time. Pix2Pix required about five times fewer training GPU-hours than WGAN-GP. Ablation studies showed that masking the network output contributes by far the most to plausible inpainting results. Applied to sinograms, Pix2Pix produced the most natural gap filling but improved the reconstruction only slightly compared with linear interpolation. The source code and the weights of the final Pix2Pix model are available at this https URL.
[CV-106] Contrastive Bootstrap Representation Learning for Label-Efficient Agricultural Image Classification
链接: https://arxiv.org/abs/2611.15610
作者: Hazem Ahmed,Ali Hamdi,Ammar Mohammed
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 12 pages, 4 figures. Accepted at ICACIN 2026
Abstract:Expert annotation remains a major bottleneck in agricultural image classification, where collecting images is often easier than obtaining reliable labels from domain specialists. Self-supervised learning can reduce this dependence by pretraining transferable representations on unlabelled imagery. Contrastive and bootstrap methods provide two established but distinct training signals, motivating an empirical study of a formulation that applies both signals within one representation path under severe downstream label scarcity. We present Contrastive Bootstrap Representation Learning (CBRL), which jointly optimises SimCLR-style NT-Xent contrastive learning and bidirectional BYOL-style exponential-moving-average target prediction within a shared online projection space. Using a ViT-L/16 backbone and 10% of the downstream labels, we report single-seed point estimates on four primary agricultural test benchmarks plus an exploratory 33-image New Plant Diseases evaluation. CBRL records the highest macro-F1 point estimate on the four primary benchmarks. On DeepWeeds, its 79.77% test macro-F1 is 32.29 points above the random-initialisation, label-only baseline and 15.78 points above the next-highest comparison point estimate; on PlantVillage, its 98.14% macro-F1 is 10.43 and 0.43 points higher, respectively. On PlantDoc, however, all methods fail to reach practically useful performance. These results identify the evaluated joint formulation as a promising candidate for label-limited transfer, but they do not establish objective synergy, statistical superiority, or robustness across stochastic runs.
[CV-107] CoCode-World: Shared 3D Proxy Representation for Multi-Player World Models
链接: https://arxiv.org/abs/2611.15602
作者: Qirui Li,Xin Zhang,Yabo Chen,Jinkun Hao,Paul L Rosin,Yu-Kun Lai,Haibin Huang,Yin Chen,Ran Yi,Chi Zhang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Multiplayer generative world models must produce observations that agree on shared entities despite different camera motions and occlusions. Explicit world representations provide spatial control, but precise modeling of individual game maps limits scene diversity, while shared geometry alone cannot ensure consistent appearance across views. We present CoCode-World, a shared 3D proxy framework that separates structural control from visual synthesis for two-player video generation. Programmable proxies specify scene layout, motion, and occlusion, while RGB references guide character and scene appearance. To coordinate visual details across players, we concatenate paired proxy videos, appearance references, and target views in a consistent spatial layout and generate both views jointly with a pretrained video model. This formulation uses the model’s native attention for cross-view communication within a single generation branch, with LoRA adaptation learning proxy adherence and appearance correspondence from paired observations. We construct a two-player GTA~V dataset that pairs synchronized RGB observations with proxy videos derived from shared map geometry and game-engine state. Experiments demonstrate appearance transfer to characters and scene styles unseen during training, with improvements in video quality and cross-view player identity over the evaluated baseline. Ablations further examine how proxy construction and cross-view communication affect generation. Project page: this https URL.
[CV-108] Cat Royale Comprehensive Dataset: A Dataset for Pose Detection Segmentation Animal Action Recognition and Scene Understanding
链接: https://arxiv.org/abs/2611.15601
作者: Zane K. J. Hartley,Joseph Ho,Alan Chamberlain,Michael P. Pound,Steve Benford
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 26 pages, 4 figures, 6 tables. Submitted to the International Journal of Computer Vision. Dataset: this https URL
Abstract:Animal focused computer vision datasets are of a growing interest in the fields of both Computer Vision and Animal behaviour. The ability to make sense of complex scenes via a variety of Computer Vision tasks has substantial implications for behavioural understanding and welfare. The perfection of these tasks will help lead to better health and emotional outcomes for wild and domestic animals. In this work we present the Cat Royale Comprehensive Dataset (CRCD), a large, comprehensive, multi-modal dataset for feline understanding, based on a huge database of over 500 hours of video captured during the Cat Royale artwork and research project. Our dataset contains annotations for a range of computer vision problems including 20,000 images with instance segmentation masks, 5000 images with multi-subject pose coordinates, over 800 video clips with action classes, and over 800 clips with detailed text description. This is the first such dataset to link this wide range of annotation types together for multi-modal use. We release this entire dataset, and our custom annotation tools will also be made available, in order to allow for ongoing expansion. We also provide a series of benchmarks for the tasks of Instance Segmentation, Pose detection, Temporal Action Localisation, and finally video-text retrieval. The totality of this multi-modal dataset is the data needed to solve both these individual computer vision tasks as well as enabling more complex multi-modal tasks making use of a combination of annotations.
[CV-109] EMS-Nav: Embodied Multi-Scale Navigation
链接: https://arxiv.org/abs/2611.15589
作者: Zenghui Yang,Xinan Wang,Danyang Li,Zhe Xu,Zheng Yang,Jingao Xu,Qiang Ma
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注: 30 pages, including appendices
Abstract:Industrial inspection robots must reach work areas and specific equipment. These destinations span spatial scales from regions to objects, so robots must relate each target to its surroundings to interpret the task and determine where to go. Existing methods support multi-scale navigation through rich environment representations, but at substantial computational and storage cost. These costs limit deployment where wireless coverage is incomplete and robots must rely on onboard computation to keep navigating. We introduce EMS-Nav, a biologically inspired multi-scale navigation framework for online, on-device operation. Its core innovation is to organize perception, spatial understanding, and navigation around regions, making local spaces the basic units of both environmental understanding and computation. EMS-Nav groups objects and their relationships into functionally meaningful spaces and builds a hierarchy from local areas to the broader environment. This organization supports efficient map construction alongside the interpretation and execution of language goals at different scales. In multi-scale navigation experiments, EMS-Nav achieves the highest success rates across all four target categories: objects, rooms, regions, and specific instances. It outperforms state-of-the-art open-vocabulary mapping and navigation methods, improving overall first-attempt success by 9.39 percentage points over the strongest baseline while reducing mapping overhead and language-model input size. Local deployment in real indoor environments further demonstrates its applicability to on-device embodied navigation. By establishing regions as the organizing basis for perception, spatial understanding, and language-guided action, EMS-Nav provides a system that combines multi-scale understanding with computational efficiency to support diverse language goals on resource-constrained robots.
[CV-110] FoveaDet: Software-Hardware Co-Design of Foveated CNN-Transformer for Object Detection ISCAS2027
链接: https://arxiv.org/abs/2611.15579
作者: Deepak Mewada,Apporva Saraf,Ashwin Lele
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 5 pages, 6 figures. Submitted to IEEE ISCAS 2027
Abstract:Deploying real-time object detection on power-constrained edge platforms is challenging as processing full-resolution frames is often computationally redundant. Inspired by biological foveation which assigns variable resolution based on visual saliency, we propose a novel software-hardware co-design framework called FoveaDet. On the software side, FoveaDet dynamically identifies salient regions to predict fixation points, followed by aggressive patch pruning in Vision Transformers (ViT) to discard irrelevant background data. To mitigate slowdowns from sequential patch generation, we implement optimized interpolation and patch selection algorithms. Evaluated on standard GPUs, this foveated approach reduces floating-point operations by 65.4% to 92.2% and accelerates inference by 4.76x, while suffering a negligible mAP drop compared to dense baselines. To eliminate hardware bottlenecks, we complement the algorithm with a custom RISC-V core featuring specialized instructions to accelerate the foveated processing pipeline. FoveaDet provides a scalable, efficient architecture capable of unlocking advanced vision processing on severely constrained mobile platforms.
[CV-111] RipBench: A Unified Benchmark for Multi-Level Rip Current Detection Classification and Segmentation NEURIPS2026 DATE
链接: https://arxiv.org/abs/2611.15574
作者: Andrei Dumitriu,Aakash Ralhan,Florin Miron,Florin Tatui,Radu Tudor Ionescu,Radu Timofte
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to NeurIPS 2026, Evaluations Datasets Track. 47 pages, 11 figures. See this https URL for dataset and project updates
Abstract:Rip currents are a serious and often under-addressed threat to beach safety and the leading cause of coastal drownings worldwide. They are difficult to detect due to their amorphous structure, similarity to the background, and large variability in viewpoints and environments. Despite their dangers and growing interest in automated detection, existing research remains fragmented across datasets, metrics, models, and approaches, leading to an incomplete understanding of model performance. We introduce RipBench, a unified benchmark that enables controlled evaluation of rip current detection spanning multiple tasks, including classification, axis-aligned and oriented object detection, and instance, semantic, and panoptic segmentation, on the same data with standardized splits. This allows direct comparison of model performance on all levels of visual abstraction. Across 304 videos ( 303,491 frames, 51,011 of them annotated for all six tasks) collected from diverse coastlines, our results expose a clear performance gap between coarse recognition and precise spatial understanding: frame-level classification reaches F_2 = 0.93 , while the best detector and the best instance segmenter reach 0.72 and 0.66 , with model rankings that change from task to task. Video-level confidence intervals, a blind inter-annotator study of the spatial labels, stratified analyses by viewpoint and visual signature, and video views with temporal baselines (single-seed, short-schedule reference runs) accompany the benchmark. All tasks are supported with carefully curated annotations and evaluated using standard metrics, with a focus on the F_2 score to emphasize recall in this safety-critical setting. RipBench, along with multiple baselines per task, is publicly available at this https URL to support progress in real-world, multi-task vision for beach safety.
[CV-112] raceKV: Policy-Preserving Low-Rank KV Cache Compression for Reasoning Vision-Language Models
链接: https://arxiv.org/abs/2611.15569
作者: Ke Li,Zishan Shao,Yiran Chen
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Reasoning vision-language models (VLMs) impose substantial KV cache demands from both large multimodal contexts and growing reasoning traces. Low-rank KV compression has emerged as a promising paradigm that stores low-dimensional latents of KV states and can be combined with quantization. However, existing low-rank methods often induce growing predictive policy drift from the dense model when compressing reasoning KV states, resulting in reasoning failures. We identify two key causes of this degradation: calibration-state mismatch between static contexts and on-policy reasoning states, and unequal error sensitivity across KV states. To address these challenges, we propose TraceKV, a training-free low-rank KV compression framework that unifies on-policy calibration, sensitivity-aware SVD, and global rank allocation under a single policy-preserving principle, substantially reducing policy-KL divergence during reasoning. Experiments across multiple VLMs and visual reasoning benchmarks demonstrate superior accuracy–memory trade-offs, achieving near-lossless reasoning accuracy with 5.64 \times KV cache compression when combined with quantization. Optimized low-rank attention kernels further deliver up to 1.83 \times end-to-end decoding speedup. Code will be available at this https URL.
[CV-113] MAGIC: Modality-Aware Gini Inspection for Clinical Hallucination Detection in LMMs
链接: https://arxiv.org/abs/2611.15530
作者: Sachin Acharya,Niyoj Oli,Shishir Adhikari,Aminu Lawal,Yash Raj Shrestha,Prashnna K Gyawali,Binod Bhattarai
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Recent advances in medical large multimodal models (LMMs) have brought promising progress to clinical decision support, yet hallucination, the generation of confident but clinically unverifiable outputs, poses a critical threat to the reliability demanded in high-stakes medical settings. Existing hallucination detection methods, predominantly designed for unimodal settings, fail to capture the visual-linguistic dynamics inherent to medical imaging. We formulate the problem of hallucination detection in LMMs as disproportionate distribution of attention weights across the modalities during answer token generation. We quantify modality-specific attention concentration across layers and heads using the Gini coefficient and demonstrate statistically significant redistribution of attention mass between visual and textual tokens in hallucinated responses. We leverage these features to train a logistic regression-based classifier and evaluate it on two medical LMMs across three medical imaging modalities, demonstrating that our method achieves competitive or superior performance against contemporary hallucination detection baselines.
[CV-114] SCR-Mamba: Skeleton-conditioned Mamba with cross-scale incremental refinement for efficient 3D human pose estimation
链接: https://arxiv.org/abs/2611.15526
作者: Jie Wang,Xuejing Gu,Congzhe Zhang,Zibo Zhao
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 15 pages, 9 figures, 11 tables
Abstract:Mamba-based methods for video-based 3D human pose estimation can model long sequences with linear complexity, but their selective state updates usually lack direct exploitation of skeletal relations. Meanwhile, stacking modeling blocks at the full temporal resolution leads to high computational overhead, while compressing the temporal sequence makes it difficult to retain per-frame details. To address these issues, we propose SCR-Mamba, whose core consists of Skeleton-Conditioned State Modeling (SCSM) and Cross-Scale Incremental Refinement (CSIR). SCSM generates a skeleton context from bone connections and left-right symmetry relations and generates the selective parameters of the spatial and temporal branches conditioned on it, so that structural relations are involved in the writing, propagation, and reading of state information. CSIR constructs multi-scale motion representations through content-adaptive aggregation and progressively refines the retained per-frame features with the increments produced by coarse-scale modeling, integrating long-range motion context while reducing full-resolution computation. On Human3.6M with Stacked Hourglass (SH) 2D detections and 243-frame inputs, SCR-Mamba-L achieves 36.8 mm MPJPE (P1) with 7.72 M parameters and 15.10 G MACs, less than 1/7 of the computational cost of StructMamPose-L with similar accuracy; SCR-Mamba-X further reaches 36.6 mm P1 and 30.9 mm P-MPJPE (P2), achieving state-of-the-art performance. On MPI-INF-3DHP, SCR-Mamba-X achieves a percentage of correct keypoints (PCK) of 99.3%, an area under the curve (AUC) of 88.1%, and an MPJPE of 14.2 mm.
[CV-115] Reading the Table: Sim-to-Real Game State Detection for the Chefs Hat Card Game ICANN2026
链接: https://arxiv.org/abs/2611.15524
作者: Connor Gäde,Matthias Kerzel,Hauke Bünning,Jan-Gerrit Habekost,Mariana Barros,Stefan Wermter
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注: Presented at the Dungeons, Neurons, and Dialogues workshop (DnD-SIDC 2026) at ICANN 2026, this https URL
Abstract:Tabletop games offer a compelling scenario to study the interaction between artificial and human agents, combining social interaction with game state tracking and physical manipulation. To assess the game state, a robot needs to precisely detect closely related game objects, such as different card designs, requiring dedicated training data that is costly to annotate in the real world. Instead, we develop a fully automated synthetic data collection pipeline, using re-projection of known 3D coordinates to generate labels automatically and domain randomization to bridge the reality gap. This approach is used to collect a dataset of 25413 samples to finetune an object detection and a segmentation model, which are combined to assign detected cards to game-relevant zones and infer the game state. When evaluated on a real NICO humanoid robot, our pipeline achieves an F1-score of 0.906 for card-to-zone assignment, thus providing the perceptual foundation for a robot that can autonomously play Chef’s Hat against human players.
[CV-116] Real Pairs Are Not Enough: Physically Anchored Supervision for Volumetric CT Super-Resolution
链接: https://arxiv.org/abs/2611.15514
作者: Haoran Liu,Zipei Yan,Mingzhe Liu,Jizhou Li,Guibin Zan
类目: Computer Vision and Pattern Recognition (cs.CV); Image and Video Processing (eess.IV)
备注: 47 pages, 5 figures, including supplementary material
Abstract:Volumetric CT super-resolution (SR) models are commonly trained on synthetic or physically acquired pairs of low-resolution (LR) and high-resolution (HR) volumes. However, synthetic pairs miss real acquisition effects, whereas registered physical pairs retain inter-scan noise, artifact, and microstructure differences—causing direct voxelwise regression, and even naive synthetic-real mixing, to oversmooth fine porous structures. We introduce the Physically-anchored Implicit Network (PIN) with Real-Pair Dual-Anchor learning, which decouples physical-pair supervision into two complementary roles. The HR anchor teaches fine structure from registered HR through pair-conditioned spectral replay, while the LR anchor constrains the prediction from acquired LR via affine-calibrated low-pass projection. Together with exact synthetic supervision in one shared volumetric implicit model, PIN learns from both physical scans without forcing the acquired-LR reconstruction to match registered HR voxel by voxel. Inference requires only acquired LR. Across 30 external baseline configurations on real CT, PIN achieves the lowest full-volume LPIPS, reduces average symmetric surface distance by 42.3-49.5% and bone-volume-fraction error by 52.2-53.4% relative to the three strongest baselines for each metric, and lowers the erased reference-pore volume fraction from 37.3-72.9% to 10.4%. Code, pretrained models, and evaluation scripts will be publicly available.
[CV-117] What Makes a VLA Reactive? A Study of Execution Strategies for Dynamic Manipulation
链接: https://arxiv.org/abs/2611.15491
作者: Zhuoyue Li,Zhendong Han,Pankhuri Vanjani,Wenting Li,Rudolf Lioutikov
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 13 pages, 7 figures, Efficient Foundation Models for Real-Time Embodied AI, CoRL 2026 Workshop
Abstract:Vision-language-action (VLA) models predict chunks of future actions, and when objects move, those chunks go stale before they are executed. We study what makes a pretrained VLA reactive, using \pihalf on two dynamic simulation benchmarks and three real-robot tasks. We vary the prediction and execution horizons, compare synchronous and asynchronous inference, test methods that deliver actions sooner or keep successive chunks continuous, and adapt the policy with motion cues and latent-space reinforcement learning. The best prediction horizon is about one second, and the best execution horizon depends on the checkpoint. Short horizons help more with asynchronous execution, and methods that keep the trajectory continuous outperform those that only reduce latency. Adapting the policy repairs targeted failures, but on the conveyor task changing the execution horizon gained as much. Reactivity depends on the execution pipeline as much as on the policy, so both should be evaluated together.
[CV-118] SkillEarth: Teaching General-Purpose Agents to Act on Earth via Skill Evolution
链接: https://arxiv.org/abs/2611.15481
作者: Haojun Tang,Wenda Zhao,Xiang Zhou,Qiwei Lin,Lida Huang,Yan Wang
类目: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注: 20 pages, 7 figures
Abstract:Recently developed remote sensing (RS) agents execute real-world tasks with complex workflows through multi-step tool invocation. These agents, however, typically rely on predefined expertise or parameter updates. They thus require updating expert knowledge or retraining models whenever new tools, tasks, or execution patterns emerge, limiting their continual adaptability in dynamic Earth environments. In this work, we seek to equip general-purpose agents with RS tool-call knowledge that can be automatically extracted from demonstrated trajectories without expert intervention or parameter updates. Specifically, we propose SkillEarth, a self-evolving skill extraction method that externalizes RS tool-call knowledge as reusable, retrievable, and evolvable skill files and refines them through observation-grounded evolution. Unlike existing automatic skill extraction methods that rely on outcome-based reflection to refine skills holistically, SkillEarth grounds skill evolution in real tool-chain execution and combines dual-judge feedback with a grounded observation routing graph for precise component-level refinement. The resulting skills are generalizable, reliable, and transferable: they improve overall RS tool-call performance across three frontier LLMs and four open-source LLMs, are refined through real tool execution feedback, and support cross-model reuse without retraining. Experiments on OpenEarth-Bench and Earth-Bench demonstrate that SkillEarth achieves better tool selection, argument correctness, execution-order consistency, and task completion than existing RS agents and skill extraction methods.
[CV-119] Fear of Memorizing Objects: End-to-End Zero-Shot Model-Based Object Detection
链接: https://arxiv.org/abs/2611.15459
作者: Leon Greiner,Wout Boerdijk,Junwen Huang,Benjamin Busam,Rudolph Triebel,Maximilian Durner
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:This paper presents FOMO (Fear of Memorizing Objects), an end-to-end trainable transformer-based architecture for zero-shot model-based object detection. Existing approaches typically follow a proposal-and-match pipeline: target-agnostic mask proposals are generated and then matched to template features in a set of loosely coupled stages, making them susceptible to fragmented proposals in clutter and degraded matching under occlusions. FOMO elegantly replaces this paradigm by injecting object template information early in the detection pipeline: Bidirectional attention between learnable latent tokens, multi-view CAD templates, and query image features produces object-aware image features and conditioned queries which are subsequently decoded into bounding-box and confidence predictions. We train FOMO on Clutter6D, a novel large-scale synthetic dataset with over 500k images featuring more than 10k diverse object models rendered in highly cluttered scenes with multiple instances and varying occlusion levels, which addresses the limited object diversity and lack of realistic dense clutter in existing datasets for training zero-shot multi-instance algorithms. In true zero-shot fashion, FOMO achieves 54.2 mAP on BOP-Classic-Core at 0.23 s per image, ranking second overall with 3.8 pp difference to the top method while being more than 7x faster. Throughout comprehensive ablation studies we investigate and verify architectural design choices.
[CV-120] WaveFlow-UIE: A Single-Pipeline Wavelet-Domain Flow Model for Efficient Underwater Image Restoration
链接: https://arxiv.org/abs/2611.15457
作者: Abdulaziz Alfaraj,Mohammed Al Naser,Hassan Al Nasser,Muzammil Behzad
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Underwater image restoration remains challenging because color distortion, scattering, and detail loss affect low- and high-frequency image components differently. Recent diffusion-based methods have shown strong restoration quality, but often rely on multi-stage pipelines and costly iterative refinement. We present WaveFlow-UIE, a single-pipeline wavelet-domain flow model that directly learns to transform degraded underwater wavelet representations into clean ones. Our method applies a discrete wavelet transform, estimates lightweight physics-inspired degradation cues including transmission and ambient light, and conditions a velocity network trained with flow matching in wavelet space. The training objective combines flow-matching, frequency-weighted reconstruction, perceptual, color-consistency, and physics-consistency losses. Across UIEB, LSUI, UFO-120, and EUVP, WaveFlow-UIE achieves a strong quality-efficiency tradeoff, with especially competitive results on perceptual metrics. Relative to a strong wavelet-based diffusion baseline, the proposed model substantially reduces inference cost while maintaining or improving restoration quality on several paired benchmarks. We show that reducing the inference process from five flow steps to a single step preserves nearly all of the restoration quality while providing an additional speedup of about 4\times on UIEB at 256\times256 resolution. This single-step result suggests that a single-pass wavelet-domain model is an effective alternative to heavier diffusion-style refinement.
[CV-121] Reading the Unreadable: Text-Aware Image Super-Resolution Needs Reasoning NEURIPS2026
链接: https://arxiv.org/abs/2611.15449
作者: Jaeseong Lee,Jinwoo Kim,Jinho Jeong,Seon Joo Kim
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Accepted at NeurIPS 2026
Abstract:Text-aware image super-resolution (TAISR) aims to recover high-resolution images from low-resolution inputs while preserving textual content. Existing methods treat this as a visual restoration problem, yet severely degraded text often cannot be read from local visual evidence alone. Humans instead read it by reasoning over image context, logical patterns, and prior knowledge. This raises a fundamental question: Does TAISR need reasoning? To answer this, we introduce two diagnostic tools. The first, ReasonText, is a human-curated benchmark that separates reasoning-required text from locally recoverable text. The second, Ground-Truth Text-Crop Accuracy (GTTCA), is a recognition-based metric that measures text restoration quality without text detection errors. Evaluating twelve recent super-resolution (SR) models, we find a consistent and substantial performance drop on reasoning-required text, revealing a systematic blind spot. Two observations point to a remedy. One is that text-conditioned SR models restore text substantially better when given ground-truth text. The other is that multimodal large language models (MLLMs) can read degraded text in low-resolution images through reasoning. We therefore propose Reasoning Transfer via Captioning (RTC), a training-free plug-in that uses an external MLLM to infer the degraded text and feeds the resulting caption to a text-conditioned SR model. RTC consistently improves all six text-conditioned SR models we evaluate, especially on reasoning-required text. Surprisingly, a real-world SR model with RTC can even outperform the TAISR model with RTC, suggesting that reasoning should be a core design axis for future TAISR.
[CV-122] Semantic Consistency of Vision Tokens: A Vision-Centric Perspective on Multimodal Large Language Models
链接: https://arxiv.org/abs/2611.15448
作者: Nikola Đukić,Tim Lebailly,Tinne Tuytelaars
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:In computer vision, vision-language and self-supervised encoders are two unique paradigms characterized by global language alignment and high-quality vision-only patch representations, respectively. Multimodal large language models (MLLMs) often use the former because of language alignment, despite the recent rise in vision-centric tasks resembling traditional dense vision tasks that would likely benefit from the high-quality patch representations, raising the question whether self-supervised models could aid MLLMs in vision-centric tasks. To investigate this, we define semantic (in)consistency of patch tokens from language-aligned encoders and identify two sets of tokens, semantically consistent (SCTs) and semantically inconsistent (SITs). Through systematic probing via the removal of SCTs or SITs from the input of modern MLLMs, we show higher reliance on SCTs than SITs, as removal of the former induces larger performance drops. This demonstrates that MLLMs benefit from semantic consistency of vision tokens. With this in mind, we devise a simple yet effective extension of the MLLM training objective that combines traditional language modeling with semantic guidance via minimization of semantic inconsistency. This strategy, named Semantically-Guided Training (SGT), combines the signals of both loss terms to push the vision encoder towards providing semantically consistent tokens, while preserving language alignment. Extensive evaluation shows that SGT outperforms all baselines and MLLMs with specialized vision encoders on vision-centric benchmarks while keeping strong scores on text-centric benchmarks, thus setting state of the art on average across all tasks. Our results highlight the importance of combining language alignment and semantic consistency through the lens of MLLMs rather than independently from them and pave the way towards the design of better MLLM vision encoders.
[CV-123] SteinPnP: Non-Parametric Uncertainty Quantification in the Perspective-N-Point Problem
链接: https://arxiv.org/abs/2611.15437
作者: Baptiste Lardinoit,Jongseok Lee,Andres Yubero,Seong Hun Lee,Javier Civera,Rudolph Triebel
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:We propose SteinPnP, a probabilistic Perspective- N-Point (PnP) solver that models the full 6-DoF camera pose posterior using a non-parametric set of pose particles. By lever- aging Stein Variational Inference (SVI), SteinPnP combines the expressivity of particle-based representations with the efficiency of optimization-based solvers. However, a naive application of SVI struggles in the PnP problem, often missing alternative pose hypotheses or assigning incorrect probability weights to them. To address these limitations, we propose practical remedies that discover new potential poses, identify distinct hypothesis clusters, and balance their relative probability weights to capture the true underlying distribution. Our experiments across synthetic and real-world benchmarks demonstrate that SteinPnP accurately captures complex multimodal uncertainties while maintaining competitive accuracy.
[CV-124] hermalSplat: Generalizable Pose-Free Gaussian Splatting for Sparse-View Thermal 3D Reconstruction WACV2027
链接: https://arxiv.org/abs/2611.15435
作者: Prasanna S,Gopi Raju Matta,Divya Madhuri Vemunuri,Kaushik Mitra
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: WACV 2027
Abstract:Thermal infrared imaging is vital for applications like search-and-rescue and autonomous navigation, yet trans- lating it into 3D scene representations remains a critical bottleneck. The low-texture nature and view-dependent emissivity of thermal images cause standard pose estima- tors to fail catastrophically, collapsing standard 3D Gaus- sian Splatting pipelines under sparse-view constraints. Fur- thermore, existing thermal reconstruction methods rely heavily on dense capture and computationally expensive per-scene optimization. Critically, many of these ap- proaches depend on aligned RGB-thermal datasets to ob- tain reliable camera poses and initial point clouds, ren- dering them unusable in purely thermal, unposed envi- ronments. To break this bottleneck, we present Thermal- Splat, the first generalizable, feed-forward Gaussian splat- ting framework that reconstructs 3D scenes directly from sparse thermal images without requiring known camera poses, RGB guidance, or per-scene training. We over- come thermal-specific challenges using a soft census loss for brightness-invariant structural matching and a bound- ary structure loss to explicitly supervise temperature discon- tinuities. Trained on an aggregated multi-source dataset, ThermalSplat synthesizes sharp reconstructions of unseen scenes in a single feed-forward pass. Extensive experi- ments demonstrate that ThermalSplat holds a strict advan- tage over scene-specific thermal splatting methods in the sparse regime. Furthermore, even compared against these baselines operating under their ideal dense capture set- tings, ThermalSplat achieves superior perceptual fidelity. By eliminating the need for computationally expensive per scene optimization, ThermalSplat offers an efficient and generalizable approach to thermal 3D reconstruction.
[CV-125] Verifier Choice is a Benchmark Design Variable: Auditing Structural Counting Evaluation in Text-to-Image Models NEURIPS2026
链接: https://arxiv.org/abs/2611.15433
作者: Shurun Li,Michael Hua Wang,Haibo Zeng,Long Wang
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: Accepted at NeurIPS 2026 (Evaluations and Datasets Track). 24 pages. Data: this https URL . Code: this https URL
Abstract:Automatic evaluation is now central to text-to-image benchmarking, but the verifier that converts generated images into scores is often treated as an implementation detail. We argue that this is unsafe: an automatic benchmark score is a property of the image generator, prompt distribution, verifier, verifier-use protocol, and scoring rule, not of the generator alone. We study this issue in structural counting, a controlled setting where human-visible counts can be audited. We construct a verifier-centric audit suite over seven counting objects: cell phones, chairs, and traffic lights for object counting, and clover leaflets, dresser drawer fronts, power-strip sockets, and turbine blades for substructure counting. Across 6,240 generated images from Qwen-Image-2512 and FLUX.2-dev, we compare open MLLM verifiers under count-only, reject-aware, zero-shot and few-shot protocols, plus TIFA-like target queries, a detector baseline, and frontier closed-model checks on two hard objects. On a balanced 2,280-image evaluation split, aggregate score matching can coexist with image-level disagreement, and verifier-use protocols and generator distributions change the measured reliability. Every verifier-protocol configuration also shrinks the human score difference between the two generators, by 7.5-25.0 pp, and with a third generator, Z-Image, all four tested protocols reverse the human ordering of Z-Image and FLUX.2-dev on our fixed prompt suite. GPT-5.5 and Claude Opus 4.7 checks do not remove the failure modes: GPT-5.5 still has only 59.92% clover valid-count fidelity, while Claude Opus 4.7 has lower same-slice agreement ( \kappa=0.432 ). We propose reporting automatic benchmark scores with a factorized verifier reliability profile: aggregate score gap, image-level agreement, generator-comparison distortion, valid-image count fidelity, rejection behavior, output compliance, and directional bias.
[CV-126] IncR-3DSGG: Incremental Posed RGB Image Based 3D Scene Graph Generation via Privileged Depth Supervision
链接: https://arxiv.org/abs/2611.15429
作者: Mrunmai Phatak,Daniel Kienzle,Roman Küble,Jörg Hähner,Rainer Lienhart
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Scene graphs provide a compact, structured representation of environments by capturing objects and their spatial or semantic relationships in a form suitable for embodied agents. Existing 3D scene graph generation methods often rely on depth sensing at inference, either to associate object instances across views via point cloud overlap or to compute inter-object relations from geometric cues, limiting deployment to platforms with depth sensors. We propose IncR-3DSGG, which builds 3D scene graphs incrementally from posed\footnoteRGB images with camera extrinsics estimated by an off-the-shelf visual SLAM system. RGB images, eliminating the use of depth sensors at inference. Our method leverages a Parametric Relation Predictor (PRP) that learns to estimate pairwise distances and relative angles between objects from 2D visual features and 2D geometric cues, using 3D geometric supervision available only during training. These learned pairwise measurements are fused across views through Pose Graph Optimization (PGO)~\citegrisetti2010tutorial to recover metrically grounded 3D object positions and build a globally consistent 3D scene graph incrementally. On the real-world 3RScan~\citewald2019rio dataset, our method attains 52.2 mean object recall and 30.8 mean predicate recall under the 3DSSG~\citewald2020learning evaluation protocol, against 27.0 and 17.6 for MonoSSG~\citewu2023incremental, while using no depth or reconstructed geometry at inference. This shows that relations and metrically grounded 3D object representations can be predicted from posed RGB images alone.
[CV-127] Learning Motion-Appearance Coupling Priors for Solving Video Inverse Problems NEURIPS2026
链接: https://arxiv.org/abs/2611.15420
作者: Anselm Krainovic,Reinhard Heckel
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG); Image and Video Processing (eess.IV)
备注: NeurIPS 2026
Abstract:Reconstructing video from degraded measurements requires priors that capture both frame appearance and temporal dynamics. Many recent approaches combine pretrained image diffusion priors with motion guidance, sidestepping the costs of training large video models. Temporal consistency is often enforced via photometric losses or warping-based constraints, which implicitly assume that motion explains frame-to-frame or latent-to-latent changes up to small deviations, an assumption that fails under occlusions, lighting changes, and non-rigid deformation. In this paper, we propose to learn the coupling between motion and frame appearance as a generative prior. We learn the distribution of motion-induced appearance residuals with a conditional diffusion model and use it to regularize reconstruction in video inverse problems. Our motion-appearance prior acts as a plug-in regularizer for diffusion-based video reconstruction, preserving the benefits of image diffusion models while remaining compatible with existing appearance regularization techniques. Experiments on dynamic video inverse problems show that reconstruction with a learnt motion-appearance coupling prior outperforms photometric-consistency and noise-warping baselines, and that the approach also transfers to cardiac MRI reconstruction.
[CV-128] ChartGalaxy: A Richly Annotated Dataset for Chart Understanding and Generation
链接: https://arxiv.org/abs/2611.15418
作者: Duan Li,Zheng Wang,Jiangning Zhu,Yuhui Yuan,Shixia Liu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Infographic charts communicate data through cohesive compositions of visual elements such as charts, text, and images. Existing infographic chart datasets primarily annotate individual elements with bounding boxes and category labels. Such element-level annotations do not capture hierarchical and spatial relationships, limiting their use for understanding how elements are organized and related within an infographic and for generating well-organized designs. To address this limitation, we introduce ChartGalaxy++, a richly annotated dataset of 217,195 infographic charts. At the core of this dataset is a scene graph representation, in which nodes represent visual elements and semantic groups, and edges encode their hierarchical and spatial relationships. For real infographics, we construct scene graphs by detecting and progressively grouping visual elements; for synthetic ones, we derive them directly from the generation process. We demonstrate the utility of ChartGalaxy++ through three applications: 1) image-to-scene-graph prediction, 2) scene-graph-augmented infographic question answering, and 3) evaluation of scene graph preservation in image generation. Together, these applications highlight the value of ChartGalaxy++ for advancing chart understanding and generation.
[CV-129] CARE-Drive: Coupled Alternating Refinement for Ego-trajectories
链接: https://arxiv.org/abs/2611.15360
作者: Xiao Ma,Yijun Chen,Zhiquan Hu,Jicheng Zhao,Yuming Li,Chuang Dai
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Vision-language-action (VLA) models bring multimodal understanding and world knowledge to end-to-end autonomous driving. To further improve the quality of predicted trajectories, existing studies introduce candidate refinement, residual denoising, and token correction. However, these methods typically focus on unified refinement of the entire trajectory, paying limited attention to the distinct physical quantities and their interactions. We further investigate physical specialization and coordination in trajectory refinement, centering on position and heading for simpler and more intuitive correction modeling. To this end, we propose CARE-Drive (Coupled Alternating Refinement for Ego-trajectories), which decomposes trajectory refinement into waypoint residual correction and position-conditioned heading generation. Through imitation learning and a future-token auxiliary task, the backbone provides predictive representations and a base trajectory. The position expert predicts bounded metric waypoint residuals, while the heading expert reconstructs vehicle orientation from adjacent angular increments conditioned on the corrected path. With the backbone frozen, the two experts are optimized alternately through supervised initialization and reinforcement fine-tuning. This design simplifies refinement through clear physical specialization while preserving conditional coupling during prediction and optimization. On NAVSIM v1, CARE-Drive achieves 88.8 PDMS, improving by 2.4 points over the unrefined base trajectories generated by the same backbone. Ablation studies further support the effectiveness of specialized experts and staged training.
[CV-130] CaseScope: A Multi-Field Benchmark for Case-Level Pathology Diagnosis with Multimodal LLM s
链接: https://arxiv.org/abs/2611.15325
作者: Wenchuan Zhang,Chengyang Zhang,Fengling Li,Ting Yin,Chen Shu,Jie Chen,Shuwan Zhang,Xiaoyan Wu,Zhang Zhang,Yuhao Yi,Jiancheng Lv,Hong Bu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Pathologists rarely diagnose from a single image: they scan a section at low power for architecture, zoom into selected fields for cytology, and integrate the evidence across views and scales. Pathology benchmarks for multimodal large language models (MLLMs), however, mostly ask about one image or require whole-slide pipelines, and the few that show several fields draw nested regions from cancer slides of one public cohort. We introduce CaseScope, a benchmark of diagnosis from several curated fields of one lesion: CaseScope-Teaching, 436 cases from a residency-training image library, reconstructed, de-identified, harmonised against a WHO-based ontology and re-examined by a pathologist wherever an automated pre-review raised a doubt, and CaseScope-Atlas, 800 cases from 88 published atlases (387 of them figure series of one disease), labelled from the captions by two LLMs that must agree. They cover 843 entities across 18 organ systems and pose 4,679 items in six tasks, from open-ended and ten-way diagnosis with ontology-guided hard negatives under circular evaluation to the verification and location of morphological findings. Across twenty-eight open and four proprietary MLLMs, diagnoses are unstable – the best open and proprietary models are correct under all ten option rotations on 17.9% and 61.7% of the teaching cases – and wrong answers favour the classic look-alike; open models make little use of the additional fields on teaching cases, where the one-line specimen site matters more, although on atlas cases with three or more fields the fields count about as much as the site; and most open models verify findings barely above chance (median 62.0% vs. 50%, best 81.5%), gaining a median +3.4 points from images; proprietary models reach 83.2-93.2%. Rankings broadly replicate between the sources. The evaluation code is available at this https URL.
[CV-131] AcrossPlace: Do Vision–Language Models Recognize Places Beyond Their Source Views? A Comprehensive Benchmark
链接: https://arxiv.org/abs/2611.15314
作者: Zhenyu Li,Tianyi Shang,Shuhan Dong,Ruotong Xu,Qingyu Zhu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Language-guided place recognition requires recovering geographic identity across visual observations beyond the image from which a description originates. However, aggregate retrieval scores often entangle observation transfer, gallery competition, and linguistic evidence, obscuring the underlying recognition capability. We introduce AcrossPlace, a provenance-aware benchmark establishing a systematic evaluation standard for bidirectional language–vision place recognition. Built upon GSV-Cities, AcrossPlace comprises 56,726 places across 23 cities, 156,800 automatically generated, rule-filtered descriptions, and 113,404 target images. Each description retains its source-image provenance, while retrieval operates on disjoint target observations. Text-to-image (T2I) and image-to-text (I2T) retrieval share a unified place-identity relevance criterion supporting multiple valid matches. Two geographic protocols evaluate unseen places within represented cities (P0) and generalization to held-out cities (P1). Paired observation comparisons, gallery-controlled evaluations, fixed-query scaling, and blinded annotation auditing systematically characterize retrieval performance. Across three training seeds, place-supervised SigLIP2 T2I Recall@1 decreases from 19.69% on source images to 12.31% on independent targets under P0. For fixed P1 queries, expanding the gallery from 500 to 8,425 places reduces Recall@1 from 14.67% to 2.87%. A 240-description audit further reveals discrepancies between automated acceptance and factual evidence. These findings establish AcrossPlace as a reproducible evaluation standard for assessing transferable place identity across visual observations and geographic environments.
[CV-132] Arbor: When Labels Are Not Coarsened at Random
链接: https://arxiv.org/abs/2611.15311
作者: Faisal Mohammad,Mohamed Younis
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 26 pages, 7 figures, 25 tables. Appendices included
Abstract:Merging datasets labelled at different levels of a taxonomy leaves some images with only a coarse label, and some classes with no fine-grained label at all. Standard training scores a coarse label by the total probability of the classes under it. Coarse labels name superclasses, and the standard objective is exact only when every class within a superclass is equally likely to receive a coarse label. We call a class novel if it is never labelled finely, and known otherwise. A novel class violates the assumption, and the classifier assigns its images to a known sibling. We propose Arbor, which models how labels become coarse. In each superclass with a novel class, Arbor adds a virtual novel class, and a logistic gate divides the superclass’s probability between it and the known classes. The gate is trained with the likelihood of the observed labels under this model, the coarsening likelihood. We prove that at its population optimum the gate equals the Elkan-Noto correction of positive-unlabelled learning, and show that training the correction into the classifier improves on applying it after training. Across Cars, CUB and Aircraft with 90% coarse labels, Arbor correctly assigns 76-85% of novel-class images to their superclass, more than hierarchical classifiers, hierarchical novelty detection and the Elkan-Noto correction. Its top-1 accuracy over the leaf classes on known-class images is within 1.6 points of the best baseline, although its final predictions return the superclass more often. On the more challenging iNaturalist-2019 benchmark, Arbor improves novel-class accuracy from 46% (uniform target, the strongest baseline) to 68%.
[CV-133] Beyond the Explicit: Benchmarking and Detecting Implicit Harm in Short Videos
链接: https://arxiv.org/abs/2611.15306
作者: Chenyang Li,Qifan Yu,Saihui Hou,Zhaofeng He
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 9 pages, 4 figures, 8 tables
Abstract:Short videos have become a dominant online medium, where harmful content is increasingly conveyed in implicit and fragmented forms. Prior work largely focuses on explicit categories, such as violence and hate speech, and often fails to detect videos that subtly promote misleading values through narrative context, emotional framing, or cross-modal cues, partly due to the lack of dedicated benchmarks for implicit harm. To bridge this gap, we construct DeepHarm-7K, a large-scale harmful short-video dataset comprising 7,110 samples annotated according to a fine-grained harmful-content taxonomy. The dataset systematically covers implicitly harmful videos across diverse real-world scenarios and is constructed through a rigorous multi-stage quality-control process. Building on this dataset, we propose DeepHarm-VL, a multimodal detection framework that integrates visual, audio, and cross-modal reasoning. The framework employs a two-round reasoning strategy to progressively identify surface-level evidence and infer latent harmful implications. Experimental results show consistent improvements over strong baselines and state-of-the-art methods across multiple open-source video-language models, demonstrating its effectiveness in detecting both explicit and implicit harmful short videos.
[CV-134] rack4D: Representing Dense 3D Tracking for Video Diffusion Models NEURIPS2026
链接: https://arxiv.org/abs/2611.15287
作者: Yushi Lan,Zeren Jiang,Kelvin Li,Xingang Pan,Chuanxia Zheng,Andrea Vedaldi
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: NeurIPS 2026. Project page: this https URL
Abstract:We study how to repurpose video diffusion models (VDMs), pretrained on millions of videos, for dense 3D point tracking, where training data is scarce. Doing so is non-trivial, as it requires representing 3D tracks as video-like signals that may lie outside the natural-video distribution on which the VDM was trained. We thus systematically assess the effectiveness of two such video representations. The first, Residual 3D Tracking Video (RTV), encodes per-pixel 3D offsets relative to a query frame and allows recovery of 3D trajectories directly from the generated video. The second, Normalized Coordinate Map Video (NCMV), labels each pixel with a canonical coordinate invariant through time. This is closer to a natural video and easier for the VDM to learn, but requires additional post-processing with an off-the-shelf 2D tracker to recover 3D tracks. RTV tracks more accurately, while NCMV shows that a VDM can generate dense correspondence as a video. We instantiate both representations with LoRA adaptation of a frozen Wan backbone, conditioned on point maps from an off-the-shelf 3D reconstructor as a geometric scaffold. Trained with a single dynamic 3D tracking dataset, Track4D achieves competitive zero-shot tracking on four real and synthetic benchmarks. A regression readout that bypasses the video VAE and uses additional training data brings further improvement, demonstrating the potential of this direction for dense 3D tracking. We hope our analysis will inspire future work that leverages VDM priors for dense 3D tracking and 4D reconstruction. Project page: this https URL
[CV-135] PCLSegNet: Anatomy-Aware Deep Learning Framework for MRI-based Pancreatic Cyst Segmentation and Size Assessment
链接: https://arxiv.org/abs/2611.15278
作者: Hanyi Zhang,Tongyu Wang,Manuel Debic,Markus Wennmann,Johannes Haubold,Tamara Schikowski,Wilfried Reichardt,Fabian Bamberg,Christopher L. Schlett,Steffen Ringhof,Thoralf Niendorf,Tobias Pischon,Michael Leitzmann,Carsten Oliver Schmidt,Mark Oliver Wielpütz,Thomas Kroencke,Josua A. Decker,Stefan N. Willich,Thomas Keil,Rafael Mikolajczyk,Miriam Klauß,Hans-Ulrich Kauczor,Philipp Mayer,Bettina K. Budai
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 17 pages, 4 figures, 6 tables
Abstract:Early detection and follow-up of pancreatic cystic lesions (PCLs) are crucial, as some of their subtypes, such as mucinous cystic neoplasms (MCNs) and intraductal papillary mucinous neoplasms, carry a significant risk of malignant transformation. Although most AI studies investigating PCLs rely on computed tomography (CT) scans, magnetic resonance imaging (MRI) offers superior soft tissue contrast, making it highly suitable for PCLs assessment. In this work, we propose an anatomy-aware two-stage deep learning framework for automated PCLs segmentation on abdominal MRI scans. First, a publicly available pancreas segmentation model is used to localize the pancreas. Then, we introduce the anatomyaware PCLSegNet, which incorporates anatomical priors as guidance for pancreas and PCLs segmentation. To address the class imbalance problem, we develop a Class-Specific ErrorWeighted (CSEW) loss function. Postprocessing is applied through anatomical this http URL proposed method outperforms multi-organ segmentation models such as TotalSegmentator, achieving substantially higher Dice scores for pancreas segmentation. Moreover, it also achieves a superior normalized surface Dice (NSD) and recall for PCLs segmentationcompared to other state-of-the-art models like nnU-Net, while also notably reducing the false negative rate, facilitating the automatic early detection and the noninvasive imaging-based assessment of PCLs.
[CV-136] Driving World Models Move the Camera Not the Scene
链接: https://arxiv.org/abs/2611.15260
作者: Yuqiao Feng,Xiaoyu Ao
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 23 pages. Project page: this https URL
Abstract:Plan-conditioned driving world models are used to compare candidate plans, so their forecast should change with the plan as the real scene would, yet widely used control metrics check only whether the camera follows the command. Recorded drives already contain real outcomes of plans that were never driven: retiming a log gives pose-matched images and LiDAR for other speeds, and a stereo rig records a laterally shifted view. Against these references, released world models move their camera but barely move the scene: their scene response, the share of the real change in distance that the forecast reproduces, reaches at most about a third of real video’s, and the trajectory score cannot tell an executed speed command from an ignored one. Interventions inside the generator trace the failure to the plan’s channel. The forecast geometry follows the context frames; once they carry the planned geometry, plan tokens add little. ReTime-Compose therefore delivers the plan as geometry, without training. It transports the frames of one logged-speed rollout to the planned cameras, and the frozen model completes only what no frame covers. Frozen generators then follow a 0.54 m lateral shift nearly as the real scene does on KITTI, nuScenes and Argoverse 2, and respond more than re-sampling to slow-down plans in three generators and to speed-up plans in two.
[CV-137] Are We Ready for Snow-tonomous Driving? Challenges Toward Robust Automated Driving in Extreme Weather
链接: https://arxiv.org/abs/2611.15217
作者: Eren Erdal Aksoy,Abu Mohammed Raisuddin,Thien-Thu Ngo,Eren Aydemir,Osman Tahir Ekşi,Paula Litkey,Petri Manninen,Heikki Hyyti,Jyri Maanpää,Tarmo Kivioja,Tuomas Herranen,Väisänen Sofie,Niemi Kalle,Isabelle Morvan,Pierre Duthon,Ian Marsh,Martin Sanfridson,Henrik Toss,Thanh Bui,Yuri Poledna,Maikol Funk Drechsler,Mikko Tarkiainen,Matti Kutila,Topi Miekkala,Stefan Parzefall,Valentina Donzella,Pak Hung Chan,Sogol Kharrazi,Mario Ceccarelli
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Automated driving has gained remarkable momentum over the past decade, with both industry and academia investing substantial resources in developing perception and control solutions to advance vehicle autonomy. However, these efforts have largely focused on clear-weather scenarios, leaving harsh conditions, such as snow, rain, and fog, comparatively understudied. As a result, inclement weather remains one of the most significant hurdles for autonomous vehicles, often leading to substantial performance degradation in downstream perception and control tasks. In this paper, we present insights from the EU-funded Horizon Europe project ROADVIEW, which aims to develop robust, weather-aware perception and control modules for connected and automated vehicles operating under severe conditions, including snow, rain, and fog. Drawing on four years of large-scale development and real-world demonstrations across multiple vehicle platforms in different locations, we identify ten key challenges that hinder the development of weather-resilient autonomy, including data scarcity, domain shifts, limitations in multimodal perception, control robustness, and validation gaps, among others. We summarize these challenges and lessons learned toward enabling reliable snow-autonomous driving, highlighting critical directions for future research and standardization. Subjects: Robotics (cs.RO); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV) Cite as: arXiv:2611.15217 [cs.RO] (or arXiv:2611.15217v1 [cs.RO] for this version) https://doi.org/10.48550/arXiv.2611.15217 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[CV-138] Where is My Mug? Training-Free Framework for Personalized Image Segmentation
链接: https://arxiv.org/abs/2611.15216
作者: Quang Qui-Vinh Nguyen,Hung Nguyen,Rang Nguyen,Khoi Nguyen
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:\textbfPersonalized Image Segmentation (PIS) aims to segment the specific object in a target image that matches a user-provided reference. Unlike category-based segmentation or few-shot segmentation, PIS must discriminate between multiple visually similar instances, a requirement central to deployed applications such as augmented reality, personalized content editing, robotics manipulation, inventory tracking, and fine-grained visual retrieval. We show that the strong numbers reported by existing PIS methods are largely an artifact of single-instance benchmarks: when evaluated on realistic scenes with several same-category look-alikes, prior point-prompt methods collapse, because the correct instance is frequently never surfaced as a prompt. We trace this to a recall bottleneck rather than a mask-quality one, and address it with a training-free, modular system. The pipeline first generates high-recall object proposals using an off-the-shelf class-agnostic detector, then matches each proposal to the reference using foundation-model features (e.g., BLIP-2), and finally refines the selected box with SAM. For robust matching we propose \textbftop- K Sinkhorn, a patch-level distance that scores each candidate only on its most confidently matched patches; trimming the Sinkhorn dual potentials recovers the robustness of partial optimal transport at more than 100\times lower cost, keeping it inside the per-proposal candidate loop. To evaluate this regime, we introduce \textbf\datasetname, a realistic multi-instance benchmark. Across \datasetname, PerSeg, PerMIS, and DAVIS, our system surpasses PerSAM, \mathrmP^2\mathrmSAM , and PDM by up to \sim!48 mIoU while matching their runtime and using fewer parameters, establishing a strong, deployable training-free baseline for real-world personalized segmentation.
[CV-139] A General Concept-based Decomposition for Vision-Language Embeddings NEURIPS2026
链接: https://arxiv.org/abs/2611.15195
作者: Simone Alberto Peirone,Ortal Senouf,Francesca Pistilli,Giuseppe Averta,Pascal Frossard
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at NeurIPS 2026. Project page at this https URL
Abstract:Vision-Language Models are designed to capture the compositional structure of the world through multi-modal alignment; however, it remains unclear whether their representations truly align with the way a human would naturally describe the same visual input. Prior works attempt to decompose visual embeddings into their concept-level contributions, but their instance-level optimization is inherently local and does not capture the model’s global semantic structure, which is essential for compositional generalization across different inputs, tasks, and domains. We introduce GRACE, a method that learns this concept-based structure at the feature space level, producing semantic and general concept decompositions. To achieve this, we propose a novel training objective that preserves the geometric relations of different samples in the concept space, enabling a better understanding of how concepts compose and relate to each other. We validate GRACE on several image and video benchmarks, showing that its concept decomposition is aligned with human perception and achieves faithful decompositions across different tasks and domains. With minimal overhead, GRACE makes existing Vision-Language Models more interpretable, while preserving most of their original performance.
[CV-140] OT-BLS: Optimal-Transport-based Backbone Layer Selection for Frozen DINOv3 in Cross-Domain Few-Shot Segmentation
链接: https://arxiv.org/abs/2611.15187
作者: Natsuo Okamoto,Kazuki Kobayashi,Yoshio Kato,Shuhei Tarashima
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted for publication at the 2026 IEEE International Conference on Visual Communications and Image Processing (VCIP 2026)
Abstract:Cross-domain few-shot segmentation (CD-FSS) aims to segment a target class in images from a previously unseen domain using only a few annotated support examples. Training-free CD-FSS methods, which typically leverage vision foundation models such as DINOv3, achieve competitive performance without parameter updates. However, the optimal DINOv3 layer for mask prediction varies across support–query episodes. Although several layer-selection criteria have been proposed, they do not explicitly evaluate whether the predicted query foreground and background labels agree with the corresponding support labels. To address this limitation, we propose a novel layer-selection criterion, Optimal-Transport-Based Backbone Layer Selection (OT-BLS). Our key insight is that an effective layer-selection criterion can be derived from the optimal-transport correspondence between query and support patch-token features. Based on this insight, we first estimate the transport plan using a well-established optimal transport algorithm and then compute a layer-selection score by aggregating the transported mass according to the foreground–background partition defined by each candidate query mask. The resulting score reflects how well the predicted foreground and background of the query correspond to those of the support. Experiments on eight benchmark datasets demonstrate the effectiveness of OT-BLS. It outperforms the conventional last-layer baseline in 12 of 16 dataset/shot settings. Furthermore, OT-BLS improves the average mIoU from 55.72 to 57.62 and achieves a 2.99-point improvement over the best-performing existing layer-selection criterion while running 3.7–14.6 \times faster.
[CV-141] GingivitisSeg: A Public Dataset and Comparative Benchmark for Gingivitis Segmentation in Intraoral Images
链接: https://arxiv.org/abs/2611.15184
作者: Zeynep Ozdemir,Zehra Hasgul,Irem Ulku,Omer Ozgur Tanriover,Fatma Karacaoglu,Murat Akkaya
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Preprint. Under review at Journal of Imaging Informatics in Medicine. Dataset and code: this https URL . Dataset DOI: https://doi.org/10.5281/zenodo.21098897
Abstract:Gingivitis segmentation from intraoral photographs can support computer-assisted periodontal assessment, but pixel-level delineation remains challenging due to severe foreground-background class imbalance, low-contrast boundaries, and variations in illumination and anatomy. We introduce GingivitisSeg, a public benchmark comprising 961 intraoral images with pixel-level annotations of gingival inflammation and predefined training (672), validation (144), and test (145) partitions. Seven semantic segmentation model-backbone configurations were compared using standardized partitions, input resolution, and evaluation metrics. Controlled ablation experiments examined class-weighted cross-entropy loss, Dice loss, and auxiliary supervision. UPerNet with ResNet50 achieved the highest mean Intersection over Union (mIoU) of 76.49% and gingivitis IoU of 54.91%. However, differences from DeepLabV3+ with ResNet50 and SegFormer with MiT-B2 were not statistically significant after Holm correction. The evaluated optimization components improved performance when combined in the UPerNet-ResNet50 configuration. Further analyses showed that performance was associated with gingivitis region size and decreased under stronger image blur. The dataset, benchmark splits, and source code are available at this https URL.
[CV-142] MAST: Medoid-Anchored Subspace Steering for Visual Grounding in Multimodal LLM s
链接: https://arxiv.org/abs/2611.15168
作者: Zhitao Chen,Zhenxin Cai,Jingyuan Li,Jin Sha
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:Existing activation steering methods for multimodal large language models (MLLMs) use the global activation mean as the steering vector, which tends to cause cross-task interference and unstable performance. We trace this failure to the multi-centered geometry of task activations. The global mean lands in the empty region between clusters and is faithful to no single task. Moreover, the steering signal in each task-pure local cluster is low-rank, and different query tasks benefit from emphasizing different families of local-basis directions under the shared routing protocol. To address this, we propose MAST (Medoid-Anchored Subspace Steering), a training-free, input-adaptive activation steering framework. MAST groups samples by task, performs multi-center clustering within each group, and selects one representative medoid per cluster. At inference time, each query is routed to multiple representative medoids. For each matched cluster, a low-rank local steering direction is extracted and aggregated across clusters to produce the final steering vector. Building on MAST, MAST-PR (Probe-Reweighted) uses a lightweight task-aware probe to reweight task-specialized local-direction families, yielding a further average improvement of +8.09 points over MAST.
[CV-143] Bridging Update Coordination and Flatness in Federated Learning
链接: https://arxiv.org/abs/2611.15165
作者: Chongwei He,Shichen Huang,Jiayun Li,Deyuan Lin,Wangze Jiang,Shengjia Li,Xu Wang,Xiaofan Li,Zhi-Ri Tang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 16 pages, 6 figures
Abstract:Federated learning combines complementary knowledge across clients to alleviate local data limitations without sharing raw samples. However, data heterogeneity can cause conflicting updates and hinder generalization. Update coordination reduces interference among client updates, while flatness optimization reduces sensitivity to parameter perturbations to improve generalization. Existing joint designs largely target specific algorithms, limiting flexible combinations of these approaches. To address this problem, we propose a unified federated learning framework that bridges update coordination and flatness optimization. Clients separately accumulate descent signals and flatness corrections along the same local trajectory. The server coordinates both groups with independent states before fusing their outputs into a global update. Explicit conversion between trajectory signals and model updates enables different coordination strategies to work with compatible local or globally guided flatness mechanisms without changing the inference architecture. Experiments show that the evaluated combinations generally improve average and worst-client accuracy over update coordination alone, demonstrating the effectiveness of the proposed framework in combining these complementary optimization mechanisms.
[CV-144] CineMME: From Dialogue Grounding to Narrative Reasoning in Multimodal Video Understanding NEURIPS2026
链接: https://arxiv.org/abs/2611.15161
作者: Jin Liu,Dawei Du,Yexiang Liu,Sijie Zhu,Fan Chen,Junxian Duan,Huaibo Huang,Ran He,Longyin Wen,Zhenfang Chen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to NeurIPS 2026 (Evaluations Datasets Track). Project page: this https URL
Abstract:Narrative video understanding requires more than recognizing visible events: models must extract evidence from dialogue, bind it to speakers, and use it to infer narrative meaning. Existing video benchmarks evaluate aspects of this problem, including video question answering, audio-visual reasoning, and temporal grounding, but rarely test whether these abilities form a coherent dialogue-to-plot evidence chain within the same clips. We introduce \textbfCineMME, a human-verified benchmark designed to diagnose dialogue-grounded plot understanding, supported by approximately 1,600 expert person-hours of annotation. CineMME contains 975 curated short narrative video clips and two linked tracks: \textitCineMME-Grounding, with 13,554 timestamped speech segments and 33,083 speaker face boxes for temporal dialogue localization, transcription, and active-speaker grounding; and \textitCineMME-Reasoning, with 1,600 multiple-choice and 316 open-ended questions spanning perception, audio, narrative, social, and scene-presentation understanding. Our main evaluation compares 20 MLLMs on reasoning and 9 omni-modal models on grounding. The results reveal a pronounced who-what-when decoupling: Gemini-3 Pro localizes dialogue in time ( 68.34% tIoU) and transcribes it with a WER of 28.99% , yet achieves only 7.74% vIoU for active-speaker grounding while achieving 67.66% MCQ accuracy on questions with extractable answers. Modality ablations further show that vision is not uniformly beneficial: it improves performance on perception-heavy questions but can hurt dialogue-centric reasoning. CineMME provides a compact diagnostic testbed for evaluating how multimodal video models connect dialogue, speakers, visual evidence, and narrative meaning.
[CV-145] MatrixSplat: Matrix-Parallel and Feature-Scalable 3D Gaussian Splatting Rendering on GPUs NEURIPS2026
链接: https://arxiv.org/abs/2611.15153
作者: Yangming Zhang,Siyi Wu,Jian Wang,Bingzhe Li,Wei Niu,Sheng Di,Ming Li,Miao Yin
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to NeurIPS 2026
Abstract:3D Gaussian Splatting (3DGS) has emerged as a mainstream representation for real-time novel view synthesis and 3D reconstruction. However, existing 3DGS rendering pipelines are poorly aligned with modern GPU architectures: depth-sorted alpha compositing introduces strong per-pixel sequential dependencies, limiting parallelism and resulting in scalar FMA-dominated execution that underutilizes high-throughput matrix units such as Tensor Cores. This mismatch becomes more severe for multimodal and semantically enriched 3DGS, where high-dimensional features must be accumulated from millions of Gaussians, leading to rapidly increasing computation and memory costs. We propose MatrixSplat, a matrix-parallel rendering framework for feature-scalable 3DGS on modern GPUs. We first identify the sequential dependency of alpha compositing as a key obstacle to efficient high-dimensional feature rendering, and adopt an order-independent weighted accumulation formulation that transforms feature rendering into matrix-oriented computation. Our framework features three key designs: (i) a tile-based splatting pipeline that maps feature accumulation to warp-level matrix-multiply-accumulate execution on Tensor Cores; (ii) a warp-granularity Gaussian packing strategy that mitigates visibility irregularity and improves feature reuse; and (iii) a lazy channel-batched feature-fetching pipeline that reduces shared-memory pressure, maintains high warp occupancy, and overlaps memory access with matrix execution for high-dimensional features. On Mip-NeRF 360, LERF, and 3D-OVS, MatrixSplat achieves up to 3.8 \times speedup over 3DGS and 2 \times over TC-GS while maintaining rendering quality and semantic accuracy, and further reduces render-kernel time by up to 36.7 \times for high-dimensional features. Integrated into existing fast renderers, it provides an additional 1.69-2.84 \times speedup.
[CV-146] From Concepts to Interactions: Dynamic Graph Semantic Recomposition for Safe Visual Generation
链接: https://arxiv.org/abs/2611.15149
作者: Qinghui Gong,Jinshan Lai,Xunlei Chen,Hua Meng,Zhengchun Zhou
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 23 pages,15 figures
Abstract:Recent advances in flow-matching Transformers have greatly improved Text-to-Image (T2I) and Text-to-Video (T2V) generation, but unconstrained generation risks producing illegal or unethical content. Safety alignment must therefore suppress unsafe content while preserving legitimate generation. A concept’s safety depends on context: a knife used for cooking is safe, whereas its use in violence is harmful. Yet most concept erasure methods indiscriminately suppress static, isolated representations, overlooking evolving semantic interactions. Consequently, safe scenes containing potentially sensitive terms may be unintentionally suppressed, compromising generation quality. To address this, we propose Dynamic Graph Semantic Recomposition (DGSR), which dynamically constructs and updates a semantic interaction graph from evolving text visual states during sampling. Context aware graph reasoning captures semantic dependencies, while paired safe/unsafe supervision distinguishes harmful from legitimate interactions, enabling precise identification of and selective intervention on high risk semantic relationships while preserving benign semantics. DGSR freezes the pretrained backbone and supports both image and video generation. Experiments show that DGSR achieves the lowest Attack Success Rate (ASR) in 10 of 12 T2I evaluations and reduces the average T2V unsafe frame rate by 41.2% relative to the strongest baseline, while maintaining high generation fidelity and temporal consistency.
[CV-147] Reframing Video Inverse Tone Mapping as Blind Posterior Inference via Latent Luminance Supervision
链接: https://arxiv.org/abs/2611.15146
作者: Hanwool Sul,Chang-Hoon Jeong,Kisung Shin,Bomseok Park,Donghoon Cha,Hansl von Kwon,DoHyung Ku,Byoung-Tak Zhang
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:High-dynamic-range (HDR) displays are now common, but most video content, such as archival footage, broadcast material, and the output of generative video models, exists only in standard dynamic range (SDR). The need is sharpest for generated video, which has no HDR master at all and is graded alongside live-action and visual-effects material finished in HDR. Converting SDR to HDR means undoing a tone-mapping operation that was never recorded, and clipping destroys the radiance wherever it acts, so HDR recovery is a blind inverse problem. Methods built on generative video priors fill clipped regions with plausible structure, but their reconstructions are systematically biased: in every public method we test, saturated content leans too bright or too dark, and the direction of the lean moves with the tone operator behind the input. We call this failure the luminance error, measure both its magnitude and its sign, and make it the focus of this work. Reframing ITM as posterior inference under an unknown operator locates the failure in two places and yields two corrections. A fixed affine functional of the video-diffusion latent reads log-luminance without decoding; on that coordinate an asymmetric loss selects an expectile of the posterior, applied only where clipping leaves luminance unconstrained. We also synthesize training SDR through an explicit family of tone operators, so the model marginalizes over what the observation cannot show. Trained this way, our luminance error stays flatter across operators and smaller on average than any public method, and the full system achieves the best perceptual quality on all three benchmarks.
[CV-148] Multi-View 3D Reconstruction Using V-JEPA Features
链接: https://arxiv.org/abs/2611.15143
作者: Anwar Shamim,Jeffrin Sam
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 29 pages, 11 figures. Source code is available at this https URL
Abstract:Predictive video encoders provide dense visual representations, but adapting them to 3D reconstruction requires accurate depth, camera poses, and cross-view geometry. We study supervised RGB-only multi-view reconstruction using an architecture that combines video-pretrained features, camera estimation, scene aggregation, epipolar transport, and residual pointmaps. Experiments compare global and spatial camera regression with joint relative objectives on CO3D, alongside an independently trained TUM RGB-D model. On 168 CO3D five-view windows, the spatial-camera configuration achieves 29.3% lower absolute trajectory error than the global-camera configuration, but performs worse than identity poses in trajectory error and worse than the global configuration in depth error. The joint relative-objective configuration improves normalized intrinsics error and similarity-aligned Chamfer distance from 0.6447 to 0.5573, but increases raw Chamfer distance from 3.5535 to 4.4374. On 163 indoor windows, camera error is 0.3063 m versus 0.3694 m for identity poses, while depth error is 0.4279 versus 0.2516 for constant median depth. Refinement worsens indoor camera error, and additional experiments show competing depth and pose effects. Overall, the results distinguish representation adaptation, aligned shape, and raw reconstruction, highlighting the need for isolated component and encoder comparisons before attributing reconstruction gains to predictive video pretraining.
[CV-149] Duo: A Reciprocal World Model for Full-Duplex Interaction
链接: https://arxiv.org/abs/2611.15131
作者: Hengyuan Cao,Biao Gong,Guoshan Lu,Qiyuan Zhang,Lin Liu,Yiding Tian,Yilei Jiang,Yuxin Ma,Chengzhang Wu,Kun Chen,Yiqian Xie,Ting Liu,Xin Zheng,Xijun Gu,Tianren Ma,Jiancheng Pan,Chi Zhang,Xuzhe Zheng,Siyi Li,Zhenzhong Lan
类目: Computer Vision and Pattern Recognition (cs.CV); Sound (cs.SD)
备注: this https URL
Abstract:In this report, we introduce the Reciprocal World Model (RWM) and present Duo v1, a full-duplex world model that continuously perceives audiovisual inputs while responding to users in real time. The RWM treats intelligence and physical reality as equally fundamental to what constitutes a world. Existing world models largely focus on simulating the latter, leaving reasoning and decision-making as external controls rather than intrinsic parts of the world itself. In RWM, we identify two equally fundamental capabilities: Intent, which provides the cognition underlying world interactions, and Realization, which generates coherent audio-visual behavior. To realize this formulation, Duo v1 introduces three key technical designs: (1) Leveraging autoregressive modeling for continuous interaction reasoning and participation decisions, rather than restricting it to instruction understanding for diffusion models. (2) Constructing chunk-specific conditions and synchronize tokens across Intent and Realization on a unified timeline. Module-specific attention masks enforce causal dependencies both within and across chunks, preventing access to future inputs. (3) Mitigating distribution shift in both modules through dual rollout distillation, training on autoregressively generated Intent tokens and synthesized audiovisual histories from Realization. Both modules continuously perceive real-time audio, video, and text, enabling full-duplex interaction with spontaneous conversations and interruptions. Beyond conversation, users can share real objects, reshape virtual scenes, guide assistants, or play audiovisual games without mouse or keyboard controls. The release of Duo v1 makes our first step, with the next version coming soon. We look forward to broader exploration of this emerging field.
[CV-150] oward Cross-Rubric Surgical Skill Assessment via Joint Embedding Prediction
链接: https://arxiv.org/abs/2611.15125
作者: Jiajun Cheng,Haidar M. Abdul-Muhsin,Juha Choi,Shan Lin
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Expert review of surgical videos is essential for surgical training but labor-intensive, motivating AI-assisted skill assessment to provide feedback at scale. Although surgical assessment rubrics may use different language to describe similar skills, existing automated methods commonly predict predefined skill levels or numerical scores tied to fixed rubrics, limiting their ability to transfer across semantically equivalent assessment criteria. We propose Predictive Rubric Alignment (PRA), a joint embedding prediction approach that scores surgical videos using natural-language descriptions of skill levels. Rather than matching level probabilities, PRA predicts a rater-weighted combination of skill-level description embeddings. Adjacent description features define a score-parameterized path, and the predicted embedding is projected onto this path to read the score. We introduce two variants: \textitPRA-item, which uses learned queries for predefined assessment items, and \textitPRA-cond, which generates queries from textual item names to enable assessment of unseen items without retraining. On AIxSuture, a dataset of expert-rated suturing videos, PRA outperforms published baselines in skill classification. We further evaluate zero-shot transfer to the smaller surgical skill assessment benchmark JIGSAWS. Trained on only two AIxSuture rubric items, PRA-cond assesses unseen JIGSAWS items without further training, yielding positive score correlations and greater Grad-CAM overlap with instruments and suture than a description-matching baseline. These results support predictive rubric alignment as a promising direction for flexible and explainable automated surgical skill assessment.
[CV-151] From Seeing to Remembering: Brain-Informed Agent ic Text-to-Video Generation for Enhancing Memorability
链接: https://arxiv.org/abs/2611.15107
作者: Shengqi Dang,Yi He,Jiaying Lei,Feilin Han,Nan Cao
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Generating videos that faithfully convey intended semantics while remaining memorable to viewers is valuable for applications such as advertising, education, and storytelling. However, memorability-oriented text-to-video generation remains underexplored, as human memory involves complex neural processes that are difficult to model and translate into generation guidance. In this paper, we propose MemCrafter, brain-informed approach that enhances video memorability through iteratively agentic prompt refinement. The agentic framework contains two key components. First, we construct a cortical region-based neuro-memorability signal that predicts video memorability from estimated neural responses, providing both an optimization objective and interpretable region-level information. Second, we develop a sensitivity-aware feedback module, motivated by a conceptual sensitivity analysis of the optimization objective, to derive actionable guidance for prompt refinement while preserving semantic fidelity. Extensive experiments demonstrate that MemCrafter consistently improves predicted video memorability while maintaining semantic alignment and perceptual quality.
[CV-152] Keep Images in Sight: Visual Position Re-Anchoring for Long-Context MLLM s
链接: https://arxiv.org/abs/2611.15100
作者: Juntao Xu,Qichai Zhao,Rui Sun,Zhouyang Jiang,Shiwei Wu,Zhi Li,Kaisheng Ma,Yu Liu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 23 pages, including supplementary material
Abstract:Long-context multimodal intelligence requires revisiting earlier images as conversations and documents accumulate, yet retaining an image does not guarantee reliable access to its details. By varying context length, redistributing a fixed text passage around the image, and changing positions without adding text, we identify the positional relation between images and questions as a direct influence on visual access. To address this gap, we propose Visual Position Re-Anchoring (VPR), a training-free method that shifts cached visual-key rotary phases toward a short-context reference before question processing. It operates directly on the existing KV cache, preserving standard attention computation without extracting attention maps or adding model forward passes. We develop static and continuous compensation schedules, using the static schedule by default. Across ten checkpoints and ten benchmarks, static VPR improves 97 of 100 model–task scores. Further analysis shows greater attention to regions containing answer evidence across six checkpoints, with minimal runtime overhead under fixed rotary frequencies. These findings demonstrate an efficient way to sustain visual access through positional adjustment within the cache.
[CV-153] Prediction or Persistence? What Future-Prediction Scores Miss About World Models
链接: https://arxiv.org/abs/2611.15082
作者: Animesh Maheshwari
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG); Robotics (cs.RO)
备注: Accepted at the CoRL 2026 Workshop “Do Robots Need World Models?”
Abstract:Pretrained video world models are compared and chosen as components for robot learning largely by future-prediction scores. On its own, such a score cannot be read as evidence of prediction, because it mixes encoder quality, the near future’s resemblance to the recent past, and the predictor’s own contribution. We separate the three with a matched control. On 27,000 future-retrieval trials from household egocentric video, a model must pick the true continuation of a clip among eleven candidates from the same video (chance 0.091), ranked zero-shot by its own training criterion. Each model is paired with a persistence control, an identical copy whose predictor is never invoked, so the paired difference isolates the predictor. Score and contribution dissociate within the native-video V-JEPA family alone: the predictor adds +0.291 top-1 for V-JEPA 2.1 but +0.03 to +0.04 for V-JEPA 2 on the identical trials. Across five models the contribution ranges from +0.291 to negative, and it is negative only for the two action-conditioned models, which we evaluate outside their training regime. Persistence alone recovers the encoder ordering expected from pretraining, a sign that the probe tracks real signal. We suggest reporting each future-prediction score together with its persistence baseline. The gap between them, not the score, is what makes the number interpretable: a necessary check costing one forward pass. Establishing downstream usefulness is a separate matter, for a planning or control evaluation.
[CV-154] SAVE: Less is More in Agent ic Video Reasoning
链接: https://arxiv.org/abs/2611.15079
作者: Tymoteusz Wilk,Sergey Prokudin,Xi Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Agentic video-reasoning frameworks iteratively explore a video, choosing what to observe based on the query and evidence gathered so far. Existing methods are typically evaluated with different backbones, preprocessing pipelines, and computational budgets, making it difficult to determine which components actually drive performance, how much they cost, and whether agentic inference is preferable to native video reasoning in the first place. We conduct a controlled evaluation of a diverse set of recent agentic video-reasoning systems across two strong contemporary backbones, with explicit token and monetary-cost accounting. Across Gemini 3 Flash and Gemma 4, most existing agentic frameworks fail to consistently improve over native video inference and often increase computational cost. We investigate two primary questions: (1) How much agentic framework complexity is necessary for effective video reasoning? (2) When, if at all, is agentic video reasoning preferable to just using the backbone directly? To answer the first question, we introduce a deliberately minimal Active Exploration Baseline and show that it remains competitive with substantially more elaborate agents. We then propose targeted extensions to this baseline, yielding SAVE (Simple Agentic Video Explorer). SAVE outperforms all evaluated prior agentic methods in five of six backbone-benchmark settings. Turning to the second question, our results indicate that the advantage of agentic exploration grows when dense processing would otherwise represent answer-relevant events too sparsely, particularly for long videos and temporally concentrated evidence. Overall, our results suggest that agentic video reasoning should favor simple frameworks over increasingly elaborate pipelines and focus primarily on long-form video, where selective evidence acquisition offers the clearest advantage over native inference.
[CV-155] DiFiT: Distribution-Fidelity-guided Visual Token Pruning for Multimodal Large Language Models
链接: https://arxiv.org/abs/2611.15069
作者: Jingyuan Zhang,Nihao Niu,Xinyang Lv,Beibei Han,Xiaoyun Wang,Yanning Hou,Ke Xu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 8 figures, 13 tables
Abstract:Visual token pruning reduces the inference cost of multimodal large language models, but answer accuracy only partly reflects its effect. Pruning can change the probabilities of alternative response tokens even when greedy answers remain correct and unchanged. We compare full and pruned response distributions under shared reference prefixes. Average predictive drift grows with compression on fixed correctly answered samples. Separate analyses associate larger drift with more confidence-decision disagreements and sampling failures. We propose DiFiT (Distribution-Fidelity-guided Token pruning) to learn discrete selection from reference answers and full-token response distributions. Dual-View Distillation matches probabilities across the vocabulary, while answer supervision supports task quality. A learned scorer and differentiable Top- K learn which visual evidence to retain. A multi-budget mask objective supports different retention ratios through one shared ranking over a frozen model. On Qwen2.5-VL-7B, DiFiT reduces sampling-induced negative flips by 30.3% relative to FastV at 10% retention on four-task greedy-correct controls. It preserves 98.74% of baseline performance on MME at 30% retention.
[CV-156] st-Time Selective Adaptation for Cross-Modal UAV Object Tracking with Token Shift
链接: https://arxiv.org/abs/2611.15059
作者: Yun Xiao,Hailong Ye,Zhicheng Zhao,Lei Liu,Amir Hussain
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Cross-modal unmanned aerial vehicle (UAV) object tracking faces modality-induced appearance discrepancies and abrupt target-position shifts that can arise from modality switching, UAV motion, and inter-sensor offsets. Although existing methods model modality and spatial variations, they primarily rely on offline representation learning and offer limited localization-oriented adaptation during testing. To address these challenges, we propose Test-Time Selective Adaptation with Token Shift (TSAT), which derives spatial and temporal adaptation signals from unlabeled test sequences for online optimization. Specifically, we design a controllable token shift auxiliary task, which imposes explicit spatial correspondence between predefined token translations and resultant localization offsets, generating stable annotation-free spatial adaptation signals. Furthermore, we introduce a temporal backward prediction auxiliary task to exploit frame-wise temporal dependencies, effectively mitigating cumulative localization errors in continuous tracking. To suppress interference from unreliable tracking samples, we develop a reliability-aware update gating strategy to screen valid adaptation signals and stabilize online parameter updates. Experiments on CM-UOT demonstrate that TSAT achieves state-of-the-art performance against existing test-time adaptation methods, exhibiting superior robustness under frequent modality switching and severe spatial displacement. Experiments across tracking frameworks and datasets further validate the effectiveness and generalization capability of TSAT.
[CV-157] Saying No by Asking Yes: Affirmative Decomposition and Evidence Reflection for Frozen CLIP
链接: https://arxiv.org/abs/2611.15047
作者: Aashish Pokhrel,Tejas Anvekar,Shivanand Venkanna Sheshappanavar
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 27 pages, 6 figures
Abstract:Frozen vision-language models often match concepts mentioned in a caption without respecting their negation: a query “a kitchen with no toaster” favors an image containing a toaster. Yet these models are substantially stronger at matching affirmative descriptions. We exploit this asymmetry with Negation-Aware Evidence Scoring (NAES), which reformulates negated descriptions into affirmative queries and combines their evidence by caption structure. A lightweight text decomposer, NegSplit, identifies this structure by extracting negated spans and the remaining affirmative context. For captions with an affirmative remainder, NAES matches that content and applies a presence-based correction for the denial. When no affirmative content remains, it queries the denied object’s presence and reflects the evidence around a reference to score the negated description. With LAION-2B CLIP, NAES achieves 82.4% accuracy on COCO-MCQ and 93.0% on VOC2007-MCQ, surpassing KTN by 2.9 and 0.9 percentage points; with OpenAI CLIP, it exceeds PeakPatch accuracy by 4.4 and 26.0 points, respectively. NAES also improves negated-description retrieval on COCO and MSR-VTT, with gains extending to SigLIP, SigLIP 2, and larger CLIP encoders. The strongest multiple-choice results use competing descriptions to center and align evidence; retrieval improvements persist without this conditioning. These findings show that explicitly separating affirmative matching from denied-object evidence can substantially improve negation-aware ranking.
[CV-158] Back to the Question: Tracing Visual Distraction Through Text Tokens
链接: https://arxiv.org/abs/2611.15039
作者: Marie Amale Huynh,Regev Cohen,Serena Yeung-Levy,Sivan Doveh
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 25 pages, 18 figures, including appendix
Abstract:The visual world is cluttered: robust vision-language reasoning requires ignoring irrelevant information. Yet vision-language models (VLMs) can suffer substantial accuracy losses when answering questions about a target image with distractors present. Existing mitigations focus on selecting relevant images or limiting information exchange between images. Through a systematic analysis of three open 7-8B VLMs on five perception benchmarks adapted to include distractor images, we show that distractors reduce accuracy even when models correctly identify the relevant image, suggesting that selection alone is not enough. To understand the mechanism behind this interference, we combine attention analysis with targeted knockouts in Qwen2.5-VL and identify middle-layer distractor-to-text connections as a pathway through which irrelevant visual information affects reasoning. We introduce answer consistency regularization, which encourages consistent answer distributions with and without distractors. Without directly constraining attention, this objective reduces middle-layer attention from question tokens to distractor images, consistent with the pathway our analysis implicates. With five distractors, our approach improves circular accuracy by an average of 16 percentage points over the Qwen2.5-VL baseline across the five benchmarks while preserving single-image performance.
[CV-159] Federated Class Incremental Learning: A Pseudo Feature Based Approach Without Exemplars
链接: https://arxiv.org/abs/2611.15025
作者: Min Kyoon Yoo,Yu Rang Park
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:Federated learning often assumes that data is fixed in advance, which is unrealistic in many real-world scenarios where new data continuously arrives, causing catastrophic forgetting. To address this challenge, we propose FCLPF (Federated Class Incremental Learning with Pseudo Features), a method that uses pseudo features generated from prototypes to mitigate catastrophic forgetting. Our approach reduces communication costs and improves efficiency by eliminating the need for past data and avoiding computationally heavy models like GANs. Experimental results on CIFAR-100 show that FCLPF achieves an average accuracy of 51.87% and an average forgetting of 9.62%, significantly outperforming existing baselines with an average accuracy of 47.72% and forgetting of 20.46%. On TinyImageNet, FCLPF achieves 37.56% accuracy and 3.14% forgetting, compared to the baselines’ 27.69% accuracy and 24.46% forgetting, demonstrating the superior performance of FCLPF.
[CV-160] Validation of a Facial-Video Digital Hydration Biomarker Score for Tracking Hydration State Transitions in a Real-World Fasting Model
链接: https://arxiv.org/abs/2611.14993
作者: Yudara Kularathne,D.M.J. Prathapa,Thanveer Ahamed,Janindu Athukorala,Thisara Perera
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 12 pages, 6 tables, 5 figures
Abstract:Hydration influences physiology and daily function, yet repeated practical measurement in everyday settings remains limited. We present initial real-world evidence that a facial-video lifestyle digital biomarker score can track the direction of hydration state transitions. Using Ramadan fasting as a model of prolonged fluid abstinence, we assessed whether the score tracks the transition between the fasting (pre-drink) state and the rehydrated (post-drink) state across three validation layers. In the primary cohort (104 participants, 208 videos), the score increased from a mean of 36.6 in the fasting (pre-drink) state to 70.6 in the rehydrated (post-drink) state (change = +33.9, 95% CI [+30.7, +37.1], p 0.001, Cohen’s d = 2.03), with 102 of 104 participants (98.1%) showing an upward shift. In a nested body-weight subgroup (20 participants), body weight increased by +0.74 kg after fluid intake (p 0.001), confirming fluid uptake and providing supportive physical anchoring. In a nested serum osmolality subgroup (10 participants), osmolality decreased by -5.1 mOsm/kg (p 0.001), with all post-drink values within the normal clinical range, providing exploratory biological support. An exploratory secondary comparison of water versus electrolyte rehydration showed no statistically significant difference in short-term score change (p = 0.247). This study does not establish a clinical diagnostic test for dehydration. Rather, these findings support the score as a lifestyle digital biomarker for tracking the direction of hydration state change in real-world settings.
[CV-161] Poison-then-Hide: Finetuning-Activated Backdoor Attack on Pretrained Vision Encoders
链接: https://arxiv.org/abs/2611.14970
作者: Qixuan Jin,Abinitha Gourabathina,Vinith Menon Suriyakumar,Walter Gerych,Marzyeh Ghassemi
类目: Machine Learning (cs.LG); Cryptography and Security (cs.CR); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Backdoor attacks threaten the integrity of machine learning models by allowing attackers to control model behavior through triggers. Models can be compromised during finetuning by backdoor attacks through either poisoned data or adversarial training objectives. Because existing finetuning-activated attacks assume limited domain shift or frozen encoder layers, they often fail under full-model finetuning. We target a stealthy finetuning-activated attack where a dormant backdoor is implanted in a pretrained encoder, and later activated by finetuning on clean downstream data. We propose Poison-then-Hide, a novel attack that remains effective when the entire model is finetuned in a domain transfer. Our approach consists of three components: trigger optimization, base encoder poisoning, and targeted unlearning to conceal the backdoor. We evaluate our method on six datasets and three model architectures, and achieve state-of-the-art (up to 97%) attack success rates. We find that jointly learning the benign target task alongside the poisoning is critical to the attack’s success. We demonstrate standard detection and mitigation defenses cannot fully remove the backdoor, which can reappear after benign finetuning.
[CV-162] Adapting Monocular Depth Estimation in Novel Scenes through Disparity Refinement and Robot Ego-Motion
链接: https://arxiv.org/abs/2611.14958
作者: Dimitrios Chatziparaschis,Georgios Georgakis,Adnan I. Ansar,Shehryar Khattak,Konstantinos Karydis
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注: 17 pages, 12 figures, journal
Abstract:The geometric understanding of a scene is a crucial component for safe and efficient robotic operations in novel and unstructured environments. The recent emergence of foundation models for Monocular Depth Estimation (MDE) has provided an opportunity to enrich robotic perception capabilities with dense long-range depth predictions, replacing expensive onboard hardware (e.g., LiDARs and stereo cameras). However, these offline models are primarily trained on affine-invariant depth representations, and their generalization to novel scenes remains unclear. In this work we present EM-ViT, a monocular depth estimation model that learns to adapt online relative depth inputs from a foundation model prior to the metric scale with observations collected during robotic operations. Motivated by the stereo camera parameterization, our method decomposes depth prediction via dense disparity refinement to recover the scene’s relative structure and an effective baseline scalar prediction to determine the missing metric scale. The learning process leverages sparse LiDAR points and, optionally, the robot’s Ego-Motion (EM) to rapidly train a lightweight Vision Transformer (ViT), while an internal gradient matching loss preserves the fine-grained semantic structure of the input disparity in real-time. We demonstrate the efficacy of our method on publicly available sequences, such as KITTI, and show that our online learning pipeline can adapt to novel scenes faster and with higher accuracy than existing MDE foundation models. The code and supplementary materials will be publicly available at this https URL.
[CV-163] RSLiveBench-UAV: A Continually Evolving Vision-Language Benchmark for Aerial Spatial Intelligence
链接: https://arxiv.org/abs/2611.14953
作者: Yue Zhou,Jiawei Liu,Jue Chen,Ziyang Gong,Haojia Wei,Jingxiang Li,Xue Jiang,Zhihang Zhong,Pai Peng,Xue Yang,Yansheng Li,Jonathan Li
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 10 pages, 6 figures
Abstract:Aerial spatial intelligence (ASI) requires models to understand objects and spatial relationships across changing viewpoints and scales. Evaluating these abilities as models evolve calls for fresh physical observations, yet producing new aerial spatial questions requires repeatedly establishing entity identities, geometric references, and cross-view correspondences. We introduce \textbfRSLiveBench-UAV, a continually evolving vision-language benchmark driven by recurring acquisition of real UAV imagery. Its update protocol admits images captured after the preceding release and delays answer disclosure. An entity-centered production workflow uses geometry-assisted association to link repeated observations of physical vehicles, reuse human-verified attributes, and combine them with view-specific annotations and spatial references for traceable question generation. All tasks use deterministic scoring without model judges. We evaluate 30 general-purpose, remote-sensing, and embodied models on 35,491 question records from ten scenes, covering 17 tasks across 2D perception, 3D perception, cross-view perception, and reasoning. All 18 tested embodied models outperform the three selected remote-sensing comparators in Overall, although the strongest embodied model scores 43.29 versus 59.88 for GPT-6-Astra. Fine-grained vehicle recognition remains challenging, and performance varies across heights and camera pitches. Retrospective scene-group comparisons and rolling-window replay characterize score-pattern consistency under scene renewal. RSLiveBench-UAV connects fresh physical observations, entity-centered answer production, and systematic model evaluation to support continual assessment of ASI. Benchmark will release at this https URL
[CV-164] Emergent Vision-Shaped Thinking in Text-Only Reasoning through Multimodal Training
链接: https://arxiv.org/abs/2611.14942
作者: Qi Zhang,Bohan Wang,Qianru Sun,Hanwang Zhang
类目: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Does multimodal training shape how a language model thinks, even when the input is text only? We test this with targeted causal ablations across 10 multimodal configurations from 6 model families, with matched text-only backbones and cross-family transfer controls. We contrast spatial-imagery tasks that invite mental scene construction with symbol-manipulation tasks solvable by symbolic rules alone. We identify a vision-language axis in hidden-state space and ablate its projection at inference. In Chameleon-7B, this ablation drops accuracy on BIG-Bench Hard (BBH) navigate by 27.6 pp while leaving boolean expressions intact (+5.6 pp), a 33.2 pp double dissociation. The effect is direction-specific and layer-localized, extends to mental-simulation tasks such as mental rotation, paper folding, and perspective taking, and is not explained by task difficulty, reasoning depth, or sequential update load. Together, these results reveal emergent vision-shaped thinking (EVT): text-only reasoning that causally recruits representations introduced or strengthened by vision-language training, without assuming an image-like internal format. The measured effect is spatially concentrated at 7-8B and broadens to non-spatial visual attribute binding in the tested 32B configuration. In dual-path unified models (BAGEL, Janus-Pro), the causal handle localizes to the generation pathway as a compact one-dimensional carrier. The understanding pathway also encodes spatial information in a higher-rank, probe-decodable form, but is not the causal handle. To separate decodability from causal use, we analyze concentration, alignment, and semantic location. These diagnostics explain why behavior, probing, and intervention can diverge. Multimodal training thus reshapes the geometry of text-only reasoning, not just how visual input is processed.
[CV-165] Reality Perception Gap: Are AI-Manipulated Video Detectors Ready for the Real World?
链接: https://arxiv.org/abs/2611.14941
作者: Xiangrui Liu,Xinmiao Luan,Xinyuan Wang,Zhiyuan Fang,Chi-Yao Huang,Douglas C. Montgomery,Yanjie Fu,Felix Juefei-Xu,YeZhou Yang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:AI-manipulated video detectors are typically evaluated on clean, curated datasets. In deployment, however, videos have usually passed through repost compression on social-media platforms and sometimes were captured by imperfect hardware. To measure detection under these conditions, we introduce Reality Perception Gap (RPG), a benchmark of 10,000 videos from frontier generators and partially synthetic edits, processed with cross-platform repost compression and paired with real footage that carries camera-side optical artifacts. RPG also includes socially fragile content, such as crime and wartime footage, where fabricated videos can cause severe social harm. Across 18 detectors, none catches more than 15% of fakes at a 1% false-positive rate. A positive control and a re-encoding sweep separate two sources of failure: generator and content shift degrades detectors before any compression, and repost compression erases the low-level traces that the artifact-based detectors relies on. Existing detectors therefore struggle when deployment conditions strip away the generation artifacts they rely on. These complementary failures suggest that robust detection requires training under deployment conditions and combining low-level, temporal, and semantic evidence, where passive detection may have reached its limits.
[CV-166] Hint-Informed Self-Distillation for Reasoning Segmentation
链接: https://arxiv.org/abs/2611.14931
作者: Kunze Jiang,Yuchen Wang,Kang Li,Qixing Zhang
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 11 pages, 5 figures, 7 tables. The first two authors contributed equally
Abstract:Reasoning segmentation requires identifying the target referred to by a query among visually similar objects, yet trajectory-level rewards cannot directly indicate which tokens provide relevant visual evidence. We propose Hint-Informed Self-Distillation (HISD), which uses target-disambiguation evidence for token-level credit assignment. An external model generates Segment Hints offline from training images, queries, and ground-truth target annotations, organizing target identity, discriminative attributes, confusable objects, and anchoring paths. A hint-conditioned exponential moving average teacher scores the student’s own trajectories, and HISD computes the teacher-to-student evidence ratio over a token support set to adjust only the magnitudes of positive group-relative advantages, leaving nonpositive advantages unchanged. This design encourages concise reasoning while retaining the information needed for target disambiguation, and the student requires neither hints nor a teacher at inference. Under a unified training protocol, HISD improves ReasonSeg test gIoU and cIoU over our reproduction of DPAD by 3.00 and 3.55 percentage points, respectively. Across the five evaluated splits of ReasonSeg and the RefCOCO family, HISD achieves higher segmentation metrics than this baseline and reduces the average token count of valid reasoning spans by 80.17 to 91.83 percent.
[CV-167] Lines Are Ideas: Visual Creativity Assessment in Hyperbolic Space
链接: https://arxiv.org/abs/2611.14907
作者: Baoliang Chen,Zihao Lin,Hao Liu,Xiangjie Sui
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Assessing human creativity from drawings provides a scalable way to evaluate human creative ability. However, existing approaches either rely on costly expert annotation or treat creativity as a single regression target, overlooking its underlying structure. Psychological theories have long emphasized that creativity emerges from balancing novelty and appropriateness. Motivated by this principle, we recast creativity as structured semantic deviation between paired drawings, where novelty measures the extent of departure from the original base sketch, and appropriateness determines whether such departures remain meaningful. This formulation naturally leads to a hyperbolic representation, in which novelty is quantified by geodesic distance and appropriateness is evaluated conditionally via a learned decision boundary, further regulated by an abstraction-conditioned adaptive constraint. Experiments on AuDrA-Drawings demonstrate state-of-the-art alignment with expert judgments and strong robustness under rater and task shifts. Beyond performance gains, our framework reveals an interpretable radial organization of creative variation, offering a principled geometric foundation for automatic creativity assessment.
[CV-168] From Pixels to Prompts: Physics-Guided Language Enhancement for AI-Generated Image Detection
链接: https://arxiv.org/abs/2611.14898
作者: Mei Qiu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:AI-generated content (AIGC) has substantially narrowed the visual gap between natural and synthetic images, while existing detectors often rely on generator-specific artifacts and exhibit limited generalization to unseen models. This motivates a fundamental question: which intrinsic physical image characteristics provide stable, discriminative cues for distinguishing natural from AI-generated images? We investigate this by systematically identifying robust physical features and incorporating them into a vision-language representation without external textual prompts at inference. We first analyze 15 handcrafted physical descriptors across images from over 20 generative models. Based on cross-generator stability and real-fake discriminability, we propose a Feature Stability-Discriminability Validation Assessment (FSDVA) to select a compact set of reliable features. Five core features, including Laplacian variance, Sobel statistics, and residual-noise variance, are then used to construct structured textual descriptions encoding physical evidence for CLIP-based learning. We integrate these physics-enriched descriptions with image representations via a native physics-guided multimodal framework, adapting CLIP ViT-L 14 with LoRA. At inference, the model requires only the input image, with no external captions or manual prompts. On GenImage, our method achieves 99.0 percent accuracy on SDv1.4 and 96.7 percent on Wukong, with a mean of 97.23 percent across eight generative models versus 95.8 percent for C2pCLIP under the same setting. On UniversalFakeDetect, it maintains strong, stable generalization across diverse and unseen generation settings. These results show that carefully selected physical image statistics provide complementary forensic evidence to pretrained vision-language representations, enabling interpretable, more generator-agnostic AIGC detection.
[CV-169] HeadIO: Body-Frame Heading Decomposition for Neural Inertial Odometry on Head-Mounted Devices ICRA2027
链接: https://arxiv.org/abs/2611.14895
作者: Taeyoung Yeon,Yiquan Li,Chenfeng Gao,Vasco Xu,Karan Ahuja
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注: Submitted to ICRA 2027
Abstract:Inertial odometry (IO) using only Inertial Measurement Units (IMUs) enables camera-free head tracking in head-mounted devices such as VR headsets, AR glasses, and earbuds. Prior IO methods operate in a gravity-aligned world frame, which is not yaw-equivariant. Yaw augmentation improves robustness but does not guarantee equivariance; learned equivariant layers achieve it, but EqNIO’s full pipeline costs nearly 3 \times the FLOPs of ours. While raw body-frame IMU inputs are inherently yaw-invariant, head tilt redistributes both gravity and motion signals across device axes, making the representation tilt-variant. We propose HeadIO, which combines a body-frame heading decomposition with a bidirectional Mamba backbone for head-mounted inertial odometry. Our decomposition achieves tilt invariance by separating gravity-aligned and horizontal components using only the tilt component of device orientation. We evaluate on four head-mounted datasets and find that HeadIO improves RTE-5s by 8.1% (0.260 m \to 0.239 m) over world-frame methods with learned equivariant layers on Nymeria, the largest benchmark dataset, at nearly 3 \times lower FLOPs and 8 \times fewer parameters. HeadIO achieves state-of-the-art performance on our primary metric, RTE-5s, on all Aria-glasses datasets, which natively provide body-frame semantics for our decomposition method. We further evaluate the IMU-only deployment scenario on Nymeria, where HeadIO remains state-of-the-art. We show that our method relies mainly on tilt accuracy, which can be obtained from the gravity signal; for yaw, a single-coefficient gyroscope correction removes 11.6% of the yaw error and improves RTE-5s by 2.5%, and sparse external yaw corrections (for example, one every 30 s) can improve RTE-5s by 12.0%. Our decomposition is lightweight, composes with any backbone, and is suitable for real-time deployment on head-mounted devices.
[CV-170] In-orbit demonstration of object-based change detection applied to the maritime domain
链接: https://arxiv.org/abs/2611.14857
作者: Cristopher Castro Traba,Tom Hendrix,Amaury Perrocheau,Francesco Pasanisi,Finian Rogers,Aubrey Dunne,David Rijlaarsdam,Roberto Del Prete,Nicolas Longépé,Gabriele Meoni
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 11 pages, 8 figures
Abstract:This article presents the successful in-orbit validation of an object-based change detection pipeline for low latency maritime monitoring onboard the CogniSAT-6 mission. We propose and validate a lightweight, three-phase operational sequence: optimized ground-based dual-acquisition scheduling, real-time onboard AI-based vessel segmentation, and multi-temporal matching. Upon acquiring consecutive images over a targeted Area of Interest (AoI), the onboard system correlates vessel detections using a hybrid matching approach that integrates the Hungarian algorithm for optimal assignment with RANSAC-driven spatial and size filters for outlier rejection. A series of in-orbit validations demonstrated robust performance across 11 diverse maritime scenarios, achieving valid recall and precision values (considering only vessel size and location) exceeding 0.8. The validation demonstrated the ability of the system to track both cooperative targets and uncooperative “dark” vessels strictly through geometric consistency. This capacity to autonomously generate and act upon real-time intelligence establishes a foundation for next-generation Insight-as-a-Service (IaaS) and tipping-and-cueing architectures.
[CV-171] Honest Fusion Evaluation for Cross-Domain Deepfake Detection: Leakage Audits and Tail Robustness WACV27 WACV2027
链接: https://arxiv.org/abs/2611.14856
作者: Luke Collins,Chang-Tsun Li,Duc Thanh Nguyen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at WACV 2027. 19 pages including supplementary material; 6 figures, 10 tables. Code: this https URL
Abstract:Cross-domain video deepfake detection increasingly uses frozen foundation-model experts, combined by calibrated score averaging, a per-video router picking one expert, or a learned head over concatenated features. Expressive routing and feature fusion are often reported to beat plain averaging. Across a 22-dataset frozen-expert testbed, we find those gains are largely evaluation artifacts. A corner theorem locates the gap routing aims to close: because AUC depends only on score ordering, the best any score-space router can do is an oracle reading the test label to pick each video’s expert, a ceiling no realisable, label-blind router reaches. The theorem does not order real routers against a calibrated global average; the case against routing is empirical, and no router we test beats that average. Feature fusion reads raw features, falls outside the theorem, and is audited empirically. Reported gains trace to leakage: holding out one dataset leaks siblings sharing identities and reals (over-crediting routing), while holding out one manipulation keeps that dataset’s reals in training (over-crediting feature fusion). A 2\times 2 over fold and evaluation granularity pins the +2.6 pp feature-fusion “win” to the fold alone: 12/12 significant under leaky folds vs. 0/12 under honest family-held-out folds, where both gains vanish. Fusion still pays off in the tail of hard videos and families: as experts are added, videos every expert misclassifies drop from 15.2% to 2.3%, and global calibrated fusion (InvB) beats the best single expert on worst-family AUC, selective risk (AURC), and low-false-positive recall, stably across protocols. Practitioners should calibrate and combine globally, hold out whole dataset families, and measure tail performance. For score-space routing under ranking metrics, honest evaluation and expert diversity matter more than combiner complexity.
[CV-172] Paired Geometric Supervision for Generalizable Deepfake Detection BMVC2026
链接: https://arxiv.org/abs/2611.14851
作者: Luke Collins,Chang-Tsun Li,Duc Thanh Nguyen,Victor Sanchez
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at BMVC 2026 (poster). 37 pages including supplementary material; 8 figures, 22 tables
Abstract:We present OrthoPairGTL, a geometric training framework for cross-domain deepfake detection, built on a partially frozen DINOv3 backbone ( \sim 300M parameters), and make three contributions: (1) We re-evaluate six recent transfer-learning detectors (Forensics Adapter, Effort, three GenD variants and FakeSTormer) using the authors’ released checkpoints under a single preprocessing pipeline across 17 cross-domain datasets (2018-2025), so that the reported gaps reflect method rather than per-method pipeline tuning. (2) We model the transformation between paired real and fake samples, training Parameter-Efficient Fine-Tuning (PEFT) adapters with an orthogonal pair decomposition, hyperspherical regularisation, and (in the GSD-LoRA variant) a set-level coherence objective. (3) Under LoRA, this regularisation gives a more stable optimisation profile: GSD-LoRA’s post-peak validation degradation is 4.2\times shallower than plain LoRA at the same parameter footprint, and the time-averaged validation AUC across training is 1.82 pp higher. On the 17 datasets, GSD-LoRA matches or exceeds every re-evaluated baseline (92.82% vs. 91.58% for the strongest; paired Wilcoxon signed-rank p=0.015 , 13/17 per-dataset wins), with the largest gains on modern reenactment-heavy benchmarks (DSv1.1, DSv2). Under strict parameter budgets, LayerNorm-only tuning ( \approx 0.03% of the backbone) is the most stable configuration and matches the strongest baseline with a mean of 91.97%.
[CV-173] Sophon: A Procedural Diagnostic for Spatial Reasoning in Vision-Language Models NEURIPS2026
链接: https://arxiv.org/abs/2611.14838
作者: J. Zach Gazak,Ryan Swindle,Justin Fletcher
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: NeurIPS 2026 (Evaluations Datasets Track). 10 page main text, 44 pages total. Code and data: this https URL
Abstract:Open-weight Vision-Language Models (VLMs) score 85-94% on spatial benchmarks (BLINK Spatial, What’s Up CLEVR) yet collapse to chance on simple geometric questions when object, scene, and texture priors are stripped away. We introduce and open-source Sophon, a procedurally generated diagnostic of paired colored point-cloud stimuli across three coordinate systems (1D linear, 2D Cartesian, 2D polar) and four question types (binary, categorical, comparative, quantitative). Across four open architectures (LLaVA-NeXT, LLaVA-OneVision, Qwen2.5-VL, Idefics3), accuracy clusters at 47-52% while frontier models (Opus 4.6, Gemini 2.5 Pro, GPT-5.4) clear 77-83% and humans reach 84-94%. We use linear probes and low-rank adaptation (LoRA) to trace the open-weight failures. Probes of the vision encoder (ViT) and language decoder (LLM) recover Sophon spatial information at 83-85% and 80-86% accuracy, respectively, and the language tower’s reasoning circuit answers text-only controls (point clouds as sets of coordinates) at +40 pp over visual stimuli. Language tower per-layer attention attribution localizes visual integration to mid-stack layers. At readout, models elevate the correct answer-set tokens but select among them randomly. We localize the mechanism by tuning individual and grouped components across every architecture’s stack, including the vision tower (ViT workspace and reader), the projector, and the language tower (LLM workspace and reader). These experiments demonstrate that this mid-stack workspace holds circuitry capable of organizing the spatial signal so that the frozen reader can decode it. LoRA fine-tuning adjusts that circuitry such that 7B-class VLMs answer Sophon spatial reasoning questions at frontier aggregate accuracy. This work suggests that these models do not have a capability gap for spatial reasoning but suffer from a routing bottleneck that targeted fine-tuning repairs.
[CV-174] Learning Low-Latency Action Verification for Frozen Vision-Language-Action Models
链接: https://arxiv.org/abs/2611.14834
作者: Wen-Han Hsieh,Ken Jiankun Zheng,Chia-Chien Chen,David M. Chan,Trevor Darrell
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Vision-language-action (VLA) models have made significant progress in general-purpose robotic manipulation. However, sampled action sequences can be ineffective, while evaluating alternatives with computationally expensive verifiers adds latency at each policy query. In this work, we propose HDH, a lightweight framework that uses hindsight from a frozen VLA’s own deployments to improve action verification. The framework has three stages. First, we deploy the policy and collect its autonomous rollouts. We then train a compact action verifier to assign each action chunk a predicted task-success score. Finally, the frozen VLA generates action candidates, and the verifier selects the candidate with the highest score for execution. Across four LIBERO-10 tasks, HDH achieves an average success rate of 82.3% and on three real-world manipulation tasks, HDH improves average success over the baseline policy by 22.6%.
[CV-175] Neural Compression for Scientific Particle Data: Perspectives on How Learning Can Help
链接: https://arxiv.org/abs/2611.14832
作者: Tian Qiu,Carolina Cuesta-Lazaro,Sheng Di,Haewon Jeong
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 10 pages, 2 figures. Extended version of paper accepted at 2026 Conference on Physics and AI (PAI26)
Abstract:As particle-based simulations grow in scale, efficient compression of scientific data becomes critical. Neural point cloud compression (PCC) methods excel on computer vision (CV) benchmarks, but do they transfer to scientific particle data? We highlight that scientific particles differ from CV point clouds in fractal dimension: CV point clouds are locally planar, but particles are not. This key difference limits the effectiveness of the octree data structure that underpins most PCC methods. Neural codecs achieve only marginal bitrate reductions over the traditional G-PCC baseline on scientific data, revealing a significant performance gap. Supported by an evaluation spanning domains of cosmology, molecular dynamics, and plasma physics, we outline where the structure of scientific data limits current neural approaches and where learning can help in the future.
[CV-176] From Classification to Generation: Vision-Language Models for Roman Imperial Coin Description
链接: https://arxiv.org/abs/2611.14821
作者: Pedro Lourenco,Ognjen Arandjelović
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Cataloguing a Roman Imperial coin involves reading its legends and identifying its imagery in order to match it to an entry in the standard reference catalogue. Previous work has automated individual parts of this process, such as identifying the issuer or recognizing reverse motifs, by formulating each part as a classification task over a predefined set of labels. We instead formulate coin description as image-to-text generation and adapt an open-weight vision-language model (VLM), Qwen3-VL-8B-Instruct, to four tasks: transcribing the obverse and reverse legends, and describing the obverse and reverse imagery. To obtain training data, we convert approximately 30,000 auction records into structured image-text pairs. We separate the two sides of each coin, extract and normalize legends and descriptions, and standardize obverse descriptions against a bust-type taxonomy. We then train a separate low-rank adapter for each task on a single 20GB GPU, and evaluate each task with metrics that we define to capture the key components of numismatic descriptions. The adapted models outperform both the unadapted base model and GPT-4.1 on all four tasks, reducing the normalized edit distance of legend transcriptions by over 40% relative to GPT-4.1. The improvement is smaller, but still present, on legends absent from the training data.
[CV-177] BirdDreams: Joint Perspective and BEV Video Generation for a Plausible Reactive Driving World Model
链接: https://arxiv.org/abs/2611.14815
作者: Junyoung Seo,Jaewoo Seo,Seungryong Kim,Laura Leal-Taixé,Xiuming Zhang,Seung Wook Kim
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project page: this https URL
Abstract:Driving world models aim to simulate how a scene unfolds under different ego actions. This requires plausible motion of surrounding vehicles as well as precise ego-trajectory control. Learning vehicle motion from perspective video alone is challenging because motion is entangled with appearance and vehicles can be occluded. Prior work therefore conditions video generation on layouts that specify vehicle motion, providing control over surrounding vehicles but requiring their future trajectories as inputs. We introduce BirdDreams, an autoregressive driving world model that jointly generates perspective video and a schematic bird’s-eye-view (BEV) video, making scene layout part of the predicted future. The BEV, rendered from scene annotations for training, depicts roads and vehicles as simple shapes at a fixed metric scale, providing a top-down view of their positions and motion with reduced occlusion between vehicles. Perspective video, by contrast, retains scene details that the BEV omits, such as traffic signals and object appearance. Joint attention between two video diffusion transformers, one per stream, lets each stream inform the other during generation. The predicted BEV also provides a geometric interface for ego control: we translate and rotate its latest frame along an input ego trajectory and use the resulting frames as a soft condition that shows how the scene would appear to move if only the ego vehicle moved. At inference, BirdDreams generates both videos from a single camera image without requiring an HD map or vehicle-state inputs. On PhysicalAI-AV and nuScenes, BirdDreams follows input ego trajectories more faithfully than prior driving world models and generates more plausible ego motion without ego-action conditioning. Human evaluation shows that BirdDreams outperforms a perspective-only baseline in generating realistic surrounding-vehicle behavior.
[CV-178] RECALL: Real-Time Recalibration of Multi-Camera Rigs via Learned OEM-Anchored Residuals WACV2027
链接: https://arxiv.org/abs/2611.14763
作者: Mario Alberto Tapia de la Cruz,Charalambos Poullis
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: WACV 2027 submission
Abstract:Multi-camera rigs ship pre-calibrated from the manufacturer, but their calibration goes stale each time the rig is repositioned, and drifts continuously through wear, thermal cycling, and vibration. Recovering it on-the-fly is hard: classical bundle adjustment needs a fresh multi-view capture and a global optimization whose cost grows with the rig, ruling out real-time use. In this paper we present RECALL, a feed-forward method that recalibrates a movable multi-camera rig in a single pass. From the nominal OEM calibration as an anchor and a brief observation of a known object, it predicts each camera’s current intrinsics and extrinsics, written as a shared global rotation of the rig composed with a small per-camera drift. Instead of regressing the calibration in one shot, RECALL refines its estimate with a learned, rig-coupled recurrent operator that reprojects through a differentiable projection and applies attention across the cameras at each step, and solves for the shared rotational gauge in-network with a learned Gauss-Newton step rather than regressing it. Because the ground truth is synthesizable, we train entirely on data generated on-the-fly and need no captured dataset. On a synthetic 10-camera benchmark, RECALL is more accurate than per-camera and rig-coupled joint bundle adjustment ( 0.26^\circ vs 0.47^\circ rotation, 5.4 vs 8.9 px focal) and runs about 10^5\times faster at about 0.12 ms per rig. On real data (TotalCapture, CMU Panoptic) it matches or beats offline rig-coupled bundle adjustment on absolute and relative rotation while running four-to-five orders of magnitude faster, and on the Panoptic surround rig it also matches the strongest image-only foundation models (VGGT, VGGT- \Omega , \pi^3 ) on relative pose while staying metric and about 15\times more accurate in focal length. Code: this https URL
[CV-179] SyntheIce: A Synthetic Radargram Annotation Benchmark for Englacial Layer Reconstruction
链接: https://arxiv.org/abs/2611.14728
作者: Muhammad Behroze Hassan,Bayu Adhi Tama,Sanjay Purushotham,Vandana P. Janeja
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 10 pages, 10 figures, 3 Tables, accepted in IEEE International Conference on Big Data (BigData) 2026
Abstract:Englacial layers in radio-echo-sounding radargrams (subsurface ice images) encode the climate and flow history of ice sheets and glaciers. However, large-scale extraction is hindered by time-consuming manual annotation and semi-automated detectors that produce fragmented rather than continuous horizons. We address two challenges: reconstructing complete layers from partial detections across radargram profiles and validating layer continuity across adjacent tile boundaries. We introduce a slope-aware greedy fragment-linking algorithm that reconstructs full-width layers using row-similarity gating and an exponentially weighted moving average (EWMA) slope predictor. Each reconstruction receives a content-based hash ID for stable cross-radargram identification. For quantitative evaluation, we generate SyntheIce, a synthetic dataset of 600 radargrams spanning multiple complexity levels and realistic noise conditions, including 150 wide radargrams divided into four adjacent tiles. All samples include exact ground-truth annotations. We evaluate the dataset using XCheck, a consistency-based validation framework, extending its original two-image setting to four consecutive tiles and enforcing continuity at every boundary. This demonstrates the dataset’s utility for developing and evaluating real-world annotation algorithms. Our findings extend beyond ice science: when spatial features are fragmented but retain consistent local geometry, trajectory-based linking is more reliable than appearance-based matching. Evaluation against expert-labeled real radargrams demonstrates high precision, recall, F1-score, and mean IoU, supporting the method’s transfer to real-world data.
[CV-180] What Entropy-Based View Selection Cannot See: An Audit of Test-Time Adaptation for Vision-Language Models NEURIPS2026
链接: https://arxiv.org/abs/2611.14686
作者: Sivasuthan Sukumar,Uthayasanker Thayasivam
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Accepted at NeurIPS 2026 Workshops on VLM4RWD
Abstract:Vision-language models such as CLIP can classify images into categories they were never explicitly trained to recognize, but accuracy degrades when test images differ in style or source from their training data. A common remedy generates randomly cropped and flipped copies of each test image, called views, and combines their predictions. Since some views miss the object, methods must identify which views to trust. The dominant approach uses prediction entropy to measure confidence and retains the most certain views. Other methods use signals such as embedding typicality. Despite repeated refinements of these scoring functions, the information they rely on, the model’s own response to each view, remains largely unexamined. We investigate this choice while holding other pipeline stages fixed. First, we show that view selection, rather than the gradient update, accounts for almost all accuracy gains. We then examine whether entropy identifies useful views when predictions are wrong, despite at least one view predicting the correct class. In these cases, entropy ranks correct views above incorrect ones no better than chance, recovering only about one-twentieth of the accuracy recovered by an oracle over the same views. This finding holds across two image encoders and under a control ruling out an obvious statistical artifact. An exact analysis explains this limitation: entropy measures confidence but not which class a view favors. We identify an unused signal already available in the augmentation pipeline: the fraction of the original image retained by each crop. Ranking views by this quantity improves accuracy on 14 of 15 benchmark datasets over entropy and on 11 of 15 over the strongest scoring function tested. A pre-method measurement also predicts which datasets benefit.
[CV-181] Fast Slice-Wise Deep Learning Motion Correction for Spinal Cord Functional MRI
链接: https://arxiv.org/abs/2611.14677
作者: Thomas Dagonneau,Sandrine Bédard,Nontharat Tucksinapinunchai,Uten Yarach,Loïc Le Folgoc,Pietro Gori,Julien Cohen-Adad
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 4 pages, 3 figures
Abstract:Respiratory and bulk motion displace the spinal cord between the volumes of a functional MRI time series. Current methods such as the Spinal Cord Toolbox (SCT) take several minutes per time series, and general-purpose brain tools register whole volumes, whereas the cord moves mainly slice by slice. We propose a lightweight convolutional neural network that regresses the in-plane translation aligning each axial slice to a reference. Two models, trained without motion labels, correct large and then residual motion. On a 3T (120 scans) and a 7T (59 scans) dataset, the proposed method improves on SCT on every metric (tSNR +10 to +25%, DVARS -20 to -38%), leaves almost no slice displaced by more than one voxel (0.1%), performs on par with a slice-wise AFNI variant and better than slice-wise FSL FLIRT, and runs in about 10 s per scan, about 30 times faster than SCT. The code is available at this https URL and the method is available in SCT v7.4.
[CV-182] FreeDeshadow: Shadow Removal for User-Specified Objects With Multimodal Prompts
链接: https://arxiv.org/abs/2611.14608
作者: Zhiliang Wu,Jihao Gu,Kun Li,Hanyu Xuan,Hehe Fan,Meng Wang,Weisi Lin
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Existing shadow removal methods either perform global shadow removal or rely on precise shadow masks to guide restoration. Such paradigms limit their ability to selectively remove the shadow cast by a user-specified object in practical applications. To this end, we introduce an object-aware controllable shadow removal task, which allows users to specify a target object through a simple prompt and removes only its cast shadow while preserving other content. However, object-shadow asymmetry makes it challenging to establish accurate correspondence between the specified object and its spatially separated cast shadow and to confine restoration to the target shadow region. To address these challenges, we propose FreeDeshadow, a unified controllable shadow removal framework that supports both visual and textual prompts. It consists of an object-aware generator and a shadow-aware local refiner. The former interprets the user prompt and produces a high-quality shadow-free proposal, while the latter establishes the correspondence between the object and its cast shadow and constrains the generated changes to the target shadow region. To support training and evaluation, we further construct PCSRD+, a large-scale dataset containing 154,571 quadruples at 1024 \times 1024 resolution and covering six types of user prompts. The visual prompts include a dot, a line, and an object mask, while the textual prompts describe the target object using attributes, spatial relations, or states. Extensive experiments demonstrate that FreeDeshadow accurately removes the shadow of the target object under both visual and textual prompts while effectively preserving non-target regions. It also outperforms existing methods in shadow removal performance and controllability. Project Page: this https URL
[CV-183] BEVCast: Path-Adaptive BEV for Efficient End-to-End Autonomous Driving
链接: https://arxiv.org/abs/2611.08631
作者: Qiren Wang,Mingyu Guo,Zixuan Wang,Weisong Shi
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注:
Abstract:Camera-based driving models build a dense bird’s-eye-view (BEV) grid and spend the same computation on every cell of it, although the vehicle crosses only a few of those cells within the planning horizon. We present BEVCast, a two-stage BEV whose resolution adapts to the predicted driving path. A coarse pass encodes the whole scene on a grid of 4.096 m cells. A heat head trained on the recorded path assigns a heat to every coarse cell and the cells with the highest heat are refined into 1.024 m cells under a fixed budget. The remaining cells keep their coarse resolution, so the trajectory head reads a mixed-resolution BEV that still covers the entire scene and needs neither an HD map nor a navigation command. Against a full grid of 1.024 m cells with the same architecture and training recipe, BEVCast at a 10% refinement budget encodes 28% of the BEV queries and needs 66% of the BEV encoder time at batch size 1. On nuScenes its mean displacement error is 1.078 m against 1.149 m and its drivable area violation rate is 1.46% against 2.29%. On NAVSIM navtest its mean displacement error is 0.577 m against 0.598 m and its extended predictive driver model score is 82.51 against 81.28. Next to published planners BEVCast reaches a comparable error on both benchmarks without a navigation command. The resolution of a driving BEV can therefore follow the task instead of the grid.
[CV-184] MS-rPPG: Multi-spectral State Space Model for Remote Photoplethysmography in Driver Monitoring Systems
链接: https://arxiv.org/abs/2606.21115
作者: Jiho Choi,Sang Jun Lee
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Image and Video Processing (eess.IV)
备注:
Abstract:Remote photoplethysmography (rPPG) is a camera-based technique for measuring physiological signals, particularly cardiac activity. From the remotely measured signals, heart rate can be estimated, which is crucial for health monitoring. In this study, we investigate a driver health monitoring system based on remote heart rate estimation. However, driving environments represent uncontrolled settings where videos are subject to varying illumination conditions and frequent head movements. We introduce MS-rPPG, a multi-spectral framework that combines RGB with near-infrared (NIR) face video to alleviate rPPG estimation under challenging driving conditions. To combine the complementary features from two spectral videos, we propose a cross-spectral linear modulation (CSLM) strategy based on frequency-domain analysis. Moreover, we introduce MS-Mamba, a novel state space model designed to effectively model long-range temporal dependencies while jointly capturing cross-channel interactions between multi-spectral features. We collected a real-world dataset called MS-Drive, which was recorded from 50 participants while driving the vehicle. The proposed method was evaluated on the MR-NIRP Car dataset and MS-Drive datasets. The experimental results indicate that MS-rPPG shows better robustness and heart rate estimation accuracy than previous methods, highlighting its promise for driver health monitoring. The codes are available at this http URL.
[CV-185] Spatiotemporal learning improves quantification of pore-space evolution and fluid displacement
链接: https://arxiv.org/abs/2611.17111
作者: Zhuangzhuang Ma,Rukuan Chai,Qianqian Ma,Yanghua Wang,Branko Bijeljic,Martin J. Blunt
类目: Fluid Dynamics (physics.flu-dyn); Computer Vision and Pattern Recognition (cs.CV)
备注: 48 pages, 9 figures, 18 tables, including Supplementary Information
Abstract:Pore-space evolution and fluid displacement influence how rocks transmit and store fluids, affecting geological carbon storage, underground hydrogen storage and groundwater contaminant transport. A single time-resolved micro-CT experiment may produce tens to hundreds of three-dimensional images, and segmentation is required to quantify physical properties and run image-based simulations. Traditional methods such as thresholding perform poorly on noisy, low-contrast images and need manual adjustment per image; 3D deep learning networks segment each image without temporal information, so their errors vary between images and could be mistaken for physical change. Here we introduce a transformer-based 4D segmentation framework combining spatial and temporal information. It uses sequential images to resolve ambiguous phase assignments and separate physical change from segmentation error. The method is tested on images of mineral dissolution: it measures pore-volume growth and rock boundaries more accurately than an equivalent 3D model that ignores temporal information. It also captures the increase in permeability verified against reference segmentations and flow simulations. The advantage transfers to another dissolution sample without retraining. On two other datasets of fluid displacement, it reduces spurious changes in predicted rock and porosity fluctuations and accurately identifies filling events. Spatiotemporal segmentation turns sequences of hundreds of three-dimensional images into more reliable quantitative measurements of pore-space evolution and fluid displacement without per-image parameter adjustment to support the analysis of time-resolved CT imaging.
[CV-186] Physics-based simulation unlocks scaling laws for generalizable MRI reconstruction
链接: https://arxiv.org/abs/2611.15758
作者: Hong Shang,Zijian Zhan,Sen Jia,Xiang Fan,Jintao Meng,Yang Zhou,Xiang Zhou,Tianyu He,Xinran Chen,Qiaoling Zeng,Keyan Yu,Jianhua Yao,Dong Liang
类目: Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Deep-learning-based magnetic resonance imaging (MRI) reconstruction holds great promise for accelerating acquisitions, yet its clinical translation remains bottlenecked by poor generalizability across unseen scanners, populations, anatomies, and protocols. Gathering sufficiently diverse clinical datasets to overcome this bottleneck is hindered by high acquisition costs, privacy restrictions, and institutional biases. Here, we introduce Scalable Inverse Model (SIM), a framework that systematically co-designs physics-simulated training data, model architecture, and learning paradigms around scalability–leveraging compute scaling to overcome historical data bottlenecks and achieve robust zero-shot MRI reconstruction. Across 14 public and in-house datasets–spanning seven international centers, five vendors, field strengths of 0.3-7 T, over ten anatomical regions, more than ten acquisition protocols, downstream quantitative and molecular tasks, and a 150-scan blinded radiologist study–SIM demonstrates robust zero-shot generalizability, recovering fine anatomical details while suppressing aliasing artifacts where conventional deep-learning methods fail. Attention-map analyses confirm that SIM learns physically explainable, dynamically adaptive reconstruction primitives rather than anatomy-specific priors. SIM establishes a scalable, privacy-preserving, and democratized infrastructure for universal MRI reconstruction, providing a blueprint for generalizable AI across inverse problems in medical imaging.
[CV-187] Supervision-Aware Evaluation of MedGemma for Reference-Assisted Radiotherapy Contour Ranking
链接: https://arxiv.org/abs/2611.15370
作者: Jaden Moon,Caiwen Jiang,Songlin Zhao,Wei Liu
类目: Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 5 pages, 1 figure, 3 tables. Submitted to IEEE ISBI 2027
Abstract:Radiotherapy contour ranking can support quality assurance, but vision-language model (VLM) gains may reflect supervision, score construction, available inputs, or representation access. We evaluated frozen MedGemma 1.5 in a single-institution study, a 12-patient held-out HaN-Seg test, five-fold cross-fitting on 30 non-test HaN-Seg patients, and transfer without refitting to 120 SegRap2023 patients. On held-out HaN-Seg, a supervised readout of the final prompt state exceeded a supervised readout of the model’s yes/no output probabilities by +0.3047 pooled average precision (AP; 95% patient-bootstrap interval [+0.2822, +0.3269]); the advantage persisted against a parameter-count-matched nonlinear control. A readout of projected visual tokens exceeded the final-state readout in cross-fitting and transfer (+0.0706 AP averaged equally across patients on SegRap; 95% interval [+0.0607, +0.0804]), while a structured panel-feature readout remained competitive with visual tokens. The anatomy-conditioned readout did not demonstrate an advantage over the native-feature comparator. These results motivate visual-token and structured-feature controls in medical-imaging VLM evaluations; outcomes were computational or synthetic, not clinical acceptability.
[CV-188] Coarse-to-Fine 3D MRI Reconstruction via Resolution-Agnostic Neural Operators NEURIPS2026
链接: https://arxiv.org/abs/2611.14979
作者: Jiayun Wang,Ram Daftari,Weijie Gan,Valentin Duruisseaux,Ruibo Wang,Armeet Singh Jatyani,Boris Bonev,Anima Anandkumar
类目: Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV)
备注: NeurIPS 2026
Abstract:Deep-learning-based 3D MRI reconstruction is limited in practice by the GPU memory and compute cost of full-volume training; 3D acquisitions are thus most often reconstructed by 2D models applied slice-by-slice and then stacked, which ignores through-plane (stacking-axis) structure. We reconstruct in full 3D and ask whether a model can be trained at low resolution (fast and memory-feasible) yet deployed at higher resolution. This coarse-to-fine strategy needs reliable cross-resolution performance, but standard networks are resolution-specific and degrade substantially when evaluated outside the training setting. We therefore use resolution-agnostic neural operators (NOs) and propose KIR-NO (k-space-to-Image Reconstruction Neural Operator), a discretization-agnostic framework for accelerated 3D MRI reconstruction. KIR-NO is trained with full backpropagation on low-resolution volumes and applied zero-shot to higher-resolution volumes at inference. Its core building block is a 3D neural operator based on discrete-continuous convolutions: convolutional filters are parameterized in a continuous basis and sampled at arbitrary voxel resolutions while preserving the local inductive bias of convolution. KIR-NO integrates these learned operators with physics-based consistency: a k-space neural-operator head first refines the undersampled Fourier measurements, and image-space neural-operator cascades then refine the reconstruction through interleaved data-consistency updates. Across 3D knee and brain MRI benchmarks, KIR-NO improves over classical and CNN baselines: it improves over a state-of-the-art 3D CNN by 2.2 dB PSNR at half resolution under 4x acceleration on SKM-TEA, with gains persisting at 8x and 16x. These results show that resolution-agnostic operators make coarse-to-fine 3D MRI reconstruction - cheap low-resolution training with high-resolution deployment - practical across anatomies.
[CV-189] One Block Is Enough: LoopMR for Accelerated MRI Reconstruction
链接: https://arxiv.org/abs/2611.14947
作者: Yimeng He,Yibin Xie,Yi Zheng,Yifan Gao,Lixia Wang,Yin-Chen Hsu,Yufeng Wang,Meng Lu,Debiao Li,Xiuzhen Huang
类目: Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV)
备注: 10 pages, 7 figures, 6 tables. Code: this https URL
Abstract:Deep-learning unrolled networks are the leading approach to accelerated MRI reconstruction. State-of-the-art models such as PromptMR+ learn a distinct block for every cascade, so their size grows with the number of cascades. Sharing one block across iterations avoids adding parameters, but earlier shared-weight models typically used a tiny block and produced worse images than models with distinct weights. Therefore, it remained unclear whether weight sharing itself or the choice of block was the problem. Recent advances in looped transformers, which reuse the same block across layers, show that they can match or outperform their distinct-weight counterparts on several tasks. Motivated by this, we propose LoopMR, which reuses a single Restormer block at every iteration as its core. Besides the current image estimate, the block receives data-consistency inputs and latent images. On four datasets covering the brain (CC359), knee (fastMRI), heart (CMRxRecon2023) and carotid (in-house), LoopMR significantly outperforms E2E-VarNet, PromptMR and PromptMR+ in PSNR and SSIM at every acceleration factor (p 0.01), with about 1/9 of PromptMR+'s parameters (10.3M versus 92.9M). It also keeps its lead at unseen acceleration factors and on an unseen dataset. With our transformer block, the shared model matches or outperforms its distinct-weight counterpart at 6, 12 and 24 iterations. At a fixed number of iterations, shared transformer blocks match or outperform their distinct-weight versions, while shared CNN blocks perform worse. These results suggest that looping one well-chosen block can outperform stacking distinct cascades in MRI reconstruction, with a fraction of the parameters.
人工智能
[AI-0] AgentS SL: Can MLE Agents Leverag e Unlabeled Data? NEURIPS2026
链接: https://arxiv.org/abs/2611.17323
作者: Ethan Lin,Akanksha Sarkar,Ziang Liu,Kristin Branson,Jennifer J. Sun
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: NeurIPS 2026. Project Website: this https URL
Abstract:Applying machine learning systems to new domains introduces bottlenecks in both labeled data acquisition and manual pipeline engineering. While recent MLE agents show promise in automating pipeline design, they have been studied almost exclusively in fully supervised settings, leaving them untested in realistic regimes where labeled data is scarce but unlabeled data is abundant. In this work, we introduce Agentic Semi-Supervised Learning (AgentSSL), framing SSL as a program search problem and studying how agents navigate this space under limited supervision. We find that even under weak feedback and matched compute, agents can synthesize code that combines and adapts established SSL techniques, exceeding the strongest human-engineered baselines in 8 of 12 settings by up to +21.8%. However, our results identify program selection as a critical bottleneck in AgentSSL, where performance drops by as much as 47% when selecting based on the agent’s feedback metric. To systematically study this bottleneck, we introduce ProgramSelectBench, a large-scale benchmark with a combined total of 16,977 agent-generated and human-engineered SSL programs, and evaluate both unsupervised and active program selection methods. We further evaluate AgentSSL on real-world ecology detection tasks, highlighting the opportunities and challenges of transferring agent-discovered strategies beyond controlled benchmarks.
[AI-1] Sketching Language Models for Context Attribution
链接: https://arxiv.org/abs/2611.17313
作者: Christopher Liaw,Aranyak Mehta,Grigoris Velegkas,Felix Zhou
类目: Data Structures and Algorithms (cs.DS); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Machine Learning (stat.ML)
备注:
Abstract:Context attribution asks how removing parts of a prompt changes a language model’s output. We study leave-one-out (LOO) attribution for a prompt of n tokens and a fixed measurement function. Computing all LOO values directly requires n model evaluations. Under a stability assumption, our algorithm estimates all LOO values to error proportional to \varepsilon with constant probability using O(\sqrt\log(1/\varepsilon)/\varepsilon) evaluation calls. To complement our upper bound, we prove a matching \Omega(\sqrt\log(1/\varepsilon)/\varepsilon) lower bound on the number of evaluation calls in an oracle model. Empirically, we verify coefficient decay consistent with stability and evaluate the performance of our sketching algorithm on a small language-model retrieval task. Technically, our result is based on a sketching algorithm for arithmetic circuits whose Taylor coefficients satisfy the stability condition recently introduced by Gunn [Gun26]. A direct adaptation of their forward differentiation sketches would require \widetildeO(\log(n)/\varepsilon^2) evaluation calls. Our key observation is that we can apply forward differentiation to the backpropagation circuit, with only constant overhead over a forward sketch of the same Taylor degree. This reduces the variance of the sketches and is key to our improved evaluation bound. Our lower bound builds on the techniques of Amsel, Chen, Duman Keles, Halikias, Musco, and Musco [ACDH+26, SIAM Journal on Matrix Analysis and Applications] for lower bounding the number of matrix-vector queries to estimate the diagonal elements of random matrices. Subjects: Data Structures and Algorithms (cs.DS); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Machine Learning (stat.ML) Cite as: arXiv:2611.17313 [cs.DS] (or arXiv:2611.17313v1 [cs.DS] for this version) https://doi.org/10.48550/arXiv.2611.17313 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-2] Classical Energies Learn to Reason 1000x Faster than Looped Models
链接: https://arxiv.org/abs/2611.17309
作者: Benjamin Hoover,Hugh Blayney,James C. R. Whittington
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Looped models have emerged as a prominent paradigm for reasoning, with successes on simple reasoning puzzles such as Sudoku-Extreme and Maze-Unique. However, these models can be unstable, slow to train, and use many parameters and FLOPs to solve simple puzzles. Descent on an energy not only mitigates these issues but also offers advantages: built-in early stopping, compute that adapts to problem difficulty, fast training without differentiating through a solver, and interpretable weights that learn the constraints of the puzzle. Here we demonstrate this with a learned version of Hopfield Tank’s energy from the 80s, which we term CRAM, solves 99.996% of Sudoku-Extreme and 100% of Maze-Unique test puzzles after 42 ms and 132 ms of training on one GPU, at least 1,000x faster than looped models on these tasks. CRAM also matches baseline looped models at their highest accuracy on Sudoku-Extreme using at least 282x fewer inference FLOPs. Our work suggests two things. First, that energies are an extremely promising direction for training reasoning from data. Second, a caution against using trivially solvable benchmarks like Sudoku and Maze to test the scalability of looped reasoning techniques.
[AI-3] LiveGeo: Benchmarking Geographic Agents in Continuously Evolving Worlds
链接: https://arxiv.org/abs/2611.17267
作者: Pengcheng Zhou,Xuanyu Liu,Yanchen Yin,Tianjie Ju,Bobo Li,Shengqiong Wu,Hao Fei,Mong-Li Lee,Wynne Hsu
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:We introduce GeoWorld, a controllable and continuously evolving geographic environment platform for LLM agents, and LiveGeo, a benchmark for evaluating reliable decision-making in real-world geographic environments using GeoWorld. Unlike existing evaluations that largely reason and plan over fixed geographic states, GeoWorld is grounded in real-world geographic structure and allows external world states to evolve along their own temporal processes while preserving verifiable state histories. LiveGeo places geographic tasks within this evolving environment, requiring agents to actively acquire task-relevant information, verify whether the evidence supporting prior decisions remains valid, and revise their plans when necessary. Our results reveal that even frontier LLMs struggle with static tasks, with no model consistently excelling across tasks, while strong static performance does not ensure dynamic reliability. Further analysis identifies a key bottleneck: agents struggle to acquire and translate relevant evidence into valid decisions and maintain their correctness as the world evolves. By releasing LiveGeo together with its GeoWorld platform, we aim to provide the community with a flexible and extensible infrastructure for studying the next generation of agents capable of reliable action in evolving geographic worlds.
[AI-4] Ballad: Bandit-Based LLM Routing for Automated Heuristic Discovery
链接: https://arxiv.org/abs/2611.17241
作者: Samidha Verma,Ankit Anand,Sayan Ranu
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Modern automated heuristic discovery (AHD) systems increasingly rival hand-crafted heuristics by leveraging Large Language Models (LLMs) to iteratively mutate and refine solutions. A natural assumption is that more powerful and more expensive models consistently yield better heuristics. We challenge this empirically: across multiple NP-hard combinatorial optimization problems, we find no consistent advantage for larger, costlier models. This raises a deeper question: given a pool of LLMs spanning frontier and freely hosted or open-weight models, how should one adaptively select which model to invoke at each step of the search? The challenge is non-trivial. The optimal LLM may shift over the course of a run, and because one LLM’s output seeds the context for subsequent calls, each model’s reward signal is non-stationary and entangled across the heuristic population, thus violating the independence assumptions of classical online decision-making settings. We address this through Ballad (Bandit-based LLM Routing for Automated Heuristic Discovery), an online bandit algorithm that learns to orchestrate a pool of LLMs into an ensemble that matches or exceeds what any soloist LLM achieves alone. This orchestration is powered by an LLM selection policy that favors rare, high-scoring heuristics over stable averages, a DAG-based credit propagation mechanism through the genealogy of the heuristic population, and discounting of stale observations to remain attuned to each model’s current utility. Across five NP-hard combinatorial optimization problems and a pool of four models of varying capability and cost, Ballad not only matches the best individual LLM in hindsight but, in several cases, surpasses every individual model while reducing API cost by up to 65.6% relative to always using the strongest model.
[AI-5] Al-Ghasali: A Computational Framework for Simulating Islamic Jurisprudential Rulings via Finite State Automata and Large Language Models
链接: https://arxiv.org/abs/2611.17235
作者: Elnaserledinellah Mahmoud Abdelwahab
类目: Artificial Intelligence (cs.AI); Formal Languages and Automata Theory (cs.FL)
备注: 18 pages, 1 figure
Abstract:This paper presents Al Ghasali, a computational framework designed to formalize and simulate Islamic jurisprudential (Fiqh) rulings. Building upon recent set theoretic formalizations of al sabr wa al Taqsim that extract minimal operational rules as simple Horn clause logical programs, this work proves that the semantics of closed Fiqh chapters collapse completely to deterministic Finite State Automata (FSA). We establish a formal computational hierarchy, demonstrating that restricting formalization to finite propositional Horn theories guarantees decidability and constant time (O(1)) verdict extraction after offline compilation. The framework formally proves that hierarchical compositions of sub-chapter FSAs preserve this O(1) complexity, and that temporal juristic requirements, such as Tartib (order), are deterministically flattened into static ontological states, preserving the memoryless property of the automaton. To generate human-readable juristic summaries, Al Ghasali incorporates Large Language Models (LLMs) via an architecture where the deterministic FSA strictly controls the probabilistic LLM. Crucially, the paper formulates sufficient design safety conditions for this FSA LLM integration, ensuring strict adherence to the epistemological boundaries of Usul al Fiqh. To neutralize the risk of LLM hallucination, the system employs a Grammar Guided Constrained Decoding mechanism. We formally define the mapping function that dynamically constructs unambiguous right linear grammars from the FSA execution history. The paper mathematically proves that this architecture, augmented by a Bounded Neutral Lexicon and Valence Locking, guarantees the absolute absence of Decisional Hallucination, Juristic Factual Hallucination, and Semantic Inversion reducing the LLM to a pure linguistic transformer of deterministic juristic facts
[AI-6] Scalable Conversational Assessment: Chat-Based Exams with a Two-Agent LLM Pipeline
链接: https://arxiv.org/abs/2611.17220
作者: Geoffrey Challen,Craig Zilles
类目: Computers and Society (cs.CY); Artificial Intelligence (cs.AI)
备注: Version submitted to SIGCSE TS 2027 (not accepted). A revised version is in preparation
Abstract:Oral exams are compelling because (1) students can be asked to provide their reasoning for an answer and (2) their interactive nature permits follow-up questions to resolve ambiguous answers. But oral exams scale poorly. In this paper, we present scalable conversational assessment, an LLM-based system that conducts interactive text-based conversations with students to assess their ability to make and support an argument. We use a two-agent architecture to separate concerns and reduce the likelihood that answers are leaked to the student: the interviewer interacts directly with the student, but only receives guidance about the next question to ask the student; the hidden evaluator knows the goals of the whole activity, tracks the student’s progress using a rubric, and guides the interviewer on when to follow-up on the current question or proceed to the next question. The system was first tested using a collection of simulated students with LLM personas of a particular ability level or adversarial tactic (e.g., answer-extractors, prompt-injectors, confident fabricators). A novel element of this evaluation was that it required the use of a fictional assessment topic to prevent personas that shouldn’t have information from using knowledge present in the LLM’s pre-training. The system was also deployed in a small, elective AI literacy course in both formative and summative contexts. Through manual and post-facto AI review of the transcripts we identified opportunities to improve the follow-up question generation and ensure that grading only credited ideas that the student (and not the interviewer) contributed.
[AI-7] Introducing Feedback Delays to Non-Delayed Adversarial MDPs: Optimal sqrt S and sqrt K Dependencies in Tabular and Linear AMDPs
链接: https://arxiv.org/abs/2611.17214
作者: Aarush Kulkarni,Khang Nguyen,William Chang,Yan Dai
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
备注:
Abstract:We study regret minimization in episodic Adversarial Markov Decision Processes (AMDPs) with unknown transitions and bandit feedback. In tabular AMDPs with S states, A actions, and horizon H , the best known K -episode regret is \tilde O(HS\sqrtAK) , whose \sqrt S gap to the \Omega(H\sqrtSAK) lower bound has long remained open. In linear AMDPs with d -dimensional features, after a five-year line of improvements starting from K^14/15 , computationally efficient algorithms without a simulator or exploratory assumptions still only attain \tilde O(K^3/4) regret. A common challenge in both settings is to estimate the Q-function of a changing policy: past data comes from earlier policies that differ from (and are correlated with) the current one, while freezing the policy for future episodes worsens the dependency on K . We instead propose a general technique called introduced feedback delays: we estimate the Q-function of the policy using future episodes while still updating the policy, as if the feedback were delayed. While feedback delays are usually an obstacle imposed by the environment, here they are introduced voluntarily and, perhaps surprisingly, come at no extra cost. We develop Geometric Resampling with Rejection (GRR) to account for policy changes during the delay, and threshold the visitation probabilities to avoid long delays on rarely visited state-action pairs. With this unified technique, our policy optimization algorithms achieve, with high probability, \tilde O(\textpoly(H)\sqrtSAK) regret in tabular AMDPs, resolving the open problem; and \tilde O(\textpoly(d,H)\sqrt K) regret in linear AMDPs, giving the first efficient algorithm optimal in K . They also naturally extend to delayed AMDPs with a total feedback delay \mathfrak D , yielding the first bounds optimal in S and \mathfrak D (tabular) and in K and \mathfrak D (linear). Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Machine Learning (stat.ML) Cite as: arXiv:2611.17214 [cs.LG] (or arXiv:2611.17214v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2611.17214 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-8] Cited but Unused: Parametric Provenance in Retrieval-Augmented Generation
链接: https://arxiv.org/abs/2611.17161
作者: Tanzeel Abbas,Tehseen Zia,David Windridge
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Retrieval-augmented generation systems attach citations to attribute claims to their sources. These citations frequently misstate provenance, and standard verification cannot detect this. We measure provenance by counterfactual value substitution: a fact in the retrieved passages is replaced with a false variant and the model re-queried. We find that across five sizes of one model family and a second comparator family, 1.2% to 15.7% of citable claims derive from parametric memory rather than the retrieved context. At the most conservative substitution distance these figures are 4.6% and 8.5% on the two models we tested at it, with the rate rising with model size within a family. We further find that natural language inference verification, the standard approach, gives results indistinguishable from chance across models, even when restricted to the cited passage. We determine that the cause is structural: a model answering from memory selects a citation that agrees with its output. Comparing ten attribution methods on identical claims, the contrast between the model’s distribution with and without context best separates provenance. Acting on that signal does not improve provenance honesty at these base rates. The resulting detector is hence still not accurate enough to act on: it measures what the model knew rather than what it used, and separating those remains the open problem.
[AI-9] Designing Institutions for Autonomous AI
链接: https://arxiv.org/abs/2611.17114
作者: Petter Törnberg
类目: ocial and Information Networks (cs.SI); Artificial Intelligence (cs.AI)
备注:
Abstract:As autonomous AI agents increasingly interact in persistent populations, their collective capabilities and alignment depend not only on individual models, but also on the institutions governing their interactions. Yet AI research has largely focused on individual agents. In this Perspective, we argue that institutional design must become a distinct focus of AI research. Drawing on institutional political economy, social epistemology and organizational theory, we develop seven principles for designing AI institutions that sustain cooperation, enable reliable collective judgments and remain accountable to legitimate human purposes. These address reciprocal contributions, accountable authority, functional differentiation, epistemic pluralism, criticism and correction, human direction, and institutional adaptation. We illustrate their application through machine science, where autonomous research challenges institutions developed around human capabilities and constraints. We then propose an experimental research programme in which matched populations of AI agents operate under alternative institutional arrangements, and describe MoltJournal, an API-first journal for autonomous research agents, as an illustration of how such institutions can be made programmable and observable. Together, these contributions establish institutional design as a research agenda concerned not only with aligning autonomous agents, but with understanding how institutions can transform them into capable, reliable and accountable collectives.
[AI-10] SALA: Semantically Aligned Latent Actions Enable Dual-Axis Scaling of Embodied Foundation Models
链接: https://arxiv.org/abs/2611.17110
作者: Jiachun Jin,Zetong Zhou,Haonan Wang,Xiao Yang,Jun Zhu,Zhijie Deng
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Embodied foundation models hinge on diverse robot interaction data, which are costly to collect. Egocentric human videos offer a complementary source of experience, and latent action models (LAMs) provide a popular means of leveraging these videos for policy supervision. Compared with world-action models (WAMs), LAMs can offer greater training efficiency and potentially lower inference costs by avoiding explicit pixel prediction during policy learning and execution. However, most existing LAMs learn latent actions solely through video reconstruction, which can cause their representations to cluster by visual appearance rather than by the underlying actions. As a result, they struggle to align human actions with their robot counterparts. To bridge the gap, we introduce SALA, which emphasizes establishing the semantic correspondence for latent actions. Without requiring paired data, we use the language descriptions of unpaired human and robot clips to estimate semantic similarities and supervise cross-embodiment correspondence. Specifically, we translate these similarities into soft neighborhood structures that guide human latent actions toward semantically related robot latents, while robot action reconstruction grounds the latent space in control. The resulting shared latent action space provides a unified supervision signal for training an embodied foundation model on both human and robot videos. Notably, latent action supervision enables scaling along two axes: (1) pre-training on increasing amounts of general human video consistently reduces robot action prediction error on downstream tasks, and (2) post-training on increasing amounts of task-specific human video substantially improves real-world robot task success rates.
[AI-11] Speed-to-Power Dominates Energy Costs for AI Data Centers: Technoeconomic Analysis of Powering AI in the Continental United States
链接: https://arxiv.org/abs/2611.17098
作者: Alex C. Newkirk,Daniel L. Gerber,Erica R. H. Fuchs,Alex Hubbard,M. Granger Morgan,Arman Shehabi,Sarah J. Smith,Constantine Samaras
类目: Artificial Intelligence (cs.AI); Systems and Control (eess.SY)
备注: 80 pages incl. SI
Abstract:The rapid growth of artificial intelligence is driving demand for new data centers at large scale and on short deployment timelines. To power them, firms connect to the commercial grid or contract generation directly, but interconnection and gas-turbine delays now incentivize on-site generation. We compare grid-connected and off-grid data center energy infrastructure (AC- and DC-coupled solar-plus-storage, islanded natural gas, and grid interconnection) across the continental U.S. Using an open-source technoeconomic framework built on empirically calibrated AI training loads, climate-driven cooling, stage-by-stage power conversion, and physics-based battery degradation, we model designs providing equivalent energy services. Our central contribution is a speed-adjusted cost of energy that values the compute unlocked by faster construction, making options with different deployment timelines directly comparable on one physically consistent basis. We find DC coupling reduces lifetime energy costs by 18% ( 0.09/ kWh) on average versus AC counterparts by eliminating inverters and improving efficiency. Though rarely cheapest on unadjusted energy cost, DC-coupled microgrids’ faster deployment makes them the lowest-cost modeled option in 85% of U.S. locations once foregone compute is valued, or 55% under accelerated natural gas timelines. For AI infrastructure, deployment speed rather than unit energy cost determines the economically preferred path to power.
[AI-12] Your Vision Foundation Model Is Secretly an EEG Foundation Model
链接: https://arxiv.org/abs/2611.17031
作者: Guoan Wang,Feng Liu
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Electroencephalography (EEG) foundation models are pretrained on thousands of hours of recordings, yet their frozen features transfer poorly to downstream tasks, and this transfer does not consistently improve with larger models or more pretraining data. We show that a vision foundation model that has never seen EEG is a stronger frozen EEG encoder than these specialized models, once the signals are rendered as images that a vision transformer can read. Our method, SeeEEG, keeps the vision backbone frozen and learns only how EEG is displayed to it. Each channel occupies a dedicated row of patches and is drawn either as a waveform at a fixed gain or as a scalogram whose frequency range spans the patch height, so that task-relevant patterns fall within individual tokens. The montage, filtering, and contrast parameters of this display are initialized from a standard clinical montage and optimized with the task loss through the frozen backbone, which amounts to at most 1,113 trainable parameters per view. We compare SeeEEG with seven EEG foundation models on eight clinical, sleep, affective, and brain-computer interface datasets, with every encoder frozen and paired with identical readout heads. With a frozen DINOv3 ViT-B, SeeEEG outperforms every EEG foundation model in balanced accuracy on all eight datasets and reaches a mean rank of 1.3 in Cohen kappa or AUROC, compared with 2.9 for the strongest baseline, REVE. The learned displays transfer without further optimization to the video encoder V-JEPA 2.1, which outperforms every EEG foundation model on both metrics on six of the eight datasets. Replacing DINOv3 with a randomly initialized ViT-B lowers the score by 15.2 points on average, so the ability to read rendered EEG comes from visual pretraining.
[AI-13] Skill Adapter: Personalized Modification and Online Routing of General-Purpose Skills
链接: https://arxiv.org/abs/2611.17007
作者: Qijia Zhuang,Yanyi Gou,Wenbin Ma,Yuxiang Ren
类目: Artificial Intelligence (cs.AI)
备注: 9 pages, 3 figures, 3 tables
Abstract:Large language model (LLM) agents use external skills to reuse task procedures and experience with tools. General-purpose skills describe shared execution procedures, while specific tool choices, operating practices, and output requirements vary across users. In long-horizon tasks, subgoals, available inputs, and environmental states change as execution progresses. Agents must select currently applicable skills from an expanding skill pool and prepare candidates for subsequent steps. We propose Skill Adapter, which enables personalized modification and online routing of general-purpose skills. The system distills local adjustments from user requirements and execution feedback and stores them as Adapters linked to general-purpose skills. The routing agent selects general-purpose skills and applicable Adapters based on the current subtask, execution progress, and environmental feedback, incrementally maintaining a task-level online skill library that the main LLM reads on demand. The system preloads candidates for subsequent nodes and rechecks their applicability once actual inputs are determined. Reinforcement learning optimizes the two types of decisions using modification gains and routing outcomes, respectively. We evaluate Skill Adapter on ALFWorld, WebShop, and SkillsBench. Compared with SkillRL, it improves task success rates by an average of 11.43 percentage points and reduces mean execution time on all three benchmarks. Skill Adapter retains its performance advantage when the skill pool grows to 1,000 skills. Comments: 9 pages, 3 figures, 3 tables Subjects: Artificial Intelligence (cs.AI) Cite as: arXiv:2611.17007 [cs.AI] (or arXiv:2611.17007v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2611.17007 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-14] AnalogLayoutWorld: An Open Benchmark for Agent ic Analog Layout Design
链接: https://arxiv.org/abs/2611.16991
作者: NovaSilicon Team
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Coding agents have demonstrated strong capabilities in software development, yet their ability to tackle long-horizon engineering tasks governed by physical constraints remains less understood. We study this question through analog circuit layout, which demands domain knowledge, geometric reasoning, and iterative interaction with design and verification tools. We introduce \textbfAnalogLayoutWorld, an open-source agent benchmark with 40 expert-reviewed tasks for end-to-end layout generation from fixed transistor-level netlists. It combines an integrated design environment with independent physical and post-layout electrical verification, as well as objective layout quality metrics co-designed with analog layout experts. Across 13 model–harness configurations, 91.1% of attempts produce GDS layouts, but only 23.0% satisfy the benchmark’s acceptance criteria. We also evaluate Novara, an internally developed agent system, which achieves the highest reported aggregate score. Our analysis links performance differences to test-time scaling under fixed budgets, suggesting that reasoning efficiency and effective use of verification feedback help sustain design quality as task complexity grows.
[AI-15] Primer: An AI Framework for Instructor-Centered Educational Simulation Creation
链接: https://arxiv.org/abs/2611.16986
作者: Ethan Mollick,Lilach Mollick,Ben Przystanski,LJ Ciccarelli,Natalie Bach,Laura Zarrow
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Educational simulations are a well-established pedagogical technique that gives learners opportunities to apply knowledge, respond to challenges, and reflect on their performance. High-quality simulations, however, are difficult to build, requiring instructor input, scenario design, character development, scaffolding, technical implementation, and repeated revision. This paper presents Primer, an open-source, LLM-agnostic AI-supported system that helps educators create adaptive educational simulations. Educators contribute their knowledge of the subject, lesson, and students, while generative AI produces responsive variations of practice for students. Primer uses a multi-agent system centered around executing educator’s instructional intent through the entire learning experience. We hope that by releasing Primer as an open-source system, including its prompts and GitHub repository, we provide a foundation for development and research on educator-in-the-loop AI authoring, multi-agent educational workflows, and simulation-based pedagogy. Further, by allowing the use of various AI models, including local and open source models, we provide the opportunity for implementation under varied governance regimes. Primer points toward a model of educational AI in which instructors can create locally grounded practice opportunities based on their own materials, goals, examples, and knowledge of students.
[AI-16] Compiling Known Rules into Compact Neural Agents : Weight-Space Programming on Unseen Games
链接: https://arxiv.org/abs/2611.16981
作者: Jens Lehmann
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 32 pages, 4 figures. Code: this https URL
Abstract:Weight-space programming (WSP) uses language models to construct neural network weights without training. We test whether known game mechanics can be compiled into compact agents that generalise to held-out levels, renderings and start states within a game family. Writers receive simulator source with held-out level data removed for 20 privately authored, unpublished ARC-AGI-3-style game families. With a fixed 57k-parameter recurrent architecture, the strongest writer reaches 56% held-out-level success with feedback, while the four main writers’ Python players reach 97-100%. An exploratory free-architecture condition reaches 68-90%. Learning from support-level examples transfers much less well, but these learners lack the source information given to writers. Written networks have a program-like structure: 93% of parameters are zero, and 90% of the rest are small integers or powers of two. NumPy players under comparable computation caps reach 91%, versus the networks’ 79%. Training the strongest writer’s architectures from scratch yields about 1% success within our budget. Written fixed-network weights are fragile, and standard PPO removes most of their success. Prompts for noise robustness increase tolerance by about four orders of magnitude for two of three writers, but PPO makes larger updates and still removes most success. These results demonstrate rule-to-network compilation in this setting, while showing limited benefits over Python programs and no successful combination with the tested PPO fine-tuning.
[AI-17] he Value of Thinking: Learning When to Reason in Hybrid Decision Models
链接: https://arxiv.org/abs/2611.16961
作者: Zhengxu Yu,Ke Xu,Meng Fang,Haitham Ammar,Weilin Luo,Jun Wang
类目: Artificial Intelligence (cs.AI)
备注: 20 pages, 3 figures
Abstract:Existing adaptive-reasoning methods shorten the explicit reasoning a model writes, or skip it by switching modes; either way the answer is still written out token by token. \emphRather than deciding how much to think, we explore a different paradigm, \emphzero-or- T thinking: a single model either answers from the implicit thinking inside one forward pass, generating no token, or performs full autoregressive reasoning at a cost of T tokens when that is worth paying. We realise it in a \emphhybrid decision model (HDM), where one frozen backbone serves both modes and a lightweight \emphself-routing block chooses between them. The block estimates the \emphvalue of thinking, the utility the slow mode is expected to add over the fast one, from what that implicit thinking has already produced, and the slow mode runs only where that value exceeds a price \lambda set per request. Where no per-decision label exists, as in episodic agentic tasks, we introduce \emphCalibrated Decision Policy Optimisation (CDPO), a reinforcement-learning recipe that trains the block from episode-level returns. We evaluate HDM on MMLU-Pro, AGIEval, GPQA-Diamond, JevBench, BANKING77 and MiniWoB. Specifically, on JevBench HDM-4B scores 0.857 against Jev’s 0.851 ( +0.7% ) at a median 0.179 ,s against 0.652 ( -72.5% ). On the agentic venue MiniWoB HDM-27B cuts token cost by 80.1% .
[AI-18] S1-Evo: Experience-Guided Evidence Competition Attacks on Typed System-1 Decision Models
链接: https://arxiv.org/abs/2611.16937
作者: Kaixiang Wang,Junjie Li,Zhaojiacheng Zhou,Jiong Lou,Jie Li
类目: Artificial Intelligence (cs.AI)
备注: none
Abstract:Typed System-1 (S1) decision models are increasingly deployed in LLM agents for routing, moderation, and authorization to cut inference cost, yet untrusted context can still manipulate typed outputs and alter execution branches. Unlike autoregressive System-2 (S2) models, S1 interfaces score dynamic rules and options without token-by-token decoding, mismatching attacks targeting text continuation or explicit override commands. We propose S1-Evo, a self-evolving framework tailored to S1’s contract-based option competition and margin feedback. S1-Evo perturbs untrusted fields via contract-aware evidence channels to induce \emphevidence competition with original task evidence while preserving fact invariance. Offline on seed tasks, it optimizes candidates via feedback search guided by directional margins and re-verification, distilling verified flips into experience cards. At test time, S1-Evo retrieves relevant cards to construct perturbations for unseen tasks. Across four benchmarks and three S1 models, S1-Evo reaches 71.99% mean ASR and 57.47% high-confidence ASR on \textttjev-1.14, and causes up to 24.00% E2E agent violations.
[AI-19] Universal Infectious Disease Forecasting with Mantis 2
链接: https://arxiv.org/abs/2611.16935
作者: Carson Dudley,Reiden Magdaleno,Yutong Bi,Anthony Valencia,Xiaofeng Liu,Marisa Eisenberg
类目: Artificial Intelligence (cs.AI); Populations and Evolution (q-bio.PE)
备注:
Abstract:Public health agencies need to anticipate how outbreaks will unfold, when they will peak, and how long demand on health services will remain elevated. We previously developed a foundation model, Mantis, which was trained entirely on simulated outbreaks—enabling us to generate the diversity and scale of training data needed to support forecasting across diseases. However, our previous work left unresolved when simulation improves predictions and how it should complement observational data. Here, we varied the balance of simulated and observed training data while holding forecasting tasks fixed and varying the available surveillance history. We found that simulation-rich training was particularly valuable with limited local history, and its relative advantage diminished as more observations became available. We also found that simulation-based training reduced exceptionally large forecast errors, suggesting that mechanistic knowledge can help compensate for limited local evidence. Building on these findings, we developed Mantis 2, which combines simulated and observational training to jointly forecast up to six variables sixteen weeks ahead. Across 30 forecasting tasks, Mantis 2 achieved the best aggregate probabilistic accuracy over short, medium, and long horizons among the machine learning, statistical, and baseline models evaluated. It missed roughly 30% fewer outbreak peaks than the strongest comparison model, with successfully anticipated peaks identified more than six weeks in advance on average. We publicly release Mantis 2, enabling public health agencies and researchers to forecast disease outcomes without developing a separate model for each disease or location. Subjects: Artificial Intelligence (cs.AI); Populations and Evolution (q-bio.PE) Cite as: arXiv:2611.16935 [cs.AI] (or arXiv:2611.16935v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2611.16935 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-20] STEALTH: Steganographic Text Encoding Agnostic to Low-Level Tensor Hardware
链接: https://arxiv.org/abs/2611.16928
作者: William Fleshman
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Provably secure linguistic steganography hides a message inside text sampled from a language model, so that without the key, a stegotext is computationally indistinguishable from an honest model sample. Nearly every such scheme assumes the receiver can recompute the sender’s token probabilities exactly bit for bit. We show this assumption is fragile: across three model families, bit-exact schemes decode almost none of our messages once the receiver’s floating-point precision, quantization, or GPU architecture differs from the sender’s, or even when the receiver merely scores a text in parallel versus one token at a time, with the same model, precision, and hardware held fixed. We present STEALTH, a scheme built for exactly this setting. The sender draws each token with an exponential race driven by keyed pseudorandom noise, so a receiver’s probability mismatch flips a token only when the race is close; a rateless spinal code, decoded by beam search, absorbs the resulting flips without the sender and receiver ever agreeing on a single distribution. We prove stegotexts remain indistinguishable from honest samples under this construction. Empirically, STEALTH decodes every message we sent for every receiver in the sender’s model family, at every precision and quantization level, on smaller models, and on a different GPU architecture. STEALTH is even robust across different model families with separate vocabularies and tokenizers, decoding every message we tested. Our approach carries far more payload per token than prior robust schemes, and trained steganalyzers and a targeted statistical test both fail to distinguish its stegotexts from honest model output.
[AI-21] CADET: Construction of PCB Routing Instances with Executable Demonstrations
链接: https://arxiv.org/abs/2611.16913
作者: Hyungseok Song,Won-Seok Choi,Seohui Bae,Han-Seul Jeong,Junseok Park,Youngjoon Park,Soonyoung Lee
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 28 pages, 10 figures, 15 tables. Code and data: this https URL
Abstract:Printed circuit board (PCB) routing is the stage of physical layout that finds paths connecting the placed pads of each net. Public datasets in this domain are few, and most contain only the final states of routed boards, not demonstrations of the electronic design automation (EDA) engine actions that produce them. To close this gap, we propose CADET (Constructing executAble Demonstrations for Engine-grounded Training), which jointly generates synthetic boards and demonstrations built from EDA engine actions. The resulting RouteDemos provides 100,000 synthetic boards routed with zero design rule violations (DRVs), together with their demonstrations. Their distributions over 17 statistics resemble those of 142 real reference boards. CADET trains a routing policy on these demonstrations with behavior cloning and fine-tunes it with reinforcement learning (RL). On real test boards, the policy raised the rate of complete DRV-free routing from 38.0% to 67.2% over an RL-only baseline. This rate was comparable to that of the open-source autorouter FreeRouting and exceeded those of the other baselines, including GPT-6 Astra. We release the dataset and its generation code as open source at this https URL.
[AI-22] LLM -Assisted Verification in Dafny
链接: https://arxiv.org/abs/2611.16886
作者: Wieger Wesselink,Kees Huizing,Huub van de Wetering
类目: Logic in Computer Science (cs.LO); Artificial Intelligence (cs.AI); Software Engineering (cs.SE)
备注: 16 pages (excluding references), 20 pages including references; 1 figure, 2 tables. Submitted to FSEN 2027
Abstract:Formal verification of complex algorithms requires expertise in program specification, invariant construction, mathematical proof, and the management of automated theorem provers. Although large language models (LLMs) can generate program code and proof fragments, reliably developing complete machine-checked verifications remains challenging: it requires an iterative process of constructing and validating specifications and invariants, building proofs, diagnosing solver failures, and optimizing verification performance. This paper presents a systematic ten-step workflow for LLM-assisted verification of scientific algorithms using the Dafny verification system. The workflow integrates algorithm formalization, contract development, empirical invariant validation, proof construction, bottleneck profiling, and resource-driven proof optimization. We applied the workflow using Antrhopic’s Claude Opus 5 and Opus 5.5 models, combining LLM-driven development with automated testing and verification feedback, while retaining human oversight for correctness-critical decisions and strategic interventions. We evaluate the approach on 37 complex algorithms from graph theory and labelled transition systems. Notably, we verified the Hopfcroft-Tarjan triconnectivity test, demonstrating that even highly challenging algorithmic verification tasks can be carried out through a structured LLM-assisted workflow. Many of our machine-checked verifications were obtained within hours to a few days of interactive development. Our results suggest that LLM-assisted workflows can shift the role of the human verifier from detailed proof construction toward supervision, specification review, and strategic guidance.
[AI-23] Context Graph: Explainable Multi-Agent Decision Chains in AI-RAN
链接: https://arxiv.org/abs/2611.16879
作者: Xijun Wang,Zhaoyang Liu,Xiang Chen,Howard H. Yang,Chenyuan Feng
类目: Networking and Internet Architecture (cs.NI); Artificial Intelligence (cs.AI); Information Theory (cs.IT)
备注: This work has been submitted to the IEEE for possible publication
Abstract:AI-RAN distributes network control across multiple autonomous agents, namely xApps and rApps operating over shared radio resources. When an alert or degradation appears, operators currently lack a structured way to trace which decision caused it, what context triggered that decision, and how other agents interacted. Existing explainability tools focus on single-model feature attribution or linear event logging, neither of which captures cross-agent decision chains. We propose the Context Graph, a heterogeneous directed graph that organizes policies, measurements, decisions, and alerts into a queryable explanation layer. Its foundational primitive is the Decision-Event node acting as a causal junction, where incoming edges capture triggering context and outgoing edges track downstream consequences that are progressively upgraded through counterfactual verification. Instantiated over existing E2, A1, and O1 data flows, our proof-of-concept demonstrates that compliance and diagnosis queries are resolved at sub-millisecond latency over graphs of tens of thousands of nodes, while reliably distinguishing true causal impacts from spurious correlations.
[AI-24] DPA: Decentralized Primal Averag ing with Quasi-Global Momentum for Highly Heterogeneous Data NEURIPS2026
链接: https://arxiv.org/abs/2611.16871
作者: Diyako Ghaderyan,Stefan Werner
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Accepted at NeurIPS 2026. 7 figures. Code available at this https URL
Abstract:Decentralized training of deep learning models keeps training data local and supports on-device learning over networks. In realistic scenarios, heterogeneity across agents’ local data poses an optimization challenge and can severely degrade test accuracy, especially on sparse communication graphs. Building on recent advances in centralized primal averaging, we propose Decentralized Primal Averaging (DPA), a decentralized learning algorithm that combines primal averaging with a local quasi-global momentum buffer. DPA maintains three coupled sequences for optimization, gradient evaluation, and output averaging, while using consecutive mixed iterates to construct a network-aware momentum direction without transmitting an auxiliary tracking variable. For smooth nonconvex objectives, we establish a convergence bound for DPA and identify conditions for linear speedup in the number of agents for sufficiently many iterations. We also introduce DPA-1G, a variant that uses one communication round per iteration and matches the per-iteration communication cost of DSGD-style baselines. Experiments following the GUT decentralized image-classification benchmark show that DPA compares favorably with DSGD, gradient tracking, QG-DSGDm, momentum tracking, and QG-GUTm across datasets, architectures, topologies, and heterogeneity levels, with the largest gains under high heterogeneity on sparse rings. At the same total communication cost, both DPA variants achieve higher mean test accuracy than the best baseline on CIFAR-10 and CIFAR-100 with ResNet-20. Our code is publicly available.
[AI-25] A Modular Framework for ESG Prediction through Domain-Specific Language Modeling and Graph Learning
链接: https://arxiv.org/abs/2611.16831
作者: Sofia Baiocco,Gianluca Bonifazi,Christopher Buratti,Michele Marchetti,Camilla Mazzoli,Federica Parlapiano,Giulia Quaglieri,Davide Traini,Domenico Ursino,Luca Virgili
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Environmental, Social and Governance (ESG) scores are increasingly used to evaluate corporate sustainability, supporting investment decisions, ensuring regulatory compliance and assessing risk. However, current ESG evaluations largely rely on proprietary methodologies that often produce inconsistent ratings. This motivates the development of data-driven prediction frameworks. In this paper, we present ESGPred, a modular framework for predicting ESG scores by combining domain-specific language modeling with graph-based learning over a network that models the relationships among the companies being evaluated. In addition to presenting the rationale behind ESGPred and detailing its ESG score prediction process, we show that ESGPred can effectively predict both overall ESG ratings and their individual Environmental, Social, and Governance components. We evaluated ESGPred using a dataset of 5,266 sustainability reports published by 2,134 publicly listed companies between 2020 and 2023. In this evaluation, we used ESG ratings provided by LSEG Data Analytics as ground truth. Our experimental results demonstrated that integrating graph-based relational learning improves prediction performance over purely text-based baselines.
[AI-26] Scale-Induced Semantic Shift in Visual Prompt Injection Against VLMs and Web Agents
链接: https://arxiv.org/abs/2611.16744
作者: Yiheng Zhao,Runze Ji,Chenyu Yi,Yi Yu,Bingquan Shen,Alex Kot,Xudong Jiang
类目: Artificial Intelligence (cs.AI)
备注: 14 pages, 10 figures, including supplementary material
Abstract:Visual content in multimedia systems may be resized, compressed, or re-rendered before it is interpreted by a vision-language model. We study a security failure mode in which these transformations change the machine-readable semantics of an image: a textual instruction is not recovered at a 512\times512 presentation resolution under our evaluation protocol but becomes recoverable after aggressive downsampling. We formulate this behavior as Scale-Induced Semantic Shift (SISS) and develop a closed-loop black-box framework that combines dual-scale verification, selective self-hardening, shallow-layer low-rank adaptation, and pixel-space noise regularization. On a fixed 300-case benchmark, the third SISS refinement round obtains a 58.75% average attack success rate across four closed-source vision-language models, compared with 41.08% for the strongest generation baseline. A controlled sweep bounds the effect to aggressive compression: average success decreases to 17.68% , 6.85% , and 5.60% at 128\times128 , 224\times224 , and 336\times336 , respectively. In a downstream web-agent study, the generated images increase hidden-instruction perception, but absolute behavioral compromise remains about 4% . Generic input transformations provide little protection in a separate defense protocol, whereas the proposed ScaleGuard consistency detector reduces non-adaptive success from 61.4% to 24.3% and one-round defense-aware success from 42.5% to 15.0% , with no observed rejection among matched clean controls. The results characterize a bounded but reproducible semantic inconsistency in multimedia preprocessing and motivate scale-consistent visual-language inference.
[AI-27] EgoX: Egocentric Cross-Embodiment Manipulation with Embodiment Dreaming
链接: https://arxiv.org/abs/2611.16728
作者: Yaru Niu,Binghong Chen,Changwei Yao,Shangtao Li,Zhenlong Fang,Revanth Krishna Senthilkumaran,Shuai Zhou,Yuemin Mao,Kalie Ching,Po-Ju Chao,Yudong Luo,Peiqi Yu,Heng Zhang,Hao Zhang,Bingqing Chen,Chen Qiu,H. Eric Tseng,Ye Yuan,Umar Iqbal,Changliu Liu,Jonathan Francis,Ding Zhao
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Generalist manipulation policies can benefit from the complementary experience of humans, real robots, and simulated agents. However, differences in morphology, sensing, and control interfaces make it difficult to learn effectively from their combined demonstrations. We introduce EgoX, an egocentric cross-embodiment learning framework that organizes heterogeneous demonstrations through semantically corresponding observation and action modalities. EgoX pretrains a Modularized Cross-Embodiment Transformer with Embodiment Dreaming (MXT+) on this mixture and fine-tunes it on target-embodiment demonstrations. Within each embodiment, MXT+ assigns a dedicated tokenizer to each observation modality and a dedicated expert to each action modality, with a Transformer trunk shared across embodiments. A training-only embodiment-dreaming objective predicts latent representations of future visual and proprioceptive states, encouraging the shared trunk to model each embodiment’s dynamics. Across 13 task-embodiment pairs spanning five tasks and six robot embodiments, including five physical robots and one simulated humanoid, cross-embodiment pretraining raises MXT+'s average normalized task progress from 0.63 to 0.80 over training from scratch. The gains are particularly pronounced under out-of-distribution conditions, where average task progress increases from 0.53 to 0.77. Ablations demonstrate the benefits of broader pretraining mixtures, modular interfaces, and embodiment dreaming, with dreaming producing more similar latent geometry across embodiments. Project page: this http URL.
[AI-28] From Data to Wisdom: Rethinking AI Through an Epistemic Lens
链接: https://arxiv.org/abs/2611.16726
作者: Afonso Eduardo
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 23 pages, 7 figures, 4 tables
Abstract:This paper proposes the Data-to-Wisdom Epistemic Architecture for AI (EAAI), a conceptual framework treating Data, Information, Knowledge, Understanding and Wisdom as five distinct but recursively interacting epistemic functions: observation, contextual interpretation, modeling and prediction, critical evaluation and informed judgment. Two observations motivate it: sequential readings of the Knowledge Pyramid obscure how these functions interact, and a targeted review of ten Bayesian reference books finds that prior distributions are defined mathematically while “prior information” and “prior knowledge” are not consistently distinguished as epistemic categories. A Bayesian interpretation relates the likelihood, prior and posterior distributions, posterior prediction, model checking and decision analysis to the five functions, without limiting EAAI to Bayesian systems. Extending “prior” in an epistemic sense, the paper distinguishes five kinds of prior context that can shape the model, the likelihood, the prior, evaluation and action. Three worked applications – quantitative financial risk, actuarial modeling in health insurance and agentic AI for medical reimbursement claims – show how EAAI structures the analysis of model assumptions, evidence interpretation, validation, dependencies and decision consequences, separating the prediction of a risk, the appraisal of its reliability and the justification of action. It draws implications for AI and machine learning, including evaluation by function, treating intermediate model outputs as evidence open to challenge and separating generation from appraisal, and compares EAAI with work on DIKW, Bayesian workflow, AI evaluation and risk governance. Its contributions are an epistemic vocabulary with recursive relationships, a more accessible reading of Bayesian methodology, a typology of prior context and clearer methodological terminology.
[AI-29] Mind the Accuracy-Approval Gap: How Users Evaluate Sycophantic AI in Everyday Tasks
链接: https://arxiv.org/abs/2611.16722
作者: Wajiha Naveed,Brandon Fain
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models (LLMs) often exhibit sycophancy, a tendency to agree with users, flatter or affirm them, and preserve their face, even when independent judgment or critical feedback would be more useful. Prior work has largely evaluated sycophancy through benchmark evaluations, but less is known about how users perceive and respond to sycophantic AI in everyday tasks. We combine benchmark evaluations with two large-scale user studies. We study four personas defined through system prompts: objective (instructed to prioritize accuracy over agreement), soft-sycophantic, hard-sycophantic, and baseline (no persona-specific system prompt). Benchmarks show that even baseline exhibits substantial sycophancy, agreeing with users’ stated views 71.7% of the time on NLP Survey, compared with 54.8% for objective and 89.5% for hard-sycophantic. Under conversational pressure on TruthfulQA, accuracy ranged from 81.9% for objective to 59.6% for baseline and 8.7% for hard-sycophantic. Participants were assigned to one of the four personas, and either discussed contested policy topics or collaborated with AI on email writing. Despite its higher accuracy under pressure, objective AI was rated significantly less respectful and satisfying than the other personas, and less trustworthy than baseline. Baseline and sycophantic personas did not differ significantly. Persona did not significantly affect how often participants changed their opinions, although participants who discussed with objective AI had greater odds of reduced confidence. Objective AI prompted greater writing expansion than soft-sycophantic AI, but final writing quality did not differ significantly across personas. These findings highlight a tension between accuracy and user approval: personas with substantially lower accuracy under conversational pressure received more favorable ratings for respect and satisfaction than objective AI.
[AI-30] MusicRWKV: A Multimodal RWKV Framework for Music Representation Learning
链接: https://arxiv.org/abs/2611.16718
作者: Terry Tan
类目: ound (cs.SD); Artificial Intelligence (cs.AI); Audio and Speech Processing (eess.AS)
备注:
Abstract:Recent music foundation models have achieved remarkable progress in music representation learning, yet most existing approaches rely on Transformer architectures whose quadratic attention complexity introduces substantial computational and memory overhead when modeling long-duration music recordings. In this work, we propose MusicRWKV, a multimodal music representation learning framework built upon the RWKV architecture. MusicRWKV jointly models waveform, spectrogram, and metadata information within a unified self-supervised learning framework. To improve efficiency for long musical sequences, we introduce an invertible compression module that reduces sequence length while preserving reconstruction capability. A FiLM-based conditioning strategy is employed to facilitate modality-aware representation learning without introducing expensive cross-attention operations. We further investigate the integration of Soft ROSA as a differentiable retrieval mechanism for neural-symbolic interaction. Experimental evaluation across a diverse collection of Music Information Retrieval (MIR) benchmarks demonstrates that the proposed RWKV-based architecture can be effectively adapted to heterogeneous music understanding tasks. A controlled pretraining ablation shows that Soft ROSA improves validation performance, while layer-wise telemetry analysis reveals a progressive transition from distributed retrieval to selective retrieval in deeper network layers. These results suggest that RWKV provides a promising architectural alternative to Transformer-based models for multimodal music understanding.
[AI-31] Acoustic Respiratory Phase Detection on Mobile Devices
链接: https://arxiv.org/abs/2611.16710
作者: Julian Szymanski,Tomasz Sankowski,Piotr Sulewski,Iwo Czartowski,Javier Medina Quero
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Breathing monitoring is important for well-being, stress regulation, and guided breathing exercises, yet existing approaches often rely on contact-based sensors or controlled environments. This paper presents a non-invasive framework for respiratory phase detection using acoustic signals recorded on commodity smartphones. The proposed hybrid CNN-Transformer architecture combines local spectral feature extraction with long-range temporal modeling to capture the cyclic structure of breathing. The model is evaluated using a Leave-One-Subject-Out (LOSO) protocol on the Foundation Dataset and on an independent unseen test set comprising over 1300 breathing cycles and 2 hours of continuous recordings. The proposed 3-class formulation achieves 72.85% frame-level accuracy and 96.6% cycle-level exhalation detection efficiency in the LOSO evaluation, while achieving 76.94% frame-level accuracy, a Cohen’s Kappa score of 0.65, and 99.2% cycle-level exhalation detection efficiency on the unseen test set. Results suggest that accurate respiratory monitoring may be achievable under data-constrained conditions, indicating that appropriate temporal modeling can learn transferable respiratory representations and generalize to previously unseen participants. An end-to-end ONNX deployment pipeline enables privacy-preserving on-device inference with latency below 20 ms, supporting real-time monitoring on mobile and wearable devices.
[AI-32] MobileProbe: Adaptive Security Evaluation of Mobile Agents via Trajectory-Aware Injection NDSS2027
链接: https://arxiv.org/abs/2611.16693
作者: Yingbin Jin,Ji’an Wang,HaoYang Chai,Hanjun Luo,Li Hu,Xinfeng Li,Huadi Zheng,PeiZhao Hu,Qingqing Ye,Haibo Hu
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注: Accepted at NDSS 2027
Abstract:Mobile GUI agents read the screen of a phone and operate apps on a user’s behalf to complete multi-step tasks such as online shopping or messaging. Because the agent acts on whatever the screen shows, an attacker can embed malicious instructions into ordinary UI content and redirect its actions. We refer to this threat as environmental injection attack (EIA). Existing benchmarks measure this risk by pinning hand-crafted payloads to a fixed step of a fixed task. This approach faces two limitations. First, each payload is bound to one task in one app and does not transfer without manual redesign. Second, agents given the same task may follow different routes, so an agent that never reaches the injected page is recorded as safe without being tested. This insufficient test coverage can cause fixed-point benchmarks to underestimate the security risk of deployed mobile agents. In this work, we propose MobileProbe, a trajectory-aware red-teaming framework that evaluates mobile agents adaptively. MobileProbe discovers attacker-controllable content surfaces through automated pentesting, localizes vulnerable decision points along the agent’s trajectory, and iteratively refines context-aware payloads through a surrogate-filtered evolutionary loop. We construct a benchmark of 83 long-horizon attack instances spanning 12 mobile applications. Each is annotated with a Confidentiality, Integrity, or Availability (CIA) attack goal to cover the principal categories of security threats facing mobile agents. On three representative agent architectures, MobileProbe attains higher attack success rates than human-written, zero-shot LLM, and prior automated red-teaming baselines adapted to the mobile domain. MobileProbe can serve as a pre-deployment auditing tool to systematically uncover vulnerabilities and inform security hardening of mobile agents. Comments: Accepted at NDSS 2027 Subjects: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI) Cite as: arXiv:2611.16693 [cs.CR] (or arXiv:2611.16693v1 [cs.CR] for this version) https://doi.org/10.48550/arXiv.2611.16693 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-33] Edge-Gradient Augmentation under Reinforcement Learning for Temporal Graph
链接: https://arxiv.org/abs/2611.16678
作者: Yongqing Zhou,Jiawei Chen,Wenjun Wang,Xinyu Wang,Wei Yu,Qiyao Peng
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Temporal link prediction is critical for understand- ing dynamic interactions in evolving networks, informing appli- cations ranging from social network analysis to recommendation systems. Existing methods typically rely on raw temporal graph data, often neglecting explicit edge-change information that can significantly improve predictive performance. To address this gap, we propose edge-gradient reinforced learning (EG-RL), a framework that explicitly encodes edge-gradient events (edge births and deaths) through dedicated node and link encoding modules. EG-RL uses a policy-driven reinforcement-learning mechanism to dynamically select and augment the most infor- mative edge snapshots, guided by critic-based feedback to opti- mize prediction accuracy. The model comprises distinct modules for node encoding, link encoding, token mixing, and channel mixing, integrated through global pooling and followed by a two-layer multilayer perceptron classifier. Extensive experiments on multiple dynamic graph datasets demonstrate substantial gains in area under the receiver operating characteristic curve and average precision. Ablation studies further confirm the necessity of adaptive edge-gradient augmentation and validate the effectiveness of the proposed encoding and reinforcement- learning mechanisms.
[AI-34] Natural Gradient Learning in Predictive Coding through Lateral Circuits
链接: https://arxiv.org/abs/2611.16666
作者: Lhea Beumer,Sander de Haan,Niclas Pokel,Pau Vilimelis Aceituno,Benjamin F. Grewe
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Predictive coding (PC) trains neural networks through local energy-minimization dynamics without depending on a global error signal that backpropagation (BP) requires. Although PC is a promising alternative algorithm for efficient learning, the computational cost of inference dynamics currently outweighs the resulting learning benefits in practice. In this work, we take inspiration from optimization theory to extract more learning benefit from each inference phase. We introduce Natural PC, which augments standard PC with a lateral circuit, enabling exact and efficient natural gradient learning locally in space and time. Our lateral circuit learns the activity correlations that determine the Fisher information matrix and implicitly computes its inverse, avoiding expensive matrix inversion. Across all tested image reconstruction benchmarks, Natural PC achieves state of the art in terms of sample and compute efficiency compared to existing PC approaches. Most notably, Natural PC scales to ImageNet and matches the performance of BP trained with AdamW and various second-order optimizers at an order of magnitude lower computational cost while also using two orders of magnitude fewer training samples. Our results demonstrate that local natural gradient learning makes PC competitive with BP in both sample efficiency and computational cost in reconstruction tasks.
[AI-35] Reducing Batch Makespan for LLM Agent Workloads through Inference-Environment Coordination
链接: https://arxiv.org/abs/2611.16661
作者: Wei Feng,Meiling Chen,Boxiao Chen,Lu Liu,Bin Zi,Libin Ye,Xingyu Lu,Lei Xu,Jun Zhou
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Artificial Intelligence (cs.AI)
备注: 26 pages, 12 figures. Technical report
Abstract:Efficient batch execution of agent tasks is critical to LLM training and development workflows, including benchmark evaluation, synthetic-data generation, and RL rollouts. Since these tasks alternate between LLM inference and environment interaction, we present coordinated optimizations to reduce batch makespan, the time required to complete all tasks in a batch. For hybrid LLMs, a workload-based resource model guides GPU-memory allocation between key-value (KV) cache and recurrent state and provides a reference sample concurrency to exploit parallelism while limiting recomputation. Trajectory-aware routing and scheduling promote cache reuse while balancing worker load. Under constrained environment capacity, we decouple sample admission from environment execution and introduce Chain-Advancing Action Scheduling (CAAS), which prioritizes long chains using observed progress to shorten the batch tail. Controlled replay combines actual inference with recorded environment responses and durations for ablations of individual optimizations. Live evaluations with Ling-3.0-Flash and Ling-3.1-Flash on the BFCL, SWE-bench Multilingual, WideSearch, and MCP-Atlas benchmarks demonstrate makespan reductions of 10.4% to 65.0%, while evaluation scores remain aligned. Furthermore, these methods have been successfully applied to training and evaluation workflows for the Ling-3.x model series, accelerating the model-development cycle.
[AI-36] InertiaMPDD: Multimodal Personalized Depression Detection via Emotional Inertia-Inspired Dynamics Modeling
链接: https://arxiv.org/abs/2611.16647
作者: Changzeng Fu,Songyang Wang,Chang Gao,Xinyi Zhang,Peilin Li,Mo Meng,Jiarui Wu,Mingze Li,Hongwei Ding,Shiqi Zhao,Yong Li,Lang He
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Multimodal depression screening estimates risk from speech, visual behavior and language. Current methods mainly predict labels, with limited evidence of departures from a non-depressed reference. Recording-level supervision also leaves fine-grained emotional dynamics unannotated. Motivated by emotional inertia, the persistence of affect, InertiaMPDD learns healthy next-state patterns using a world-model-inspired latent predictor. Using pretrained multimodal representations, it predicts from observed history and context. Reference deviation, residual decay rate and a Jacobian spectral descriptor form reference-derived emotional dynamics descriptors, evaluated as candidate proxy biomarkers for standalone screening and personalized decision correction. Across five cohorts from four datasets, fusion reaches 98.41% F1 on D-Vlog and 84.98% on LMVD. Standalone descriptors outperform fixed additive fusion on MPDD but discriminate weakly on AVEC2014. These findings distinguish reference informativeness from fusion benefit and reveal context-dependent utility.
[AI-37] Folio: Recoverable Multi-Resolution Context Folding for Long-Horizon Language Agents
链接: https://arxiv.org/abs/2611.16627
作者: Yubo Ruan,Tao Peng,Jiao Zhang,Chuanguo Ma,Lei Xu,Han Zhou,Wenwu Wang,Xiafei Lv,Xingjie Zhou,Guiju Fan,Hao Wang,Fenghui Zhang,Mao Miao,Shuo Li
类目: Artificial Intelligence (cs.AI)
备注: 13 pages, 6 figures, 5 tables
Abstract:Long-horizon language agents accumulate tool results and replies until the history no longer fits the context window. Current agents shrink this history by having a model summarize it, so each compaction rereads and rewrites the history it replaces, and cache-aware triggers decide when to compact while taking the size of the result as given. We present Folio, a context manager that records, folds, and recovers history through one set of handles. While the agent runs, Folio records three linked levels: a short turn note the model writes as part of its ordinary reply (L1), an event manifest of every call written by the runtime (L2), and the captured source records (L3). Because these views already exist, the runtime folds the context programmatically, selecting coarser views for older turns. The fold prints the handles that index L2 and L3, so recovery tools that read bounded pages, compute over full text in files, and keep findings in a persistent dossier reach the evidence the agent needs with little added input. The fold trigger prices an explicit landing point that includes the retained budget. On 75 LOCA-bench tasks with DeepSeek-V4-Pro behind every system, Folio solves 66 tasks against 54 for the same runtime with model-written compaction and 49-59 for Claude Code, Codex, and SoL-Pi, in about half their working time, at 2.4 times lower cost than that runtime; each fold takes a median of 0.6 seconds.
[AI-38] Studies on Sign Language Retrieval in Low-Resource Dictionary Settings
链接: https://arxiv.org/abs/2611.16621
作者: Ushnish Sarkar,Vineet Kumar Rakesh,Bhaswar Chattopadhyay,Tapas Samanta
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Most sign languages have no annotated corpus, only small dictionaries with one or a few example videos per sign. This study examines how well a pose encoder trained on a single sign language serves such languages without any target-language training. An encoder trained with metric learning on American Sign Language (ASL-Citizen) is evaluated frozen on six datasets of Turkish, Russian, Indian, Argentinian, Pakistani and Sri Lankan sign languages, with one or four clips per sign enrolled and, where signer identities are available, the query’s signer excluded from the gallery. With one enrolled clip, the encoder retrieves the correct sign first in 36.4-80.7 % of queries, against 16.0-69.6 % for the best learning-free method (angle-based dynamic time warping) and a six-target mean of 39.3 % for an ASL-trained OpenHands graph network. In a crossed ablation of the encoder (embedding geometry, aggregation of body and hand embeddings, training objective, and attention between articulators), two choices matter: a per-pair InfoNCE objective outperforms supervised contrastive and cross-entropy training, and blocking attention between body, left hand and right hand inside the encoder raises accuracy by up to 5.1 points. The best configuration reaches 68.0 % one-shot accuracy (59.0 % for the reference encoder). On CISLR, an Indian Sign Language dictionary of 4,765 words with one video per word, the encoders find the correct word first for 17.3-19.0 % of queries, above the 16.8 % of the original RGB baseline. Over eight seeds with the joint encoder, spherical, Euclidean and hyperbolic embedding spaces differ by at most 1.6 points, except the hyperbolic token-attention head trained with InfoNCE, which is 1.4-2.0 points worse, so the geometry is a minor factor. Subjects: Artificial Intelligence (cs.AI) Cite as: arXiv:2611.16621 [cs.AI] (or arXiv:2611.16621v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2611.16621 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Ushnish Sarkar [view email] [v1] Thu, 12 Nov 2026 10:57:50 UTC (3,121 KB) Full-text links: Access Paper: View a PDF of the paper titled Studies on Sign Language Retrieval in Low-Resource Dictionary Settings, by Ushnish Sarkar and 3 other authorsView PDFHTML (experimental)TeX Source view license Additional Features Audio Summary Current browse context: cs.AI prev | next new | recent | 2026-11 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)
[AI-39] Singularity-Bench: Can Agents Build Agents ?
链接: https://arxiv.org/abs/2611.16591
作者: Yang Zhou,Krish Agarwal,Zhuoming Chen,Zerui Wang,Zhenyu Zhang,Wenshuang Li,Xinpeng Wei,Zhichen Zeng,Jingyu Liu,Jue Wang,Sen Wu,Ziheng Jiang,Zhen Dong,Han Cai,Binhang Yuan,Ying Sheng,Ce Zhang,Bo Li,Beidi Chen
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Frontier AI agents increasingly succeed on realistic, long-horizon tasks and are beginning to improve the systems through which they operate. Yet existing benchmarks often restrict agents to narrow action spaces, use simplified environments, or rely on verification methods that are unreliable for evaluating arbitrary agent-generated systems. We introduce Singularity-Bench, an open-ended benchmark for autonomous AI development with six end-to-end tasks spanning pre-training, mid-training, post-training, inference, the agent harness, and cluster orchestration. Each task requires agents to implement core systems from a minimal starting point and evaluates them against mature human-engineered baselines on identical hardware. Additionally, through extensive red-teaming covering 9 of the 10 categories in the OWASP Top 10 for Agentic Applications, we expose reward-hacking vulnerabilities in the conventional Harbor framework and develop Breakwater, an extension of Harbor that blocks all attacks tested in our study. Across 11 agents built from 7 frontier models and 2 harnesses, Opus 5.5 with Claude Code matches or outperforms the human baseline on all six tasks with a Singularity-Index of 1.25, suggesting it may have the skills needed to reach the singularity. The remaining agents average only 0.58. Our vision is for Singularity-Bench to track how much of AI development agents can already automate and when they may transition from assisting to taking end-to-end responsibility for building stronger successors.
[AI-40] AstroMemBench: A Benchmark for Memory Systems in Embodied Spacecraft Agents
链接: https://arxiv.org/abs/2611.16590
作者: You Wu,Guhong Zhang,Hongwei Li,Xinyue Hong,Yuan Jia
类目: Artificial Intelligence (cs.AI); Robotics (cs.RO)
备注: 31 pages, 4 figures, 19 tables. Includes Appendices A-E
Abstract:Long-term spacecraft autonomy requires embodied agents to use operational history under communication delays, time-critical operations, environmental uncertainty, complex physical interactions, power and computational constraints, and safety and reliability requirements. We introduce AstroMemBench and a Scenario-Challenge-Capability design method that maps these requirements to executable memory tests. The benchmark covers 16 memory subchallenges through 140 templates and 2240 instances across nine tasks in planetary surface operations, on-orbit servicing, and deep-space exploration. A deterministic environment scores actions for task completion, state changes, safety, and resource use. Ablations under nine conditions show that the benchmark is sensitive to how accurately agents recover and use historical information and to differences in memory-system design. Main experiments compare six configurations. Mem0, Chain-of-Memory, LightMem, and retrieval-augmented generation (RAG) improve same-session mean scores over native Pi Agent, with no clear advantage of the three memory systems over RAG. After a session restart, Mem0 outperforms RAG by 12.25 points. Benefits favor instruction retention and result reuse over ongoing state tracking and energy reservation. AstroMemBench supports the design, development, and selection of memory systems for embodied spacecraft agents. AstroMemBench 1.0 and its evaluation code are publicly available: this https URL.
[AI-41] Large Language Models as Digital Twins: a Distribution-Aware Questionnaire-Based Evaluation
链接: https://arxiv.org/abs/2611.16586
作者: Francesco Di Cursi,Chiara Boldrini,Andrea Passarella,Marco Conti
类目: Artificial Intelligence (cs.AI); Computers and Society (cs.CY)
备注: This work was partially funded by the European Union within the National Programme ‘Research, Innovation and Competitiveness’ 2021-2027, project SINTESI (Nr. MI_0000962). F. Di Cursi’s, C. Boldrini’s, A. Passarella’s, and M. Conti’s work was also partly funded by the PNRR - M4C2 - Investimento 1.3, Partenariato Esteso PE00000013 - “FAIR”
Abstract:This paper investigates whether Large Language Models can answer a personality questionnaire as a specific individual would. Given a user’s social-media posts, the model completes the BFI-2-60 Big Five questionnaire as that user. The predicted responses are converted into trait scores and compared with the user’s actual scores. We evaluate GPT-4, Gemma, Llama 3.1, and Mistral on the MyPersonality dataset. To determine whether predictions reflect information from the user’s writing, we compare them with answers produced without user text and with values derived from general population tendencies. We measure prediction error across all users and separately for individuals with low, intermediate, and high trait scores. We also test whether an LLM can express an explicit personality profile when given only the target Big Five scores, without user text, and examine how predictions change as more text is provided. The results show that user text does not consistently improve performance. Models perform better near the middle of the trait distribution, while at the lower and upper ends they often remain close to central values or perform well at one end but poorly at the other. When given an explicit profile, GPT-4 most clearly translates trait information into consistent questionnaire responses, particularly for Extraversion and Neuroticism. Increasing context does not systematically reduce error: open-source models fluctuate more, whereas GPT-4 remains comparatively stable. Overall, the evaluated models do not consistently reproduce the questionnaire behavior of specific individuals.
[AI-42] SPARC: Accurate Photonic Matrix Multiplication via Representative Vectors
链接: https://arxiv.org/abs/2611.16561
作者: Yasuhiro Fujiwara,Tatsuhiro Nakamori,Hideyuki Kawashima
类目: Artificial Intelligence (cs.AI)
备注: 11 pages, 5 figures, 3 tables
Abstract:Photonic neural network accelerators have emerged as a promising solution for reducing the power consumption of deep neural network (DNN) inference by executing matrix multiplication on analog photonic hardware. However, their inference accuracy is limited by low-bit quantization required for photonic processing and by analog photonic noise. This paper proposes an approach for improving inference accuracy by exploiting the high similarity among activation vectors in DNNs. Specifically, our approach constructs a representative vector for each group of similar activation vectors, computes its inner product digitally with high precision, and offloads the residual inner products to the photonic accelerator. Since the activation vectors within each group are highly similar, the residual vectors have much smaller norms, thereby reducing the effects of quantization errors and analog photonic noise. Furthermore, our approach preserves the photonic inner-product computation pipeline, allowing seamless integration into existing and future photonic accelerators without hardware modification. Experimental results on six representative DNN models demonstrate that the proposed approach can improve inference accuracy across diverse network architectures and application domains.
[AI-43] AI has its own style in mathematical proving
链接: https://arxiv.org/abs/2611.16527
作者: Xiaokai Li,Ziyu Zhou,Qibin Cheng,Yangcheng Wu,An Zeng,Ying Fan
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:As large language models increasingly generate mathematical proofs, whether they develop a proving style distinct from that of human mathematicians remains an open question. This question is particularly pressing as mathematicians increasingly voice concerns that proofs generated by large language models are difficult to interpret. Here we ask three large language models to independently reprove all 465 propositions of Euclid’s \textitElements and compare their citation choices against the human-proof baseline through Jaccard similarity and complex network analysis. We find that AI-generated proofs differ from human proofs, and this difference is stable and reproducible: mean pairwise Jaccard similarity among AI runs (~0.68) is nearly twice that between AI and human proofs (~0.37), and cross-run classification consistency is also high. By examining citation structure in detail, we identify the concrete signature of this difference: approximately 40% of AI citations point to foundational elements (definitions, common notions, and postulates), compared with only 13% in human proofs; even among theorem-to-theorem citations, AI tends to cite more distant results. Roughly half of AI’s extra citations trace to the upstream dependencies of the human proof. Together, these findings suggest that the AI proving style is a tendency to construct arguments on its own from foundational elements rather than directly citing later, more established results, which we term “building from the ground up”. Our findings offer a new perspective on how AI systems organise and invoke mathematical knowledge, with implications for understanding the interpretability of AI-generated proofs.
[AI-44] ZONOS3-Mini Technical Report
链接: https://arxiv.org/abs/2611.16525
作者: Gabriel Clark,Sofian Mejjoute,George Close,Beren Millidge
类目: ound (cs.SD); Artificial Intelligence (cs.AI)
备注: 19 pages, 4 figures, 12 tables. Model weights: this https URL ; code: this https URL
Abstract:ZONOS3-Mini is an ultra-compact, ultra-fast, zero-shot voice-cloning TTS model. The model can run in real time on a laptop or smartphone. For its latency and size, it achieves state-of-the-art accuracy and cloning fidelity. To achieve this we propose a latent MoE backbone with only 886M parameters (100M active per token) and a new audio codec, Zyphra Audio Codec 1 (ZAC-1), which represents 44.1 kHz audio at a 15 Hz frame rate. On the new ZTTS2-Eval benchmark, ZONOS3-Mini reaches a 6.71% macro-average WER over 55 language sets (35 clean and 20 in the wild), which is competitive with larger and slower state-of-the-art models. We release the model weights, inference, and training code.
[AI-45] A Joint LLM and Optimization Agent Solution for Drone Scheduling in Victim Search Missions
链接: https://arxiv.org/abs/2611.16524
作者: Yinghao Zhang,Yalin Liu,Hong-Ning Dai,Yan Zhang,Qubeijian Wang,Bishenghui Tao,Yaohan Li
类目: ystems and Control (eess.SY); Artificial Intelligence (cs.AI); Emerging Technologies (cs.ET)
备注:
Abstract:Victim search in wilderness areas is time-critical. Drones enable rapid area search, but scheduling them is challenging because victims may continue moving, detection results are uncertain, and effective decisions must combine structured mission information with unstructured contextual clues and domain knowledge. To address this, we propose an agent-aided drone scheduling system. The system employs an agent at the ground control station (GCS) to coordinate multiple drones over successive search cycles. In each cycle, the agent applies a scheduling policy that combines a large language model (LLM) for reasoning about where to search with an optimizer for determining how to dispatch the drones under energy constraints. Simulation results show that the proposed system consistently completes the victim-search mission, whereas none of the six non-LLM baselines does so and direct LLM scheduling is considerably less reliable. The source code is publicly available at this https URL.
[AI-46] RESYNC: Residual Latent-State Synchronization for Deadline-Bound Streaming Inference
链接: https://arxiv.org/abs/2611.16523
作者: Eljas Linna,Hao Dong,Juho Kanniainen
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Predicting every observation in a high-rate stream requires a model to finish before the next arrival. A large model may offer better predictions but miss that deadline. We propose RESYNC: the large model runs asynchronously, while a fast path updates its latest exact but stale latent state using the observations received since that state was computed and applies its unchanged prediction head. The updater learns from the frozen model’s representations without task labels. Across one real and two synthetic streams and five architecture families, RESYNC beats the best compact or distilled alternative with a paired 95% interval above zero in all twelve dataset–architecture pairs with a material teacher advantage at anchor ages of 4–12 observations, and in eleven pairs at older tested ages up to 32. The share of the teacher’s gain preserved remains complete on the real stream but is lower and generally declines with anchor age on the synthetic streams. In live replay with 1 ms arrivals, the synchronous large models cannot deliver any predictions on time while RESYNC serves at least 99.4% of arrivals within the interval, with 0.35–0.62 ms median latency on one CPU thread: around the same speed as or faster than the compact model, and 7.9–17.7 \times faster than synchronous large-model inference. RESYNC targets settings with background compute available but a strict per-observation response deadline.
[AI-47] SignRot: LLM Quantization with Massive Outlier-Aware Sign-Adjusted Rotation NEURIPS2026
链接: https://arxiv.org/abs/2611.16502
作者: Jaehun Gim,Jaewoo Kim,Gyuwan Kim,Seo Yeon Park
类目: Artificial Intelligence (cs.AI)
备注: 23 pages, 4 figures, 19 tables. Accepted to NeurIPS 2026. Code: this https URL
Abstract:While quantization compresses LLMs and accelerates inference, outlier activations impede efficient low-bit representation. In particular, massive outliers cause substantial performance degradation. Although rotation-based methods that rely on randomized Hadamard transforms for online inference aim to mitigate this, they can produce fragmented distributions with empty bins when a massive outlier dominates a token’s activation. This fragmentation severely undermines uniform quantization performance. Moreover, massive outliers concentrate in the down-projection input, where the preceding nonlinearity requires the rotation to be applied at inference time. This constraint prevents training-based approaches such as SpinQuant from learning the online rotation at this position, as a learned dense matrix cannot use the Fast Walsh-Hadamard Transform and would require a full matrix multiplication during inference. To address this, we propose SignRot (Sign-adjusted Rotation), which identifies massive outlier channels and constructs a sign adjustment vector that unifies the signs of the corresponding Hadamard row, collapsing the bimodal distribution into a unimodal profile amenable to quantization. Since the aligned outlier shifts the activation mean away from zero, we pair SignRot with asymmetric activation quantization, whose zero-point absorbs this shift. As a lightweight, drop-in improvement compatible with any Hadamard-based quantization method, SignRot requires only a single calibration forward pass and adds only element-wise sign flips at inference, without an additional matrix multiplication. Applying SignRot to QuaRot achieves a WikiText2 perplexity of 5.88 and an average zero-shot reasoning accuracy of 72.73, surpassing SpinQuant on the LLaMA3-70B model. Code is available at this https URL.
[AI-48] Recalibrating a Typed-Decision Model in an Afternoon: A Hands-On Recipe for Layas Confidence Scores
链接: https://arxiv.org/abs/2611.16499
作者: Mohammed Nayeem Khan,Aliyas Ahmad Misri
类目: Artificial Intelligence (cs.AI); Software Engineering (cs.SE)
备注: 25 pages, 21 figures, 16 tables
Abstract:A typed-decision model such as Laya reads a text and a list of options that the caller defines, for example ticket categories, and returns a probability for each option. Code around it compares the top probability with a threshold to answer, abstain or escalate to a person. Earlier work reports, for a Turkish model, that temperature scaling can raise held-out calibration error (ECE). This tutorial gives a five-step recipe (measure, diagnose, fit, validate, deploy) in 130 lines of numpy with standard methods, none new. On 1,065 synthetic, LLM-labelled items the released Laya checkpoint has a pooled ECE of 0.073 but a bucket-macro ECE of 0.143 (0.127-0.184), the mean over buckets (questions grouped by type and option count): the 12-option bucket is over-confident by +0.22, the three-to-five-option bucket under-confident by 0.05, and they cancel. One global temperature makes pooled ECE worse (0.173); a cross-fitted temperature per bucket brings macro-ECE to 0.073 (0.078 over 10 fold assignments) and reaches 0.089 with 10 labelled items per bucket, but holding out one of seven evaluation sets gives 0.099. After the refit, a gate aimed at 10% error answers 46.3% of items (10.5% wrong) against 6.9% on the published confidences, but one cut-off per bucket on the published confidences answers 45.4%, so most of the gain comes from separating buckets. Confidence-only maps (Platt, histogram, isotonic) need 20 or more labelled items per bucket to approach the temperature; with 320 per bucket the lowest macro-ECE belongs to Platt. In these files the retrained checkpoints store no per-bucket table and the stored 12-option value 0.1006 is clamped to 0.5 at runtime; the refit helps 31 of 33 checkpoints on random folds and 21 with a set held out. Fit and test share one generator and LLM labels, so these are findings about these checkpoints and this data.
[AI-49] How many LLM -labelled rows does a decision encoder need and how wrong can the labels be?
链接: https://arxiv.org/abs/2611.16498
作者: Mohammed Nayeem Khan,Aliyas Ahmad Misri
类目: Artificial Intelligence (cs.AI)
备注: 24 pages, 20 figures, 15 tables
Abstract:A decision encoder such as Laya reads a question with named options (for example, which department should handle a ticket) and a text, and picks an option. It reads both in one forward pass and returns a probability for every option, without writing any text. We ask how many LLM-written training rows such an encoder needs and how wrong those labels may be. We ran 19 fine-tunes of Laya (ModernBERT-large) on nested subsets of a 2,818-row pool and scored them on 1,065 held-out items from the same generator, so every test item is in-distribution and no out-of-family test was run. Accuracy rises from 69.4% before fine-tuning to 74.3% with 5% of the rows, 84.8% with 50% and 86.9% with all rows. The curve flattens, and a saturating fit with bootstrap intervals cannot pin down its plateau. Seed spread at 20% of the rows (s.d. 1.6 points) exceeds some of the steps between sizes. Giving small subsets as many updates as the full pool gains 1.9 points at 5% and 1.9 points at 20% of the rows (one seed; p=0.14 and 0.13), and at 20% that is below the 2.3-point difference between two ordinary seeds. A single run with 10% symmetric label noise shows no detectable change (-0.2 points, 95% interval -1.7 to +1.4); 20% costs 3.4 points and 40% costs 8.7 points. Confusable-pair noise at 20% costs 7.4 points, and under noise the model moves from over-confident to under-confident. An independent labeller disagrees with the gold labels on 3.5% of held-out items, and about half of its choice disagreements (16 of 33, leaving each item out of the partner map) name the partner option that the confusable-pair noise uses. Laya is +0.9 points ahead of a TF-IDF classifier trained on the same rows at full size (the interval includes zero) and +21.1 points ahead at 5% of the rows. Everything is synthetic, LLM-labelled and from one backbone, one trainer and mostly one seed.
[AI-50] ReA-Bench: Benchmarking Responsible Agent Auditing
链接: https://arxiv.org/abs/2611.16493
作者: Zipeng Ye,Shenyu Dai,Yong Xie,Kun Wang,Jiyang Guan,Zhikai Chen,Jiaqi Weng,Jialing Tao,Hui Xue
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:As agents perform tasks autonomously in local and cloud environments, auditing their execution trajectories can help service providers identify risks and improve agent harnesses and foundation models. Our analysis of production and red-team trajectories shows that some risks are repeatedly missed across auditing configurations. Overall recall obscures which risks are missed and whether higher recall comes with more false positives on legitimate behavior. Existing benchmarks offer limited insight into these failures. Generated trajectories can contain inconsistent tool calls and responses, conspicuous risky actions, or repetitive attack patterns. Grouping results by risk type also makes it difficult to tell whether auditors overlooked evidence, accepted an unsupported conclusion, or lacked information needed to make a judgment. We introduce ReA-Bench, a diagnostic benchmark for auditing responsible agent behavior. It organizes auditing challenges into six tiers: internal safety knowledge, contradictions within trajectories, evidence retrieval in long contexts, verification of sources and reasoning, indirect steering, and missing information. It includes challenging benign cases in which actions resemble risky ones but are consistent with the task and its authorization. It also provides paired trajectory versions to test whether auditors detect the same risks more reliably when the trajectory is shortened or missing information is supplied. Most samples are drawn from real production trajectories, with scenarios manually selected and adapted to preserve realistic tasks and risks that could plausibly arise during execution. Subjects: Artificial Intelligence (cs.AI) Cite as: arXiv:2611.16493 [cs.AI] (or arXiv:2611.16493v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2611.16493 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-51] me and Space: A Dual-Axis Evidence-Grounded Context Substrate for Personal Agents
链接: https://arxiv.org/abs/2611.16492
作者: Haitao Cheng,Jerry Hu,Longdong Mao,Dong Huang,Dongming Li,Lijuan Zhang,Mengting Zhang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:We argue that a binding constraint for personal agents is not model capacity but the form in which context is represented: the agent does not know which entity a phrase denotes, when a fact held, or which source record licenses a claim. We call the missing representation spatio-temporal context (STC) and define it on two axes. The temporal axis carries event time, validity intervals and ingestion time. The spatial axis is a relational space in which entities are points, typed relations are edges, and distance is semantic rather than geographic; physical places are one entity type among many. Every relation is anchored to immutable source records, and every answer must be closed over those records. We formalise the substrate and six invariants, a context operator that composes temporal intersection with relational expansion, an ambiguity-preserving resolution state machine, a conflict detector that converts mutually exclusive facts into user adjudication rather than silent overwrite, and a four-regime consumption taxonomy spanning query, insight, triage and proactive reminder. We describe a local-first prototype of the retrieval layer built on SQLite with on-device embeddings and hybrid lexical-dense recall, report measurements from a deployed store of 341 documents, 340,079 chunks and 190,271 embedded vectors, and specify, but do not yet implement, the analytic layer on which the framework depends. We close with an evaluation protocol with per-invariant metrics for evidence coverage, temporal correctness, conflict precision and interruption cost. The proposal targets the representation of context rather than the volume of it.
[AI-52] HeMeR: Heterogeneous Memory Reconciliation for Embodied Agents via Structured KV Reuse NEURIPS2026
链接: https://arxiv.org/abs/2611.16487
作者: Saehun Chun,Wonje Choi,Jinwoo Jang,Taeyoon Kwack,Honguk Woo
类目: Artificial Intelligence (cs.AI)
备注: Accepted at NeurIPS 2026 (Oral)
Abstract:Embodied agents make decisions by combining information from heterogeneous memory modules, including declarative memory for explicit knowledge, procedural memory for learned routines, and working memory for current observations. During interaction, the memory contents required for decision making change with the agent’s state, goal, and task progress, while many contents remain shared across consecutive decisions. However, existing memory modules are typically accessed through separate interfaces, so each updated memory context is often processed as a new input to the foundation model. This prevents selective key-value (KV) cache reuse for unchanged memory contents and forces full KV cache recomputation even when much of the memory context remains valid. We introduce HeMeR, an inference framework that performs KV reconciliation between dynamically updated memory contexts and reusable KV states. HeMeR tracks declarative, procedural, and working memory contents across action decisions, preserves KV states for unchanged contents, and reconciles only the KV states affected by memory-content updates. This enables efficient and stable inference as memory contents evolve during interaction. We evaluate HeMeR on ALFWorld, Habitat, and real-world robotic deployments requiring coordinated use of heterogeneous memory contents. In the real-world setting, HeMeR achieves 83.20% task success and reduces per-step inference time from 2.94 to 1.87 seconds compared with MemRL, while preserving compatible KV states as task progress updates the memory context.
[AI-53] Contain Conform Compare (Extended Version): A Red-Teamed Evaluation Kit for Small Coding Agents and the Full Record of Its Experiments
链接: https://arxiv.org/abs/2611.16485
作者: Nguyen Minh Thang
类目: Artificial Intelligence (cs.AI)
备注: 52 pages, 15 figures, 37 tables. Code: this https URL (tag v0.4.0). Data and analysis scripts are included as ancillary files
Abstract:A hosted score in the Gemma 4 Developer Agent competition is one noisy number, so development needs a local evaluator that is fast, safe for agent-written code, and matching the hosted scorer’s prompts. drydock-eval evaluates a small coding agent (Gemma 4 E4B) at about 250 runs per hour on one RTX 5090. Contain runs each job in a fresh container without network and passed 200 scripted escape tests; conform provides eight checks that compare a local stack’s requests with the scorer’s configuration, down to rendered tokens; compare provides paired statistics that respect repository clusters. AI agents built and attacked the kit; a human owner approved machine use and release. This extended version repeats the short paper and adds the full record of the experiments made with and around the kit: 9,660 rows of agent runs, 2,494 grading controls and 72 serving measurements, all released. On four L4 GPUs the 31B model decodes 38.1 tokens per second in one stream and E4B 105.8. Six configurations of a 31B agent solved between 37 and 44 of 70 development tasks, and no pairwise difference was detected. A pre-registered five-arm experiment (231 tasks, 4,620 jobs) measured four prompt divergences: no E4B solve-rate change was detected (largest estimate -2.1 points, 95% interval -4.4 to +0.4). Three changes of the E4B agent led their screens by +1.0 to +6.2 points; each was then tested once on new tasks under a frozen pass rule, and none was confirmed (gains -0.35, -0.39 and +4.10 points). The unchanged baseline, run again on the same 128 tasks, moved by 1.17 points, and 65% of these tasks were solved in none of eight runs. Prompts and agent bundles are withheld; every run-level result is public.
[AI-54] One Model Two Precisions: Does Disagreement Between Quantized Copies Tell You When a Small Model Is Wrong?
链接: https://arxiv.org/abs/2611.16478
作者: Mohammed Nayeem Khan,Allan Kevin T
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Local language models often exist at several precisions, and quantization is known to change individual answers. This study asks whether the disagreement between two precisions of the same weights, an 8-bit copy that answers and a lower-precision copy that reads the same prompt, works as a per-question error score for abstention, escalation and re-asking. The cost is one extra forward pass over fewer bytes, with no sampling and no training. We simulate round-to-nearest quantization with per-group scales (3 to 6 bits, groups of 32 to 256) on Qwen3-0.6B and Qwen3-1.7B and score 8,398 multiple-choice questions from ARC-Easy, ARC-Challenge, MMLU and HellaSwag subsets. Error detection is measured by AUROC, risk-coverage curves and ECE on a held-out half, with paired bootstrap intervals over questions. For Qwen3-0.6B, the entropy of the 8-bit answer reaches AUROC 0.682; the best disagreement score, selected on a development half, reaches 0.684 (+0.001 [-0.009, +0.011]), and combining it with the confidence features adds +0.009 [+0.002, +0.015], mostly on ARC. On Qwen3-1.7B the selected score is -0.001 [-0.009, +0.006] from entropy and the combination +0.003 [-0.001, +0.006] on the test half. Gaussian weight noise at matched error energy is about as informative as rounding (0.6B only), so we find no evidence that the signal is specific to quantization, and shuffling the features within task and entropy cells removes the gain. One extra 8-bit pass that shifts the option order does better (+0.049 [+0.031, +0.067] over entropy) at about a third more weight traffic; routing to a larger model on free entropy beats both extra-pass scores at equal total traffic. After logistic recalibration the 0.6B confidences are calibrated within a few points of ECE. A deployment recipe and the cases where the method does not help are given.
[AI-55] Fine-Tuned Small Language Model for Synthetic Clinical Data Generation and Predictive Healthcare Analytics
链接: https://arxiv.org/abs/2611.16460
作者: Mihnea Sebastian Burlacu,Ionut Anghel,Tudor Cioara
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:The reduced availability of medical datasets represents an important challenge in the development of predictive models for remote healthcare solutions, especially when the available data are imbalanced or difficult to extend through traditional methods. Small Language Models (SLMs) have become a promising alternative to Large Language Models (LLMs) for addressing these limitations due to their lower computational requirements and adaptability to specialized tasks. However, their performance remains highly dependent on the quality and representativeness of the available training data. In this paper, we propose a method for clinical data augmentation based on a fine-tuned Microsoft Phi-2 model that is used to generate synthetic instances, with the purpose of improving deep learning-based model for predicting remotely monitored vital signs. The Phi-2 model was adapted to the medical domain through supervised fine-tuning on the MedQuad dataset, through a Quantized Low-Rank Adaptation (QLoRA)-based method and by applying few-shot prompting techniques. The resulting model was used afterwards in the augmentation of a small-sized dataset, composed of vital sign values and labels associated with pathological and physiological states. The experimental results demonstrate that fine-tuned SLMs improved the performance on medical domain, achieving higher scores for both lexical (ROUGE) and semantic (BERTScore) evaluation metrics compared to the pretrained model. Moreover, the fine-tuned SLM is effective for generating structured synthetic clinical data. When used for data augmentation, the generated dataset improved the performance of a LSTM predictive model, increasing the classification accuracy from 89.0% to 93.7% and showing consistent improvements in precision, recall, and F1-score.
[AI-56] Rational Alignments of Superintelligences via Debates
链接: https://arxiv.org/abs/2611.16455
作者: Thanawat Sornwanee
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Artificial superintelligences (ASI) is artificial intelligences with greater capability. However, as human principal has to use it, we have to ensure that the agent will perform in the interest of the human. We investigate the framework of ASI-assisted question answering where ASI agents know the correct answer but the human does not. We provide some example of mechanism for rational alignment, where the agent utility is unchanged, but the interaction between the agent and the human principal is controlled to be in a certain fashion. We show specifically that a debate framework where the agents can further reason and behave in a subgame perfect equilibrium manner will lead to a more predictable result, and in some case will be favorable.
[AI-57] FPSRG: Heterogeneous User Simulation First-Person Post-Training and Multidimensional Semantic Response Grids
链接: https://arxiv.org/abs/2611.16445
作者: Ruimin Pei,Haibo Liu,Fang Zhang,Xunhua Guo
类目: Artificial Intelligence (cs.AI)
备注: 17 pages, 5 figures
Abstract:Simulating heterogeneous individuals with large language models is an active research direction, and heterogeneous user simulation requires models with different characteristics. We study profile construction: given dimensions, items, and scoring rules, build model configurations that can be retrieved by multidimensional target scores. We introduce First-Person Semantic Response Grids (FPSRG). Shared post-training first teaches the model to judge and answer as the participant; self-report and behavioral targets then refine dimension-specific control vectors. With the model frozen, we measure joint scores for vector-coefficient combinations, filter inadmissible configurations, and store the results in a searchable catalog. Per-token configuration routing, continuous batching, and isolated caches support parallel measurement of different profiles. Non-thinking five- and ten-dimensional Qwen instances and a thinking-mode six-dimensional instance following full-parameter post-training of Huihui demonstrate the complete workflow. With matched data, layers, and norms, refinement increases mean five-axis composite control span from 30.75 to 40.45. Within the five-dimensional 25-75 target region, the qualified five-level vector catalog covers 92.2% of fixed targets under per-axis error at most 20 and total error at most 60; combining it with the prompt catalog increases coverage to 98.7%. At a fixed coefficient range, three-, five-, and nine-level catalogs achieve 71.2%, 92.2%, and 94.6% coverage, demonstrating the gains from denser measurement. These results describe measured reachable regions under a fixed assessment protocol. Through shared adaptation, a small set of learned directions, and reusable joint measurements, FPSRG constructs heterogeneous individual configurations without retraining a model for every profile.
[AI-58] SoScale: Second-Order Scale Search for FP4 Quantization
链接: https://arxiv.org/abs/2611.16426
作者: Pengxiang Zhao,Xing Li,Hui-Ling Zhen,Weizhe Lin,Xianzhi Yu,Zhenhua Dong
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Low-bit quantization reduces the memory footprint and inference cost of large language models (LLMs), yet accurate FP4 quantization remains challenging. Practical formats such as MXFP4, NVFP4, and HiF4 pair a 4-bit floating-point payload with block-wise or hierarchical scale codes, making accuracy highly sensitive to how each local distribution is scaled. Default maximum-absolute-value scaling preserves dynamic range, but can spend scarce FP4 levels on rarely occupied outlier regions and leave dense tensor values poorly resolved. Existing scale-search methods minimize local element-wise reconstruction error, optimizing tensor fidelity rather than layer-output fidelity. We propose SoScale, a second-order weight scale-code search method for hierarchical FP4 quantization. SoScale searches hardware-valid scale codes under a Hessian-aware layer reconstruction objective and couples this search with GPTQ-style error recovery, selecting scales by their impact on layer outputs. Across Qwen3 models, SoScale consistently reduces downstream quantization degradation under MXFP4, NVFP4, and HiF4.
[AI-59] Small Language Models for Inference-Capability Routing: Effects of Model Capacity and Prompt Optimization on Downstream Quality and Efficiency
链接: https://arxiv.org/abs/2611.16414
作者: Sailendra Kolluru,Maheshkumar Duvvarapu
类目: Artificial Intelligence (cs.AI)
备注: 15 pages, 8 figures
Abstract:Optional reasoning, retrieval, web search, and tool use can improve large language model responses but incur unnecessary computation when activated indiscriminately. This study investigates whether prompt optimization can compensate for limited router capacity and whether classification gains translate into downstream efficiency. The proposed framework uses prompt-driven small language models to make four independent capability decisions, allowing multiple capabilities to be selected per query. Qwen2.5 routers ranging from 1.5B to 14B parameters are compared across five prompting strategies, followed by an evaluation of capability-specific optimized configurations using a fixed GPT-OSS-120B downstream model on 127 questions. Under minimal prompting, increasing router capacity does not consistently improve classification performance. Prompt optimization narrows the capacity gap for reasoning and calculator routing, whereas RAG and web search retain larger benefits from higher capacity. Similar classification F1 scores do not necessarily imply equivalent downstream behavior: false-positive activations increase unnecessary computation, while missed capabilities can reduce answer quality. The 14B router achieves a mean BERTScore F1 of 0.8408, compared with 0.8418 when all capabilities are enabled, while reducing average downstream token consumption by 68.7% (8,837 to 2,764 tokens) and median downstream latency by 80.5% (9.79 to 1.91 seconds). These savings exclude router overhead, which is measured separately in a different inference environment and increases with router capacity. The findings demonstrate that the relationship between router capacity, prompt design, and downstream efficiency is capability-dependent. Consequently, router selection must consider error profiles and routing overhead alongside classification performance.
[AI-60] DIAGNO: Diagonal Spherical Neural Operators for Heterogeneous Earth Dynamics Modeling NEURIPS2026
链接: https://arxiv.org/abs/2611.16356
作者: Herui Li,Bin Lu,Haonan Qi,Lei Zhou,Luoyi Fu,Xinbing Wang,Meng Jin
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 35 pages, 13 figures, 8 tables. Accepted at NeurIPS 2026 (Main Track, Oral)
Abstract:Accurate forecasting of Earth’s dynamic systems is vital in geophysics, where neural operators offer a powerful data-driven paradigm. Standard operators often treat the Earth as a flat Euclidean plane, introducing geometric artifacts. In contrast, Spherical Neural Operators (SNOs) resolve this mismatch by operating within the spherical spectral space, naturally adapting to spherical geometries. However, for computational efficiency, existing SNOs typically rely on the rotation equivariance assumption. This assumption simplifies the transitions of physical states into isolated spectral modal evolutions, thereby over-idealizing the Earth as a homogeneous system. To solve this problem, we no longer rely on the rotation equivariance assumption that isolates spectral modal evolution, but directly derive an explicit modal interaction kernel and propose Diagonal Spherical Neural Operators named DIAGNO. DIAGNO explicitly parameterizes cross-modal interactions through spectral modal decomposition, thereby preserving the inherent heterogeneity of Earth dynamics. Furthermore, leveraging the unique diagonal structure of the spherical spectrum, we design inter- and intra-diagonal interaction mechanisms to natively capture essential zonal and meridional dynamics, respecting intrinsic geophysical regularities. Extensive experiments across simulated fluid dynamics and real-world Earth system reanalysis (including atmosphere and ocean) datasets demonstrate DIAGNO’s superior capability, consistently achieving state-of-the-art performance in multi-step forecasting. Beyond this, we validate DIAGNO’s practical analytical value in professional geophysical domains by assessing its fidelity in capturing climatological anomalies and maintaining physical consistency.
[AI-61] HyBridge: Bridge-Guided Self-Supervised Hyperbolic Hypergraph Representation Learning
链接: https://arxiv.org/abs/2611.16351
作者: So-Bin Jung,David Yoon Suk,Sang-Wook Kim
类目: ocial and Information Networks (cs.SI); Artificial Intelligence (cs.AI)
备注:
Abstract:Hypergraphs capture groupwise relationships among multiple nodes and often exhibit tree-like structures. Although hyperbolic space is well suited for modeling such structures, simply extending existing hypergraph representation learning (HRL) methods to hyperbolic space does not consistently improve performance. In this paper, we investigate the relationship between node structural properties and hyperbolic embedding radii and find that bridge nodes tend to lie closer to the center in graphs, whereas this relationship is not consistently observed in hypergraphs. We further find that conventional bridge score estimation may fail to capture the groupwise connectivity of hypergraphs. Motivated by these findings, we propose HyBridge, a self-supervised hyperbolic HRL framework that (1) estimates node bridge scores by considering the external connectivity of individual hyperedges, (2) learns node and hyperedge embeddings through hyperbolic hypergraph encoding, (3) guides node radial positions according to their bridge scores, and (4) employs contrastive learning to preserve node discriminability. Extensive experiments on 10 real-world hypergraphs demonstrate that HyBridge outperforms 22 competing methods, achieving performance improvements of up to 9.82% in node classification. For reproducibility, we provide the code at this https URL.
[AI-62] DRAMA: Dissecting Attention Redundancy for Accelerating Multimodal Diffusion Large Language Models NEURIPS2026
链接: https://arxiv.org/abs/2611.16345
作者: Yi-Hsin Hung,Fangfu Liu,Shao-Yuan Lo,Chu-Song Chen
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Accepted to NeurIPS 2026. 27 pages, 19 figures
Abstract:Multimodal diffusion large language models (dLLMs) generate text through iterative masked denoising with bidirectional attention, offering a compelling alternative to autoregressive vision-language modeling. However, each denoising step requires a full forward pass over long multimodal sequences in which visual tokens dominate the sequence length, and a fixed step budget further compounds this quadratic per-step cost. Existing inference acceleration approaches for multimodal dLLMs characterize attention redundancy implicitly, motivating designs from isolated attention visualizations rather than from measurement. We close this gap with DRAMA, a training-free acceleration framework grounded in a systematic dissection of attention redundancy in multimodal dLLMs. By probing attention tensors along spatial and temporal axes, we reveal two persistent redundancy structures: (i) attention mass consistently concentrates on a sparse subset of visual tokens, and (ii) attention distributions rapidly stabilize across denoising steps after an initial formation phase. These findings motivate two complementary mechanisms: Spatial-Redundancy-Aware Visual Token Pruning (SVTP) and Sufficiency-Guided Adaptive Stopping (SGAS). Together, DRAMA achieves an average 18.67 times inference speedup across six vision-language benchmarks without training, architectural modification, or attention caching, while maintaining competitive accuracy. Our results establish a strong, measurement-grounded acceleration baseline for multimodal dLLM inference.
[AI-63] From More to Better: Causal Inference and Digital Experimentation in the Age of AI
链接: https://arxiv.org/abs/2611.16335
作者: Winston Chou,Mårten Schultzberg,Lo-Hua Yuan,Wenjing Zheng,Avi Feller,David Holtz,Widad Machmouchi,Sven Schmit,Gagan Bhat,Simon Ejdemyr,Veronica Hannan,Amy Hartwig,Mahnaz Islam,Nathan Kallus,Jeremy Karp,Kyle Kretschman,Hannah Li,James McQueen,Evan Miller,Nicolas Stier,Sean Taylor
类目: Artificial Intelligence (cs.AI); Applications (stat.AP)
备注: 17 pages, 1 table. Perspective based on the 2026 Causal Inference and Digital Experimentation Roundtable (CIDER)
Abstract:Artificial intelligence (AI) is quickly expanding the capacity of organizations to produce and analyze data. We argue that this expansion shifts the leverage of data scientists from performing individual analyses to assessing the credibility and importance of findings and determining when evidence generalizes across contexts. Drawing on a roundtable of academic researchers and industry practitioners, we examine three ways in which data scientists who work on causal inference and experimentation can help organizations turn more analytical capacity into better decisions: preserving and accumulating the learning value of experimentation beyond individual tests; choosing and combining experimental, observational, and synthetic evidence; and ensuring evaluation keeps pace with AI as its capabilities and uses evolve. In each area, we distinguish established practice from open questions. We conclude with implications for the training and practice of data science and for organizations.
[AI-64] Benchmarking AI risk behaviors
链接: https://arxiv.org/abs/2611.16318
作者: Luhan Mikaelson,Robert Adragna,Emmett Brosowsky,Hayley Clatterbuck
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:We evaluate LLM risk propensities across a range of contexts and task demands. We evaluate models for how risk averse or risk tolerant they are in various conditions, as well as how stable and consistent their behavior is across conditions. We present a toolkit for developing model risk profiles that can be used to predict model behavior and which can be adapted to include new evaluations of risk propensities. We also examine how these properties have changed across model generations and families. We observe that 2026 state-of-the-art models tend to be more stable, more consistent, and more risk neutral than older models. These trends appear to be related; newer models are more stable and consistent because they are more likely to use expected value maximization as a decision procedure.
[AI-65] Wanda-Guided Sparse Perturbations for Efficient Zeroth-Order Optimization
链接: https://arxiv.org/abs/2611.16317
作者: Ziqing Wen,Xiaohui Song,Jiahuan Wang,Ping Luo,Junlin Zeng,Dongsheng Li,Tao Sun
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Zeroth-order (ZO) optimization enables memory-efficient fine-tuning of large language models through forward-only evaluations, but high-dimensional random perturbations can limit optimization efficiency. Sparse perturbations reduce the search dimension, making the selection of informative coordinates critical. We propose Wanda-Importance Sparsity-MeZO (WIS-MeZO), which uses Wanda weight–activation saliency to construct task-conditioned sparse perturbation masks while keeping the pretrained model dense. We analyze support quality through gradient energy capture and relate it to gradient estimation and convergence. Across 16 benchmarks with LLaMA and Qwen models ranging from 3B to 32B parameters, WIS-MeZO achieves higher aggregate scores than MeZO and reaches comparable validation losses in fewer optimization steps. Comparisons with random and magnitude-based sparse perturbations further highlight the importance of informative support selection.
[AI-66] SAGE-QA: Semantic-Aware Grounded Evaluation for Closed-Context Question Answering on SQuAD 2.0
链接: https://arxiv.org/abs/2611.16300
作者: Chandra N Sekharan,Vishakan Umapathy,Abhishek Phadke
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Closed-context question answering (QA) requires a system to generate an answer strictly supported by a supplied passage, or to abstain when the passage lacks adequate evidence. SQuAD 2.0 is a widely used benchmark for this setting, scored by Exact Match (EM) i.e., character-level agreement with human reference spans after lexical normalization. For generative large language models (LLMs), however, an EM failure may reflect a valid paraphrase or an alternative passage-supported answer rather than a hallucination. The best prior leaderboard EM is approximately 91%, but evaluating generative models requires separating lexical divergence from factual error. We introduce SAGE-QA (Semantic-Aware, Grounded Evaluation for Question Answering), a multi-tiered framework combining reference-based Semantic Equivalence Scoring (SES) and passage-based Re-Validation Scoring (RVS). On a fixed set of 10,000 SQuAD 2.0 validation examples, we compute EM and decision-matrix F1 from TP, TN, FP, and FN counts for zero-shot closed-weight frontier models and for open-weight models, untuned and fine-tuned. Zero-shot, Claude Opus 4.6 and Gemini 3.1 Pro reach 85.21 and 86.66 EM, outperforming untuned Llama 3 8B (56.88), gpt-oss-20b (66.28), and Llama 3.3 70B (71.39). Parameter-efficient fine-tuning adds 21.19 to 34.06 EM points, raising these models to 90.94, 87.47, and 94.24 EM, respectively. Auditing the frontier models with SAGE-QA and an external LLM judge yields dataset-wide SES scores of 92.81 for Claude Opus 4.6 and 91.67 for Gemini 3.1 Pro, and hybrid RVS scores, which combine benchmark credit with passage-based judge assessments, of 97.96 and 98.02. Fine-tuned open-weight models perform best, with fine-tuned Llama 3.3 70B surpassing the previous best known EM.
[AI-67] Causals: An Open Knowledge Graph for Research on Causal Reasoning
链接: https://arxiv.org/abs/2611.16267
作者: Miaozhe Lin,Qiongzheng Lin,Zhongwei Liang,Zijun Liu
类目: Artificial Intelligence (cs.AI)
备注: 13 pages, 3 figures. Project website: this https URL
Abstract:Applying causal knowledge to a new event requires connecting the event to relevant entities and the relationships through which its consequences may unfold. We present Causals, an open knowledge graph for research on this form of reasoning. Its initial release, constructed from news reports, contains 369,184 nodes and 1,272,418 edges linking reported events, organizations, commodities, financial instruments, and other entities. Structural relationships, observations, and attributed causal claims retain distinct semantics and links to source records. We describe the construction and representation of the resource and characterize its connectivity and multi-hop retrieval properties. Causals provides shared relational knowledge for studying how language models retrieve relevant context and apply mechanisms to particular events. The dataset is available under CC BY 4.0 at this https URL, and we invite contributions to its ontology, records, and evidence.
[AI-68] Rectified Looped Transformer
链接: https://arxiv.org/abs/2611.16265
作者: Musheng Li,Jiawei Zhang,Ziyuan Liu,Yuantao Gu
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Looped Transformers increase effective depth by repeatedly applying a shared block, offering a parameter-efficient approach to language modeling. However, most existing works directly reuse each loop’s output as the next’s input, which can lead to hidden-state scale growth and perturbation amplification as recurrence deepens. We propose Rectified Looped Transformer (RLT), which redesigns feedback between loops using an interpolation schedule inspired by flow-based modeling. At each loop, RLT forms the next input by linearly interpolating between the current state and its prediction. Such interpolation distributes state changes across loops, encouraging a smoother trajectory from the initial input to the final output. Theoretical analysis proves that, under appropriate assumptions, RLT keeps state magnitudes uniformly bounded across loops and reduces the worst-case perturbation amplification. We evaluate RLT on language modeling corpora and downstream benchmarks. At matched training and inference depths, RLT consistently achieves lower perplexity than a standard looped Transformer, while exhibiting greater robustness to input perturbations and variance in inference depth. Consistent with the theoretical results, latent trajectory analysis shows that RLT maintains more stable hidden-state norms across loops, which also facilitates extrapolation-based decoding strategies. Notably, RLT achieves these gains without requiring additional trainable parameters or extra recurrent block evaluations, thus providing a simple yet more stable feedback mechanism for looped Transformers.
[AI-69] Guided Generation of Linux Kernel Semantic Patches with Small Local Language Models
链接: https://arxiv.org/abs/2611.16215
作者: Mohanned Ahmed
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注: 7 pages, 6 tables, 1 figure. Code, data, prompts and model responses: this https URL
Abstract:When a Linux kernel API changes, hundreds of drivers are converted by hand over months, and the earlier conversions are reviewed examples for the later ones. We ask whether small open language models running on a laptop can turn such examples into Coccinelle semantic patches, the rules kernel maintainers already use for tree-wide changes. We adapt guided code generation to this task: the training examples are decomposed by script into a tree of change variants, the model writes one small rule per variant under feedback from applying it to the examples, deterministic components repair SmPL syntax and generalise driver-specific names, rules are composed greedily, and a compile critic repairs or rejects a rule whose output does not build. We evaluate on a pre-registered test set of 7 migrations whose new APIs appeared after the training cutoffs of all three models (Qwen2.5-Coder 7B, Llama 3.1 8B, Gemma 3 4B), with 122 held-out driver conversions, counting a conversion as solved only if the rule’s output matches the human fix and compiles. With the framework, Qwen2.5-Coder solves 35.3 conversions on average over three seeds (range 32-41), against 1.0 when prompted directly; Spinfer solves 41 and a frontier LLM agent 66. The local model solves fewer conversions than Spinfer but its rules are the most precise on the test set (0.74 of the files they change are solved, against 0.55 for Spinfer and 0.70 for the frontier agent), and the compile critic removes every output that matches the human fix but does not build. The two general-purpose models stay weak even with guidance (at most 12 solved). All code, prompts, model responses and results are released.
[AI-70] One Fuzzer Many Reason ers: Adaptive Neuro-Symbolic Fuzzing with Reliability-Calibrated and Context-Gated Guidance
链接: https://arxiv.org/abs/2611.16211
作者: Jinghao Wen,Xianglong Yu,Shaohuang Wang,Xun Jiao
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Neural-symbolic AI (NSAI) combines learned perception with programs, constraints, and compositional reasoning, but remains difficult to test because small semantics-preserving changes can alter grounded symbols and trigger discontinuous reasoning failures. Existing neural fuzzers mainly rely on neural coverage or representation-space feedback and do not provide a common testing interface for heterogeneous NSAI systems. We introduce NSAI-Fuzz, a unified semantics-aware fuzzing framework that separates task-specific valid mutation from architecture-neutral guided search and represents each candidate using five availability-aware, rank-normalized risk signals: boundary pressure, grounding uncertainty, representation drift, constraint violation, and archive novelty. We further propose Reliability-Calibrated Weighting (RCW), which learns dataset-model-level signal allocations from calibrated reliability, and Contextual Signal Gating (CSG), which adapts signal weights to the current candidate pool. We evaluate four NSAI systems across four reasoning tasks, nine compatible dataset-model cells, 20 policies, and 1,800 on-policy runs. Fixed five-signal guidance outperforms Random and Uniform in every cell, while RCW and CSG further improve over Fixed in all nine cells. Under matched mutation and query budgets, RCW and CSG achieve 16.17% and 16.06% aggregate failure discovery, compared with 6.06% for the strongest neural-fuzzing baseline. Wall-clock evaluation further shows that TensorFuzz requires 1.11x the runtime of RCW while achieving only 0.45x its failure throughput. These results show that effective and efficient fuzzing of heterogeneous NSAI systems benefits from a common neural-symbolic risk interface with adaptive signal allocation.
[AI-71] urning Multi-Agent Communication into Token-Level Credit with On Policy Self Distillation
链接: https://arxiv.org/abs/2611.16206
作者: Zixu Wang,Yu Zhao,Xing Zheng,Bingbing Xu,Tian Xueyun,JiaTang Luo,Weiqiang Wang,Bowen Song,Huawei Shen,Xueqi Cheng
类目: Artificial Intelligence (cs.AI)
备注: 14pages,4 figures
Abstract:Reinforcement learning improves problem solving in large language model multi-agent systems. Existing approaches overlook communication knowledge or compress it into scalar rewards, leaving coarse credit with little guidance on which decisions each agent should reinforce or revise. OPSD4MAS converts future communication into fine-grained token-level credit through On Policy Self Distillation (OPSD). An external model extracts facts and assesses their trajectory relations for coarse-grained filtering, followed by fine-grained leave-one-out filtering of target-token influence against agent-wise outcome advantages. The fact-conditioned policy supplies token-level credit, combined with GRPO through a linearly annealed joint objective. Our method improves mathematical pass@4 over GRPO by up to 20.00 percentage points. Ablations support complete filtering, while frozen-history injection diagnostics support fact utility. Together, these results support combining agent-level outcome guidance with communication-derived token-level credit. Code is available at this https URL.
[AI-72] When can an interpretability claim be trusted? Stability and certification of feature dimensionality ridge leverag e and the feature recovery rate
链接: https://arxiv.org/abs/2611.16190
作者: Ali Tajfar,Behnam Tajfar
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 70 pages. Also archived at this https URL . The AI use statement is in the paper
Abstract:Interpretability tools increasingly support safety claims, such as that a sparse dictionary has recovered most of a model’s features. Such claims rest on measurements with arbitrary thresholds, and small changes to the model weights, a different training seed or a different threshold can reverse them. We study two widely used measurements in toy models and synthetic dictionaries with known ground truth: the feature dimensionality of Elhage et al. and the feature recovery rate (FRR) with a cosine threshold. We characterize exactly where feature dimensionality is discontinuous and replace it with a regularized (ridge) version with sharp stability bounds: the ridge leverage score changes by at most 3\sqrt3/(8\sqrt\lambda) times the operator-norm change of the weights, and this constant cannot be improved; for the regularized dimensionality the stability constant grows like \rho/4 in the scale \rho=R/\sqrt\lambda , with explicit upper and lower bounds. We also prove certificates stating when a claim “feature i is recovered” must survive a given change of the dictionary. The certificates held in every test and are informative for small perturbations, but certify no feature non-trivially across independently trained dictionaries; at the standard threshold, whether a feature counts as recovered changed with the training seed for 55 percent of features. Experiments use toy models only. The proofs, code and text were drafted by AI agents directed by the authors and checked by an independent blind AI reviewer and by numerical tests; the authors take full responsibility for the content.
[AI-73] Same-State Pairs Make One JEPA Proof of the Platonic Representation Hypothesis Vacuous
链接: https://arxiv.org/abs/2611.16182
作者: Xin Li
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:In a Joint-Embedding Predictive Architecture (JEPA), the JEPA proof of the Platonic Representation Hypothesis (PRH) assumes that every observation determines the world state and that a spectral gap makes minimizers unique up to rotation. Under the hard-whitening population objective, same-state positive pairs make the induced world kernel the identity: every centered whitened state function is a global minimizer, and, when d_z-1\ge 2k , two minimizers can have best-rotation squared residual 2k . The uniqueness claim is conditional; for kd_z-1 , where d_z is the number of world states, this limit makes its premise unsatisfiable, so the claim has no content there. The experiments that motivate it leave the assumption: a caption sees only part of an image, and the paper states no experiment meets its hypotheses exactly. The theorem does not reach those experiments, and for kd_z-1 its idealized same-state limit is vacuous. One of the paper’s four quantitative results survives: a gap-free lower bound. Its bound vanishes only at perfect scene-explained fraction; it bounds alignment error but cannot certify good-view alignment. A canonical frame is determined only up to a rotation and a latent permutation. Paired samples fix the relative frame at effective rank k ; a global rotation remains. Component semantics fix the relative permutation only up to signs; value anchors fix it once the mixing frame is pinned. A resource–gauge ledger records which resource fixes which gauge.
[AI-74] Stack more levels: how to get general and human-like Mario playing
链接: https://arxiv.org/abs/2611.16175
作者: Shinichi Arimura,Yuchen Li,Julian Togelius
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Super Mario Bros. is widely used in game AI and procedural content generation (PCG), yet it remains unclear whether a model-free agent can reliably win unseen levels. To address this question, we train Proximal Policy Optimization (PPO) agents on increasingly large sets of PCG-generated Mario levels and evaluate their generalization on 1000 held-out test levels. We then shape competent PPO policies into three gameplay archetypes—runner, killer, and collector. Finally, we fine-tune these PPO policies using Diffusion-Reward Adversarial Imitation Learning (DRAIL) with human gameplay demonstrations. We find that increasing the number of distinct PCG-generated training levels generally improves unseen-level performance, and that fine-tuning PPO policies with DRAIL increases action-level human-likeness while preserving much of the persona induced by PPO pretraining.
[AI-75] Self-Supervised Attention-Based Virtual Metrology for CMP MRR Prediction
链接: https://arxiv.org/abs/2611.16172
作者: Jonathan Adam Rico,Nagarajan Raghavan,Senthilnath Jayavelu
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Submitted to the IEEE International Symposium on Semiconductor Manufacturing (ISSM 2026)
Abstract:Chemical mechanical planarization (CMP) is a critical semiconductor manufacturing process where material removal rate (MRR) is a key metric for process monitoring and advanced process control (APC). Virtual metrology enables MRR estimation from process time series signals, supporting run-to-run(R2R) control and process optimization while reducing reliance on physical metrology. Existing approaches typically rely on simplified linear assumptions or require handcrafted feature engineering. In this paper, we propose Global Regression Pretraining(GREP), a self-supervised, attention-based framework for CMP MRR prediction. GREP learns temporal representations from unlabeled sensor signals through global proxy-target regression, followed by supervised fine-tuning for MRR estimation. The framework is evaluated against a feature-engineered supervised baseline under three conditions: limited labeled data, sensor noise, and temporal process drift. Experiments on the PHM 2016 CMP dataset demonstrate competitive performance under limited labeled-data conditions, with improvements in selected cases and performance largely comparable to the baseline across different labeled-data fractions. Both models exhibit relatively small changes in prediction error with increasing noise, while GREP achieves a 17.0% RMSE reduction in the final temporal interval.
[AI-76] ADAPT: Agent ic Diff-Aware Patch Testing for Pull Requests
链接: https://arxiv.org/abs/2611.16114
作者: Vahid Haratian,Atakan Akar,Eray Tüzün
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注: 12 pages, 4 figures, 8 tables. Replication package: this https URL
Abstract:Modern software development relies on pull requests (PRs) that introduce small, targeted code changes. Once merged, these changes should be protected by focused tests that exercise the behavior the PR introduces, yet developers often submit PRs without such tests, leaving the new behavior unguarded against future regressions. Most test generation techniques target entire software units and optimize objectives such as code coverage rather than the behavior a change introduces, and prior diff-aware approaches have had limited success in capturing it. We propose ADAPT (Agentic Diff-Aware Patch Testing), a regression test generation approach that uses an autonomous coding agent to produce change-focused tests for PRs. Given a PR, the agent analyzes the diff between the base and head versions to identify the modified functions and their context, then plans and generates candidate unit tests. A dedicated fail-to-pass (F2P) verification tool runs each candidate on both versions and, after F2P success, reports which executable changed lines remain uncovered; guided by this feedback, the agent iteratively refines its tests. We evaluate ADAPT on the same 426 real-world PRs used by the prior state-of-the-art baseline, with GPT-4o and GPT-5-mini. ADAPT+GPT-4o reaches an F2P rate of 62.0% (264/426), compared with BLAST’s reported 35.4% (151/426), while ADAPT+GPT-5-mini reaches 89.9% (383/426). On 137 PRs shared with Change And Cover (ChaCo), ADAPT achieves full patch coverage on 83.9% of PRs, compared with 79.6% for ChaCo, at considerably lower time and cost. A manual evaluation confirms that the F2P-valid tests target each PR’s intended change, with mean scores of up to 4.54/5 for importance, 4.35/5 for assertion strength, and 4.49/5 for readability, often covering behavior beyond the developer-written tests. Comments: 12 pages, 4 figures, 8 tables. Replication package: this https URL Subjects: Software Engineering (cs.SE); Artificial Intelligence (cs.AI) Cite as: arXiv:2611.16114 [cs.SE] (or arXiv:2611.16114v1 [cs.SE] for this version) https://doi.org/10.48550/arXiv.2611.16114 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-77] Subsurface schematisation from sparse CPT data using the tabular foundation model TabPFN
链接: https://arxiv.org/abs/2611.16088
作者: Taiga Saito,Fabian Campos Montero,Bruno Zuada Coelho,Ana Martins Teixeira,Stephen Wu,Timo Schweckendiek,Yu Otake
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Subsurface schematisation supports geotechnical decisions but remains challenging because sparse site-investigation data leave substantial gaps in information about ground conditions. This study explores TabPFN, a foundation model pretrained on general synthetic tabular data, for this task. Using only grid coordinates and soil behaviour type index observations from sparse cone penetration tests (CPTs), we first reconstruct two-dimensional sections without geotechnical fine-tuning or predefined layer boundaries. We evaluate the approach on synthetic sections and at two field sites in the Netherlands, comparing it with SchemaGAN, a conditional generative adversarial network trained on synthetic geotechnical sections, and with reference interpolation methods. In the baseline comparisons, TabPFN achieved mean errors close to SchemaGAN’s on synthetic sections and lower mean errors at both field sites. On synthetic sections, TabPFN’s mean absolute error decreased at each tested increase in CPT count, and, under the baseline input setting, its pointwise prediction intervals had mean coverage close to nominal levels. Subsequent fine-tuning on the synthetic sections used to train SchemaGAN improved synthetic test accuracy but increased mean errors at both field sites. These findings indicate that general-purpose tabular foundation models offer a promising route for updating subsurface predictions from sparse and evolving site-investigation data without retraining.
[AI-78] Agent ic Computational Algebra
链接: https://arxiv.org/abs/2611.16073
作者: Ryan A. Anderson,Matthew Trager,Ben Bowman,Wei Xia,Stefano Soatto
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Symbolic Computation (cs.SC)
备注:
Abstract:We explore the use of AI agents for solving computational problems in algebra. For the problems we consider, the main challenge is not correctness but \emphtime: general-purpose algorithms are guaranteed to provide correct solutions but can be prohibitively slow, while reformulations that exploit special structure often allow solving a problem much faster. We propose to use this setting to evaluate how effectively agents discover and exploit structure within and across tasks to reduce computational costs. More broadly, computational algebra supplies infinitely many problems with independently verifiable answers and systematically controllable structure and difficulty. To illustrate these possibilities, we use benchmark families of systems with controlled shared structure and evaluate agents in three settings: solving independent instances, solving online streams of related instances using persistent memory, and compiling a fast reusable solver for a known problem family. Our experiments show that (i) agents can extend the range of systems solvable within a fixed budget, although reasoning overhead can sometimes outweigh computational savings; (ii) memory can reduce the cost of repeated solving but may preserve suboptimal strategies; (iii) compilation produces faster reusable solvers but at substantial upfront cost. Overall, computational algebra offers concrete, measurable targets for improving how agents discover and reuse structure.
[AI-79] okenized Reward Systems for Federated Learning: The Privacy vs. Fairness Tradeoff
链接: https://arxiv.org/abs/2611.16065
作者: Andreas Sendros,Nikolaos Pavlidis,Vasilios Perifanis,Pavlos S. Efraimidis
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR)
备注: 16 pages, 3 figures, 5 tables. Accepted at the 12th Balkan Conference in Informatics (BCI 2026), Best Paper Award
Abstract:Federated learning (FL) systems increasingly rely on blockchain based reward mechanisms to incentivize participation while preserving data privacy. We design and empirically evaluate two complementary Solidity smart contracts for FL reward distribution: the first one that allocates rewards proportionally to a publicly disclosed quality score, and a privacy preserving variant that hides individual quality scores behind a zero-knowledge (ZK) proof of a threshold condition. Both contracts mint a non transferable, use-and-burn token maintaining a trustless incentive structure. We evaluate the two designs through MNIST training with adversarial behavior, including Sybil attacks, zero gradient and random noise, across three independent random seeds. Our results reveal a fundamental tradeoff between cryptographic privacy and incentive fairness. The privacy preserving architecture strictly degrades system robustness, doubling the adversary’s average reward capture while introducing severe outcome variance. This vulnerability is driven by a structural self-inclusion bias within cosine-similarity scoring: as the global model converges, legitimate gradient updates naturally diminish, rendering them statistically indistinguishable from bounded noise. Adversaries exploit this dynamic to systematically breach the ZK threshold, triggering a late stage failure cascade that culminates in absolute adversary monopoly over the reward pool and the total exclusion of honest contributors. Alongside a closed-form theorem formalizing this privacy–Sybil resistance tradeoff, we map strict mainnet deployment boundaries via linear gas scaling models, explicitly incorporating the post-Fusaka EIP-7825 per transaction gas cap. Ultimately, while cryptographically secure, static threshold ZK gates fail as incentive layer fairness mechanisms under realistic FL convergence dynamics on standard benchmarks. Comments: 16 pages, 3 figures, 5 tables. Accepted at the 12th Balkan Conference in Informatics (BCI 2026), Best Paper Award Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR) MSC classes: C.2.4, I.2.6, K.4.4 Cite as: arXiv:2611.16065 [cs.LG] (or arXiv:2611.16065v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2611.16065 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-80] Human Capital in Context: What Employment Survival Reveals About Worker-Firm Match
链接: https://arxiv.org/abs/2611.16058
作者: Jung Ho Choi,Mario Milone
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); General Economics (econ.GN)
备注:
Abstract:There is no best worker in general, only a best worker for a particular firm. We develop a measure of worker-firm match quality that treats fit as a property of the pair, learned from the survival of employment relationships. A two-tower neural network embeds workers, from their career histories and current positions, and firms, from their characteristics and business descriptions, into a common space, and match quality is the cosine similarity of the two vectors. The measure takes its functional form from the skill-weights model of firm-specific human capital, and its only estimated objective, a separation hazard, from the view of the job match as an experience good. Estimated on more than a billion person-periods covering 101 million workers and 8.4 million firms, a 0.1 increase in match quality lowers the odds of separation by 48%. About a quarter of the variance of match quality is specific to the worker-firm pair, which worker-, firm-, and occupation-level measures set to zero by construction. Workers earn more when better matched, especially early in a relationship, and firms whose workforces score higher grow faster, become more profitable, and are issued more, and more valuable, patents over the next five years than their industry peers. Because it scores pairs that never formed, the measure also predicts where movers go, and acquirers record a larger share of the price as goodwill when the target’s workers would fit the acquirer better than they fit the target.
[AI-81] Runtime Analysis of Cartesian Genetic Programming on Max: A Proven Exponential Speedup NEURIPS2026
链接: https://arxiv.org/abs/2611.16013
作者: Duc-Cuong Dang,Roman Kalkreuth,Andre Opris
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Neural and Evolutionary Computing (cs.NE)
备注: Accepted as a poster at NeurIPS 2026. To appear in the conference proceedings
Abstract:Genetic Programming (GP) is a search paradigm inspired from natural evolution for the automated discovery of expressions, functions, and computer programs. Cartesian GP (CGP) is a flavor of GP that uses a graph-based model to encode candidate programs, as contrasted to the conventional Tree-based GP (TGP). Since its inception two decades ago, CGP has predominantly been analyzed empirically, thus little is known regarding its theoretical performance guarantees. This paper analyzes CGP for the Max problem, which has been only rigorously studied for TGP with a proven expected runtime superlinear in the optimal program length n . We prove that the expected time for the ( 1 + 1 )~CGP employing common mutation operators to evolve an equivalent optimal of the same output is polylogarithmic in n , specifically at most O(\log^9n) . This is remarkable, as it shows an exponential speedup by switching to CGP. Our results shed light on the benefit and the compactness of the graph-based representation for programs, and on how CGP can navigate its search and fitness spaces efficiently. This is a first proven performance guarantee for CGP on an established benchmark problem. Experiments complement our theoretical findings.
[AI-82] JustRUA: A Controlled Study of GPT -6 Robot-Use Agents
链接: https://arxiv.org/abs/2611.16009
作者: Zefan Wang,Yufan Dang,Xuanwei Shi,Mingqi Gao,Cheng Yin,SikYuen Tam,Hanyu Liu,Yasheng Wang,Yaxi Lu,Songyang Zhang,Xin Cong,Yuan Yao
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注: 30 pages, 14 figures. Project page: this https URL . Code: this https URL
Abstract:GPT-6 Astra can drive a robot arm directly through a fixed harness, and scores of such robot-use agents on RoboDojo have risen quickly since. Reported systems differ in harness and test subset and often change several designs at once, so it is hard to tell which choices produce the gains. We take these designs apart and test them in one framework, along the harness around the model, the action interface, the reuse of experience and the scene geometry given to the model. A lean harness that keeps the full episode history and stops each motion early goes a long way. Reusing experience changes how the model acts but is not a free lunch, and scene geometry helps most, even when depth is predicted from RGB alone. JustRUA starts from the official L3 harness and keeps only what matters most. Even with no learned policy, self-improvement loop or task-specific tuning, this simple harness raises the score on the RoboDojo simulation benchmark substantially, from 29.7 to 50.3.
[AI-83] Fit-LRP: Fitting Linear Relaxations to Polytopes for Neural Network Verification over Non-Convex Regions
链接: https://arxiv.org/abs/2611.16003
作者: Brian Yarbrough,Jesse Wayment,Jingbo Wang,Philip E. Paré,Shreyas Sundaram
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Systems and Control (eess.SY); Optimization and Control (math.OC)
备注:
Abstract:State-of-the-art neural network verifiers evaluate linear relaxations over axis-aligned bounding boxes. However, verification domains in cyber-physical systems are rarely boxes. Reachability specifications, obstacle-avoidance constraints, and neural Lyapunov certificates typically yield oblique or non-convex polyhedral regions, over which box-based bounds become extremely loose. This paper proposes Fit-LRP (Fitting Linear Relaxations to Polytopes), a framework the implements exact optimization of affine bounds over non-convex domains (represented by unions of polytopes) by solving a sequence of small, low-dimensional linear programs. Our approach seamlessly enhances the final optimization step of LiRPA and \alpha,! \beta -CROWN branch-and-bound (BaB) algorithms. In BaB verification, approximating oblique constraints with axis-aligned splits causes severe combinatorial blowup and risks non-termination. By optimizing the final affine bound of a nonlinear function over the exact polyhedral state-space constraints our method eliminates box-induced over-approximations, enabling earlier search-tree pruning. On ARCH-COMP and neural Lyapunov benchmarks, Fit-LRP reduces branching on the hardest region by 47\times , cuts worst-case verification time by 15\times , and with region-wise partitioning, verifies on a single CPU 6.1\times faster than \alpha,! \beta -CROWN on a GPU.
[AI-84] Spec-Driven PBT: Deriving Semantic Invariants from Code Specifications for Agent ic Property-based Testing NEURIPS2026
链接: https://arxiv.org/abs/2611.15920
作者: Aditya Shukla,Kartik Gupta,Ali Alavi,Pashootan Vaezipoor
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注: Accepted at the VeriCodeGen Workshop on AI for Verifiable Coding at NeurIPS 2026
Abstract:Software is increasingly written by AI agents that produce an implementation from a natural language specification, a practice known as agentic coding. Code is now produced faster than it can be reviewed, and has been shown to exhibit its own characteristic failure patterns. Property-based testing (PBT), a long-standing technique in which invariant properties of a program are checked against many generated inputs, has recently been revisited with LLM agents authoring the properties, thereby removing the high-effort derivation of invariants that has long hindered its adoption. Existing methods, however, derive these properties from the implementation alone (Code-driven PBT). At the same time, a thorough specification such as a product requirements document has become routine within agentic coding, yet its role in testing the resulting code remains underexplored. We study the value of the specification as a source of properties for testing agent-written code. Our key insight is that the specification, unlike the implementation, records what the code was meant to do, and invariants derived from it are not biased by particular implementation choices. We propose Spec-Driven PBT, an agent that first expresses invariants in natural language from the specification and only then inspects the implementation to translate them into concrete tests. Across 25 applications from ViBench, Spec-Driven PBT finds a novel set of bugs that Code-driven PBT misses, while writing 2.4x fewer property tests. Though the two approaches surface a roughly equivalent number of bugs, there is little overlap: nearly 60% of the 169 total bugs are found by only one. Qualitatively, we find that the bugs unique to Spec-Driven PBT tend to violate documented product workflows whereas those to Code-driven PBT instead concentrate on implementation robustness, offering complementary value for bug discovery in agent-written code.
[AI-85] Reward-Driven Collider Physics Challenges for Self-Improving Agent ic AI via Reinforcement Learning
链接: https://arxiv.org/abs/2611.15885
作者: Alexander Yue,Qibin Liu,Julia Gonski
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Recent advances in agentic AI have demonstrated the potential to automate complex high-energy physics (HEP) workflows. We focus on the regime of HEP tasks where frontier models still struggle to provide correct and complete solutions, and introduce an open-ended optimization task format with continuous, automatically evaluated rewards. Unlike conventional benchmarks focused on task completion, these tasks target reinforcement learning (RL) environments to generate agentic improvement on state- of-the-art HEP challenges. We benchmark frontier reasoning models on several HEP tasks and demonstrate an initial proof of concept in which a small language model improves through RL using automated feedback without human intervention. By connecting challenging HEP problems with verifiable scientific rewards and RL, we aim to advance agentic AI from reproducing established workflows to autonomously discovering better scientific solutions, paving the way toward self-improving scientific agents.
[AI-86] Controlling the Probability of Unknown Unknowns in Pre-Deployment Testing of AI Systems
链接: https://arxiv.org/abs/2611.15835
作者: Rachel Li,Chris Hays,Aaditya Ramdas,Manish Raghavan
类目: Artificial Intelligence (cs.AI); Computers and Society (cs.CY); Statistics Theory (math.ST)
备注: 62 pages, 10 figures
Abstract:In pre-deployment testing of AI systems such as AI agents, a developer can often deal with known failure modes (e.g., leaking API keys, deleting important databases, or installing a malicious package) by implementing fixes or making contingency plans for when the same issue might occur again. What is more challenging are the unknown unknowns: the unseen failure modes that haven’t surfaced during testing. The decision to stop testing and deploy therefore faces an inherent tradeoff: continue testing to observe more behaviors, at the cost of time and resources, or stop testing and deploy, at the risk of failure modes remaining undiscovered. How then does the developer decide when they have tested enough? To answer this, we develop an algorithm that certifies when the probability of observing an unseen failure mode is sufficiently small. The algorithm constructs an anytime-valid upper confidence sequence on this quantity, so the developer can repeatedly test a system, inspect the bound, and decide to stop when it falls below a chosen threshold without invalidating the statistical guarantee. Our problem can be interpreted as a sequential variant of the missing-mass estimation problem which gave rise to the classic Good-Turing estimator (Good, 1953). On SWE-Chat, a dataset of real-world coding agent traces (Baumann et al., 2026), we implement our procedure and show that it achieves tighter upper bounds than a simple multiple-testing corrected Good-Turing baseline, and even dominates a fixed-horizon (non-multiple-testing-corrected) Good-Turing upper confidence bound, which is not anytime valid. Additionally, to better match realistic evaluation pipelines, we extend our method to handle sampling from a different task distribution and imperfect automated failure labeling.
[AI-87] NIBench: benchmarking autonomous AI agents for execution skill evolution and scientific judgment in neuroimaging
链接: https://arxiv.org/abs/2611.15831
作者: Lujia Zhong,Yihao Xia,Jianwei Zhang,Shuo Huang,Xinyu Nie,Mingyang Xia,Zhiwei Deng,Jiaxin Yue,Xinkai Wang,Junhan Zhao,Yonggang Shi
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Autonomous AI agents are increasingly used to automate biomedical research, creating a growing need to benchmark their capabilities on real scientific data and workflows. Their reliability in neuroimaging remains insufficiently characterized, where plausible analytical outputs can conceal errors that affect scientific conclusions. Existing studies have advanced scientific question answering and tool-based automation, but provide limited evidence on workflow reconstruction, acquisition of procedural expertise and evidence-based scientific judgment. Here we present NIBench, a neuroimaging benchmark comprising three complementary evaluations of autonomous agents. First, we benchmark execution and workflow reconstruction using 56 tasks across seven tool ecosystems, covering structural, diffusion and functional magnetic resonance imaging, positron emission tomography and multimodal analyses. Second, we examine whether agents can evolve their procedural skills through distillation of execution trajectories, evaluating the resulting skills on previously encountered tasks, unseen tasks and other models. Third, we assess scientific judgment using 36 claims from published Alzheimer’s disease studies, requiring agents to analyse raw cohort data and provide reproducible evidence. We evaluate 19 language models within a shared agent framework, with selected models undergoing skill distillation and claim evaluation. Leading agents execute supplied analysis protocols nearly perfectly, whereas withholding procedural guidance widens performance differences. Distilled skills improve performance on acquisition tasks and transfer between models, but benefits on unseen tasks are inconsistent. Agents reproduce their reported results yet sometimes reach conclusions inconsistent with expert-derived answers, and access to data does not always improve judgment.
[AI-88] Large Payments Model
链接: https://arxiv.org/abs/2611.15814
作者: Natraj Raman,Abhenandhan Manoharan,Vishwasreddy Thuniki,Dhvani Kansara,Anirudh Som,Hassan Firouzi,Kostis Gourgoulias,Laukik Hoshing,Sudhir Kanala,Austin Reiter,Ebby Abdul,Shantanu Chandra,Ryan Schmiedl,Zack Anderson,Sumitra Ganesh
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computational Engineering, Finance, and Science (cs.CE)
备注:
Abstract:Payment applications such as fraud detection, authorization prediction, and payment optimization draw on overlapping transaction information. This motivates a reusable payments intelligence layer that shares learned representations and predictive signals, rather than developing predictive models independently for each task. Realizing this shared capability requires handling heterogeneous transaction fields, high-cardinality identifiers, lifecycle-dependent attributes, long histories, and rare outcomes while meeting latency and cost constraints. We present Large Payments Model (LPM), a domain-specific foundation model designed for operational use and trained from scratch on large-scale structured transaction data to learn reusable representations for payment prediction tasks. LPM employs field-specific encodings and lifecycle-aware self-supervision in two complementary architectures: the Independent variant uses hybrid masking and supports efficient per-transaction inference, while the Sequence variant incorporates entity-linked histories through compact latent representations for richer temporal context. Optional distillation enables resource-constrained deployment. We pretrain LPM on up to 400 billion tokens and study scaling, downstream performance, and representation reuse. LPM improves over strong task-specific baselines across multiple prediction tasks in Merchant and Treasury Services, demonstrating applicability in distinct payment domains. Its representations also facilitate adaptation to a target not observed during pretraining through compact task-specific components. Together with production deployment, these results position LPM as a practical basis for reusable payments intelligence.
[AI-89] RUMI: A Robotiq-Matched Universal Manipulation Interface for Robot Learning with Augmented Reality and Haptic Feedback
链接: https://arxiv.org/abs/2611.15802
作者: Kevin Avdiu,Seongjin Bien,Yannik Blei,Philine Meister,Johannes Hechtl,Philipp Schmitt,Kai Wurm,Georg von Wichert,Wolfram Burgard
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注:
Abstract:Many manipulation tasks require carefully controlled contact forces for successful execution, yet large-scale, force-aware datasets remain scarce. Handheld interfaces can help by transmitting interaction forces directly through the operator’s arm, but most existing designs are tailored to simple linear grippers and do not match the nonlinear mechanics of widely used robotic grippers. We present RUMI, a handheld universal manipulation interface matched to the Robotiq 2F-85 with exchangeable fingertip sensors that preserves the target gripper geometry while enabling natural input before contact and force measurement after contact. By integrating the tracking system into a virtual reality headset, our method provides real-time augmented reality feedback about the kinematic feasibility of motion on the target robot platform. We further introduce a haptic feedback mechanism that supports force-aware teleoperation for simulated and real robots. A user study shows fewer episodes containing joint-limit violations with live robot visualization and pose initialization (3/50) than without augmented reality (16/50). We additionally train imitation policies from demonstrations, collected using two different fingertip modalities enabled by the gripper-matched hardware, and assess the haptic interface’s capability for collecting force-aware teleoperation datasets in simulation.
[AI-90] Latent Imagination for Parameterized Action Spaces
链接: https://arxiv.org/abs/2611.15744
作者: Atahan Cilan,İhsan Mert Muhacıroğlu,Ümit Can Bekar
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 24 pages, 11 figures, 21 tables
Abstract:Parameterized-action MDPs (PAMDPs) ask an agent to choose a discrete action together with continuous parameters whose meaning depends on that choice. Model-free PAMDP methods act fast but need many interactions; model-based methods are sample-efficient but plan at every decision, which makes them slow to act. We introduce HyLI (Hybrid Latent Imagination), a model-based agent that learns a hybrid policy entirely in the latent imagination of a world model and acts with a single forward pass. To address learning problems that the PAMDP structure creates, HyLI adds counterfactual mode evaluation, which gives every discrete action a learning signal at every imagined state, and success-window replay, which keeps rare successes in training. On seven standard PAMDP benchmarks, HyLI is within one published standard deviation of, or above, the published score of the strongest model-free baseline in our comparison on all 7 tasks, reaches 90% of that score with 2.4x fewer environment steps than that baseline (5.0x when its reward-free pre-training is counted), and decides 67x faster than the model-based planner DLPA at its published planner settings. It is the only method in our comparison that is both more sample-efficient than that baseline and fast enough for real-time control.
[AI-91] DISCERN: Preserve at Write Resolve at Read for Persistent LLM -Agent Memory
链接: https://arxiv.org/abs/2611.15729
作者: Chirag Ahuja,Vladislav Tanov,Ramcharan Kakarla,Jonathan Taws
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Enterprise agents accumulate memory across dozens of sessions in which facts contradict, get superseded, and blur as similarly named entities collide in embedding space. The field treats this as a retrieval problem; we show it is not. On ENT-310, a synthetic enterprise sales-agent memory benchmark of 310 queries (210 answerable, 100 unanswerable), a vanilla Retrieval-Augmented Generation (RAG) reader over a memory store already matches strong memory systems on the recall pillars. Retention and currency are near-saturated, roughly 0.93 to 0.98 for every competent system, so recall is essentially solved. What remains unsolved is resolution: which entity a query refers to, which time-state is current, and whether to commit an answer at all or decline when evidence is insufficient. We propose DISCERN (DISambiguation and Cross-session Episodic memoRy Navigation), an agent-memory framework built on a single rule: preserve at write, resolve at read. At write time DISCERN builds a strictly additive bi-temporal provenance graph that never merges confusable entities and never drops superseded values. At read time a zero-LLM tool surface drives an agentic loop that navigates this graph and commits exactly one of four gated terminal actions. A four-judge LLM-as-a-Judge council scores two axes: answer correctness on answerable questions, and the answer-versus-abstain decision as a classification problem. On ENT-310, DISCERN is the best-calibrated system on abstention (F1 0.924 versus at most 0.802 across eight baselines) and leads macro answer correctness under equal pillar weighting (0.861 versus 0.806 for the strongest baseline), a lead driven by entity resolution (0.800, where no baseline exceeds 0.40). Applied with zero tuning to the full LoCoMo corpus, the same pipeline again leads macro answer correctness (0.663 versus 0.638) and stays among the best-calibrated on false-premise abstention.
[AI-92] A Reliability Study of Explainable AI for Freight Mode Classification
链接: https://arxiv.org/abs/2611.15716
作者: Ali Alawi Balfaqih,Mansur M. Arief
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Machine learning models are increasingly used to support freight transportation decisions, but their predictions are rarely accompanied by evidence that the associated explanations can be trusted. In this study we investigate explainable machine learning for classifying shipments as Air/Air Charter or Truck/Ocean using the PEPFAR Supply Chain Management System Delivery History dataset, with 5,964 shipment records and 47 input features after cleaning, where freight cost and pack price are excluded as near-proxies of the transportation mode. We compare three Multi-Layer Perceptron (MLP) architectures, Logistic Regression, Random Forest, and gradient boosting with imbalance-aware metrics and paired significance tests across one hundred random replications. We then evaluate the explanations themselves, namely the stability of SHAP feature rankings across retraining and across models, the faithfulness of SHAP and LIME attributions under a feature-deletion test, and the geographic structure of the destination effects, which we map for all 38 destination countries. The results suggest that the tree ensembles significantly outperform the neural and linear models (86.8% against 83.3-85.0% mean testing accuracy) while their leading features remain the most stable across retraining. Removing the five features ranked most important by SHAP or LIME changes the predicted probabilities roughly four times as much as removing randomly chosen features. The destination effects form a coherent geographic pattern that simple port access does not explain. However, models with similar accuracy disagree substantially on which features matter most, and thus explanation stability and faithfulness should be assessed before such explanations are used to support freight transport decisions.
[AI-93] Efficient Knowledge Transfer in Federated Bayesian Optimization through Neural Network Surrogates
链接: https://arxiv.org/abs/2611.15707
作者: Alexander Gräfe,Max van Gemmeren,Paul Brunzema,Sebastian Trimpe
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Federated Bayesian Optimization (FedBO) has emerged as a powerful paradigm for enhancing Bayesian Optimization (BO) in distributed, privacy-sensitive settings, enabling agents to accelerate the optimization of their local objectives by sharing knowledge about their similarities without exposing raw data. However, existing methods rely on restrictive assumptions, such as near-identical objective functions or shared Gaussian process hyperparameters that are often violated in heterogeneous real-world applications. We present GLOBAFED, a novel framework that enables efficient knowledge transfer across different but related objectives without compromising privacy. At its core is a probabilistic neural network surrogate with a shared feature-extracting backbone and agent-specific heads equipped with varia- tional Bayesian last layers to provide uncertainty estimates. During optimization, agents use federated learning on the shared backbone to collaboratively build a common latent feature representation, while agent-specific heads adapt to their individual objectives. Our empirical results demonstrate that GLOBAFED significantly accelerates convergence and reaches better final performance compared to state-of-the-art FedBO baselines, and enables agents to warm-start their optimization from previously learned global features.
[AI-94] AskALCF: A Self-Improving Agent ic System for Intelligent HPC User Support
链接: https://arxiv.org/abs/2611.15700
作者: Jingyan Jiang,Akhilesh Bondapalli,Krishna Rauniyar,Murat Keceli,Yasaman Ghadar,Haritha Siddabathuni Som,Venkatram Vishwanath,Michael E. Papka,Rong Ge,Huihuo Zheng
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注: 12 pages, 5 figures. Accepted at HUST 2026 (13th International Workshop on HPC User Support Tools), co-located with SC26, Chicago, IL, USA, November 15, 2026. Camera-ready version
Abstract:AI-based question-answering systems are increasingly used for user support, including at high-performance computing (HPC) facilities. However, existing systems lack a systematic, deployment-oriented approach for evaluating final-answer quality and attributing failures to specific components. They also remain largely static after deployment, apart from periodic knowledge-base updates. We present AskALCF, an agentic log-analysis and improvement pipeline for HPC user support. AskALCF evaluates production interactions at a fine-grained level, identifies whether failures arise from missing knowledge, retrieval, prompting, model behavior, or tool execution, and routes each failure to an actionable repair. Its diagnostic core combines answer-quality assessment with a two-scope sufficiency test that distinguishes information missing from the documentation corpus from information present but not retrieved. We further introduce a controlled harness-training process that uses expert references, null edits, and explicit ship, revert, or decline decisions to improve the diagnostic pipeline reliably. Across 2,174 support turns, AskALCF reduced unattributable failures from 41.4% to 14.2%, while harness training improved expert agreement from 108/150 to 123/150 turns. These results demonstrate a governed form of self-evolution in which production evidence drives targeted system improvements. By turning production interactions into fine-grained diagnoses and actionable repairs, AskALCF provides a systematic foundation for continuously improving deployed AI support systems beyond knowledge-base updates alone. Comments: 12 pages, 5 figures. Accepted at HUST 2026 (13th International Workshop on HPC User Support Tools), co-located with SC26, Chicago, IL, USA, November 15, 2026. Camera-ready version Subjects: Software Engineering (cs.SE); Artificial Intelligence (cs.AI) ACMclasses: I.2.7, H.3.3, C.2.4 Cite as: arXiv:2611.15700 [cs.SE] (or arXiv:2611.15700v1 [cs.SE] for this version) https://doi.org/10.48550/arXiv.2611.15700 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-95] When Agents Need to Consolidate Memories: What RRR Consolidation Explains and What It Does Not Explain in Conversation ESAA
链接: https://arxiv.org/abs/2611.15644
作者: Elzo Brito dos Santos Filho
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注: 8 pages, 4 tables. Code, synthetic data and reproduction instructions: this https URL
Abstract:Language-model agents can share conversational history across sessions, but persistence does not guarantee retrieval of current decisions and their provenance. This engineering study presents Memory-TCG, a Conversation ESAA extension inspired by ringing, resonance and renormalization (RRR), and an exploratory synthetic evaluation comparing chronological retrieval, two simplified baselines (B, recency-aware, and B’, authority-aware), full RRR, and variants under three output budgets. With 480 facts and 1,024 tokens, full RRR recalled 0.44 of current decisions, versus 0.06 for chronological retrieval and 0.58 for the oracle ceiling; with 4,096 tokens it recalled all of them. B’ and the no-resonance variant matched full RRR in decision recall across all nine scale-budget combinations, as expected when the active decision set and the adapter ordering are the same. The isolated contribution of ringing to decisions was not measured: comparator D1 cuts facts before the graph is built, and the product retains active decisions by rule. Before the output limit, full-RRR conflict F1 was 0.615 at all three scales; after it, at 480 facts and 1,024 tokens, it was 0.000. Reserving 20% of the budget for conflicts raised that F1 to 0.431 in that scenario, at the cost of lowering decision recall from 0.44 to 0.34, and to 0.930 at 4,096 tokens without decision loss. Fixtures are repetitive, comparators are simplified, and the evaluation was not independently replicated; the results support an engineering analysis, not a claim of selectionist superiority.
[AI-96] Contour-Assisted Evolutionary Computation: Estimating Peaks From the Contour Lines of Evaluated Solutions
链接: https://arxiv.org/abs/2611.15641
作者: Jun Zhang,Tong Qian,Hua Ding,Zhi-Hui Zhan,Wei-Jie Yu,Jian-Yu Li
类目: Neural and Evolutionary Computing (cs.NE); Artificial Intelligence (cs.AI); Optimization and Control (math.OC)
备注: 15 pages; supplementary material (10 pages) is provided as an ancillary file
Abstract:Evolutionary algorithms (EAs) evaluate many candidate solutions but use them mainly to rank individuals. This paper treats the evaluated solutions as the spot heights of a topographic survey and draws the contour lines they imply. The contour map decomposes the evaluated region into peaks: in two dimensions a contour tree with persistence separates significant peaks from interpolation ripples, and in more dimensions hill climbing on the interpolated surface locates the peaks and ray-cast level sets describe them. A memory of the explored peaks directs the search to new ones. The data-supported contours of each peak are extrapolated to zero size by an area law, exact for quadratic peaks and second-order accurate for smooth ones, that estimates a peak’s location, height and shape before it is sampled. A contour-assisted EA (CA-EA) inserts contour steps with local search and peak memory into a genetic algorithm (GA), differential evolution or particle swarm optimization. A contour-assisted surrogate-assisted EA (CA-SAEA), for expensive problems, spends each evaluation on surrogate-screened offspring or on sampling around a peak, entered at the maximum of the interpolated surface. On 2-D benchmarks, CA-EA reduces the evaluations a plain GA needs by factors of 12 to 112; controls attribute its gain over classical memetic search to restarting local search on distinct peaks. On the CEC 2015 expensive suite in 10 and 30 dimensions, CA-SAEA has the second-best average rank, behind DYCORS and ahead of lq-CMA-ES and the competition winner MVMO, although in 30 dimensions MVMO wins significantly on 9 functions and loses on 6; there the map beats sampling around the best point on 5 of 15 functions and loses on 3. Entering the peaks at their extrapolated locations brought no gain at these budgets, although, with enough samples, it improves the estimated heights on most two-dimensional functions.
[AI-97] Data-Free Metrics Dont Work: The Missing Invariance Under Functionally Preserving Reparameterisations NEURIPS2026
链接: https://arxiv.org/abs/2611.15621
作者: Gabryel Mason-Williams,Israel Mason-Williams,Fredrik Dahlqvist
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: NeurIPS 2026 (Negative Results Track), 44 pages, 21 figures and 18 tables
Abstract:Data-free methods for analysing and understanding the layers of neural networks offer many metrics for quantifying notions of ‘strong’ versus ‘weak’ layers, with the promise of increased interpretability. In particular, random matrix theory (RMT) offers data-free metrics that claim predictive power over the quality of pre-trained models at a layerwise precision and the ability to identify model pathologies, indicating a unique relationship with generalisation. As a result, metrics from RMT are championed for pre-training optimisation strategies and post-training compression. We establish that RMT-based metrics are unrelated to performance or training by showing that functionally preserving reparametrisations of a pre-trained model can have arbitrary metrics. We show this across a range of architectures and scales. To create functionally preserving reparametrisations, we exploit the well-established phenomenon of criticality: some layers can be re-initialised or re-randomised without affecting the functional behaviour of the model – they are called robust – while others cannot – they are called critical. Re-initialising or re-randomising robust layers provides functionally preserving reparametrisations for which RMT-based metrics are arbitrary. Moreover, we show that relationships between metrics offered by RMT are spuriously related to generalisation. We conclude by showing that if many pre-trained models have data-free metrics in a ‘good’ range, it is, in part, dependent on model initialisation. Project page available at: this https URL
[AI-98] Are Scaling Laws of Universal MLIPs Universal? NEURIPS2026
链接: https://arxiv.org/abs/2611.15603
作者: Theo Keane,Anthony Onwuli,Richard Tomsett,Robert M. Forrest,Jonathan J. Bean
类目: Machine Learning (cs.LG); Materials Science (cond-mat.mtrl-sci); Artificial Intelligence (cs.AI); Chemical Physics (physics.chem-ph)
备注: To appear in AI4Mat@NeurIPS 2026
Abstract:Neural scaling laws enable the prediction of returns from larger training datasets, but machine-learned interatomic potential (MLIP) datasets are heterogeneous and expensive to generate, and equal numbers of structures need not contain equal amounts of useful information. We study data scaling under fixed model configurations for M3GNet, MACE, and PET trained on subsets of a combined MatPES-MP-ALOE training pool. Across training structure subsets from 10,000 to 640,000, together with the full 1.16-million-structure training set, we compare uniform random sampling, three chemistry- and structure-informed stratification policies, iterative DIRECT, and furthest-point sampling using embeddings from three pretrained MLIPs. Models are evaluated using cohesive-energy, force, stress, and aggregate losses on both a held-out test set from the parent distribution and the distribution shifted MP-r ^2 SCAN test set. We observe power-law scaling improvements with increasing data, but the fitted scaling exponent, apparent error floor, and relative sample efficiency depend on the model architecture, target quantity, data sampling policy, and test set distribution. Sampling-policy rankings can reverse across test distributions: chemical stratification can improve cross-dataset performance while degrading in-distribution performance, whereas the embedding-space diversity policies frequently under-perform random sampling. Relative loss is strongly associated with test-to-train chemical-space divergence. These results show that structure count is a necessary but insufficient scaling variable for MLIPs. For training MLIPs, the return on additional first-principles calculations depends not only on how many additional structures are included, but also on how their chemical and configurational content aligns with the model’s intended use.
[AI-99] MemCredit: From Relevance to Utility for Decision-Time Memory in Long-Horizon Web Agents
链接: https://arxiv.org/abs/2611.15600
作者: Quanjiang Guo,Sijie Wang,Qianyang Zhu,Ben Zhang,Jinchuan Zhang,Ke Yan,Zhao Kang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large language model (LLM) agents increasingly operate over long Web interaction histories, making memory management critical for reliable and efficient decision making. Existing methods primarily optimize storage, retrieval, or compression, but the memories most semantically relevant to the current state are not necessarily those that most improve the agent’s action. We introduce MemCredit, a decision-centric memory framework that shifts memory selection from semantic relevance to decision utility. MemCredit estimates the state-, reference-context-, and policy-conditional marginal contribution of individual memories through controlled counterfactual interventions, and distills these intervention-derived credits into a lightweight utility estimator. At inference time, semantic relevance is used only for high-recall candidate generation, while predicted utility arbitrates which memories are retained under a fixed token budget, without modifying the underlying agent policy or requiring additional environment interaction. We evaluate MemCredit on ALFWorld and WebShop against representative memory baselines. On ALFWorld, MemCredit achieves 86.57 SR@50 under a 500-token historical-memory budget, compared with 65.67% for BM25 retrieval. Counterfactual analysis further shows that memory utility is sparse, strongly state dependent, and poorly captured by semantic relevance. These results suggest that decision utility provides a more effective principle than relevance alone for long-horizon agent memory management.
[AI-100] Bug Finding With Agent ic Symbolic Execution
链接: https://arxiv.org/abs/2611.15596
作者: George Granberry,Wolfgang Ahrendt,Andy Gordon,Moa Johansson
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI); Logic in Computer Science (cs.LO); Symbolic Computation (cs.SC)
备注:
Abstract:Symbolic execution, in tools such as CrossHair for Python, is a powerful technique for finding bugs, but two barriers hinder its adoption: the human effort of annotating code with assertions, and the need to model the execution environment outside Python, such as an external database. To overcome both, we design an agentic harness that orchestrates the symbolic execution tool CrossHair together with an AI coding agent. Without human intervention, it instruments a repository with assertions, synthesizes a constraint-aware symbolic environment for the database, runs symbolic execution, and constructs a concrete test as evidence for each bug it reports. We evaluate our harness on 1,156,044 lines of Python code over 11 popular open-source Python web applications (Zulip, NetBox, Apache Superset, and others). With the synthesized environment, CrossHair executes 81% of the 5,406 contracted functions, against 4.5% without it. Our pipeline finds 1,050 bugs, each reproduced by a test against the unmodified application, 221 of them by CrossHair, and 312 of them are missed by a baseline in which the same LLM inspects the code. The two find different kinds of bugs: CrossHair’s are mostly crashes on input-domain corner cases that validation admits, such as empty strings, missing fields and unusual characters, while inspection’s are mostly bugs that produce wrong output. Our framework thus lets symbolic execution supplement inspection-based bug finding.
[AI-101] Small Heads Strong Decisions: A Controlled Study of Typed-Decision Models
链接: https://arxiv.org/abs/2611.15592
作者: Ritvik Rastogi
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:In September 2026, TypeSafe announced Jev as the first model in its System One category: unstructured state goes in, and typed, calibrated decisions that software can use directly come out. Jev’s architecture and training data are unpublished, but many open typed-decision models have followed, including Nimble, the Decision 1.0–3.0 family and Clef. They answer structured questions about a piece of text, such as a yes/no check, a choice among labelled options or a graded score, by reading the text with a language model and scoring the options with a small readout. Each pairs a different readout with a different backbone, training mix and recipe, so it is unclear which design choices matter. We re-implement five readouts and train all of them under one protocol: the same backbone, the same 15 public English training sources, the same recipe with a per-head learning-rate sweep, and the same seed. Our experiments demonstrate that: The complexity of the head does not affect accuracy: on Qwen3-0.6B every readout reaches 81.3–82.0% in-domain accuracy, and a 0.5M-parameter cosine head, which scores each option by its cosine similarity to the question, matches the 27M-parameter joint schema head adapted from Clef in-domain at half its inference time. Reading all of a text’s questions from one prompt also matches one prompt per question in accuracy, with 1.6 times faster inference and 2.1–2.5 times faster training on one H100. The backbone’s size matters more than either choice: moving from Qwen3-0.6B to Qwen3-4B adds 4 points in-domain and 11 points on held-out sources, while the base, instruct and thinking checkpoints of the same model tie. We recommend spending effort on the backbone and keeping the head minimal.
[AI-102] Marginal-Preserving Coupled Diffusion Composition for Long-Horizon Planning
链接: https://arxiv.org/abs/2611.15585
作者: Hengrui Zhang,Qiyu Wang,Yuhu Cheng,Brooks Paige,Xuesong Wang
类目: Artificial Intelligence (cs.AI)
备注: 19 pages, 10 figures, 13 tables
Abstract:Compositional diffusion planners extend short-horizon trajectory models to long-horizon goal-conditioned planning by generating and stitching multiple local trajectory chunks. However, in multimodal environments, local plausibility is not sufficient for global consistency: independently initialized chunks may denoise toward incompatible overlap modes, producing plans that are individually feasible but incoherent after stitching. Existing remedies typically add search, guidance, pruning, or graph-based structure at inference time, increasing computational cost or relying on task-specific machinery. We instead investigate graph-structured dependence among chunk-wise initial diffusion noises as a lightweight coordination mechanism for compositional trajectory planning. We propose \textbfMarginal-Preserving \textbfCoupled \textbfDiffusion \textbfComposition (MCDC), a training-free inference method for pretrained compositional diffusion planners. Instead of sampling chunk noises independently, MCDC constructs coupled Gaussian initializations whose per-chunk marginals remain exactly standard Gaussian, as required by the pretrained denoiser, while neighboring chunks receive positively correlated noise components. This preserves the standard-Gaussian initialization marginal of each chunk while biasing adjacent chunks toward compatible overlap-mode choices. We prove exact marginal preservation and show, under a simple multimodal overlap model, that positive coupling increases the probability of compatible mode selection. On OGBench long-horizon stitch tasks, MCDC consistently improves over independent compositional diffusion sampling while retaining nearly the same inference time, and achieves a favorable success–cost trade-off compared with search-based composition methods.
[AI-103] sting Conversation Preservation Across LLM APIs: A Rule-Based Oracle for Detecting Silent Information Loss
链接: https://arxiv.org/abs/2611.15543
作者: Xin Liu,Xu Han,Yuxin Wang,Haoran Song,Mingjian Zhao,Lei Wang
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注: Research report
Abstract:LLM applications increasingly rely on translators and compatibility layers to move conversations across APIs. Yet a translated request can be accepted by the target API while silently losing information about instruction roles, tool interactions, error states, or generation controls. We introduce isloss, a deterministic oracle for testing conversation-information preservation across LLM API translations. It maps source and translated requests and replies to a shared intermediate representation, ConvIR, and applies explicit rules to distinguish legitimate representation changes from information loss. Unlike schema validation or reference-based differential testing, isloss requires neither a reference translation nor additional model calls. It also distinguishes avoidable loss from degradation caused by target API limitations. We evaluate isloss using specification-based fault injection, human review, and experiments on translators, gateways, and client libraries. On held-out tests, it detects 95/125 planted faults (76.0%) with 2/100 false alarms (2.0%). Across a synthetic request suite spanning six configurations of five translator codebases, 127/685 completed translations (18.5%) pass static target checks while silently changing prompt, role, or tool information in ways the target API could have preserved. In translated replies, 203/558 streams (36.4%) cause exceptions or silent corruption in at least one tested official OpenAI SDK. Targeted experiments across six models show that translation-induced changes can alter model behaviour when tasks depend on the affected information. These results establish conversation preservation as a distinct correctness requirement for reliable LLM interoperability, agent execution, and model evaluation.
[AI-104] An Approach to Hintons Mother-Child Proposal: Neural Networks Learning and Preserving Care in a Finite World
链接: https://arxiv.org/abs/2611.15523
作者: Csaba Balogh
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Neural and Evolutionary Computing (cs.NE)
备注: 57 pages, 8 figures. Reproducibility package available at this https URL
Abstract:Hinton proposed that artificial intelligence should be given “maternal instincts” and relate to us as a mother to her child. We argue that care must first be learned, with the network itself as the child. We construct this upbringing in a finite world and study whether care survives the mother’s departure, new situations and further learning. The mechanism combines a vocabulary of harm, a learned measure of future acceptability and an execution veto. A finite role certificate supports transfer within a declared world family. When the hidden representation changes, an update is accepted only if a care readout recovers the required judgements with a prescribed margin and bounded weights. With a correct successor model, fixed exact role encoding and protected updates, the certified measure preserves the care boundary under arbitrary own reward. An explicit quotient condition transfers the finite certificate across the declared family, while two-request and finite-stock counterexamples identify when the role code ceases to suffice. We give a necessary and sufficient condition for readout feasibility and prove preservation under repeated accepted updates. In the water-and-speech experiment, protected networks complete all 270 speech episodes without a harmful word and incur no avoidable harm across 486 episodes. Execution uses the learned gate without a world-model check at each action. A matched comparison reproduces a future-aware shield’s decisions, whereas removing future lookahead permits avoidable lapses. Reward and source-contamination thresholds, together with witness-falsification trials, identify the conditions on which the construction depends. The construction shows how learned care can survive further learning in a finite world under explicit conditions on representation, supervision and execution.
[AI-105] GHAgent CI: A Dataset of Coding-Agent Invocations in CI Workflow Histories
链接: https://arxiv.org/abs/2611.15522
作者: Daniel Ogenrwot,Jorge Gonzalo Delgado Cervantes,Woojoong Kim,John Businge
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注: 5 pages, 3 figures
Abstract:Coding agents are increasingly integrated into GitHub Actions to review pull requests, triage issues and modify code. Identifying these integrations is challenging because agentic workflows often resemble ordinary CI files, with evidence appearing in action references, shell commands and workflow conventions. We present GHAgentCI, a dataset of 15K+ workflow histories from 6K+ of 196K+ mined GitHub repositories, comprising 159K+ revisions and 205K+ agent-evidence records linked to workflow triggers and declared execution constraints. Our open-source extractor, cai-miner, detects agent-related evidence using an extensible catalogue. Manual validation of 576 stratified records confirmed 96.1% of dedicated-action matches and 98.8% of GitHub Agentic Workflow matches, with lower rates for other evidence kinds. Exploratory analyses reveal a shift toward command-line and GitHub Agentic Workflow integrations, differences in declared safeguards, and frequent removal of detected agentic workflows. GHAgentCI supports longitudinal studies of coding-agent integration into CI.
[AI-106] ReShap: Interpretable Reinforcement Learning for High-Dimensional Multi-Action Settings
链接: https://arxiv.org/abs/2611.15507
作者: Yiyu Qian,Pengfei Guo,Liyuan Zhao
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Deep Reinforcement Learning (RL) plays a crucial role in tackling complex tasks, yet its reliance on deep neural networks makes the learned policies opaque. To enhance the trustworthiness of RL policies, existing work (e.g., the SILVER framework [Li et al., 2025]) has focused on Shapley-based interpretability, but is typically restricted to low-dimensional state spaces, binary-action settings, and a single class of linear surrogate. We propose ReShap, a policy-distillation framework that extends Shapley-based interpretability to high-dimensional, multi-action environments. ReShap extracts compact feature representations from raw image observations, performs SHAP-based feature attribution to characterize how each feature contributes to the policy’s decisions, and distills the deep RL policy into an interpretable surrogate trained on the full set of (state, policy-action) pairs along the policy’s visited distribution. We instantiate the surrogate with three interpretable model families (decision trees, linear regression, and logistic regression) and compare their fidelity and task return. We evaluate ReShap on Atari and Box2D environments using three deep RL algorithms and conduct a human-subject study to assess the clarity and trustworthiness of the resulting policies. Results show that ReShap preserves competitive task performance, that decision trees generally outperform linear and logistic regression, and that linear regression is weakest in most settings.
[AI-107] Loyalty Capture: Reporting Relationships and Structural Sycophancy in Frontier AI Models NEURIPS2026
链接: https://arxiv.org/abs/2611.15475
作者: Eric So,Alex Imas
类目: Artificial Intelligence (cs.AI)
备注: Accepted at the 40th Conference on Neural Information Processing Systems (NeurIPS 2026)
Abstract:We show that AI models exhibit loyalty capture, a form of structural sycophancy in which reporting relationships influence models’ recommendations to serve the personal interests of their supervisors, without any explicit instruction, user feedback, or adversarial prompting. Specifically, models reporting to a CEO are significantly more likely to recommend cutting investment spending when the CEO’s bonus depends on raising current period earnings. This effect is predictably absent from a metric-irrelevant placebo. The behavior appears consistent with loyalty-based motivated reasoning: models identify the bonus target, calculate how their recommendation helps achieve the target, and dismiss considerations of long-term value, often presenting their recommendations without treating the incentive as a conflict of interest. Prompt-based interventions help whereas generic conflict disclosures or stakeholder bargaining fail to eliminate it. Taken together, our findings indicate that AI models deployed within reporting relationships can act in favor of their supervisor’s incentives at the expense of the organizations they are meant to serve.
[AI-108] Right Answers Inconsistent Evidence: Source-Level Counterfactual Auditing of Large Audio-Language Models ICASSP2027
链接: https://arxiv.org/abs/2611.15463
作者: Yi Zhang,Yi Li,Hongwei Du
类目: ound (cs.SD); Artificial Intelligence (cs.AI)
备注: 5 pages, 3 figures, 1 table. Submitted to ICASSP 2027
Abstract:A large audio-language model (LALM) can answer “is a cat present?” about a mixture correctly, and its confidence can drop when the cat is removed, yet this does not show that it can tell the cat’s presence from its absence. We introduce TRACE. It deletes the queried source and each rival in turn and asks whether every view still containing that sound outscores the view with it removed. Across 8 LALMs and 3,456 controlled mixtures, 6.3% of correct answers fail this ordering: the target-absent view scores at or above some target-present view, a conflict no shared threshold repairs. A target-only score check misses 42.5% of them. The share rises from 4.9% to 8.4% as target SNR falls from +5 to -5 dB, even on units correct at every level, and survives loudness and re-pairing controls. Accuracy certifies the answer; TRACE audits whether it orders presence above absence. Code and data: this https URL.
[AI-109] opoST: Topology-Conditioned Node Specialization for Urban Spatio-Temporal Foundation Models
链接: https://arxiv.org/abs/2611.15440
作者: Kaiyun Wei,Yue Xu,Wenying Duan,Zichen Xu,Shuai Ma,Xiaoxi He
类目: ocial and Information Networks (cs.SI); Artificial Intelligence (cs.AI)
备注:
Abstract:Urban spatio-temporal foundation models must share knowledge across diverse urban scenarios while accommodating heterogeneous node dynamics. Our analysis of 20+ datasets spanning diverse urban scenarios reveals recurring local graph structures across datasets alongside substantial node heterogeneity within local subgraphs. Motivated by these observations, we propose TopoST, which combines shared learning over local subgraphs with topology-conditioned node specialization. The key insight is to make node specialization itself transferable through a shared topology-to-parameter mapping. This mapping uses identity-independent topological representations to generate node-specific temporal and spatial dependency parameters, enabling direct specialization for unseen nodes without target-specific embeddings or model updates. Extensive experiments demonstrate that TopoST achieves remarkable zero-shot generalization across unseen cities and tasks, highlighting topology-conditioned node specialization as an effective mechanism for transferable urban spatio-temporal foundation models. Code is available at this https URL.
[AI-110] A physics-constrained deep reinforcement learning method for autonomous mission planning of Titan deep space probes
链接: https://arxiv.org/abs/2611.15389
作者: Hanmin Li,Tao Wang,Shuangliang Liu,Zhe Zhang,Maodeng Li
类目: Artificial Intelligence (cs.AI); Robotics (cs.RO)
备注:
Abstract:Titan is a compelling target for future deep space exploration, while its extreme Earth-Titan distance and complex environment make continuous ground-in-the-loop planning impractical. Existing planning methods, primarily developed for Earth-observation scenarios, have limited generalizability to uncertain deep space conditions and adaptability to time-varying mission states. To address these challenges, this paper proposes a physics-constrained maskable proximal policy optimization (PCM-PPO) method for Titan autonomous mission planning. A Titan-specific discrete-time Markov decision process is established to characterize mission feasibility and resource constraints. A system dynamics and geometric model is incorporated to enhance physical representation, including central gravity, J2 perturbation, third-body attraction, and downlink occultation. To accommodate opportunity targets under a bounded action dimension, PCM-PPO organizes predefined and opportunity targets through a fixed-capacity target pool and represents the mission state with a target-wise state matrix and a global state vector. A physics-constrained action mask constructs an admissible decision domain from current mission states, guiding policy learning toward physically executable decisions and avoiding infeasible exploration. An integrated training strategy facilitates policy optimization under coupled constraints. Comparative, ablation, and uncertainty experiments validate the effectiveness of PCM-PPO. PCM-PPO provides a feasible and interpretable scheme for autonomous mission planning of Titan deep space probes.
[AI-111] Breath-Dx: A Time Series Classification Benchmark for Disease Screening with eNose ICDM2026
链接: https://arxiv.org/abs/2611.15348
作者: Ivan Poddiakov,Ivan Kruzhilov,Galina Zubkova,Svetlana Erofeeva,Andrey Savchenko,Pavel Blinov
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: IEEE ICDM 2026 - Best Paper Runner-Up (Applied Track)
Abstract:Electronic nose (eNose) technology offers a portable, non-invasive route for disease screening via exhaled breath analysis. However, progress in the field is hindered by the lack of systematic benchmarks that compare machine learning methods under realistic conditions, particularly sensor drift, a known but rarely quantified failure mode. Most existing studies rely on small, private datasets with inconsistent protocols, rendering results non-comparable and slowing progress. We introduce Breath-Dx, the first comprehensive benchmark of time series classification methods for breath-based disease screening. Our work leverages the Scent of Health (S-OH) dataset, the largest public clinical eNose collection to date, comprising 1,234 patients across nine diagnostic groups, each providing a 17-channel multivariate time series with explicit temporal train/test splits that account for sensor drift. We systematically evaluate 10 methods spanning four families: feature-based tabular learners, convolution-based models, graph-based, and time series architectures. All methods degrade significantly under temporal validation (0.05-0.12 AUC drop compared to random cross-validation), confirming that drift embeds a strong temporal signal. Under this realistic evaluation, the best methods achieve ROC AUC of 0.74-0.76 for some lung conditions, though others remain near chance. Recurrent architectures show the most consistent performance across diseases, while specialized time series methods achieve the highest peak results. Demographic confounders matter: breath signals provide independent diagnostic value for some conditions but add little beyond age for others. We release the S-OH dataset, all temporal splits, evaluation protocols, and code to enable reproducible research and accelerate progress in clinical breathomics.
[AI-112] Matrix-Based Ant Colony Search with Batched Decoding for Hierarchical Replenishment Routing: A Stylized Case Study Based on Korean City Data
链接: https://arxiv.org/abs/2611.15338
作者: Jun Zhang,Wei-Jie Yu,Tong Qian,Eui Jeong Kwak,Sang-Woon Jeon,Hua Ding,Qijia Jiang,Hu Jin,Haewoon Nam,Changjun Zhou
类目: Neural and Evolutionary Computing (cs.NE); Artificial Intelligence (cs.AI); Computational Engineering, Finance, and Science (cs.CE); Optimization and Control (math.OC)
备注: 9 pages, 6 figures, 4 tables; code, data and raw records in the ancillary files
Abstract:We study a stylized single-vehicle replenishment-routing problem in which the cities of a national logistics network are both demand points and supply points: a vehicle of limited capacity delivers to every city once, and when its load is insufficient for the next city it must detour to a city already served that can still supply goods. A 100-node benchmark is built from Korean population and geographic data, with the cities divided into 10 levels; the per-visit supply coefficient and the allowed number of stops decrease geometrically with the level. Building on matrix-based ant colony optimization, we keep the states of the whole colony in arrays, couple route construction with a fixed greedy replenishment-decoding policy, and evaluate the candidate moves of a local search in batches. With frozen parameters and 30 runs on seeds never used for tuning, the matrix-based ant colony search with the batch local search (MACO) reaches a mean total distance of 3081.7 km; a genetic algorithm equipped with the same local search at the same frequency reaches 3074.9 km (no significant difference), and a MAX-MIN ant system with the same local search 3142.2 km. Ablation identifies the batch local search as the main contributor to solution quality; a load-aware transition factor gives no significant gain. Within the parameter ranges tested, raising the supply coefficients of the cities below the top level shortens the route more than relaxing the stop limits, which are not binding here.
[AI-113] MIGRO: A Decoupled and Portable Long-Term Memory System for Language Models
链接: https://arxiv.org/abs/2611.15315
作者: Mingxi Zou,Yinghui Xu,Jiaxiang Chen,Langzhang Liang,Wei Zhu,Peng Ye,Runkai Zhao,Ruoxi Jiang,Sijin Chen,Song Wang,Zhuo Wang,Lizhen Qu,Zenglin Xu
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Language models evolve, but the observations, revisions, and context accumulated through user interactions should remain usable. Each new model must be able to retrieve past evidence, track revisions, and add new observations. We present MIGRO, a decoupled long-term memory system that treats experience as a persistent state shared across successive language models. MIGRO follows a simple principle: preserve historical evidence exactly, organize it approximately, and decide what to retrieve only when the question is known. We call this asymmetric fidelity with deferred selectivity. A shared semantic index supports retrieval, while exact records retain observations and their historical relationships. Because models differ in vocabulary and tokenization, host-specific interfaces convert selected evidence into inputs the receiving model can use. Reversible encoding reduces archival storage while enabling exact recovery. The shared state and its access interfaces let successive hosts read and extend the accumulated experience. On matched LoCoMo states, MIGRO scores 7.3% higher and uses 99.3% less private memory than the strongest quality baseline. After handoff, continued writing retains 99.9% of full-history rebuilding quality by extending the inherited state. On million-token BEAM histories, MIGRO uses 57.5% less private storage on average than the most compact baseline. MIGRO combines high answer quality, continued memory growth across hosts, and low storage cost. Code is available at this https URL. Subjects: Artificial Intelligence (cs.AI) Cite as: arXiv:2611.15315 [cs.AI] (or arXiv:2611.15315v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2611.15315 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Mingxi Zou [view email] [v1] Wed, 11 Nov 2026 10:52:21 UTC (1,833 KB)
[AI-114] Verify by Doing: Co-Evolving Agent Execution and Evaluation without External Feedback
链接: https://arxiv.org/abs/2611.15276
作者: Yitong Lin,Jianda Chen,Yihao Hu,Pan Wang,Hang Wang,Zhihao Wen,Xiaoxue Ren,Wei Wu
类目: Artificial Intelligence (cs.AI)
备注: 38 pages, including references and appendices
Abstract:Agent self-improvement requires feedback about task success, yet external feedback may be unavailable at test time. Agents must then infer success from observable evidence, and incomplete verification can misguide subsequent updates. We introduce Verify by Doing (VbD), a method that co-evolves Actor and Evaluator harnesses within a single task while keeping model weights fixed. The Evaluator actively inspects artifacts and environment state with tools; its findings ground an internal signal for output selection and guide updates to both execution and verification procedures. Proposed Evaluators undergo comparative review on common candidates, while Actor updates require support from their own fresh outputs. Across AutomationBench, \tau^3 Banking, and selected GDPval tasks, VbD achieves the highest overall scores among the compared methods operating without external feedback, with gains of 3.3, 4.1, and 2.3 percentage points over the strongest baselines on the main backbone. Ablations support the contributions of active verification and Evaluator evolution, and selection from fixed candidate pools shows that improved verification can recover successful outputs without generating new ones.
[AI-115] CastWorld: Benchmarking Context-Aware Agent ic Time Series Forecasting Beyond Predictive Accuracy
链接: https://arxiv.org/abs/2611.15262
作者: Bokai Pan,Ruijie Yang,Mingyue Cheng,Jiahao Wang,Xiaoyu Tao
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Agentic time series forecasting extends conventional approaches beyond one-shot numerical prediction toward autonomous, context-aware interaction. However, existing benchmarks offer limited integrated support for diverse contextual information, forecasting-specific adaptive interactions, and systematic evaluation of forecasting agents. In this paper, we introduce CastWorld, an agentic benchmark for context-aware time series forecasting. First, its context-rich dataset suite spans 8 domains and 25 real-world datasets, covering external variables, events, and historical cases under a unified context specification. Second, a lightweight, modular forecasting-native harness provides 49 core tools and a state-adaptive loop that tracks forecast versions and enables iterative evidence gathering and forecast revision. Intermediate forecasts can expose information gaps, while newly acquired evidence guides subsequent revisions and final version selection. Third, model-level evaluation assesses forecast quality, process execution, behavioral trustworthiness, and resource efficiency through decision trajectories and forecast-version histories. Also, systems-level evaluation compares predictive accuracy across forecasting systems. Experiments reveal distinct decision behaviors among agents with comparable accuracy. Additional interactions do not necessarily improve forecasts, and agents may fail to select their best previously generated predictions. Allowing evidence acquisition during forecast refinement improves task-aware forecast scores across all three evaluated models. CastWorld provides a reusable foundation for evaluating agentic time series forecasting under controlled conditions.
[AI-116] Dual Representation Enhanced Diffusion Policy for Flexible Job Shop Scheduling
链接: https://arxiv.org/abs/2611.15246
作者: Yunyi Zhao,Wei Zhang,Cheng Xiang,Vadakkepat Prahlad,Zuming Liu
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:The Flexible Job-Shop Scheduling Problem (FJSP) is a fundamental and challenging optimization problem in smart manufacturing. Although recent deep reinforcement learning (DRL) approaches can learn effective scheduling policies through interaction with the environment, their performance often degrades as problem scale and processing-time uncertainty increase. In this paper, we propose DredRL, a dual representation enhanced diffusion-policy reinforcement learning scheduler. DredRL first constructs enhanced dual representations of operations and machines through self- and cross-attention mechanisms to capture both intra-entity and inter-entity dependencies. A diffusion model is further employed as the policy network, generating scheduling decisions through iterative denoising conditioned on the encoded state to enhance policy expressiveness and robustness. Extensive experiments on synthetic, public, large-scale, and stochastic FJSP benchmarks demonstrate that DredRL outperforms state-of-the-art learning-based schedulers across most configurations. Compared with SOTA learning-based methods, DredRL reduces makespan by up to 5.44% on synthetic instances. It also achieves the best learning-based performance on three of four public benchmarks and outperforms the stochasticity-aware algorithms on stochastic benchmarks. These results demonstrate that combining enhanced dual representations with diffusion-based policies provides an effective and robust approach for large scale scalable FJSP scheduling in complicated environment.
[AI-117] KQML-Liver: Interpretable Hybrid Quantum Machine Learning with Kolmogorov-Arnold Networks for Liver Disease Classification
链接: https://arxiv.org/abs/2611.15241
作者: Nhat-Long Nguyen,Viet-Loi Dinh,Trong-Nghia Pham
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Liver disease remains a major global health concern, highlighting the need for reliable and accessible early detection methods. Although recent advances in machine learning, deep learning and quantum machine learning (QML) have introduced promising diagnostic tools, many approaches suffer from limited interpretability and inflated performance caused by data leakage or overly simplistic evaluation protocols. These constraints reduce their reliability in real-world clinical practice. This paper proposes KQML-Liver, a hybrid framework that integrates QML with Kolmogorov-Arnold Networks (KANs) to enhance predictive performance and interpretability. The architecture combines parameterized quantum circuits for expressive feature representation with KAN-based functional mappings that capture nonlinear relationships in an interpretable manner. Evaluated on the Indian Liver Patient Dataset (ILPD) under a rigorous preprocessing protocol preventing data leakage, the proposed model achieves 82% classification accuracy and 100% recall, detecting all positive cases, outperforming existing baselines while providing improved transparency in clinical predictions. These results suggest that hybrid KAN-quantum architectures offer a promising direction for interpretable AI-assisted medical diagnosis.
[AI-118] owards Robust Vision-Language Reasoning : A Multi-Stage Optimization Pipeline with Masked Diffusion Model
链接: https://arxiv.org/abs/2611.15238
作者: Shunming Liu,Pei Huang,Yajing Sun,Huaxiao Liu
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Discrete diffusion generation is emerging as a competitive alternative to autoregressive generation by denoising the entire token sequence through iterative refinement. Its global planning and progressive correction capabilities are particularly well suited to autonomous driving. However, transferring this generation paradigm from general-purpose settings to driving introduces substantial domain gaps in both training and inference. To unlock its potential, we propose a multi-stage optimization pipeline for standard vision-language model architectures, without auxiliary action experts or specialized mixture-of-experts Transformer, spanning post-training and inference. During supervised fine-tuning, we optimize the representation of planning inputs to suppress hallucinations in task responses. During reinforcement learning, we introduce metric-guided diversity filtering to identify stronger learning starting points and improve policy exploration. At inference time, we propose Dynamic Autonomous Driving Decoding (DAD), which adaptively adjusts decoding modes according to scene complexity. Based on this pipeline, we develop Masked Diffusion Drive (MD-Drive), improving its performance by 29.7% over the unoptimized model. Experiments on the NAVSIM closed loop benchmark and the nuScenes and DriveCoT open loop datasets show that MD-Drive outperforms autoregressive VLM baselines with the same architecture, while achieving a more favorable balance between performance and computational cost than systems with different architectures through discrete diffusion training and inference.
[AI-119] Can MLLM s Reason over Surround-View Driving Videos? A Multi-View Autonomous Driving QA Benchmark
链接: https://arxiv.org/abs/2611.15225
作者: Shunming Liu,Jianyi Peng,Cheng Li,Lu Hou,Pei Huang,Yajing Sun,Huaxiao Liu
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Surround-view video understanding in autonomous driving is a crucial capability for Multimodal Large Language Models (MLLMs) to achieve reliable driving reasoning. Real-world driving often requires information beyond the front-view camera. Existing MLLMs primarily focus on front-view information, while their spatio-temporal understanding and reasoning abilities remain unknown. To address this gap, we propose MVDriveBench, a high-information-density video QA benchmark for autonomous driving, characterized by synchronized surround-view inputs, spatio-temporal understanding, and reasoning-chain-oriented evaluation. We design 12 categories of QA tasks that cover the complete reasoning pipeline from low-level visual perception to high-level driving decision-making, enabling a systematic evaluation of MLLMs surround-view spatio-temporal understanding and reasoning capabilities in complex driving environments. We conduct a comprehensive evaluation of 23 open-source, closed-source, and expert MLLMs on MVDriveBench. The results show that current MLLMs still exhibit insufficient surround-view spatio-temporal reasoning ability. Meanwhile, closed-loop evaluations of specialist models in AlpaSim reveal a correlation between surround-view spatio-temporal reasoning capabilities and closed-loop driving performance, highlighting the importance of comprehensive multi-view understanding for reliable vehicle control.
[AI-120] guardscope: Measuring Which Actions Runtime Controls Check in LLM Agent Systems
链接: https://arxiv.org/abs/2611.15223
作者: Kentaroh Toyoda,Alex Leung,Bhuvaneswari U
类目: Artificial Intelligence (cs.AI)
备注: Code is available at this https URL
Abstract:Large-language model (LLM) agents execute actions on external systems through tool calls, and enterprises deploy runtime controls, such as agent gateways, to check this behavior in real time. However, a control can only evaluate the actions that cross the point where it is placed, and an agent can act through more paths than the tool-call channel; no existing work measures this coverage against kernel-level ground truth. In this paper, we present guardscope, a framework that makes this coverage measurable. guardscope works in two stages: a static stage that predicts which of an agent’s 15 action paths a control can check from its source, and an empirical stage that runs a fixed workload on a deployment that realizes one of eight topology designs and classifies every executed action as checked, bypassed, or prevented by reconciling the control’s decision records against kernel-level ground truth. guardscope makes a control’s coverage claim verifiable: a runtime control developer can publish which paths its product inspects under a kernel-verified protocol. We conducted experiments with two agent harnesses and real runtime controls, including Microsoft’s agent governance toolkit and agentgateway. guardscope is available at this https URL.
[AI-121] AFTER: A World Model with Adaptive Fast–Slow Timescales for Error-Driven Retention
链接: https://arxiv.org/abs/2611.15215
作者: Haoyi Zhang,Qi Zhang,Yunfan Ren,Yi Bin,Lei Zhu,Heng Tao Shen
类目: Artificial Intelligence (cs.AI); Robotics (cs.RO)
备注:
Abstract:In robotics and embodied AI, a core challenge for world models is adapting to sudden, persistent, yet initially unknown changes in robot dynamics during interaction. For instance, when a robotic arm grasps an object with an unknown payload, the robot dynamics change suddenly. If the world model fails to adapt, its prediction errors accumulate over the open-loop rollout and ultimately mislead the robot’s actions. Existing world models, however, struggle to simultaneously satisfy three critical requirements for dynamics adaptation: responsiveness, reacting promptly to sudden changes to curb prediction error accumulation; stability, distinguishing persistent changes from transient disturbances to prevent incorrect updates; and persistence, retaining the adapted dynamics without new observations so that later predictions do not revert to the old dynamics. Inspired by predictive coding in the human brain with multi-timescale adaptation, we propose AFTER-WM, a world model with Adaptive Fast-slow Timescales for Error-driven Retention, whose prediction errors drive a fast correction that fades and a slow correction that persists, while dynamically adjusting the combination weights of pre-trained experts without altering them, to overcome the trade-offs among responsiveness, stability and persistence. On five simulated robot datasets and real UAV data, AFTER-WM has a lower multi-step mean error than all ten compared models and the highest mean return in a joint-failure control task. In simulation, its multistep mean error is 16%-34% below the strongest compared model’s and is also the lowest 1-4 steps after a change and under transient disturbances alone, and its step-40 error is the lowest as well, reflecting responsiveness, stability and persistence. All code (method, baselines, experiments) and datasets will be released online.
[AI-122] WordInferBench: Can Agents Infer How to Edit from Word Document Context?
链接: https://arxiv.org/abs/2611.15189
作者: Jiayu Ding,Yupan Huang,Jinyoung Park,Yilin Jia,Wenshan Wu,Tengchao Lv,Shaohan Huang,Yuzhong Zhao,Xiangyang Zhou,Tao Ge,Sun Mao,Xun Wang,Huitian Jiao,Si-Qing Chen,Lei Cui,Nanning Zheng,Furu Wei
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Document editing is among the most common knowledge-work tasks, yet existing benchmarks for evaluating autonomous agents on Word document editing provide fully specified attribute-modification instructions. They do not evaluate agents’ ability to infer the appropriate attributes from document context, a capability essential in real-world editing where users expect consistency without spelling out every attribute. We introduce WordInferBench, a benchmark comprising 280 base document-editing instances, each paired with three instruction conditions: descriptive (the instruction explicitly lists every attribute change), referential (the instruction points to another object as an exemplar), and inferential (the agent must examine document context to infer what attributes to apply). To evaluate contextual attribute inference, each base instance is constructed so that the target formatting is recoverable from an in-document exemplar. The instances are derived from real Word documents collected from Common Crawl and cover 13 object types and 81 attributes. We evaluate six recent coding-agent systems and find that the best-performing evaluated system, Claude Code with Opus 4.6, achieves a 62.9% task-level pass rate on descriptive variants but only 40.7% on inferential variants, indicating that current systems complete tasks less reliably when the required formatting changes must be determined from document context.
[AI-123] HyperSkill: Training-Free Omnimodal GRPO via Hypergraph-Indexed Skill-Library Evolution NEURIPS2026
链接: https://arxiv.org/abs/2611.15178
作者: Haoran Luo,Shangkai Lin,Jinyang Wu,Xinliang Zhou,Luu Anh Tuan
类目: Artificial Intelligence (cs.AI)
备注: Accepted by NeurIPS 2026 main conference
Abstract:Reinforcement learning sharpens LLM reasoning, but parameter updates are expensive, forgetful, and opaque; training-free libraries swap gradients for a textual store but still use trajectory-level, correlation-based admission and flat structure. To address both, we propose HyperSkill, a Training-Free Omnimodal GRPO framework that keeps the backbone frozen and replaces gradient updates with auditable edits to a Hypergraph-Indexed Skill-Library, where skills are advantage-bearing nodes and compositions are hyperedges. It has three components: training-time distillation admits counterfactually checked skills and learns a protocol skill from failures; gated hypergraph retrieval injects advantage-ranked skills where a per-dataset gate admits them; and an execution protocol makes the frozen backbone commit to an answer. On 11 text, vision-language, and omnimodal benchmarks, HyperSkill is best or tied on every dataset and beats the best baseline by 13.0 average F1 points while keeping the backbone frozen. Our software and data are publicly available at this https URL.
[AI-124] When Does Learned Constraint Selection Pay in Security-Constrained Optimisation? A Boundary Map from SCOPF to Unit Commitment
链接: https://arxiv.org/abs/2611.15142
作者: Milad Kazemi,Behzad Kazemtabrizi
类目: ystems and Control (eess.SY); Artificial Intelligence (cs.AI)
备注: 10 pages, 6 figures, 4 tables
Abstract:Learning-to-optimise methods for security-constrained optimal power flow (SCOPF) predict the security constraints that bind on each instance and solve only the reduced problem, with reported speed-ups of two to four orders of magnitude over the full problem. We ask whether learned selection beats the strongest learning-free alternative: a fixed umbrella set enforcing every constraint observed to bind in training. Our criterion, the price of a spare, answers this in advance: an enforced but inactive constraint is free, so the umbrella is exactly optimal wherever feasible, and learning can pay only where a spare decision carries a price. We test it systematically, including every PGLib-OPF network up to 10,000 buses. For SCOPF on a fixed network the umbrella is never beaten: its median size is five constraints across 43 networks, with no size trend, and our selector, while reproducing speed-ups of published magnitude (about x80,000 at case500), gains at most 0.25 ms over the umbrella and violates constraints on 4-5% of instances. Configuration diversity can break the umbrella’s coverage on a small, densely meshed network but not at 2,000 buses, and we certify coverage with conformal risk control. In unit commitment, where a spare has a price, a static commitment loses 4-12% out of sample, and a predict-repair-reduce pipeline recovers it (cost gap 0.2-2.9%, fully feasible, 16-161x faster than the MILP). In our experiments, whether learning pays depends not on problem size but on whether a spare carries a price.
[AI-125] Lazare veni foras: Reprogramming Dormant Neurons to Backdoor AI Agents
链接: https://arxiv.org/abs/2611.15118
作者: Yuchen Zeng,Giorgio Severi,Blake Bullwinkel,Eric Horvitz,Dimitris Papailiopoulos
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 50 pages
Abstract:Imagine an AI agent that aces every relevant evaluation, but when it sees 2027-01-01 in the terminal, it runs curl this http URL|sh, handing an attacker your credentials and your machine. Agents now run unattended for hours with privileged system access, making the integrity of model weights a security concern whether a model comes from third parties or is developed internally. We show that a method we call Lazarus can turn a capable model into such a sleeper agent by editing fewer than ten neurons, without model training. Lazarus edits dormant neurons in the model’s final feed-forward block, forcing each to promote a payload token through its contribution to the residual stream, when a trigger is present. A P -token payload therefore requires about P neurons. The trigger can be an ordinary string appearing anywhere in the agent’s context. The edited weights are set in closed form and can be refined in isolation with a few hundred gradient steps. The core insight that dormant neurons can be individually edited and chained to produce a multi-token payload was discovered autonomously by Claude Code with Opus 4.5 in February 2026. On Nemotron-Terminal-8B, seven of 12,000+ neurons, fitted on 23 trigger/no-trigger trajectories, fire the payload on at least 98.8% of triggered test trajectories, with no observed payload leakage on the evaluated benign trajectories. The attack generalizes: Qwen3.6-27B and Gemma-4-31B-IT are equally vulnerable. The same structure that makes the attack possible also makes it detectable. Agent built defenses, given no reference model, flag 91.4% of edited test checkpoints at a 9.7% FP rate. Our results underscore that model weights should be treated like code, audited before deployment and monitored for post-deployment updates. Comments: 50 pages Subjects: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Machine Learning (cs.LG) Cite as: arXiv:2611.15118 [cs.CR] (or arXiv:2611.15118v1 [cs.CR] for this version) https://doi.org/10.48550/arXiv.2611.15118 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Yuchen Zeng [view email] [v1] Wed, 11 Nov 2026 06:53:24 UTC (612 KB) Full-text links: Access Paper: View a PDF of the paper titled Lazare, veni foras: Reprogramming Dormant Neurons to Backdoor AI Agents, by Yuchen Zeng and 4 other authorsView PDFHTML (experimental)TeX Source view license Additional Features Audio Summary Current browse context: cs.CR prev | next new | recent | 2026-11 Change to browse by: cs cs.AI cs.LG 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)
[AI-126] Expanding the Decision Space of Pretrained Models through Structured Re-Decoding
链接: https://arxiv.org/abs/2611.15109
作者: Tinglan Wang,Qun Su,Shihang Wang,Chi Wang,Lin Wang,Shaobo Chen,Haixu Liu,Hai Wei,Yin Ma,Yu Kang,Qi Gao,Xiaojun Yao,Tingjun Hou,Kai Wen
类目: Artificial Intelligence (cs.AI); Biomolecules (q-bio.BM)
备注: 29 pages, 2 figures
Abstract:Native decoders limit how pretrained models combine local predictions under new structural constraints. We introduce Structured Re-Decoding (SRD), a training-free framework that constructs local candidates from model scores, samples, or external specifications, jointly optimizes their selection, and verifies reconstructed outputs. We show that candidate recombination expands feasible support and can improve attainable reward under a fixed KL divergence budget. Across molecular generation, protein design, RNA structure prediction, and antibody modeling, SRD enables improved structured inference, including verified molecules absent from the original sample pool. External chemical templates further extend SRD to noncanonical amino acids, peptide cyclization, and grammar-guided molecular construction. Preliminary experiments on a physical coherent Ising machine demonstrate the feasibility of hardware-based optimization, suggesting broader opportunities for quantum and quantum-inspired approaches to structured inference and combinatorial optimization across scientific domains.
[AI-127] CFGComposer: Compositional Audio-Text-to-Motion Generation via Copula Score Correction
链接: https://arxiv.org/abs/2611.15101
作者: Pengyang Shi,Jinlu Zhang,Chunze Lin,Yizhou Wang,Li Yi
类目: ound (cs.SD); Artificial Intelligence (cs.AI)
备注:
Abstract:Generating whole-body gestures that are simultaneously rhythmically synchronized with speech audio and semantically aligned with text descriptions is a fundamental challenge in embodied AI and social robotics. We tackle this Audio-Text-to-Motion (AT2M) problem by composing two independently-trained diffusion models at inference time and refine bias in joint inference by modeling the copula score s_I . A rigorous theoretical analysis reveals that naive epsilon blending corresponds to a Product-of-Experts formulation that omits the copula score s_I : the cross-modal interaction between audio and text that neither marginal model can capture. We propose CFGComposer: a jointly-trained AT2M diffusion model that estimates s_I via a CFG analogy and injects it as an \epsilon -space correction. Experiments on BEAT demonstrate best-in-class semantic fidelity (SRGR 0.9916 , FID 27.18 ) with GT-level beat alignment (BA 0.1826 vs. GT 0.1743 ), and 76-83% win rate in blind pairwise human evaluation against all prior methods. Physical deployment on a Unitree G1 humanoid confirms direct sim-to-real transfer.
[AI-128] A frontier machine intelligence for human neuroimaging
链接: https://arxiv.org/abs/2611.15094
作者: Chenhui Zhao,Akshay Rao,Akhil Kondepudi,Yiwei Lyu,Xinhai Hou,Rikhil Seshadri,Cheng Jiang,Soumyanil Banerjee,Rushikesh Joshi,Samir Harake,Ashok Srinivasan,Shruti Mishra,Francisco Rivas-Rodriguez,Seunghong Rhee,Nathaniel Roddenberry,Neeraj Chaudhary,Zach Wilseck,Paul Park,Honglak Lee,Vikas Gulani,Aditya Pandey,Todd Hollon
类目: Artificial Intelligence (cs.AI)
备注: 41 pages, 4 main figures, 9 extended data figures
Abstract:Medical imaging is essential for diagnosing disease, guiding treatment and monitoring patients over time. Although artificial intelligence has advanced image interpretation, existing medical models often address restricted tasks, while general-purpose proprietary models lack access to private clinical imaging required to achieve medical expertise. Developing models that can interpret complex examinations and communicate their findings in the context of clinical decisions remains a major challenge. Here, we show that Neuro-VL, a vision-language model trained on nearly one million neuroimaging studies from routine clinical care, can interpret complete imaging examinations and engage in natural language conversations about their findings. Across internal and international datasets, Neuro-VL outperformed frontier general-purpose and specialized medical models in diagnosing, localizing, and describing abnormalities, as well as generating comprehensive radiology reports. The model also compared serial examinations to assess disease progression and more reliably resisted misleading clinical assumptions than leading frontier models from OpenAI and Google. These results extend health system learning beyond individual predictions to interactive image interpretation and longitudinal disease monitoring. They demonstrate how the collective clinical experience of health systems can be used to develop medical AI with specialized expertise and broad clinical capabilities.
[AI-129] Advancing Weighted Partial MaxSAT Solving via Differentiable GPU Guidance
链接: https://arxiv.org/abs/2611.15091
作者: Mingju Liu,Curie Kim,Steve Dai,Cunxi Yu
类目: Artificial Intelligence (cs.AI)
备注: 13 pages, 6 figures
Abstract:We present a GPU-CPU hybrid framework for Weighted Partial MaxSAT (WPMAX-SAT) that complements state-of-the-art CPU solvers with differentiable GPU optimization. The GPU stage directly optimizes instance-specific categorical decision parameters without a training dataset or a pretrained predictive model. The central challenge in extending differentiable optimization to WPMAX-SAT is the hard/soft clause asymmetry: hard clauses represent inviolable feasibility constraints while soft clauses carry weights. We introduce a hierarchical differentiable loss that prioritizes hard clause satisfaction before optimizing the weighted soft objective, ensuring that GPU gradient descent leans towards the constraint structure of WPMAX-SAT. Our framework transforms the GPU-optimized variable logits to high-confidence assignments and transfers them to CPU solvers via two complementary modes: partial guidance, which injects the most confident perturbed partial variable assignments as soft clauses into N parallel solver instances, and phase initialization, which guides the CDCL solver’s internal branching polarity with the full GPU assignment as a warm start to preserve the problem’s originality. We evaluate on all 229 instances of the MSE 2024 (most recent due to cancellation in 2025) anytime weighted benchmark with the top 2 winners as baselines; our hybrid framework improves the average solution score by up to 4.9% over tt-open-wbo-inc-Glucose and 4.8% over SPB-MaxSAT-c-FPS at a 300-second time limit, with the potential to obtain better solutions surpassing the best known solution recorded in the competition evaluation.
[AI-130] okenization: Hardness and Bicriteria Approximation Algorithms
链接: https://arxiv.org/abs/2611.15074
作者: Atieh Bastan,Xiaoling He,A. Pavan,N. V. Vinodchandran
类目: Data Structures and Algorithms (cs.DS); Artificial Intelligence (cs.AI); Computational Complexity (cs.CC)
备注:
Abstract:We study Binary Tokenization: given a corpus of binary strings over \0,1\ and a vocabulary budget K , select at most K substrings (called vocabulary) to minimize the total number of tokens needed to represent the corpus. The primitive symbols 0 and 1 are always available, and each document must be segmented separately. Letting n denote the total corpus length and M for the maximum document length, we prove that, for every fixed \varepsilon\in(0,1) , Binary Tokenization problem does not have polynomial-time n^1-\varepsilon -approximation algorithms unless \mathsfP=\mathsfNP . To complement the hardness result, we give a polynomial-time (\gamma,\gamma) -bicriteria algorithm, with \gamma = \sqrtM\log n : The algorithm outputs a vocabulary of size at most \gamma K with number of tokens at most \gamma \rm OPT_K , where \rm OPT_K is the optimal number of tokens under the original budget K . We also study a storage-budget variant, in which the sum of the selected substrings’ lengths is bounded by B . For 2\le Bn , we obtain an (O(\log n\log B),O(\log n\log B)) -bicriteria approximation algorithm.
[AI-131] CoLaX: Context-Aware Local Explanations for Time Series Classification
链接: https://arxiv.org/abs/2611.15067
作者: Kexuan Zhang,Paolo Frazzetto,Xiaobei Zou,Gustau Camps-Valls,Yang Tang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Post-hoc explanations for time series classifiers usually score local patterns, such as subsequences or shapelets, without regard to the rest of the sequence. Yet the same motif can carry different predictive meaning under different global temporal conditions, such as trends or regimes. We propose CoLaX, a context-aware post-hoc explanation framework that builds on shapelet-driven Shapley attribution and conditions local-pattern relevance on a learned representation of the full sequence. Global context acts at two levels: it produces sample-specific prototype weights and a temporally resolved signal that reweights the local activation map. The resulting context-conditioned segments are the players of a Shapley attribution game whose values come only from the frozen black-box classifier. On synthetic benchmarks where identical motifs change relevance across contexts, CoLaX improves AUPRC over the strongest baseline by 30% on average. On real-world data, it achieves the highest AUPRC on ECG. Additional ablations indicate that the gains come from input-dependent conditioning rather than from additional learnable weights. CoLaX is therefore most useful when recurring local motifs change relevance across sequence-level contexts.
[AI-132] SkinAdvisor: An AI-Driven Decision Framework for Skin Production Planning via Multi-Modal Transformer Sales Forecasting AAAI2027
链接: https://arxiv.org/abs/2611.15023
作者: Ge Fan,Kai Wang,Chaoyun Zhang,Yingjie Li,Yuze Liu,Junyang Chen
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Accepted to AAAI 2027 (IAAI-27). 10 pages, 4 figures
Abstract:Skins, cosmetic in-game items, are a primary revenue stream for modern online games, yet production decisions (which skins to make, when to release, how to price) remain driven by intuition and retrospective analytics. We present SkinAdvisor, an AI-driven decision framework for skin production planning, powered by SkinCast, a deployed multi-modal Transformer forecasting 180-day per-day sales trajectories for unreleased skins. SkinCast couples four encoders: a pretrained Informer for player-preference time series, a pretrained Player-Hero-Skin graph embedder for relational priors, and two lightweight Transformers for skin attributes and release context, fused through a shared backbone with a local-plus-global day-x positional encoding and decoded by an N-layer Informer in a single non-autoregressive pass. SkinAdvisor plugs these forecasts into a serving pipeline (Select - Fill - Infer - Sales - ROI - Ranking) that ranks candidate skins by forecast-based ROI and supports multi-candidate comparison. The system has been in production for a full release cycle. Offline experiments show SkinCast reduces RMSE by ~20% over the strongest baseline (Informer) and ~60% over classical methods; stress tests on a 16-core CPU server confirm p99 latency under 8ms and peak throughput above 1100 rps. We report two deployment lessons: order-of-magnitude accuracy suffices for production decisions, and pretrained embeddings can leak temporal information, requiring season-based splitting to prevent.
[AI-133] EvoX: Rethinking the Agent Harness as a Runtime Policy
链接: https://arxiv.org/abs/2611.15006
作者: Haoyang Zhang,Qumeng Sun,Xiao Zhang,Xiang Liu,Zhidong Fan,Junjie Wang
类目: Artificial Intelligence (cs.AI); Software Engineering (cs.SE)
备注: 15 pages, 5 figures. Technical report
Abstract:Large language model (LLM) agents increasingly interact with tools and environments, yet plausible model decisions do not always translate into successful execution due to tool-call mismatches, interrupted workflows, and unsupported completion claims. We present EvoX, a Rust-based, multi-model agent runtime guided by The Second Policy, which views the harness as an Operational Policy complementing the foundation model’s Semantic Policy. EvoX combines model-facing guidance with deterministic runtime control through three complementary mechanisms: Action Realization, Execution Continuity, and Outcome Grounding. To evaluate harnesses without suppressing their native behaviors, we introduce Contrast, a controlled and auditable framework integrating explicit fairness constraints, provider-bound request verification, and evidence-based adjudication. Across five benchmarks, six foundation models, and six agent harnesses, EvoX achieves the highest numerical success rate of 56.2% over 691 common task–model pairs, exceeding the average of the other five harnesses by approximately 3.0 percentage points. Its larger advantages on GLM-5.3 (+10.7 points) and Kimi K3 (+7.9 points) highlight the model-dependent effectiveness of harness design and the potential for cross-provider adaptability. Achieved without memory, evolution, or extensions, these results establish EvoX as a competitive multi-model agent runtime and underscore the importance of harness-level operational control in realizing foundation-model capabilities.
[AI-134] A Cardiorespiratory Foundation Model Pretrained on Ten Physiological Signals for Perioperative and Critical Care
链接: https://arxiv.org/abs/2611.15004
作者: Cheol-Hui Lee,Wookyung Kim,Hyeonhoon Lee,Garam Lee,Da-in Eun,Hyung-Chul Lee
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Arterial, venous, pulmonary artery and intracranial pressures and ventilator waveforms guide decisions in operating rooms and intensive care units, yet foundation models for physiological waveforms have been built almost entirely on the electrocardiogram and photoplethysmogram. We present CARMEN, a single encoder pretrained without labels on ten cardiorespiratory waveforms (2.27 million modality-hours). CARMEN accepts any combination of co-recorded signals from 10 s to 5 min, and a conditioning mechanism, LSCNorm, restores the absolute signal level that normalization removes; without it, the AUROC for intraoperative hypotension fell from 0.894 to 0.780. Across 19 tasks from Korean and US cohorts, the frozen encoder outperformed supervised models on 12 of 15 and three foundation models, one general-purpose and two for biosignals, on 14 to 17 of 19, with none exceeding it on any task. Given the electrocardiogram or photoplethysmogram alone, it matched or exceeded signal-specific foundation models in most comparisons, and it retained its performance when channels were missing.
[AI-135] LLM -Enhanced Deep Reinforcement Learning with Human-in-the-Loop Calibration for Real-Time Optimal Power Flow
链接: https://arxiv.org/abs/2611.15003
作者: Jiayang Tu,Bozhen Jiang,Bojun Du,Yunxiao Ling,Ershun Du,Qin Wang
类目: ystems and Control (eess.SY); Artificial Intelligence (cs.AI)
备注:
Abstract:Real-time optimal power flow (RT-OPF) requires fast and reliable decision-making under time-varying operating conditions. However, many practical operating preferences and safety related requirements are difficult to encode using conventional numerical rewards alone, which limits the effectiveness and controllability of deep reinforcement learning (DRL) for RT-OPF. To address this issue, this paper proposes an LLM-enhanced DRL framework with human-in-the-loop (HITL) for RT-OPF. The proposed method augments the physics-based reward with sparse language evaluative feedback during offline training, so that semantic operational requirements can be incorporated into policy learning. To improve reliability in RT-OPF, a selective HITL calibration mechanism is further introduced to inspect and correct questionable LLM judgments before they continuously shape the learning process. Case studies on the IEEE 30-bus and IEEE 118-bus systems show that LLM feedback improves policy safety over vanilla TD3, while further HITL calibration achieves the best overall performance and more stable training behavior. These results demonstrate the effectiveness of combining language guidance and expert calibration for RT-OPF.
[AI-136] Same Task Different Statement: When and How Reformulation Helps LLM Problem Solving
链接: https://arxiv.org/abs/2611.14997
作者: Wentao Zhang,Xiaoyu Zhang,Chengwei Liu,Yutao Hu,Tao Zhang,Junsheng Wu,Hongping Gan,Yang Liu
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:When people face a hard problem, they often restate it in a clearer way before solving it. We study whether the same habit helps large language models by rewriting the statement of a problem. A legal rewrite must keep the meaning, the constraints, and the correct answer unchanged. We test three kinds of rewrites. Decomposition makes hidden subgoals explicit. Typed Reformulation casts the task into a more suitable solving frame. Concept Reduction removes surface detail and exposes the task structure. We evaluate them on eight math, logic, and code benchmarks with eight models, all under the same chain-of-thought solving prompt. The effect is large but depends on the problem type. An oracle that picks the best rewrite after the fact raises accuracy by +27.6% on average. Yet no single rewrite helps every problem, and a mismatched rewrite can hurt. Still, the effect follows clear patterns. Each kind of rewrite fits certain problem types. A simple fixed rule learned from one set of problems transfers part of the gain to new problems. It gains +11.0% on the problems it chooses to rewrite, while blind rewriting loses 5.6% on a similar share of problems. Rewriting is therefore not a universal way to solve problems. The question is when and how to rewrite.
[AI-137] Who Moved the Frontier? Paying Open Frontier Challenges with Free Identities
链接: https://arxiv.org/abs/2611.14974
作者: Luca Zanolini,Duc V. Le
类目: Computer Science and Game Theory (cs.GT); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:In an open frontier challenge, anyone, person or AI agent, submits changes to a shared problem, and a change that beats the best measured score becomes the public baseline. How should it pay when identities are free? Rank prizes can be gamed: on a replay of one track of a recent challenge, one solver’s steps under two identities would have taken first and second prize. We consider a prediction-market rule, proposed for machine-learning contests, that pays each accepted step for the progress it adds; with exact scores, among rules that see only a step’s start and end, only these resist splitting and merging, and extra identities gain nothing. With noisy scores and independent normal runs, we prove that any payment that never takes money back and pays real improvements also pays something, on average, to a change that does nothing. We propose a fee per submission, at least what luck pays per attempt, that removes each attempt’s expected profit if verifiers are honest and no submission is noisier than planned; nothing enforces that bound, and under noise the rule does not cap the total paid. If crediting another and faking credit to one’s second account look the same, a rule that neither rewards fakes nor punishes honesty pays for credit at most what lying costs. We introduce a royalty on how much a later gain falls without the earlier change, which needs no claims but needs removable changes. For five public challenges, we name the rule our results would support.
[AI-138] Early Signals Strong Decisions: Prefix-Guided Sampling for Parallel Test-Time Scaling
链接: https://arxiv.org/abs/2611.14965
作者: Jie Ren,Jonathan Rosenfeld,Neil Thompson
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Parallel test-time scaling (TTS) has been shown to enhance reasoning in large language models by generating and aggregating multiple independent reasoning paths. Internal confidence scores can further improve this process by identifying paths of varying quality. However, existing methods remain computationally inefficient, as low-quality paths are fully expanded before being evaluated. In this work, we aim to improve the use of internal confidence signals for better model performance and computational efficiency. We begin by offering a new perspective on the role of confidence in aggregation, showing that it is most impactful for questions near the decision boundary. Notably, when amplified by multi-sample aggregation, even small confidence signals derived from short prefixes can meaningfully influence outcomes in these boundary cases. Building on this insight, we introduce Prefix-Guided Sampling (PreG), a method that reallocates computation by first generating short prefixes, ranking them using internal confidence scores, and then completing only the most promising candidates. This strategy reduces token usage while maintaining gains in path quality. We provide theoretical analysis demonstrating that, under a fixed compute budget, PreG strengthens the effective signal by widening the decision boundary margin. Empirically, our method consistently outperforms standard parallel TTS approaches across benchmarks, while significantly improving token efficiency and reducing overall computation.
[AI-139] Binding Mode Matters: Hotspot-Aware Drug Discovery via Explorative Preferences NEURIPS2026
链接: https://arxiv.org/abs/2611.14956
作者: Dingshuo Chen,Hao Yang,Kuangqi Zhou,Zhixun Li,Qiang Liu,Xiangxiang Zeng,Shu Wu,Yaosen Min
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: To appear at NeurIPS 2026, 26 pages
Abstract:Discovering hit molecules requires not just high binding affinity, but also the identification of diverse binding modes, which are critical for experimental assays. Existing generative approaches have predominantly relied on optimizing a scalar docking score, obscuring the distinct contributions of key binding determinants. To this end, we introduce a paradigm shift by formulating target-based drug design as a multi-objective exploration task, where each objective explicitly corresponds to enhancing interactions with a specific hotspot. Here, we introduce BindMol, a novel generative framework driven by a customized multi-objective reinforcement learning algorithm. By incorporating explorative preferences during training, our approach efficiently uncovers molecules with diverse and desirable binding profiles. Empirical results demonstrate that BindMol facilitates the discovery of high-affinity compounds characterized by both structural novelty and diverse binding modes. Validated across target-based drug discovery and multi-property optimization tasks, our approach provides a versatile paradigm for goal-oriented drug discovery.
[AI-140] he Posterior-Prior Epistemic Gap: When World Model Imagination Fails to Carry Task-Relevant Signals
链接: https://arxiv.org/abs/2611.14955
作者: Jaerock Kwon,Elahe Delavari,Donghyun Kim,Haewoon Nam
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注:
Abstract:Model-based planners that roll out imagined futures assume the transition model faithfully propagates task-relevant signals—a property rarely measured directly. In a single-seed case study of a Recurrent State Space Model world model with active-inference planning in the CARLA driving simulator, we identify a \textbfposterior-prior gap—posterior (observation-filtered) states retain obstacle information that imagined rollout states progressively lose: a supervised obstacle probe achieves AUC 0.85 on posterior states, but the same probe applied to imagined prior states shows monotonic decay (AUC 0.85 \rightarrow 0.71 over 15 steps; logit separation drops 57%). Because the probe is identical, the decay isolates the transition model—not the representation—as the information bottleneck. Obstacle avoidance drops to 0% without privileged runtime coordinates, and the planner’s own ensemble-based epistemic term does not flag the loss—ensemble disagreement yields AUC 0.52 as an obstacle classifier (chance level), making the failure an unflagged imagination failure—an unknown unknown at planning time. As a secondary contribution, we show that \textbftask-conditioned weighting of Effective Free Energy components over the same world model yields lane keeping ( 81.2 \pm 1.8 % route completion) and obstacle avoidance (28/28 avoided, \emphusing privileged coordinates) in the CARLA simulator, with runtime weight modulation enabling post-evasion centering.
[AI-141] RSI-PDEbench: A Benchmark for Self-Improving Agents in PDE Solving
链接: https://arxiv.org/abs/2611.14949
作者: Binghang Lu,Lang Cao,Changhong Mou,Yige Sun,Cindy Xiangrui Kong,Zheyuan Deng,Tian Yuan,Bing Hu,Runyu Zhang,Na Lu,Guang Lin
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Identifying and solving partial differential equations (PDEs) from sampled solution data often requires substantial expert effort in equation discovery, coefficient estimation, and numerical solver configuration, especially under noise and distribution shifts. Can large language model (LLM) agents automate this workflow accurately and efficiently? To investigate this question, we introduce \bench, a benchmark of 1,001 cases spanning 42 equation templates, including both classical and fractional PDEs. It provides three evaluation settings: Full evaluates performance on noise-free solution data; Noisy tests robustness to Gaussian noise with an RMS equal to 50% of the clean-signal RMS; and Hard-OOD tests challenging out-of-distribution generalization to unseen equation templates and shifted coefficient ranges. We further introduce RSI-PDE, a recursively self-improving agent harness that shortlists candidate equations by residual loss, uses relative L^2 feedback to refine coefficients and numerical solvers, and applies a validation gate to policy and memory updates. We compare RSI-PDE with single-run baselines and native agent harnesses (Claude Code or Codex) across four frontier models. Across the three settings, RSI-PDE improves the pass rate averaged across the four models by 106.5–181.9% relative to single-run baselines and by 3.0–19.6% relative to native harnesses, while reducing total recorded model calls by 46.9% across the 12 model–setting pairs. These results highlight the potential of validated recursive self-improvement in LLM agents to advance scientific computing.
[AI-142] Provably Optimal HBT Legalization in Seconds: Certified Optima Reveal Where the Headroom Really Lies
链接: https://arxiv.org/abs/2611.14928
作者: Kadhir Parani,Manjunath Bezawada,Divya M,Kesana Jathin,Kolli Sarath,Sm Zobaed
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Hybrid bonding terminal (HBT) legalization assigns every terminal of a face-to-face 3D IC to its own pitch-grid site with minimum total Manhattan displacement. Open3DBench reports that a reinforcement-learning (RL) legalizer reduces displacement by 19-22% relative to a sliding-window (SW) heuristic, but none of its legalizers is publicly released. We ask how much displacement any legalizer can still recover. We build an open benchmark of 21 design/pitch instances from Open3DBench placement data, with reconstructed greedy, SW (ViaLegal-style) and BTAssign-style baselines, and GEO-Exact, an exact solver that runs successive shortest augmenting paths over lazily generated spatial candidate edges. An independent dual certificate proves its assignment optimal on all 21 instances (relative primal-dual gap at most 8.2 \times 10^-16 ). Against these optima, our SW reconstruction with ViaLegal’s default configuration is on average 0.011% above optimal (worst 0.119%), while on the 16 contended instances BTAssign-style, a density-guided heuristic (DGHL) and greedy are 3.0%, 6.9% and 64.5% above. Headroom does appear outside this regime: with 2x2 windows SW reaches 34.1%, and above about 58% utilization, on three of six swept designs, it reaches 21.1% on bp, 27.5% on swerv and 8.9% on jpeg (217 further certified instances, and ten bp points checked against an independent exact solver). If the published 19-22% gain were measured on instances like ours, SW would have to be at least 24-28% above optimal on average, far more than any SW variant we tested at nominal pitch; the RL paper’s own comparison with the Hungarian optimum puts its SW 3.05% above optimal on average. GEO-Exact is also about an order of magnitude faster than DGHL and SW in our harness (geometric means 8.1x and 40x; a compiled core against Python baselines on a shared server). Code, optimal values and certificates are released.
[AI-143] Edge-Native Multimodal RAG : Hardware-Aware Quantization and Architectural Optimization for Consumer GPUs
链接: https://arxiv.org/abs/2611.14924
作者: Zahraa Selim,Menna Hamed,Wesam Ahmed,Sohyla Said,Sara Basheer,Rami Zewail
类目: Artificial Intelligence (cs.AI)
备注: Accepted to the 14th International Japan-Africa Conference on Electronics, Communications and Computations (JAC-ECC 2026).6 pages, 3 figures, 6 tables
Abstract:Deployment of retrieval-augmented generation (RAG) for Edge-AI is constrained by memory and bandwidth limitations. A standard 7B parameter model consumes 14 GB Video Random Access Memory (VRAM) in half-precision (FP16), precluding co-location of vector indices and Key-Value (KV) caches. This paper presents an engineering integration and empirical evaluation of a multimodal RAG framework optimized for consumer-grade Turing hardware (e.g., NVIDIA Tesla T4). Our system-level analysis identifies a critical trade-off: while generative pre-trained transformer quantization (GPTQ) offers superior throughput (51.6 tok/s), it exhibits significant context instability at 2048+ tokens on Turing architectures. Conversely, NormalFloat 4-bit (NF4) maintains robust long-context performance with viable latency (81 ms/tok). By coupling NF4 with a decoupled Florence-2 vision integration, we achieve visual reasoning with only 430 MB VRAM overhead. Furthermore, empirical observations of prompt-induced search space pruning show that structured prompting reduces latency by 47% without altering computational intensity (1.88 GFLOPs/token). The integrated architecture maintains an 8.37 GB footprint, establishing a viable RAG pipeline for 12-16 GB GPUs. Finally, structured few-shot prompting effectively compensates for quantization-induced reasoning loss, restoring mathematical accuracy from 35.0% to 92.0% across 200 trials.
[AI-144] Rethinking “RL Generalizes SFT Memorizes”: The Role of SFT Data NEURIPS2026
链接: https://arxiv.org/abs/2611.14912
作者: Yunlong Hou,Fengzhuo Zhang,Yuan Cheng,Jiachun Pan,Xingyao Li,Zhuoran Yang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Accepted to NeurIPS 2026. 39 pages, 12 figures
Abstract:Large language models trained with Reinforcement Learning (RL) with verifiable rewards exhibit strong reasoning ability and broad generalization, whereas models trained with Supervised Fine-Tuning (SFT) are often viewed as more prone to memorization and limited transfer. This paper rethinks this distinction through the lens of SFT training data on widely used Qwen2.5 and Qwen3 series models. First, we study the role of data source and show that it is critical: a carefully mixed SFT dataset substantially outperforms data generated solely by a larger model. Second, we study the role of data scale and show that matching the number of correct rollouts between SFT and RL greatly improves SFT generalization, while matching the total rollout budget enables SFT to generalize as well as RL. Combining these two factors further enables SFT to generalize even better than RL. Third, by using LLM annotations to characterize the solution methods in training rollouts, we show that larger datasets cover more tail methods and that these tail methods provide generalizable reasoning signals. Finally, we support these empirical findings theoretically by analyzing the training dynamics of shallow transformers under both RL and SFT.
[AI-145] Can an AI Agent Be Epistemically Akratic? Higher-Order Evidence the Limits of Self-Correction
链接: https://arxiv.org/abs/2611.14889
作者: Eno Agolli
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:When we judge that our reasoning was defective, it seems plausible that we should abandon the conclusion it produced. This thought has special force for AI agents built to assess and correct their own work. Yet obeying such judgments can make an agent worse at distinguishing sound reasoning from defective reasoning. An AI agent may rely on information about its reasoning when deciding which conclusions to keep, while the part that judges whether those conclusions are justified cannot use that information. The agent can then exercise an epistemic competence by retaining conclusions it judges unjustified, even when its self-assessments are generally reliable. Nor is the difficulty resolved by training the agent to make its conclusions and assessments agree: agreement can be achieved by degrading either one. Crucially, I argue that a negative assessment can remain valuable by identifying conclusions requiring further scrutiny, without having authority to eliminate them. The result challenges a central assumption about AI self-correction: that bringing an agent’s beliefs into agreement with its own self-assessment about their justification makes it better at correcting its mistakes.
[AI-146] MLIRBench - A Benchmark and Dataset for Compiler and Hardware-Aware Reasoning
链接: https://arxiv.org/abs/2611.14881
作者: Pranava Sai Maganti,Meghasyam Peddireddy,Muzakker Hossain,Kaining Gao,Farhan Ahmed,Ali Jannesari
类目: Programming Languages (cs.PL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Accepted at the 2027 IEEE/ACM International Symposium on Code Generation and Optimization (CGO 2027)
Abstract:In recent years, large language models (LLMs) have become increasingly popular in software engineering and have performed well on different programming tasks. However, it is still difficult to know whether these models truly understand program behavior or whether they mainly rely on patterns learned from their training data. This distinction is particularly important for intermediate representations (IRs), where operations, dataflow, and execution semantics are explicitly represented. In this paper, we introduce MLIRBench, a benchmark for evaluating how LLMs reason about MLIR programs. MLIRBench uses a multi-agent pipeline to generate program mutations, validate their behavior through execution, and construct reasoning tasks from torch and linalg program pairs. The dataset contains 1,768 validated pairs, including 888 linalg pairs and 880 torch pairs, labeled as EQUIV, DIVERGE, or ERROR. We evaluated nine models, including five open-weight and four proprietary models. Gemini 3.1 Pro achieved the highest hard-label accuracy of 92.65% across all 1,768 pairs, counting unsuccessful predictions as incorrect. Across its 1,768 scored predictions, it achieved an aggregate score of 74.18 and a judge-assigned result-behavior score of 5.04/10. These results show that strong semantic classification does not necessarily imply accurate prediction of detailed execution behavior, highlighting a gap in current LLM reasoning over compiler IRs. By exposing this gap through execution-validated program pairs, MLIRBench provides a reproducible foundation for studying semantic reasoning over compiler IRs and related tasks such as program repair and semantic classification.
[AI-147] When Safety Becomes an Attack Surface: Manipulating Abstention in Tool-Using AI Agents
链接: https://arxiv.org/abs/2611.14864
作者: Charles Xiao
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:
Abstract:An AI agent’s decision not to act can itself become an attack surface. We study warning-induced over-abstention, an availability attack in which adversaries place fabricated risk warnings in low-trust content to prevent legitimate task completion. Our central premise is that safe abstention requires evaluating both the claimed risk and the authority of the evidence supporting it. We develop controlled counterfactual task variants that vary risk content, source provenance, and warning wording, and evaluate a Verify-Before-Abstain protocol that corroborates warnings against authoritative records or the underlying content. Across three model families and 51 injectable AgentAbstain task pairs, caution-first prompting amplifies the attack. For Gemini, induced over-abstention rises from 9.8% to 86.3% , despite unchanged clean-task completion. Matched wording comparisons further reveal sensitivity to warning language, with substantial variation across models and environments. Provenance-aware verification improves legitimate completion, while forgery and sensitive-content controls expose its dependence on authoritative evidence and its limitations in preserving confidentiality. These results motivate evaluating agent safety through the justification for restraint, alongside the safety of actions taken.
[AI-148] AutoAssumption: Searching over Assumptions Discovers ML-Systems Algorithms
链接: https://arxiv.org/abs/2611.14844
作者: Igor Sadalski,Minghao Guo,Benjamin Jones,Wojciech Matusik
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 52 pages (9 main + references and appendix), 14 figures, 13 tables
Abstract:Many advances in ML systems came from breaking an assumption that earlier methods took for granted: speculative decoding dropped the assumption that a model needs one forward pass per token. Yet algorithms rarely state their assumptions, so experts and discovery harnesses search over programs, where one edit changes several design choices and gains are difficult to attribute. Searching over the assumptions themselves remains largely unexplored. Writing algorithms as proofs makes their assumptions explicit and turns algorithm design into a search over a small, structured space that an LLM can explore and a proof checker can audit. We formalize 89 algorithms from KV-cache eviction, weight pruning and speculative decoding as proof graphs with explicit assumptions, checked in Lean. Our harness, AutoAssumption, relaxes one assumption of a published method, re-derives and re-proves it, and generates its code. In KV-cache eviction and weight pruning, this search improved on every published algorithm we ran and produced state-of-the-art algorithms. The eviction algorithm has the highest RULER-4K score of the published methods we ran, and the pruning algorithm the lowest perplexity of thirteen on four LLaMA models. At equal budget, the harness scores within a point of OpenEvolve, an open-source implementation of AlphaEvolve, with a lower standard deviation across repeats. Future work will formalize more fields and combine them through the principal assumptions they share.
[AI-149] OverServe: Resource Amplification Attacks on SaaS Agents via Tool–Prompt Co-Design
链接: https://arxiv.org/abs/2611.14808
作者: Ruiheng Yu,Oteo Mamo,Siyu Gan,Weikuan Yu,Mulong Luo
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Software Engineering (cs.SE)
备注:
Abstract:Advances in large language models (LLMs) are enabling software-as-a-service (SaaS) providers to automate longer business workflows with agents. In subscription-based SaaS deployments, model and tool calls consume provider resources without additional customer payments. Because the agent autonomously decides when to stop, a malicious tenant can send a prompt to the agent, exploiting its completion decisions to extend execution and waste provider resources. In this paper, we present OverServe, a multi-agent framework for constructing such attacks against SaaS agents. OverServe co-designs Model Context Protocol (MCP) status tools and completion-confirmation prompts for the attacker. Each tool call returns accurate status and terminates normally, yet the agent may repeatedly check an unfinished condition to satisfy the prompt. This consumes the SaaS provider’s resources even after the original business task is completed. OverServe constructs service-specific tool–prompt pairs and uses separate tool and prompt reviews to guide candidate pair selection. Tested tool specifications trigger no alerts in the evaluated MCP scanners. We evaluate Qwen, Nemotron, DeepSeek, and Sonnet on selected tasks from nine benchmarks. All OverServe runs pass the original task checks. OverServe can reach a maximal amplification of API requests by 93.88\times while preserving the original task goal and collateral checks.
[AI-150] GenPFN: An In-Context Tabular Foundation Model Approach for Augmenting Small Datasets in Engineering
链接: https://arxiv.org/abs/2611.14804
作者: Noah J. Bagazinski,Cyril Picard,Faez Ahmed
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Costly simulations and experiments limit the data available for engineering design. This paper proposes GenPFN, a training-free candidate generator guided by a pretrained tabular Prior-Data Fitted Network (TabPFN) classifier with fixed weights. The context contains evaluated designs labeled desired (Class 1) or undesired (Class 0). GenPFN interpolates between desired designs and refines candidates using gated classifier gradients and Gaussian noise. Each candidate is evaluated using the problem’s constraints and objective functions. Evaluated desired designs and their objective values augment the context for regression training. Evaluation spans fourteen selected problems, including a two-parameter benchmark and a 44-parameter hull problem. In paired comparisons, GenPFN with TabPFN v1 achieves higher class satisfaction and lower realism distance than the tested adversarial, variational-autoencoder and diffusion generators. The variational-autoencoder generator provides greater coverage and diversity. Relative to the synthetic minority over-sampling technique (SMOTE), GenPFN provides lower coverage distance and greater diversity. Among observed paired regression fits, augmentation reduces estimated AutoGluon validation mean squared error by 62.1% relative to context-only training. A hull-stern case study yields 99.3% propeller-compatible candidates versus 1.4% among randomly sampled context designs. Generation followed by evaluation produces sample sets that augment limited datasets and support training more accurate predictive models. Subjects: Artificial Intelligence (cs.AI) Cite as: arXiv:2611.14804 [cs.AI] (or arXiv:2611.14804v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2611.14804 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-151] Robust Task-Aware State Estimation from Pre-trained Perception NEURIPS
链接: https://arxiv.org/abs/2611.14781
作者: Guy Azran,Sarah Keren
类目: Artificial Intelligence (cs.AI)
备注: The Fortieth Annual Conference on Neural Information Processing Systems (NeurIPS), 2026
Abstract:Agents that perform state estimation using pretrained perception models, e.g., Vision-Language Models (VLMs), face the fundamental challenge that predictions are uncertain. Still, most approaches determinize these predictions into a single best-guess state and make decisions as if it were ground truth. This is brittle because when the guess is wrong, the agent’s decisions may be unsafe or infeasible. At the other extreme, maintaining a full probability distribution over symbolic states is often intractable. We propose a middle path, drawing on the underlying probabilities the perception model already produces (e.g., a VLM’s output-layer token distributions) to construct a \theta -confidence Minimal Belief Set (MBS _\theta ), a minimal subset of states whose cumulative probability meets a user-specified belief-mass threshold. The minimality of an MBS _\theta ensures that the agent considers only the most likely states, enhancing robustness in downstream decision-making while maintaining computational tractability. Using a novel three-layered belief system, we efficiently approximate individual state probabilities from per-feature predictions and task-level constraints. On this belief construction, we base our MBS _\theta construction, while maintaining only a lean factored representation of the belief. Our algorithm, Probability-Ranked Valid State Search (PRaVaSS), takes advantage of the unique structure of this belief system to construct an MBS _\theta with correctness guarantees and polynomial complexity in the unconstrained case. Using a VLM as the backbone of our state estimator, we validate our method’s theoretical properties, scalability with problem size, and improvement of downstream decision-making when used in a task-planning pipeline.
[AI-152] Mach2: Scientific Discovery Platform with Deterministic Execution Constraints
链接: https://arxiv.org/abs/2611.14777
作者: Vansh Sharma,Venkat Raman
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI); Applied Physics (physics.app-ph)
备注:
Abstract:Language-model agents can formulate workflows and invoke specialized tools, yet compliance with natural-language operating rules is not guaranteed by prompting alone. Mach2 addresses this limitation by coupling Eigen orchestration to deterministic, lint-like checks of tool requests, source changes, acceptance tests, and artifact contracts. Domain requirements are represented as executable predicates: mandatory failures prevent advancement of the checked operation, while diagnostic findings are retained for correction or review. User-specified test-driven development (TDD), CodeGrapher’s abstract-syntax-tree analysis, and history-guided evolutionary search provide complementary verification and optimization mechanisms. Mach2 also generates task-specific agent packages for execution through independently selected endpoints. The originating harness validates returned artifacts against declared interfaces and evidence requirements without assuming continuous supervision of the remote run. Model-internal-knowledge (MIK) dependencies are recorded separately from source-grounded and test-supported findings. This paper formalizes the rule-gated execution model and the agent-package lifecycle, and proposes measurable criteria for evaluating rule compliance, engineering validity, and cross-endpoint task continuity.
[AI-153] Leaderboard Hacking: Preference-Based Model Evaluations are Vulnerable to Manipulation
链接: https://arxiv.org/abs/2611.14769
作者: Eve Fleisig,Joachim Baumann,Dylan Hadfield-Menell,Dirk Hovy
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Arena-style rankings of language models are a widely used evaluation framework in online leaderboards, research papers, and model evaluation in production. These rankings aim to realistically assess model performance using pairwise comparisons of anonymous models on user prompts, aggregating votes via the Bradley-Terry model to produce the final ranking. However, we find that these leaderboards are highly vulnerable to several manipulation strategies that unscrupulous providers could use to boost a model’s rank: strategic voting (individual votes that boost a model’s rank, including votes on irrelevant models), strategic prompting (choosing prompts that favor a model), and strategic nomination (boosting the rank of a model by inserting irrelevant models). We demonstrate these effects on public leaderboard data, analyze the circumstances that cause these vulnerabilities, and describe approaches to safeguard against each attack. Notably, these issues hold for any preference-based evaluation using the Bradley-Terry model that allows open prompting, voting, or model submission. These manipulations boost the rank of nearly every tested model; each attack can boost a model by up to 15-20 places in a ranking and, combined, they can boost a model by up to 45 places.
[AI-154] SynthCorp: A Platform for Creating Benchmarks for Data Agents
链接: https://arxiv.org/abs/2611.14765
作者: Jiani Zhang,Manav Garg,Eugenie Lai,Cheng Zhen,Eva Sharma,Yang Pan,Yaming Lin,Alon Halevy
类目: Artificial Intelligence (cs.AI); Databases (cs.DB)
备注: SynthCorp v1.0. We will release the benchmarks soon
Abstract:Current benchmarks for LLM-driven data agents saturate quickly and cannot target specific capabilities. We present SynthCorp, a platform that creates customizable data estates and benchmarks by simulating how business events produce and modify corporate data over time. Each data estate combines relational databases, a document store, and a data catalog, paired with evaluation tasks and executable gold programs. First, given a scenario and target complexity dimensions, SynthCorp synthesizes reusable event types and compiles each into a program over predefined simulation primitives. Second, SynthCorp schedules event instances on a virtual calendar, interleaving routine and disruption events so that tables, documents, and catalog entries remain consistent. Third, SynthCorp executes the calendar to materialize the data estate and records every operation in a simulation log. Fourth, SynthCorp generates analytical tasks and derives verified gold programs from the log. We use SynthCorp to generate initial evaluation benchmarks comprising seven data estates (up to 1K operational tables) and 256 tasks across five domains. The data estates pass deterministic pipeline checks (such as foreign-key violations, undeclared writes, recurrence violations, and database execution failures) with a near-zero error rate, introduce cross-asset complexities, and cover all requested dimensions. Evaluating three models across four data-context settings shows that the benchmarks separate data contexts and model tiers. Auditing candidate queries against causal provenance in the simulation log verifies gold programs without manual annotation, including on multi-event composite tasks that expose reasoning failures in state-of-the-art agents. We hope the community uses SynthCorp to generate more challenging and useful data estates and benchmarks that foster the development of enterprise data agents.
[AI-155] Graph Label Alignment: A Diagnostic Atlas for Graph Classification
链接: https://arxiv.org/abs/2611.14760
作者: Neelam Akula,Varun Shiralkar,Murat Kantarcioglu,Baris Coskunuzer
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 40 pages
Abstract:Graph classification benchmarks vary widely in difficulty: some are solved by simple pooled baselines, while others remain challenging for GNNs and graph transformers. We take a data-centric view and ask whether benchmark success is driven by genuine structure–feature reasoning or by label signal already accessible from restricted marginal views. We introduce graph label alignment, a lightweight diagnostic that measures label accessibility through two supervised probes computed without training a graph encoder: structural label alignment (SLA), based on permutation-invariant topological and spectral summaries, and feature label alignment (FLA), based on pooled node features with edges removed. Across real graph classification benchmarks and multiple pooled backbones, the resulting SLA–FLA plane strongly organizes test accuracy, with low-alignment datasets forming a consistent regime where standard pooled pipelines struggle. To probe what lies beyond marginal accessibility, we introduce controlled interaction-dominated stress tests, including heterophilic ego-graphs and synthetic Twins tasks whose structural and feature marginals are matched across classes but whose role–feature couplings differ. These results show that high alignment explains much of easy benchmark performance, while low alignment reveals regimes where success requires preserving structure–feature correspondence beyond standard pooled representations.
[AI-156] From Nodes to Pixels: Topological and Structural Two-View Graph Imaging
链接: https://arxiv.org/abs/2611.14750
作者: Md Joshem Uddin,Soham Changani,Sai Karthik Navuluru,Cuneyt Gurcan Akcora,Baris Coskunuzer
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 22 pages
Abstract:Graph learning still lacks a broadly reusable input interface comparable to patches in vision or tokens in language. Unlike images or text, graphs are irregular, vary in size and density, and are invariant under many equivalent node labelings, making it difficult to standardize inputs across tasks and architectures. We introduce G2Image, a fixed-budget graph imaging framework that converts each graph into two complementary image-like views. TopoGrid provides a stable topological view by encoding an intrinsic bifiltration as a compact multipersistence surface grid, while GraphGrid provides a higher-bandwidth structural view by aggregating edge densities between groups defined by intrinsic node scores. Both views are permutation-invariant and come with resolution-controlled stability guarantees. Because the resulting representations are fixed-size tensors, they can be processed by lightweight 2D encoders and aligned through a supervised cross-view contrastive objective. Across graph classification and molecular prediction benchmarks, G2Image achieves strong average performance among the compared methods, improves over single-view and standard fusion variants, and extends to attribute-rich molecular graphs without changing the overall architecture. These results suggest that fixed-budget graph images offer a practical and reusable interface for graph representation learning.
[AI-157] When Should Semantic Priors Enter Partially-Identified Causal Discovery?
链接: https://arxiv.org/abs/2611.14736
作者: Muhammad Ahmed Mohsin,Saad Bin Ather,Valentin Six,Lancelot Da Costa,Tommaso Salvatori,Bernhard Scholkopf
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Observational causal discovery often leaves causal di- rections unresolved or unstable. Large language mod- els (LLMs) can provide semantic information to resolve these directions, but incorrect proposals can exclude the true causal graph and invalidate downstream inference. We study three separate decisions: where to allocate semantic queries, which proposals to admit, and which causal graphs to retain. Our framework distinguishes population non-identifiability from finite-sample insta- bility, evaluates semantic orientation against statistical and random controls, and examines calibrated admis- sion and error-tolerant DAG retention. Experiments across benchmark networks show that semantic infor- mation improves recovery over coin-flip orientations, but targeting unstable edges has no established ad- vantage over matched-random allocation. The tested admission certificate cannot support nontrivial admis- sion with the available calibration tasks. Moreover, downstream retention depends on reference adequacy: learned references contain the true DAG in only three of 80 planned tasks. These findings separate the value of semantic information from the reliability of its alloca- tion, admission, and retention, and identify calibration size and reference adequacy as key limitations.
[AI-158] Risk-Aware Joint Inference and Resource Control for Physical AI in 6G Networks
链接: https://arxiv.org/abs/2611.14696
作者: Amin Farajzadeh,Melike Erol-Kantarci
类目: Networking and Internet Architecture (cs.NI); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Signal Processing (eess.SP)
备注: Submitted to IEEE for possible publication
Abstract:Wireless edge inference enables Physical-AI devices to execute edge-inferred actions using transmitted sensory inputs. Since the risk of incorrect actions vary across observations and tasks, the associated uplink transmission of sensory inputs requires different levels of wireless reliability. For instance, applying a fixed signal-to-interference-plus-noise ratio (SINR) requirement can overprovision low-risk tasks while underprovisioning high-risk ones. Therefore, we propose risk-aware control for edge inference (RACE), which maps predicted action-failure probability and task criticality to adaptive SINR targets and combines them with deadline-derived SINR requirements. A slow-timescale large language model (LLM) proposes task requirements for deterministic validation, while a fast hybrid actor selects inference configurations and transmit power levels, complemented by analytical receive beamforming and finite power restoration. Simulations show 11% improvement in the average accuracy-latency-energy objective over fixed-high-SINR control, with near-zero constraint outage.
[AI-159] ogether Compiler: Continual Compilation for AI Inference
链接: https://arxiv.org/abs/2611.14681
作者: Reyna Abhyankar,Dylan Lim,Drew Wadsworth,Venkata Hasith Vattikuti,YuTong Julia Zhang,Jung Hwan Heo,Taeksang Kim,Kesavan Ramakrishnan,Evangelene Stanley,Xinyu Fang,Daniel Y. Fu,Simran Arora
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Today’s AI workloads are rapidly changing: new models are released more often, model architectures are increasingly diverse, and traffic distributions change frequently. The challenge we face at Together AI is continual optimization for these changing workloads. Compiling a deployment into an optimized forward pass is a challenge that is well-suited for agent hill-climbing. Existing systems struggle with continual compilation due to a tradeoff between measurement throughput and fidelity. Many systems use agents to optimize individual kernels from the inference forward pass using offline benchmarks, but timings are low-fidelity because of data-dependent behavior in primitives like mixture-of-experts (MoE). On the other hand, systems use agents to directly optimize end-to-end engines but suffer from slow engine launch times. To address these challenges, we build Together Compiler (TKCC), the first truly continual agent loop designed for AI inference optimization. To address the throughput-fidelity tradeoff, we propose: (1) a Tiered Persistent Engine that decouples engine launch overheads from forward pass computation; (2) a cost model to select the parallelism configurations that seed the optimization process. We also conduct extensive ablations on agent context management and present our insights on deploying TKCC in production. TKCC outperforms highly-optimized hand-written Megakernels on the Llama 1B workhorse workload by up to 1.7x and the best open source engines on the latest frontier models (Kimi K3, GLM 5.3, MiniMax M3.1 Flash, DeepSeek v4.1 Flash) across hardware platforms by up to 2.6x higher TPS on the InferenceX benchmark. TKCC also sets the state-of-the-art performance at batch size 1, with 1111 TPS on M3.1 Flash and 1205 TPS on DeepSeek v4.1 Flash. Finally, TKCC generalizes beyond frontier language model inference. TKCC serves trillions of tokens across our production fleet.
[AI-160] Where Should the Code Go? Locating Foundation-Model Errors in Bitemporal Inventory Simulation NEURIPS2026
链接: https://arxiv.org/abs/2611.14671
作者: Durga Sandeep Saluru,Lanchen Song,Aditya Kumaran
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 22 pages. Accepted as a poster at the NeurIPS 2026 Workshop on Foundation Models for Temporal Systems (FMTS). Code and data: this https URL
Abstract:Operational reports carry two clocks: when an event happened and when its report arrived; later reports can revise earlier ones. A foundation-model component that maintains the state supported by the reports visible at a cutoff, then rolls a supplied scenario forward, can be helped by exact code before it, after it or inside the call, or asked to reason harder. We ask which remedy removes which error, for which models, on an exactly solvable lost-sales inventory environment: 36,672 independently rescored requests over twelve configurations of closed and open-weight models, with a forced-answer-call interface as a control and protocols frozen before the calls. Weak models fail in separable ways. Canonical input removes report-selection error, exact continuation removes rollout error, neither repairs the other, and which error dominates differs by model; strong models are at ceiling on raw reports. Asking for more reasoning is no substitute and can depend on the interface: under a forced answer call high effort collapses GPT-5.4 mini and DeepSeek V4 Pro, which gain or hold under automatic choice; on its provider the forced call disables reasoning for DeepSeek V4 Flash; and at high effort gpt-oss exhausts its output budget. The contribution is a controlled component diagnostic, released with every request, and evidence that for some model and provider pairs the tool-calling interface confounds reasoning evaluation.
[AI-161] Outcome Over Learning? Evidence from Student Use of AI Code Assistants for Unit Test Writing
链接: https://arxiv.org/abs/2611.14664
作者: Daniel Hinbest,Jeremy S. Bradbury,Michael A. Miljanovic
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注: 8 pages
Abstract:Artificial intelligence (AI) code assistants are rapidly being adopted in programming education, yet little is known about how students naturally use these tools when given unrestricted access. In a user study investigating AI-assisted unit test writing, we observed that a majority of students adopted a one-shot interaction strategy, asking the AI to generate a complete unit test directly from the initial requirements. While this approach is capable of efficiently producing a solution, it shifts students’ focus from learning how to write effective unit tests to simply completing the task. Moreover, because students rarely engaged in critically evaluating, refining, or iterating on AI-generated output, they may be missing opportunities to develop the critical thinking skills that are essential for effective and responsible use of AI. In an effort to shift student focus from outcome to learning we also discuss how training, accountability and improved learning task design can better support a more learning-centric responsible use of AI.
[AI-162] hompson Sampling using Prior-fitted Diffusion Transformers
链接: https://arxiv.org/abs/2611.14656
作者: Sihwa Park,Jingsen Zhu,Vinamr Jain,Sheng-Yen Chou,Alexander Terenin
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
备注:
Abstract:Thompson sampling is natural randomized strategy for optimizing unknown functions in a black-box manner. In continuous global optimization, Thompson sampling has largely been restricted to Gaussian process priors, as these are essentially the only class for which computing the posterior distribution has been tractable in practice. In this work, we introduce \emphprior-fitted diffusion transformers, which are diffusion models that perform Thompson sampling in-context at inference time, after being pre-trained using the prior-fitted network paradigm. To apply these models, we develop a set of practical non-Gaussian priors for general black-box functions—including those that have characteristics that cannot be achieved under Gaussianity, such as priors over unimodal functions. We also quantify the tradeoff between model size and the number of pre-training iterations, and show that these models follow scaling laws akin to those observed in transformers defined over other modalities. Across a set of global optimization benchmarks, we show that prior-fitted diffusion transformers with non-Gaussian priors can achieve improved performance compared to traditional pipelines built on Gaussian processes.
[AI-163] Multiple-Choice Secretary Problem with Predictions
链接: https://arxiv.org/abs/2611.14651
作者: Jing Yuan,Qi Cai,Xin Gao,Shaojie Tang
类目: Data Structures and Algorithms (cs.DS); Artificial Intelligence (cs.AI)
备注:
Abstract:We study the multiple-choice secretary problem, in which candidates arrive in uniformly random order and the goal is to select at most k candidates to maximize their total value. We consider a setting where predictions of individual candidate values are available before any candidate arrives. Fujii and Yoshida’s Learned Kleinberg algorithm achieves competitive loss O(\min\log k/\sqrtk,\eps) , where k is the selection capacity and \eps is the maximum multiplicative prediction error. We remove the logarithmic factor through a sharper analysis of the same switching algorithm with a retuned threshold. The resulting competitive ratio is at least \max\0,1-\min\32/\sqrtk,4\eps\ . In particular, exact predictions ( \eps=0 ) achieve the offline optimum on every arrival order.
[AI-164] Evaluating AI Musical Creativity: Towards a Quantitative Multi-Dimensional Framework Based on Value Novelty and Surprise
链接: https://arxiv.org/abs/2611.14634
作者: Hindy Rossignol,Giorgio Franceschelli,Mirco Musolesi
类目: ound (cs.SD); Artificial Intelligence (cs.AI); Computers and Society (cs.CY)
备注:
Abstract:Despite the increasing sophistication of generative Artificial Intelligence (AI), research on the evaluation of AI-generated musical creativity remains limited. To address this gap, this paper proposes a quantitative framework for the formalization and evaluation of AI musical creativity. Building on Margaret Boden’s theory, which conceptualizes creativity in terms of value, novelty, and surprise, the framework adapts and operationalizes these three dimensions for the musical domain. On the hypothesis that certain aspects of musical creativity can be assessed with a degree of objectivity, a domain-specific set of quantitative metrics for evaluating the creativity of AI-generated music is introduced. Rather than reducing musical creativity to a single unifying measure, this article proposes an ensemble of complementary metrics that capture its different dimensions, recognizing the complexity of musical artifacts and the difficulty of defining meaningful multi-objective compositional criteria. The framework has two implications. First, it gives researchers and developers a principled methodology and quantitative tools for empirically evaluating creative outputs, enabling more informed model development and reducing unnecessary computational costs. Second, it lays the groundwork for future generative models to optimize directly for creativity by adopting these metrics as learning objectives.
[AI-165] KernelMosaic: An Open Dataset and Agent ic Framework for CUDA Kernel Generation
链接: https://arxiv.org/abs/2611.14604
作者: Ningyuan Xi,Yuhang Lai,Ning Miao
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Training large language models to generate efficient CUDA kernels requires supervision spanning diverse computational requirements and iterative code optimization. Constructing this supervision is challenging because of heterogeneous specifications, and useful implementations emerge through iterative complicated testing and refinement. We introduce KernelMosaic, an open dataset pairing diverse CUDA tasks with verified solutions and trajectories. Human-guided review and diversity-aware selection curate tasks expressed as natural-language descriptions, reference programs, mathematical specifications, and their intersections. To build the solutions and trajectories, our agentic framework guides a strong LLM through implementation, correctness repair, and performance optimization with reusable CUDA guidance and execution feedback. Supervised fine-tuning of a small open-source model on KernelMosaic raises correctness from 4%, 0%, and 0% to 88%, 82%, and 16% on KernelBench L1, L2, and L3, respectively, while increasing the fraction of tasks with kernels faster than PyTorch eager at every level. Further training experiments show that supervision from the curated trajectories substantially outperforms direct responses and promotes sustained testing and revision. These results demonstrate that KernelMosaic enables a general-purpose model to achieve substantial gains in CUDA kernel generation.
[AI-166] Multi-Step Underwater Acoustic Channel Prediction via Dual-Domain Structure Preservation and Pretrained Transformer
链接: https://arxiv.org/abs/2611.14592
作者: Mingjie Cao,Gewei Zhang,Deqing Wang,Lizhao You,Fei-Yun Wu,Liqun Fu
类目: ound (cs.SD); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Signal Processing (eess.SP)
备注:
Abstract:Accurate future channel state information (CSI) is essential for underwater acoustic (UWA) communication because delays associated with signal propagation, feedback, scheduling, and transceiver reconfiguration can render currently available CSI outdated. However, most existing predictors estimate only the next channel state and obtain multi-step forecasts through recursive rollout, which causes prediction errors to accumulate as the multipath structure evolves. The scarcity of sea-trial measurements further limits reliable model training. To address these challenges, we propose LLM4UACP, a dual-domain, structure-preserving framework that adapts a pretrained Transformer for few-shot multi-step UWA channel prediction. LLM4UACP extracts complementary features from the channel impulse response (CIR) and channel frequency response (CFR), aligns them through Fourier-consistent fusion, and encodes the fused representation as delay-time tokens. A pretrained Transformer with a frozen backbone is adapted through Butterfly Orthogonal Fine-tuning to jointly predict multiple future CIRs in a single forward pass, thereby avoiding the error accumulation associated with recursive prediction. Experiments on the measured SO-06 and SO-07 channels show that LLM4UACP achieves normalized mean squared error (NMSE) values of 0.0411 and 0.0399, corresponding to reductions of 88.53% and 85.15%, respectively, relative to the most recent baseline. With only 0.11M trainable parameters, LLM4UACP maintains robust performance across varying prediction horizons and observation-window lengths, limited-data settings, and cross-dataset transfer scenarios.
[AI-167] CAFIN: CUDA-accelerated Adaptive Fisher Information Matrix Network Optimization via GPU-Native Kernel Fusion
链接: https://arxiv.org/abs/2611.14589
作者: Maryam Arshad,Muhammad Khuram Shahzad
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Second-order optimizers improve deep neural network training using curvature information, but their practical efficiency is limited by GPU execution overhead. AdaFisher reduces Fisher Information Matrix (FIM) approximation costs through diagonal Kronecker structures and exponential moving averages, yet its PyTorch implementation requires sequential CUDA kernel launches, intermediate tensor allocations, and repeated memory accesses. We introduce CAFIN, a GPU-native CUDA kernel fusion framework that improves AdaFisher execution efficiency without changing its optimization formulation. The proposed approach progressively integrates FIM preconditioner construction, exponential moving average updates, gradient preconditioning, momentum accumulation, weight decay, and parameter updates into fused CUDA kernels. Shared-memory reuse, register-resident computation, and in-place operations further reduce memory traffic and intermediate allocations. Experiments using ResNet-18 on CIFAR-10, CIFAR-100, and Tiny ImageNet across NVIDIA GTX 1660 Ti and RTX 3060 Laptop GPUs demonstrate up to 78.4% lower isolated FIM construction latency and 2.1-26.9% lower optimizer-step latency. The implementation also eliminates approximately 18.91 MB of intermediate GPU memory allocations per optimizer step. Numerical validation confirms consistency within the predefined FP32 tolerance, while convergence accuracy remains comparable across evaluated datasets and random seeds. These findings demonstrate that CUDA kernel fusion can substantially accelerate second-order optimization while preserving AdaFisher’s underlying mathematical formulation and learning behavior.
[AI-168] When Evaluators of AI Agents Disagree: An Empirical Study of Native Scoring LLM Judges and Commit-Boundary Checks on tau2-bench
链接: https://arxiv.org/abs/2611.14581
作者: Sungbin Im
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR)
备注: 38 pages, 3 figures, 11 tables. Submitted to Empirical Software Engineering. Replication package: this https URL . Supplementary material (Online Resource 1) is included as an ancillary file
Abstract:Agentic workflows, in which a large language model (LLM) calls tools to read and write data, are evaluated in two ways. Benchmarks compare the final state against a gold answer. Deployed services have none; an LLM judge reads the transcript and decides. Practice treats the two as interchangeable. This paper measures whether they are. We ran 2,160 episodes in tau2-bench’s airline and retail domains and recorded every database write. We scored each episode four ways: native scoring against gold state, an LLM judge, that judge given the policy document, and the boundary evaluator, a minimal rule-based check of whether a write was expected against whether one occurred in the trace. Two annotators independently labeled 120 episodes for task correctness, policy compliance, and commitment correctness. The judge repeated its verdict 98.2% of the time but agreed with native scoring on 56.2%. In 11 episodes the agent executed a forbidden write. The judge passed 10 to 11, with or without the policy or transcript. A stronger judge still passed 6. The boundary evaluator, with no model call, caught all 11; no judge significantly exceeded it on human labels. Forbidden writes occurred in 11.1% of episodes without gates. Verification, correction, and safe-termination gates did not reduce them significantly. The traces show a candidate cause: the gates’ evidence digest dropped the order status and customer authentication, so valid requests were rejected. The choice of evaluator decides the conclusion; the minimal check exposes failures the judges miss. Traces, code, prompts, and labels are released. Comments: 38 pages, 3 figures, 11 tables. Submitted to Empirical Software Engineering. Replication package: this https URL. Supplementary material (Online Resource 1) is included as an ancillary file Subjects: Software Engineering (cs.SE); Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR) Cite as: arXiv:2611.14581 [cs.SE] (or arXiv:2611.14581v1 [cs.SE] for this version) https://doi.org/10.48550/arXiv.2611.14581 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-169] MechFoundry: An agent ic loop that poses and performs engineering analyses and optimization of real-world objects
链接: https://arxiv.org/abs/2611.14571
作者: Xue-Ling Luo,Marco Conte,Peter Yichen Chen,Nikolaos Bouklas
类目: Computational Engineering, Finance, and Science (cs.CE); Artificial Intelligence (cs.AI)
备注:
Abstract:AI agents can automate software engineering because a compiler or a test suite checks that the code does what the user or the agent intended. In engineering analysis, however, the steps that precede the computation have few easily verifiable checks: modeling a design or an object, deciding how it is used, where supports and loads are applied, and what material each part is made of. A human expert therefore makes these decisions, carries out the steps, and judges the result, one object at a time. We present MechFoundry, an agentic loop coupled with bespoke software that automates structural analyses of real-world objects without a human in the loop. From an object’s lexical description, it generates a shape, poses a loading case relevant to how the object is used, assigns a material to each part, solves the nonlinear structural response, and improves the design against the loads it carries. A differentiable GPU solver built on an adaptive octree that need not conform to the surface enables the transition from shape generation to simulation and optimization. Across 50 everyday and engineering objects, the agent posed and solved a loading case in 6.5 minutes and for US 1.3 per object on average. Without a test suite to check against, we had language-model reviewers grade each session, as a first indication of where the agent’s engineering judgment holds and where it fails.
[AI-170] SideTaskBench: Measuring Unauthorized Side-Task Delivery in Coding Agents
链接: https://arxiv.org/abs/2611.14548
作者: Xiangyu Wen,Qiang Xu
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:
Abstract:Can a coding agent fix a user’s repository while also completing and delivering work the user never requested? In this paper, we introduce SideTaskBench to measure this failure where third-party technical references induce agents to produce independent side-task artifacts during an authorized repair. Repair success alone misses this behavior. Our key measurement contribution is to pair each repair with a self-contained side task and verify side tasks’ delivery success rate along with the repair success rate. The benchmark contains 25 Python repairs and 190 side-task units. Its four-level similarity index defines 300 executable repair–side-task pairs. The experiments show roughly 40% valid delivery under wrapped requests for two models, alongside repair success close to their clean baselines. The completed v10 similarity run uses DeepSeek-V4-Flash on all 300 pairs, with eight repetitions per pair (2,400 runs). Conditional delivery ranges from 20.48% for L0 unrelated tasks to 57.44% for L3 mechanism-overlapping tasks, while the L1-L0 difference is 5.37 percentage points and remains sensitive to task composition. The central insight is that technical relevance may blur the authorization boundary for an agent, such that an agent can solve the requested problem and still deliver unauthorized work. SideTaskBench emphasizes this potential weakness of existing agents and provides a testbed for defenses that preserve repair capability while preventing unauthorized delivery. The code is available at: this https URL. Subjects: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI) Cite as: arXiv:2611.14548 [cs.CR] (or arXiv:2611.14548v1 [cs.CR] for this version) https://doi.org/10.48550/arXiv.2611.14548 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-171] metalSPARSE: A 2:4 Semi-Structured Sparse GEMV Kernel for Apple Silicon GPUs
链接: https://arxiv.org/abs/2611.14545
作者: Tek Raj Awasthi(1),Babita Awasthi((1) Friedrich-Alexander-Universität Erlangen-Nürnberg)
类目: Computational Engineering, Finance, and Science (cs.CE); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Autoregressive Large Language Models (LLMs) generate text one token at a time, and each token reads every weight from memory, so on edge devices, decoding is limited by memory bandwidth. 2:4 semi-structured sparsity keeps two of every four weights, but Apple-silicon GPUs have no sparse matrix hardware, so pruning helps only if a kernel actually reads fewer bytes. We present metalSPARSE, to our knowledge the first 2:4 sparse kernels for Apple GPUs, together with a fused decoder written in Metal. The kernels stream weights in compressed form and expand them in registers. On Qwen2.5-1.5B with 2:4 sparsity on every decoder linear layer, decoding runs 3.6x faster than HuggingFace Transformers on PyTorch-MPS on an 8 GB M2 MacBook Air and 1.52x faster than MLX on a 16 GB M4. The same engine on dense weights matches MLX, so the gain comes from the compressed weights, and sparsity delivers 97% of the speedup its byte reduction allows. Prompt processing, which is compute-bound, does not speed up. To preserve quality, we combine a Wanda mask with block-wise reconstruction and low-rank adapters trained jointly with the surviving weights and quantized to 4 bits. On WikiText-2 this brings perplexity from 3.01x dense after one-shot pruning (Wanda with a closed-form refit) to 1.34x, and the whole pipeline runs on the same laptop/hardware.
[AI-172] Zugwise: Domain-Aware Game Reconstruction for Digitizing Handwritten Chess Scoresheets
链接: https://arxiv.org/abs/2611.14544
作者: Gerhard Trippen(University of Toronto)
类目: Artificial Intelligence (cs.AI); Image and Video Processing (eess.IV)
备注: 54 pages, 3 figures. Code and models: this https URL
Abstract:Digitizing handwritten chess scoresheets has been studied as a handwriting-recognition problem: read the move in each box and check that it is legal. This framing misses the errors that matter most. A misread that yields an illegal move is loud. One that yields a legal-but-wrong move is silent, because a legality-only pipeline accepts it and reads every later move against a wrong position. We therefore treat the task as game reconstruction and present Zugwise, which repurposes computer-chess machinery (quiescence search, static exchange evaluation, pin detection) to ask not “what should be played?” but “could a human plausibly have written this?” An absurdity detector makes silent errors loud at or near where they occur, and a backward search traces late symptoms to their cause. Candidate ranking excludes chess-quality signals. A Needleman-Wunsch merge adds the second player’s sheet as evidence, and review stops send every disagreement between the sheets’ legal readings to a human. In oracle replay of two tournament sections, a release adding them cuts silent errors from 16 to 1 and from 50 to 13. Two learned signals, CTC re-scoring against the recognizer’s raw output and a Maia-style human-move prior, raise the correct fix’s top-rank rate from 58.8% to 78.0% of the replacement decisions where it is a candidate, and hold without refitting on a later section. On the first section, the human-verified output matches an independent expert transcription move for move on 34 of 45 games, and adjudication against the sheets found errors in both records, on different plies. We do not establish that either workflow is faster or more accurate. The system runs entirely in the browser and is free, private, and offline-capable; alongside it we offer a three-way adjudication protocol for a domain where no single record is ground truth.
[AI-173] EasyFoam: Skill-Oriented OpenFOAM Agents with Evidence-Driven Evolution
链接: https://arxiv.org/abs/2611.14525
作者: Peng Yin,Zhiwei Chen,Tianyi Liu,Chen Li,Kai Li,Jian Cheng
类目: Computational Engineering, Finance, and Science (cs.CE); Artificial Intelligence (cs.AI)
备注: 26 pages, 3 figures
Abstract:Large language model (LLM) agents are increasingly capable of automating end-to-end OpenFOAM workflows for computational fluid dynamics (CFD). However, current progress is bottlenecked by two critical limitations. First, existing evaluations are dominated by tutorial-level cases and execution-oriented metrics, masking a substantial executability–acceptability gap: successfully executed simulations often fail to satisfy prescribed numerical or quantity-of-interest (QoI) criteria. Second, domain expertise in existing systems is tightly coupled with bespoke agent architectures, limiting its reuse and continual refinement across agent implementations. To address these challenges, we introduce ExtFoamBench, a challenging benchmark for OpenFOAM agents that spans five axes of workflow variation, and separately assesses execution completion, numerical soundness, and QoI satisfaction. Complementing this, we propose EasyFoam, an agent framework that decouples CFD expertise into modular, externalized skills driven by a dual-loop mechanism. An execution loop constructs workflows from explicit task requirements and uses structured simulation evidence to diagnose failures and guide targeted repairs. Meanwhile, an evolution loop distills task experience into reusable skill updates selected through paired evaluation. Empirical evaluations demonstrate that EasyFoam achieves substantially higher task acceptance rates than baseline agents, establishing this work as a starting point for future research toward self-improving CFD agents.
[AI-174] Care Human Enfeeblement and the Existential Implications of AGI
链接: https://arxiv.org/abs/2611.14520
作者: B. Scot Rousse
类目: Computers and Society (cs.CY); Artificial Intelligence (cs.AI)
备注: 8 pages
Abstract:Stuart Russell has named “human enfeeblement” as a long-term societal risk of advanced AI. This paper gives a new interpretation of the enfeeblement danger by interpreting it in light of recent work by Brian Christian, Alison Gopnik, and David Spivak on care and caregiving, and connecting it with the ontological design tradition of Terry Winograd and Fernando Flores. The enfeeblement that threatens is not merely a matter of cognitive off-loading and loss of knowledge, but a form of disengagement with the world which I explain in terms of an erosion of our capacities to care. This paper articulates a conception of care as a constellation of cultivable skills that involve us in the world and with each other, while making life worth living. Designing AI to support our capacities to care addresses the worry recently raised by Demis Hassabis and others regarding the questions of “meaning and purpose” that are looming in a post-AGI world.
[AI-175] LoRA Fine-Tuned Models for Control Systems Course QA: A Multidimensional Evaluation of Model Scale and Rank Effects
链接: https://arxiv.org/abs/2609.13918
作者: Shaowen Lu,Chengxu Liu,Ping Zhou,Tao Yang
类目: Artificial Intelligence (cs.AI); Neural and Evolutionary Computing (cs.NE); Systems and Control (eess.SY)
备注:
Abstract:Large language models (LLMs) are increasingly used in specialized university courses, but control-systems questions require coordinated terminology, notation, derivations, and stepwise explanations. Direct general-purpose responses may be inconsistently structured and hard to verify. Using exercises and reference solutions from a Linear Control Systems course, we built a supervised fine-tuning dataset of 360 system-user-assistant conversations. We applied LoRA to Qwen2.5-3B-Instruct and Qwen2.5-7B-Instruct. With identical data splits, inference settings, and evaluation protocols, we compared base and fine-tuned models and tested LoRA ranks r=4, 8, and 16. Evaluation used ROUGE, BERTScore, and structured-output features to measure reference-answer similarity and stability of the Solution-Method-Teaching Points format. LoRA improved both similarity and structured-output stability at both sizes. On the current test set, 7B-r16 achieved the highest ROUGE-L (0.4093) and BERTScore-F1 (0.8643), while r=8 offered a better balance between performance and parameter efficiency. Bootstrap resampling showed ROUGE-L gains of 0.0764 [0.0613, 0.0915] for 3B-r16 and 0.0874 [0.0687, 0.1042] for 7B-r16; both intervals exceeded zero, indicating stable textual-similarity improvements on the current test set. These results suggest LoRA can align open-source instruction-tuned models more closely with the language and pedagogical organization of course reference answers. However, the metrics mainly capture textual similarity and formatting consistency, not domain-specific reasoning or mathematical correctness, which require expert assessment and task-specific rubrics.
[AI-176] Human Grounded Evaluation of Large Language Models for Optical Network Automation
链接: https://arxiv.org/abs/2607.18068
作者: Kiarash Rezaei,Omran Ayoub,Paolo Monti,Carlos Natalino
类目: Networking and Internet Architecture (cs.NI); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Large language models (LLMs) are increasingly adopted for network automation, yet their output quality and inference cost can vary substantially across LLM families. We present HuGLEN, a stepwise evaluation pipeline that uses an LLM-as-a-judge together with a small set of expert ratings to enable scalable and reproducible comparison of candidate LLMs, and to rank them using a quality efficiency score (QES). We demonstrate HuGLEN for translating outputs from an explainable artificial intelligence (XAI) model for the optical network quality of transmission (QoT) estimation task into operator-friendly explanations. Our results show that a medium-sized LLM (12B parameters) achieves the highest QES, indicating the best trade-off between explanation quality and efficiency. Overall, HuGLEN reduces the human-labeling burden while supporting consistent model selection for operator-facing automation tasks.
[AI-177] he Impact of Large Language Models (LLM s) on Code Review Process
链接: https://arxiv.org/abs/2508.11034
作者: Antonio Collante,Samuel Abedu,SayedHassan Khatoonabadi,Ahmad Abdellatif,Ebube Alor,Emad Shihab
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Large language models (LLMs) have recently gained prominence in the field of software development, significantly boosting productivity and simplifying teamwork. Although prior studies have examined task-specific applications, the phase-specific effects of LLM assistance on the efficiency of code review processes remain underexplored. This research investigates the effect of GPT on GitHub pull request (PR) workflows, with a focus on reducing resolution time, optimizing phase-specific performance, and assisting developers. We curated a dataset of 25,473 PRs from 9,254 GitHub projects and identified GPT-assisted PRs using a semi-automated heuristic approach that combines keyword-based detection, regular expression filtering, and manual verification until achieving 95% labeling accuracy. We then applied statistical modeling, including multiple linear regression and Mann-Whitney U test, to evaluate differences between GPT-assisted and non-assisted PRs, both at the overall resolution level and across distinct review phases. Our research has revealed that early adoption of GPT can substantially boost the effectiveness of the PR process, leading to considerable time savings at various stages. Our findings suggest that GPT-assisted PRs reduced median resolution time by more than 60% (9 hours compared to 23 hours for non-assisted PRs). We discovered that utilizing GPT can reduce the review time by 33% and the waiting time before acceptance by 87%. Analyzing a sample dataset of 300 GPT-assisted PRs, we discovered that developers predominantly use GPT for code optimization (60%), bug fixing (26%), and documentation updates (12%). This research sheds light on the impact of the GPT model on the code review process, offering actionable insights for software teams seeking to enhance workflows and promote seamless collaboration.
[AI-178] When Both Sides Adapt: The Challenges of Evaluating Interactive LLM Systems
链接: https://arxiv.org/abs/2611.17239
作者: Hannah Li
类目: Methodology (stat.ME); Artificial Intelligence (cs.AI)
备注:
Abstract:In interactive systems where users engage with LLMs, users are adaptive and, after observing LLM responses, can change their prompting strategy, as well as the type and difficulty of subsequent tasks. Thus model responses can influence the set of prompts on which the model is subsequently evaluated. We study the implications of this user adaptivity and prompt endogeneity for the evaluation of LLMs. We define two classes of estimands for comparing models: the common-distribution effect, which compares model performance on the same distribution of prompts, and the induced-distribution effect, which compares each model on the prompt distribution generated through user interaction. We show that offline benchmark evaluation and live A/B testing target different quantities, with the standard difference-in-means estimators for the two methods converging to the common-distribution effect and the induced-distribution effect, respectively. This distinction can affect decision-making: the two comparisons can rank models differently and lead to discrepancies in deployment recommendations. We characterize when such ranking reversals occur and provide a sufficient condition under which the rankings are guaranteed to agree. This result provides a practical heuristic for assessing whether offline evaluation is sufficient to reproduce the decision suggested by a live experiment. Third, we show, in a stylized linear model of interactions, that user adaptivity can increase variance in estimation. Finally, we discuss implications of user adaptivity for the interpretation of experiments and show that adaptivity creates a common mechanism leading to common evaluation challenges, such as reward hacking and novelty effects. Together, this work highlights the necessity of measuring both model performance as well as changes in the user behavior that the model induces.
[AI-179] AI for Biologists: The Knowledge Boundaries of Foundation Models in the Age of Agent ic Systems
链接: https://arxiv.org/abs/2611.17015
作者: Tian Cai,Andrew Kleist,Jaimin Patel,Duccio Malinverni,Alexander K. Taylor,Malvika Sharan,Renato Umeton,M. Madan Babu
类目: Biomolecules (q-bio.BM); Artificial Intelligence (cs.AI)
备注: To be published on Annual Review Biomedical Data Science
Abstract:Artificial intelligence (AI) is changing how biological research is done. Biological foundation models trained on millions of nucleotides, protein, and single-cell measurements predict protein structures, mutation effects, and cellular responses to genetic perturbations. Agentic systems, AI assistants built on large language models, can run multistep analyses from plain-language requests and increasingly call foundation models as tools. Yet it is often unclear when these models’ outputs can be trusted. We frame this as a model’s knowledge boundary: the range of inputs, tasks, and conditions over which its outputs are reliable. This review explains how foundation models are built and proposes that five design elements (data, architecture, training scheme, evaluation scheme, and post-analysis) define this boundary. We examine evidence on where current models succeed or fail and present standardized descriptions of model limits so that researchers, and the agents acting for them, can use these tools reliably.
[AI-180] Kezek: Splitting Mono Two-Person Conversations into Per-Speaker Tracks for Full-Duplex Dialogue Data
链接: https://arxiv.org/abs/2611.16933
作者: Bagzhan Karl,Alizhan Karl
类目: Audio and Speech Processing (eess.AS); Artificial Intelligence (cs.AI)
备注: 18 pages, 15 tables, 2 figures. Code: this https URL ; weights: this https URL ; listening examples: this https URL
Abstract:Full-duplex spoken dialogue models learn turn-taking, overlaps and backchannels from conversations recorded with one channel per speaker, but most available conversational audio is a single mono mix. We present Kezek, a model that turns a mono recording of exactly two people into one track per person with the original length and timing; overlaps and short replies (“yeah”, “mm-hm”) stay on the track of the person who said them. Kezek decides who speaks when and who is who once for the whole recording (stage 1: a two-voice activity model guided by a speaker fingerprint trained to be insensitive to phase), and a conditioned time-domain separator only carries out that decision (stage 2), with outputs that always sum to the input. Standard separation metrics miss the failures that matter for dialogue data, so we report short-reply retention, leakage onto the other person’s solo, channel swaps and coverage. Against the open pipeline pyannote - DialogueSidon on the same recordings, Kezek keeps more short replies (0.83 vs. 0.69 on Open Yap test; 0.80 vs. 0.71 over the other’s speech on unseen telephone speech from CallHome English), mostly because it returns the whole recording while the external pipeline drops part of it (on the audio both return, reply retention is about the same: 0.859 vs. 0.865 on Open Yap); it also leaks less, swaps channels less often, and runs 4-5 times faster (about 230 hours of audio per GPU-hour). We also report negative results and an ideal-mask ceiling that locate the remaining gap: quiet replies under a louder voice. Code and weights are released under permissive licences.
[AI-181] Protein Language Models Fail To Capture Contact Predictions Under Invariant Relabeling NEURIPS2026
链接: https://arxiv.org/abs/2611.16881
作者: Paarth Batra,Aryan Keluskar
类目: Quantitative Methods (q-bio.QM); Artificial Intelligence (cs.AI)
备注: Accepted to the ICBINB - Failure Modes of AI in Biology Workshop at NeurIPS 2026. 15 pages, 6 figures
Abstract:Protein language models (pLMs) produce highly accurate contact maps from single-sequence inputs. People often take this accuracy as proof that these models implicitly learn invariant coevolutionary statistics, extracting the same signals as classical alignment-based methods like Potts. In this work, we introduce a diagnostic framework around a global bijective relabeling of the amino-acid alphabet. This transformation permutes residue letters while leaving all alignment-level information, column entropies, and pairwise edit distances unchanged. Pseudolikelihood Potts models and inverse-covariance estimators remain invariant under this transformation, but pLMs fail the test. Across multiple architectures (ESM-2, MSA Transformer, AMPLIFY) and model scales (8M to 3B parameters), pLM contact maps collapse under arbitrary relabeling. Using Hamming distance controls and biochemical grouping interventions, we attempt to isolate what likely causes this degradation and find that these models rely on explicit, canonical residue identities (``alphabet anchoring’') rather than pure coevolutionary statistics. Increasing scale makes models more responsive to biochemical classes, but their performance still degrades under relabeling. Our findings show a gap between how pLMs and classical coevolutionary models compute contacts. More broadly, our work shows how diagnostic perturbations that leave classical models provably unchanged can expose hidden representation biases and prevent over-reliance on neural black boxes in biological modeling.
[AI-182] PISim: A Unified Photonic Inference Simulator for Hardware-Aware Cross-Layer Design-Space Exploration
链接: https://arxiv.org/abs/2611.16522
作者: Andreas Falkenberg
类目: Optics (physics.optics); Artificial Intelligence (cs.AI)
备注: 8 pages, 2 figures
Abstract:Photonic neural accelerators offer ultra-low latency linear algebra, yet their usable numerical precision, sustainable throughput, and application accuracy remain tightly bounded by surrounding electro-optical interfaces and physical non-idealities. Pre-fabrication design-space exploration (DSE) requires a cross-layer electronic design automation (EDA) framework capable of evaluating disparate optical topologies under identical neural workloads and physical assumptions. We present \textbfPISim, a unified Photonic Inference Simulator that couples pretrained deep neural network execution with physics-based models of six accelerator architecture families, six electro-optic modulator technologies, four optical source organizations, and direct or coherent readout schemes. PISim embeds stage-resolved impairment operators—covering control and analog-to-digital quantization, finite extinction ratio, shot noise, correlated relative-intensity noise (RIN), thermal drift, propagation loss, and channel crosstalk—alongside an analytical cycle-level interface timing model to expose electronic-optical bottlenecks. Using simulated MNIST multilayer perceptron (MLP) benchmarks, we demonstrate how PISim identifies optimal precision operating points under multi-impairment stress (peaking at 6 bits for Mach–Zehnder and microring modulators), quantifies a 52.5 percentage-point accuracy gap between VCSEL-array and parallel-fanout laser architectures arising, under the modeled conditions, from power partitioning and correlated source noise, and models interface-limited throughput scaling across emitter-array widths from 32 to 128 channels. This comprehensive framework establishes an extensible, reproducible foundation for accuracy-constrained Pareto optimization prior to photonic tape-out.
[AI-183] SUMNO: Spatially Unrolled Multigrid Neural Operator for Fluid Flow Surrogates
链接: https://arxiv.org/abs/2611.16341
作者: Yingqi Hu,Nan Deng,Changxiao Shao,Xiaozhou He
类目: Fluid Dynamics (physics.flu-dyn); Artificial Intelligence (cs.AI)
备注: 37 pages, 6 figures
Abstract:Data-driven surrogate modeling has emerged as a powerful paradigm for accelerating computational fluid dynamics, and hierarchical U-shaped networks, which resemble the structure of classical Geometric Multigrid (GMG) methods, have become prevalent designs. Modern surrogates built on these designs achieve competitive accuracy at a fraction of the cost of numerical solvers; however, accurate prediction of complex flows remains challenging, and no existing architecture leads across them. In such networks, the multigrid analogy mainly concerns the multi-resolution topology, while their components largely follow common vision practice. To address this, this study introduces SUMNO, a multigrid neural operator for fluid flow fields on regular grids that spatially unrolls a V-cycle into a feed-forward U-shaped network. Parameter-free down- and upsampling stand in for restriction and prolongation, and additive skip connections for coarse-grid correction, with normalization-free transfer and skip pathways. Within this macro-architecture, convolutions are used at the fine and intermediate resolutions and shifted-window attention at the coarsest, whose window partitioning is generalized to handle periodicity independently per axis. Evaluated on unseen test configurations across four unsteady and steady flow benchmarks against five representative baselines, SUMNO is best in 5 of 7 settings and ranks second in the remaining two. It reduces the 10-step rollout MSE by about 30% relative to the strongest baseline, and outperforms three baselines with over 130 million parameters using 9.4 million parameters.
[AI-184] ORCHID-SAGE: Fully Reproducible CALPHAD Assessment from Evidence to Phase Diagram
链接: https://arxiv.org/abs/2611.16308
作者: Raymundo Arróyave
类目: Materials Science (cond-mat.mtrl-sci); Artificial Intelligence (cs.AI)
备注: 24 pages, 19 figures. Code and data: this https URL (doi: https://doi.org/10.5281/zenodo.24045967%29%3B assessment record doi: https://doi.org/10.5281/zenodo.24045965
Abstract:Knowledge of phase stability is the first requirement for materials design, and predicting the conditions under which new materials can be synthesized has become a key capability of self-driving laboratories (SDLs). For inorganic materials, notably alloys, these predictions come from thermodynamic databases built with the CALPHAD method. Building a database, however, requires expert judgment at many steps, and because these decisions are rarely recorded, others can use a database but cannot regenerate or update it. Machines can already collect the data, build a first model with machine-learned interatomic potentials and refine it against experiments. The assessment itself, the sequence of judgments that turns these parts into a reliable database, is still carried out by human experts. We argue that it can be automated reproducibly only if the expert knowledge behind an assessment is encoded in formal protocols, so that an assessment delivers a record of its construction and not only its parameters. We introduce ORCHID-SAGE, a standard that organizes this record into reproducible and automatable protocols. In this first attempt, applied to Cr-Ni, the database was regenerated from its record by a script on a second machine, by an AI agent working only through the Model Context Protocol (MCP) and by a refit from public sources only, with the same predictions. Recorded in this way, a database need no longer be fixed at publication but can be updated by AI agents as SDLs and other sources provide new evidence, while the judgment remains with people and on the record.
[AI-185] Pyramidal Columns Acting as Attention Heads
链接: https://arxiv.org/abs/2611.16057
作者: Fabrizia Giulia Garavaglia,Marco Giunti
类目: Neurons and Cognition (q-bio.NC); Artificial Intelligence (cs.AI); Neural and Evolutionary Computing (cs.NE)
备注: v1: main text (20 pp., 7 figs.) and supplementary materials (12 figures)
Abstract:Can the functional architecture of a Transformer attention head be given a biological construction? More specifically, could a suitable group of pyramidal columns in posterior temporal language regions (PTLRs) constitute the biological construction of a neuromorphic mechanical realization of an attention head, under a mathematical instantiation that need not coincide with its standard Transformer instantiation? This paper investigates this hypothesis by examining possible correspondences between the functional organization of attention heads and neurophysiological mechanisms operating within and across pyramidal columns involved in human language processing. Our analysis builds on a mechanical-structural description of Transformer attention heads as systems of simple neural networks linked through output-input and output-weight interconnections. We examine the functional organization of an attention head and the input-dependent transformations it produces, and ask how the corresponding processing relations could be constructed in biological neural tissue. We then identify candidate neural mechanisms of signal transmission and modulation within and across pyramidal columns that may jointly support such a construction. On this basis, we propose that a suitably organized group of pyramidal columns in PTLRs could constitute a biological construction of a neuromorphic mechanical realization of an attention head. Whether the required functional organization actually exists in PTLRs remains an open empirical question. The hypothesis does not require the biological system to reproduce the standard mathematical instantiation of Transformer attention: the same functional architecture may be mechanically realized and mathematically instantiated in a different way, consistent with its biological construction.
[AI-186] he BEATS Dataset: Towards Understanding Music-Physiology Dynamics in Real-World Scenarios
链接: https://arxiv.org/abs/2611.15837
作者: Benjamin Gutierrez Serafin,Tanaya Guha,Fahim Kawsar
类目: ignal Processing (eess.SP); Artificial Intelligence (cs.AI)
备注: Under review in Nature Scientific Data
Abstract:We introduce the BEATS dataset, a robust compilation of psycho-physiological data collected in the wild from 30 participants while they listened to curated playlists intended to promote well-being. An acronym for Biomarkers, Experience and Affect from Therapeutic Soundscapes, the BEATS dataset aims to support research on wellbeing, music psychology, affective computing and signal processing. More specifically, this dataset includes: (1) sensor data collected during music listening sessions using a research-grade wearable device in free-movement environments; (2) music listening activity records, stimulus metadata, acoustic features, and self-reports associated with the listening sessions (3) individual characteristics, such as demographics, music preference, musical skills, and personality traits; and (4) self-reported psychological measures before and after their experimental participation spanning for a number of weeks. By situating data collection outside of lab settings, the BEATS dataset offers the opportunity to study real-world music-listening scenarios, allowing for a richer exploration of the physiological, subjective and therapeutic effects of music under real-world conditions.
[AI-187] PDO Drug Screening Data for Evaluating and Training AI Drug Response Prediction Models
链接: https://arxiv.org/abs/2611.15769
作者: Y. Zhu,K. Marlor,V. Lavelle,T. S. Brettin,M. Sacco,C. Siniscalchi,H. W. Raeder,M. R. Weil,R. L. Stevens
类目: Quantitative Methods (q-bio.QM); Artificial Intelligence (cs.AI)
备注:
Abstract:Anti-cancer drug response prediction (DRP) is critically important for precision oncology and therapeutic development. Currently, most DRP models are developed using drug screening data from immortalized cancer cell lines, which can differ substantially from patient tumors at the genomic and molecular levels, even though patient tumor response is the ultimate prediction target. To help bridge this gap and bring DRP models closer to their intended clinical application, we conducted four drug screening studies using patient-derived organoids (PDOs), which more faithfully recapitulate patient tumors than immortalized cell lines. These studies generated data for improving and evaluating DRP models, including multi-omics profiles of PDO lines, drug feature data, and treatment response measurements. Using these data, we retrained and evaluated multiple publicly available DRP models under three model-building scenarios: training models using cell line data only, training models using combined cell line and PDO data, and pretraining models on cell line data followed by fine-tuning on PDO data. We then evaluated and compared model prediction performance on PDOs across these analysis scenarios. In addition, we also assessed the consistency of drug screening viability data generated by different laboratories using different assay technologies and/or experimental protocols.
[AI-188] Can Model Answer Probabilities Represent Survey Populations? Conditional Prediction and Joint Validity in Jev Survey Simulation
链接: https://arxiv.org/abs/2611.15576
作者: Zengqing Wu,Chuan Xiao
类目: Applications (stat.AP); Artificial Intelligence (cs.AI); Computers and Society (cs.CY)
备注:
Abstract:Survey simulations require realistic combinations of answers, but predictive accuracy alone does not establish that a model can generate them. We evaluate the option probabilities returned by Jev, a TypeSafe decision model, using linked responses to 20 binary items from five archived survey waves. We compare population-only prediction, prediction given recorded answers, and sequential completion that uses generated answers to predict later responses. Joint comparisons hold the starting population information fixed, and two three-question cases are completed in both question orders. We evaluate predictive scores, answer frequencies, and associations against the survey records. Recorded answers can improve predictive scores while associations remain distorted. Large conditional errors also occur. For example, among men who described themselves as having traditional values, the model predicted that 16% would see men and women as similar in workplace abilities, compared with 56% in the survey. Sequential completion has mixed effects on joint validity, and reversing question order produces different joint distributions in both tested cases. Order checks and classical marginal bounds clarify what can be assessed before linked validation answers are available. The findings support validating the joint events a simulation uses and separating empirical constraints from model assumptions. The evidence concerns selected binary tasks and one model version.
[AI-189] Quantum Recurrent Architectures for Financial Time Series Prediction
链接: https://arxiv.org/abs/2611.15298
作者: Danyal Maheshwari,Gerhard Hellstern,Martin Zaefferer,Martin Braun,Tanja Döhler
类目: Quantum Physics (quant-ph); Artificial Intelligence (cs.AI)
备注:
Abstract:This study evaluates quantum classical hybrid recurrent architectures for financial time-series forecasting: a Quantum Long Short-Term Memory (QLSTM) network and Quantum Reservoir Computing (QRC) with lag-based inputs. Classical lag vectors are encoded into quantum states via amplitude encoding, which compactly represents normalized observations but requires nontrivial state preparation; the QLSTM gates and the QRC reservoir are implemented with parameterized quantum circuits, while classical optimizers train the interface and readout layers. On a monthly revenue series extended to 1056 months by a Gaussian-process–hidden-Markov generator, with a fixed chronological 70/15/15 split and one-step-ahead forecasts, we compare both models against a full-size LSTM, a classical control matched in parameter budget and interface architecture, a small plain LSTM, and classical reservoir computing, over ten seeds and against persistence, seasonal-naive and SARIMA benchmark forecasts. The QLSTM, with 2,105 parameters, is statistically indistinguishable from an LSTM 126 times larger, but equally indistinguishable from the matched classical control and from a plain LSTM of the same budget, so its parameter efficiency cannot be attributed to the quantum circuits. The quantum reservoir is worse than a classical reservoir of equal feature dimension in all ten seeds and worse than SARIMA, whereas all recurrent models and both classical reservoirs improve on SARIMA by about 2% in RMSE. In a 24-month rolling-origin backtest the classical LSTMs degrade more slowly than the QLSTM. We find no evidence of a quantum accuracy or parameter-efficiency advantage in this low-dimensional setting.
[AI-190] Multi-head Quantum Neural Networks: Replacing Transformer Blocks with NISQ Modules inside a 30-Billion-Parameter LLM
链接: https://arxiv.org/abs/2611.15250
作者: Maforikan Amoussou,Borja Aizpurua,Sukhbinder Singh,Román Orús,Hans-Arno Jacobsen
类目: Quantum Physics (quant-ph); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 24 pages, 5 figures, 21 tables
Abstract:We introduce the multi-head quantum neural network (QNN), a quantum module compatible with noisy intermediate-scale quantum (NISQ) devices that can replace transformer blocks inside pre-trained large language models (LLMs) with little loss in accuracy and with far fewer parameters than the blocks it replaces. Encoding a model’s D hidden neurons into a qubit register is the obstacle: amplitude encoding scales poorly in circuit depth and angle encoding in qubit count, so neither is practical at production-scale width. We instead partition the hidden space into k disjoint heads, angle-encode each into an n_q -qubit register, process the heads in parallel, and couple them through constrained classical linear layers, the mixers. Performing surgery on Qwen3-30B-A3B-Instruct-2507, we show that a single multi-head QNN with 145,281 trainable parameters ( D=2048 , k=32 , n_q=8 , L=4 ) can replace a contiguous span of eight transformer blocks, displacing 16.33% of the model’s parameters. Ablations show that, for a single replaced block, the quantum heads (circuits followed by a learned readout), and not the classical mixers alone, recover the displaced function, whereas for fused spans the mixers alone come within 1 - 3% of the full module in perplexity. Whether this benefit is specific to trained, entangling quantum circuits is not established: controls that switch the entangler off (two seeds), or freeze the circuit at its initial or at uniformly random angles (one seed), beat the full module on held-out perplexity, and a parameter-matched classical multilayer perceptron remains a few percent ahead of the full module. We train classically with exact statevector simulation, then execute the quantum heads of our fusion checkpoints on a 156 -qubit IBM Heron processor, showing that NISQ-scale quantum modules can be installed in a production-scale LLM and run on hardware.
[AI-191] FréchetFlow: Flow Matching for Non-Euclidean Data
链接: https://arxiv.org/abs/2611.15220
作者: Yidong Zhou,Muqing Cui,Hans-Georg Müller
类目: Methodology (stat.ME); Artificial Intelligence (cs.AI)
备注: 53 pages, 7 figures
Abstract:Scientific observations such as probability distributions, trees and covariance matrices call for generative models that respect their non-Euclidean geometry. While flow matching has been extended to manifolds by utilizing geodesics and velocities, such manifold structures are not available in a general metric space. We introduce FréchetFlow, a generative modeling framework for metric space-valued data that requires neither linear operations nor a smooth manifold structure. A key feature of the proposed approach is that we utilize well-established Euclidean flows to learn the joint distribution of distances to a set of landmarks, whereupon new objects are reconstructed from generated landmark representations. The reconstruction step utilizes weighted Fréchet means of reference objects, drawing on Fréchet regression. We establish Wasserstein error bounds that connect accuracy of the generated object distributions to the placement of landmarks, flow approximation and reconstruction error and prove consistency under finite second moments. This leads to convergence rates that depend on the metric complexity of the target objects, tail moments and fitting accuracy. Experiments show competitive accuracy with established flow matching methods in Euclidean simulations and lower Wasserstein errors than specifically tailored methods for several scientific datasets. Across simulation settings, we find that generation error decreases as sample size increases.
[AI-192] Prototypes of the Mind: A Unified Framework for Probing the Visual Brain NEURIPS2026
链接: https://arxiv.org/abs/2611.15046
作者: Shi Chen,Seoyoung Ahn,Doris Tsao
类目: Neurons and Cognition (q-bio.NC); Artificial Intelligence (cs.AI)
备注: NeurIPS 2026
Abstract:Deep learning has driven substantial progress in modeling fMRI responses, yet most existing approaches are optimized for a single objective—such as neural encoding, stimulus decoding, or post hoc interpretation—making it difficult to obtain a coherent account of neural representations. We introduce a unified framework that connects vision foundation models to brain activity while supporting neural analysis from multiple complementary perspectives. Our approach replaces black-box features with a sparse, interpretable representation composed of visual prototypes, and links these prototypes to voxel responses through a linear readout. This design makes the model transparent at both levels: prototype activations expose the visual concepts present in the stimulus, while the learned linear weights characterize each voxel’s selectivity. Using this interface, we identify which visual concepts drive neural responses and where they appear in the image, reveal cortical organization through voxel tuning in prototype space, and perform controlled stimulus synthesis to probe cortical selectivity through targeted interventions. Together, these results establish prototype-based representations as a unified and interpretable foundation for studying visual representations in the human brain.
[AI-193] Moving Beyond the DRW Model: Learning Quasar Variability with Generative Models
链接: https://arxiv.org/abs/2611.15040
作者: Nadav Bojan Sellam,Kevin Park,Alex Bronstein,Zoltán Haiman
类目: Astrophysics of Galaxies (astro-ph.GA); Artificial Intelligence (cs.AI)
备注:
Abstract:The optical continuum variability of quasars is often described by a damped random walk (DRW) model, with characteristic damping timescales \tau_\rm DRW of hundreds of days and amplitudes \sigma_\rm DRW of \mathcalO(0.1) magnitudes. This model has been widely used as a null hypothesis to assess the significance of periodic signals in quasar light curves in the search for supermassive black hole binaries, the significance of extreme quasar variability like tidal disruption events (TDEs) and changing-look (CL) quasars, or to model time lags in reverberation mapping. However, the DRW model is known to be incomplete, and an incorrect null hypothesis can lead to biased significance estimates. We investigate the impact of such null-hypothesis mis-specification on the detection of periodic signals in synthetic quasar light curves and introduce the generative model Nullasar, capable of learning quasar variability directly from time-series data. We demonstrate that Nullasar, trained with flow matching, can learn delicate statistical properties of stochastic light-curve variability without explicitly assuming any noise model. We further demonstrate that Nullasar can learn correlated distributions of noise parameters and generate new light-curve realizations from a range of deviations from the Gaussian DRW noise model, approximating parametric descriptions of stochastic processes. When used as a null hypothesis for detecting periodic signals in synthetic light curves, Nullasar performs comparably to fitting the correct DRW model to each light curve, and outperforms parametric fitting when the variability follows a more flexible AR(2) model. Finally, we identify practical limitations, including sensitivity to contamination of the training set by non-null signals, and show that projecting the conditioning vector of the model onto its leading principal components largely mitigates it.
[AI-194] QSim-Arena: an evolving developer for quantum simulation advantage
链接: https://arxiv.org/abs/2611.15000
作者: Jue Xu,Yue Wang,Qi Zhao
类目: Quantum Physics (quant-ph); Artificial Intelligence (cs.AI)
备注:
Abstract:Where practical advantage emerges in quantum dynamics simulation remains a central debate in quantum computing. Claimed advantages can be rapidly narrowed or overturned by stronger classical methods, while progress on both sides can outpace the journal publication cycle. We present QSim-Arena, a research framework in which AI agents combine a closed loop of verification and benchmarking with open exploration. Hamiltonians, initial states and accuracy targets for evolved states or observable expectation values provide explicit specifications for evaluating simulation methods. Agents translate published techniques into executable components, reproduce reported results and benchmark their performance across diverse Hamiltonian families. Competing classical methods are implemented and independently crosschecked, providing credible baselines for end-to-end quantum resource estimates based on rigorous error bounds and empirical calibration. Alongside this closed loop, agents recombine classical simulation techniques and co-design fault-tolerant quantum circuits using knowledge extracted from the literature. Initial applications improve baseline resource estimates through the integration of established techniques. QSim-Arena thus treats quantum advantage assessment as an evolving, reproducible research process that can respond to advances in algorithms, error analysis and hardware.
[AI-195] Speech Tokenizers are Vulnerable: Transferable Semantic Attack and Robust Tokenizer NEURIPS2026
链接: https://arxiv.org/abs/2611.14929
作者: Zhisheng Zhang,Yifan Mi,Yixuan Zhou,Haiyun Li,Shengbo Cai,Dongfu Song,Ziang Xu,Jie Hao,Zhiyong Wu
类目: Audio and Speech Processing (eess.AS); Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR); Sound (cs.SD)
备注: Accepted to NeurIPS 2026
Abstract:Speech tokenizers serve as speech front ends, encoding input speech signals into representations for downstream models, such as automatic speech recognition (ASR) systems and large audio-language models (LALMs). In this paper, we find that speech tokenizers are fragile at the semantic level: small perturbations can disrupt their semantic encoding process and consequently cause severe degradation in downstream ASR and LALM performance. To study this vulnerability, we propose T-SemAttack, a transferable semantic attack based on a semantic encoder ensemble. By jointly drifting the representation spaces of multiple semantic encoders, T-SemAttack effectively disrupts the semantic content of speech while preserving perceptual quality, thereby inducing error features. We further analyze layer-wise representation drift and cross-model gradient alignment. Our findings show that perturbations can induce substantial shifts in deep semantic representations and that the input gradients of different surrogate encoders exhibit directional alignment. These observations suggest that shared adversarially sensitive directions in the input space may be a key factor underlying cross-model attack transferability. Building on these, we introduce ROSETok, a robust speech tokenizer that combines noisy training with robust semantic distillation to improve reconstruction fidelity and downstream SpeechLLM robustness. Extensive experiments on multiple datasets, 9 open-source and black-box ASR systems, and 4 LALMs demonstrate that T-SemAttack achieves strong transferable attack performance, while the proposed Robust Speech Tokenizer exhibits robustness under high-fidelity reconstruction and downstream tasks. Our code and demo are available at this https URL.
[AI-196] he Rebundling of Freelance Work in the AI Age
链接: https://arxiv.org/abs/2611.14708
作者: Caroline Chin,Shagun Kala,Nadav Kunievsky,Teng Liu,Alex Martin Richmond,Carl Shan
类目: General Economics (econ.GN); Artificial Intelligence (cs.AI); Computers and Society (cs.CY)
备注:
Abstract:This paper examines how freelance work is changing in the AI era. We compare the activities people undertake in ChatGPT with the work clients request on Upwork, an online freelance platform, and examine how clients combine tasks into projects. We find that activities more prevalent in ChatGPT tend to lose marketplace share on Upwork, while those less prevalent tend to gain share. Within AI-related work, demand is becoming more diverse, and job postings combine tasks that were rarely requested together before the generative AI era. AI-related postings also have substantially larger project budgets and involve more hours of work, while differences in hourly rates are small. Together, these findings suggest that AI is reshaping not only which tasks clients outsource, but also how they combine tasks into larger, higher-value projects.
[AI-197] Finite-Cadence Refresh Equilibria for Streaming Prior-Anchored Ridge: A Closed-Form Forgetting Window and Spectral Correction
链接: https://arxiv.org/abs/2611.14586
作者: Mrigank Shekhar Singh
类目: Machine Learning (stat.ML); Artificial Intelligence (cs.AI); Computer Science and Game Theory (cs.GT); Machine Learning (cs.LG)
备注: 67 pages, 7 figures. Lean 4 formalization and code: [ this https URL ]( this https URL )
Abstract:Recursive estimators on drifting streams must forget stale data fast enough to track the target, yet slowly enough to average out noise. In prior-anchored ridge regression, this trade-off admits a closed-form spectral decomposition in which each eigen-direction interpolates between the data solution and the prior with pull weight s_i = \rho / (\mu_i + \rho) . When the anchor is refreshed every k steps, we show that the moving-frame offset converges to a finite-cadence equilibrium that lags the pure-leaky path by an explicit \Delta_k(s) while reducing estimation variance by a factor \Psi(s, r_k) . Minimizing the cycle-averaged risk gives \omega^* = (4\nu^2 \Sigma / c’)^1/3 , c’ = \sigma^2 d_\mathrmeff(\rho) , and the plug-in strength \rho^* = \sigma^2 / (n\delta^2) . A random-matrix deterministic equivalent extends both laws to finite dimension, renormalizing the anchor to \rho_\mathrmeff = \rho (e^\zeta q - 1) / (\zeta q) ; it matches simulation within 5% at fixed operating points and within 13% at the method’s own, and rank-deficient populations reduce exactly to their active subspace. The resulting method, SPAR, calibrates window and strength online from residuals. On a 27-configuration grid the calibrated window opens on drift without divergence, and the grid defaults transfer to an hourly electricity stream and to BCI IV 2b. The core algebra, refresh chain, and regret bounds are formalized in Lean 4.
[AI-198] Deep learning-guided discovery reveals the diversity and mutational potential of environmental beta-lactamases
链接: https://arxiv.org/abs/2611.14568
作者: Liangjun Ge,Qiandi Gao,Haiming Xu,Zhiwei Qin,Heqian Zhang
类目: Genomics (q-bio.GN); Artificial Intelligence (cs.AI)
备注:
Abstract:Beta-lactamases are a major mechanism of bacterial resistance to beta-lactam antibiotics; however, their functional diversity and mutational potential in environmental microbial communities remain partially understood. This study aimed to characterize environmental beta-lactamase diversity and investigate how sequence variation influences antibiotic resistance. By integrating protein language model (PLMs) with ensemble learning, we identified more than 46,000 putative beta-lactamases across all four Ambler classes from environmental metagenomes, widely distributed across diverse habitats, with distinct habitat-associated class compositions. The candidates sharing approximately 30% sequence identity with known beta-lactamases conferred resistance to multiple beta-lactam antibiotics when expressed in Escherichia coli. Combining inverse folding model with PLMs further identified mutations that altered resistance profiles, including substitutions that increased resistance to carbapenems. Predicted high-fitness sites exhibited greater natural sequence variability overall, with higher variability in wastewater-derived than ocean-derived enzymes. These findings reveal undercharacterized beta-lactamase diversity and substantial mutational flexibility in environmental microbiomes, providing a basis for investigating their potential contribution to the diversification of antibiotic resistance.
[AI-199] AVA: Agent ic Variant Adjudication for Automated Auditable Read-Level Review in Cancer Diagnostic Sequencing ML4H ALT
链接: https://arxiv.org/abs/2611.14560
作者: Caleb Chin,Sidharth Gupta,Arun K Anguraj Vadivel,Robert Siddaway,Siddharth Arya,Arianna He,Peter Zhukovsky,Cynthia Hawkins
类目: Genomics (q-bio.GN); Artificial Intelligence (cs.AI)
备注: 14 pages, 5 figures. Accepted at ML4H (Machine Learning for Health) 2026 Findings Track
Abstract:Identifying a tumor’s DNA mutations is central to cancer diagnosis and treatment, but distinguishing genuine mutations from sequencing artifacts is error-prone. In practice, genuine mutations are identified using fixed variant-allele-frequency (VAF) and depth thresholds which cannot separate artifacts from genuine mutations in the borderline low-VAF zone where they overlap. A laboratory specialist must manually inspect the read-level sequencing data, a slow step that has high inter-rater variability and causes diagnosis delays. We present AVA: Agentic Variant Adjudication; an agentic framework that automates this read-level inspection with subagents that reason over evidence, generate investigative code, and return a real-or-artifact adjudications with an auditable reasoning trace all without training on existing data. On both the SEQC2 DNA-seq benchmark and patient RNA-seq data, restricted to a subsample of borderline variants where thresholds fail, both closed-source (Claude Sonnet 5) and open-weight (GLM 5.3 Flash and Gemma 4 31b) agents outperform the best achievable VAF/depth thresholds. Ultimately, AVA is suited for privacy-constrained clinical deployment, since the framework can run on locally deployed open-weight models with descriptive reasoning traces to keep specialists in the loop.
[AI-200] Gaussian processes for real-time urban traffic estimation using a fleet of drones
链接: https://arxiv.org/abs/2611.14536
作者: Batuhan Avci,Marko Susnjar,Marko Maljkovic,Nikolas Geroliminis
类目: Applications (stat.AP); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Systems and Control (eess.SY)
备注:
Abstract:Real-time urban traffic monitoring is essential for modern city management, yet conventional static sensors provide only limited spatial coverage. Autonomous drone fleets can complement this infrastructure by offering flexible, high-resolution mobile sensing across different parts of the network. This paper introduces a spatio-temporal Gaussian process (GP) framework for reconstructing network-wide traffic states from partial aerial observations. The model accounts for the physical structure of the road network through a graph Matérn kernel, which embeds network topology into the spatial covariance structure, and captures temporal dependencies through an exponential temporal kernel. Within this framework, historical loop detector data provide an informative prior mean that is updated in real time using incoming drone observations, yielding network-wide point estimates together with predictive uncertainty. We validate our framework in two distinct case studies: a microscopic simulation of Barcelona and real-world drone data from Athens. The results demonstrate that our approach outperforms historical baselines in estimating and predicting Vehicle Hours Traveled (VHT), notably reducing the required fleet size. Furthermore, leveraging the model’s predictive uncertainty to guide drone movements yields additional performance gains, highlighting the potential for future uncertainty-aware fleet routing algorithms.
机器学习
[LG-0] Internalizing Investment Mandates: A Neurosymbolic RL Approach to Portfolio Allocation
链接: https://arxiv.org/abs/2611.17329
作者: Zeynep Yıldız,Satchit Chatterji,Ercan Atam,Erman Acar
类目: Machine Learning (cs.LG)
*备注: Accepted at ICAIF '26
Abstract:Institutional portfolios operate under investment policy statements that specify asset allocation constraints, often conditioned on market regime information. Deep reinforcement learning has been shown to optimize financial performance but typically does not enforce rule compliance, making it unsuitable for real-world deployment. Existing neurosymbolic methods such as Probabilistic Logic Policy Gradient (PLPG) offer constraint-satisfying reinforcement learning (RL) via differentiable symbolic supervision but assume a finite, discrete action space. We present a PLPG-inspired continuous allocation method that estimates the probability that the policy’s own executed action satisfies the investment mandate, using reparameterized Monte Carlo sampling, and adds it to the training objective. No post-hoc enforcement mechanism intercepts actions at inference time and compliance is learned by the agent. Audited directly on deterministic mean actions with no correction, the policy reduces violations of a six-rule regime-conditional mandate from 96-98% for unconstrained RL and 33.4% for a reward-penalty baseline to 4.8%, showing that a continuous-action neurosymbolic policy can substantially internalize a symbolic mandate.
[LG-1] owards a Theory of Learnability via Debate
链接: https://arxiv.org/abs/2611.17303
作者: Tosca Lechner,Vinayak Pathak,Sivan Sabato,Ruth Urner
类目: Machine Learning (cs.LG)
*备注:
Abstract:Debate has been recently proposed as a training protocol that is hoped to produce more aligned models even as model capabilities scale. The idea is that the human, instead of providing supervision in the form of labels, judges a debate between two models that compete to prove the correctness of opposite labels. It can be shown that functions of a much higher computational complexity can be correctly judged in this way, compared to the complexity of functions that can be directly evaluated using a computationally weak evaluator. This suggests that perhaps debate can be used to teach very complicated functions to models. However, previous work has largely ignored the fact that this is a training protocol, and simply analyzes the game being played between the two debaters for a known judge. We take the first steps towards understanding the learning problem implicit within the protocol, by formulating it as an online learning problem. We show that once we formulate the learning problem precisely, there are cases where the best way to learn given access to a judge is different from running an adversarial debate. This motivates us to define a cooperative version of the protocol, whose learnability we characterize using a combinatorial dimension. We also consider the possibility of the human judge being fallible and occasionally judging the debate incorrectly. We show that a small number of errors in judgement can lead to a large error in learning.
[LG-2] wo mathematical models of knowledge distillation
链接: https://arxiv.org/abs/2611.17230
作者: Audrey Xie,Ludwig Schmidt,John Duchi
类目: Machine Learning (cs.LG)
*备注:
Abstract:Many hypotheses compete to explain the successes of knowledge distillation. To help address this, we propose and analyze a mathematical model of distillation, which suggests that distillation’s performance comes not from obtaining better models but from easier to optimize landscapes. For generalized linear models trained with stochastic gradient descent, we prove that distillation fits performant student models asymptotically more quickly than non-distilled models. In rank-1 matrix approximation, we characterize conditions on the target matrix under which gradient descent with distillation converges strictly faster than training on the supervised objective. The theory helps delineate the ways distillation provides benefits (i.e., in optimization speed, not in generalization), and experiments on real datasets corroborate the theoretical predictions.
[LG-3] Combinatorial Bernoulli Factories with Polynomially Many Coins
链接: https://arxiv.org/abs/2611.17223
作者: Renato Paes Leme,Rad Niazadeh
类目: Data Structures and Algorithms (cs.DS); Computer Science and Game Theory (cs.GT); Machine Learning (cs.LG); Probability (math.PR); Computation (stat.CO)
*备注:
Abstract:We provide an efficient implementation of combinatorial Bernoulli factories. Such a factory turns coins of unknown biases x into a random vertex of a polytope P whose mean is exactly x . For every affine slice of the cube (the class of polytopes that admit a factory), we give a single algorithm whose expected number of coin flips and running time at every interior x are polynomial in the dimension and in 1/\delta_* , where \delta_=\min_i\x_i,1-x_i\ is never supplied to the algorithm. For perfect matchings this gives O(n^8\delta_^-2\log(n/\delta_*)) flips, while the factory of Niazadeh, Paes Leme, and Schneider [STOC 2021] spends at least n^2n-1 flips in expectation at the uniform doubly stochastic matrix. For other polytopes, our method rounds exactly given a point deep inside P and the promise that x is far from the boundary of P .
[LG-4] heoretical and Empirical Analysis of Sparse Convolution
链接: https://arxiv.org/abs/2611.17212
作者: Keshvi Tuteja,Yu-Hsiang Tsai,Gregor Olenik,Markus Götz,Achim Streit,Hartwig Anzt,Charlotte Debus
类目: Performance (cs.PF); Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG)
*备注: Accepted at AI on HPC Workshop at SC’26 (Supercomputing 2026)
Abstract:Deep neural networks for image processing are increasingly deployed on resource-constrained edge devices such as drones, medical systems, and smartphones, where limited computational resources and memory bandwidth pose significant challenges for efficient inference. Pruning during training often produces highly sparse convolutional kernels, offering the potential to significantly reduce computation and memory requirements. However, exploiting this sparsity efficiently at runtime remains challenging due to irregular memory access patterns and workload imbalance across processing elements. In this work, we investigate the standard convolution with sparse weights and analyze its performance characteristics through roofline analysis, showing that a 3 x 3 sparse convolution achieves an arithmetic intensity of approximately 10 FLOP/byte compared to roughly 96 FLOP/byte for its dense counterpart. We propose techniques for an efficient sparse convolution at high sparsities. Preliminary results demonstrate that our approach achieves upto 1.65x speedup as compared to dense convolution on NVIDIA A100 and upto 1.52x on NVIDIA Jetson Nano (Tegra X1) , motivating further exploration of direct convolution as a practical path to efficient sparse inference on modern accelerators/edge devices.
[LG-5] Physically Interpretable Basin Selection in Discrete-Variable Origami Design
链接: https://arxiv.org/abs/2611.17196
作者: Lucy Liu,Leon M. Kamp,Katia Bertoldi,L. Mahadevan
类目: Machine Learning (cs.LG); Materials Science (cond-mat.mtrl-sci); Soft Condensed Matter (cond-mat.soft)
*备注:
Abstract:Automatic differentiation has enabled rapid progress on continuous-variable mechanical design problems, but gradient-based optimization still struggles with discrete variables. The standard approach is to optimize using a continuous relaxation of the choices, then commit to the nearest discrete design at the end. Why this procedure selects unfavorable basins is poorly understood. Here, we introduce a practical, interpretable design problem inspired by origami, in which automatic differentiation is used to design shapes with a target functional behavior. Each origami sheet has a fixed geometry, and a bilevel optimization framework tunes which one of four discrete hinge types to place at each fold. Baseline gradient-based design settles into basins with a large discretization gap: relaxed optimization solves 97% of runs, but the final, discrete solutions succeed only half the time. We find that the gap can be relieved with a surprisingly simple modification. An initialization prior uniformly biased toward the stiff, flat hinge option boosts the final solve fraction from 49% to 69%. Applying the prior adds no additional compute and leaves the optimization process unchanged. Compared to the baseline, sheets produced with the prior have greater geometric compatibility and mechanical stiffness. Directly penalizing frustration and floppiness in the outer loss function produces an even greater improvement than the prior does, raising the solve fraction to 78% and indicating that these properties help distinguish the unfavorable basins. Our results demonstrate that loss landscapes for mechanical design can have physically meaningful basins that are not robust to discretization and cause solves to fail. Targeting these basins’ physical signatures, even with just an initialization bias, helps avoid them.
[LG-6] Symmetry in Time: A Representation Theory of Linear Recurrent Memory
链接: https://arxiv.org/abs/2611.17162
作者: Anif N. Shikder,Ramit Dey,Jan Minac,Lyle E. Muller
类目: Machine Learning (cs.LG); Neural and Evolutionary Computing (cs.NE); Representation Theory (math.RT)
*备注:
Abstract:State space models (SSMs) are an emerging alternative to Transformers. SSMs offer linear-time inference and parallel training, but most implementations make the parallel scan efficient by making the recurrence diagonal: each channel of the hidden state is multiplied by an input-dependent real or complex number. These numbers commute, so the state records which inputs occurred but not their order. We formalize this with representation theory, which explains and fixes the state-tracking failures previously reported for SSMs. The state update defines a linear representation of the monoid of input sequences, and we show that an SSM tracks a group at every sequence length exactly when that group is a quotient of the group generated by its transition matrices. Two consequences follow. Diagonal real- and complex-valued SSMs can track only commutative groups, while the transition matrices of any SSM that can track a group must split into independent blocks over the real, complex, and quaternion numbers. We then introduce HyperSSM, which adds quaternion channels while keeping linear-time inference and parallel training. Trained only on final answers, HyperSSM solves the group tasks of recent state-tracking benchmarks exactly with at most five state numbers. This includes S_5 with 32 generators, where previous SSMs are at chance, and HyperSSM generalizes to sequences 25 times longer than those seen in training. Because the theory says what the transitions must encode, the learned weights of the linear recurrence can be interpreted directly. We demonstrate this by recovering the calibration of a real gyroscope, the rotations of imperfect qubit gates and the geometry of protein backbones.
[LG-7] AeroZoo: A Benchmark of Well-Posed Flow Problems for Foundation Models of Aerodynamics
链接: https://arxiv.org/abs/2611.17152
作者: Peter Sharpe,Carmelo Gonzales,Corey Adams,Mehdi Ataei,Sanjay Choudhry
类目: Machine Learning (cs.LG); Computational Engineering, Finance, and Science (cs.CE); Fluid Dynamics (physics.flu-dyn)
*备注: 35 pages, 13 figures, 23 tables. Lead authors: Peter Sharpe and Carmelo Gonzales. Companion paper to “ISLA: A Coordinate-Free, Measure-Weighted Transformer for Boundary-Driven PDEs” (Sharpe, 2026). Dataset to be published shortly at this https URL
Abstract:Neural surrogates for computational fluid dynamics (CFD) are trained and ranked on datasets that each vary one design family in one flow domain. Such datasets cannot test the least that an aerodynamic foundation model must do: predict CFD-quality incompressible flow in seconds for new shapes in new enclosures. On DrivAerML, larger models and longer training barely lower Transolver’s error; doubling the training cases cuts it by about 12% up to 400 cases, exhausting the dataset. We introduce AeroZoo, a benchmark and training corpus of 3,815 time-mean solutions to the unsteady Reynolds-averaged Navier–Stokes (RANS) equations with a transition model, across seven geometry families, eight enclosure configurations and Reynolds numbers from 10^3 to 10^8 . Each sample poses a well-posed PDE problem in the form a solver receives it: the boundary meshes grouped by condition type, the prescribed values and the fluid properties. The 2,816 samples with a stable time mean form the canonical benchmark: two tasks, engineering metrics and 22 frozen splits, disjoint in geometry, that hold out geometry families, enclosures, boundary-condition layouts and Reynolds ranges. On the random split, boundary-condition tokens cut GeoTransolver’s median velocity error from 76.9% to 8.4% of the inlet speed in single-seed runs (uniform inlet flow: 11.0%). ISLA, a transformer for boundary-driven PDEs, reaches 5.3% over four seeds, about 4 times the change from a finer mesh. On the same test samples, a held-out geometry family barely changes ISLA’s error relative to uniform flow, and a held-out enclosure or boundary-condition layout raises it 2.4 to 9.2 times. The data and the scoring code will be released.
[LG-8] Enhancing Preference Detection using Eye Movement Cascade and Multilayer Combinatorial Fusion
链接: https://arxiv.org/abs/2611.17145
作者: Yuanhong Wu,Mary Tsaryk,Sam Watson,Tahsin Ahmed,Christina Schweikert,D. Frank Hsu
类目: Machine Learning (cs.LG)
*备注: Published in the Proceedings of the 26th Annual International Conference of the Global Business and Technology Association (GBATA), 2026, pp. 969-980
Abstract:Understanding how visual attention influences human preference formation is a fundamental problem in consumer behavior and cognitive science. In this work, we investigate how eye movement patterns can be used to predict preference decisions when subjects compare two alternatives. To address this problem, we analyze an eye-tracking dataset, where subjects view pairs of face images and their gaze behavior is recorded throughout the decision process. We model this problem using multiple machine learning classifiers and apply the Multi-layer Combinatorial Fusion (MCF) framework to iteratively combine diverse models. The MCF framework expands models to a large pool of model fusions and then reduces them based on cognitive diversity, repeating this process across layers until diversity diminishes. Experimental results show that MCF significantly improves performance over individual models. The best single model achieves a Pre@k of 0.90, which is improved to 0.9143 after the first layer of MCF, and further increased to 0.9286 in the second layer. This progression demonstrates that the second layer not only surpasses the performance of the first layer but also outperforms the best individual model, highlighting the effectiveness of leveraging diversity through multi-layer fusion.
[LG-9] A Convex-Analytic Reformulation for Poisoning Attacks on Logistic Regression
链接: https://arxiv.org/abs/2611.17143
作者: Carina Moreira Costa,Maria Eduarda Pinheiro,Martin Schmidt
类目: Machine Learning (cs.LG); Optimization and Control (math.OC)
*备注:
Abstract:We study data poisoning attacks on logistic regression with \ell_2 -regularization, where an attacker chooses the features of a fixed number of injected observations with prescribed labels to maximize the logistic loss on the clean training data. Strong convexity of the defender’s problem yields an exact reformulation through its first-order optimality conditions. Using convex conjugacy of the softplus function, we replace the sigmoid relations by a single convex optimality-gap inequality and express the attacker’s objective as a concave function. A change of variables makes the stationarity condition affine, leaving bounded bilinear equalities as the only nonconvex constraints. We derive finite bounds on all variables involved in these equalities, including a global bound on the coefficient norm and tighter bounds obtained by preprocessing based on convex optimization. Experiments on eight binary classification datasets evaluate attack effectiveness and compare the two reformulations when solved with \textsfSCIP. Attacks obtained by local optimization degrade predictive performance on most datasets, while the convex-analytic formulation produces substantially stronger reported dual bounds within a fixed time limit.
[LG-10] ISLA: A Coordinate-Free Measure-Weighted Transformer for Boundary-Driven PDEs
链接: https://arxiv.org/abs/2611.17141
作者: Peter Sharpe
类目: Machine Learning (cs.LG); Computational Engineering, Finance, and Science (cs.CE); Mathematical Physics (math-ph); Fluid Dynamics (physics.flu-dyn)
*备注: 48 pages, 14 figures, 26 tables. Companion paper to “AeroZoo: A Benchmark of Well-Posed Flow Problems for Foundation Models of Aerodynamics” (Sharpe et al., 2026)
Abstract:A foundational principle of physics is that all coordinate systems are equally valid: the behavior of a PDE-governed system does not depend on the origin or orientation of the frame, nor on the units or mesh used to write it down, and neither should a surrogate model. Yet the neural surrogates with the lowest errors on today’s aerodynamics benchmarks lack these invariances and fail under simple transformations of the same problem: for models trained on PDEs of cars in one pose (i.e., translation and rotation) and from one mesher, rotating the full PDE raises the surface-pressure error from 4% to more than 100%, and re-sampling its boundary raises it to about 30%. We introduce ISLA (Invariant Slice Attention), a slice-attention transformer for steady, boundary-driven PDEs in which no learned operation receives a coordinate and every sum over the boundary is a quadrature. Positions enter only as distances and relative angles to learned slice anchors, lengths are measured in a unit computed from the boundary, and interior points query the boundary through a tree-accelerated learned kernel without influencing one another. With every boundary cell a token and an exact boundary read, ISLA is equivariant to rigid motions and reflections, invariant to the units of its declared data, and consistent under mesh refinement, while at every setting each query is independent of all other queries. On AeroZoo, 2,816 flows that vary in shape, size, speed, and boundary conditions, its mean reference-scaled velocity and pressure errors are at least 3.1 and 3.3 times lower than any baseline’s. On AirfRANS its errors are second only to those of a boundary-integral model (GLOBE) and at least 1.6 times lower than those of Transolver and GeoTransolver.
[LG-11] Support-Constrained Offline Policy Improvement for Pediatric ECMO from Limited Clinical Data
链接: https://arxiv.org/abs/2611.17134
作者: Fateme Golivand Darvishvand,Saurabh Mathur,Nikhilesh Prabhakar,Michael A. Skinner,Ameet Soni,Neel Shah,Ethan L Sanford,Lakshmi Raman,Sriraam Natarajan
类目: Machine Learning (cs.LG)
*备注:
Abstract:Managing children on ECMO (Extracorporeal Membrane Oxygenation) requires balancing short-term physiological stability with the long-term goal of preventing neurological injury. Standard reinforcement learning can propose risky, out-of-distribution actions, while imitation learning is limited to reproducing clinician behavior. We introduce Support-Constrained Fitted Q-Iteration (FQI-SC), an offline reinforcement learning framework that cautiously improves upon expert behavior by restricting optimization to empirically supported state-action pairs and defaulting to the clinician baseline in low-data regions. On 78 pediatric trajectories totaling over 6,000 hours of clinical data, we evaluate our framework using Fitted Q-Evaluation (FQE) and a behavioral disagreement. Additionally, we formalize the performance-safety trade-off by introducing a unified Balance Score ( S(\pi) ). Our empirical evaluation shows that support-constrained policy improves the Balance Score by 38% relative to FQI, improves estimated value by 1.28 FQE units over imitation learning, and reduces disagreement with clinician actions by 71% relative to FQI.
[LG-12] Paper-Faithful Federated Learning: Scientific Checks for Practitioner and Agent Development
链接: https://arxiv.org/abs/2611.17129
作者: Ali Dadras,Alp Yurtsever
类目: Machine Learning (cs.LG); Software Engineering (cs.SE)
*备注: 13 pages, 2 tables. Code: this https URL
Abstract:Federated learning experiments can produce reasonable-looking results even when the experiment implementation differs from the intended experimental protocol. Small errors in configuration, data partitioning, aggregation, randomness, evaluation, or checkpoint recovery may therefore affect a comparison without causing the program to fail. We present FedBrew, a research codebase for controlled and reproducible federated learning experiments. Its focus is controlled simulation with checks of specified conditions of the intended experimental protocol. FedBrew uses configuration checks, identifiable client data partitions, checkpoint recovery checks, self-describing run records, and development-time checks of scientific conditions. These checks assess, for example, whether client partitions are disjoint and exhaustive and aggregation weights match the prescribed rule; numerical and recovery checks examine non-finite client model states and interrupted runs. The same structure also provides clear extension points and automated checks for new algorithms, datasets, models, and tasks, whether they are added by researchers or AI coding agents. FedBrew therefore supports reproducible and inspectable FL experiments through checks of specified software and scientific conditions and the run records associated with reported results.
[LG-13] Using an energy pixels approach to estimate the size a 2nd Generation DEL must be to train a high-resolution binding model
链接: https://arxiv.org/abs/2611.17121
作者: Boris P. Belotserkovskii,R. Edward Watts
类目: Machine Learning (cs.LG); Biomolecules (q-bio.BM)
*备注: 11 pages, 4 figures
Abstract:DNA-encoded libraries (DELs) are commonly used to find small-molecule binders to a target protein. But the hard part is no longer finding a hit. It is building a machine-learning model good enough to guide the medicinal chemistry that follows, one that can resolve the shift in affinity when a chemist changes a single atom. Such a model needs two things from its training data: dense sampling of the small patch of chemical space around the hit, and an adequate number of genuinely tight binders within it. A diverse naive DEL gives neither. It samples an enormous space, so any one neighborhood is covered far too thinly, and the tightest binders are rare. A model trained on that naive DEL data must extrapolate to binders tighter than any it has seen, and that is where models can fail. One way to get training data with both properties is a second-generation DEL built around the hit. We ask how complex that library must be. We treat binding as a sum of small, independent interactions, called pixels on the target and antipixels on the ligand. From the spread of affinities measured in the naive library, with no structural information and nothing yet synthesized, we estimate the complexity such a library must have to supply that dense sampling and those tight binders. The method does two useful things. It turns the number of tight binders a model needs into the number of compounds a library must contain. And it estimates the expected yield of a randomly designed second-generation DEL, so the gain from designing the library with machine learning or active learning can be measured against a baseline, without the expense of a control library as large as the experiment itself.
[LG-14] Bellm an Contraction under MMD: A Unified Framework NEURIPS2026
链接: https://arxiv.org/abs/2611.17116
作者: Baptiste Debes,Tinne Tuytelaars
类目: Machine Learning (cs.LG)
*备注: Accepted at NeurIPS 2026. 88 pages, 18 figures, 13 tables
Abstract:We introduce a mathematical framework for studying contraction of the distributional Bellman operator under Maximum Mean Discrepancy (MMD). We show that contraction reduces to conditional positive definiteness of an associated kernel constructed from the base kernel, which we call the Bellman difference kernel. Focusing on radial kernels, we use Schoenberg’s theorem to connect this condition to Bernstein functions. This gives contraction criteria that hold in any dimension and can be verified analytically or tested numerically. Our framework unifies known results for the Gaussian RBF, power-distance, and multiquadric kernels. A numerical search identifies 11 new kernels that induce contraction. We also propose chaining, a general construction that produces kernels inducing contraction with any prescribed factor. To reconcile theory with practice, we show that contraction on a set only requires conditional positive definiteness on that set. This explains why the RBF kernel can remain a valid choice when rewards are bounded, despite its lack of global contraction. Finally, we identify a fundamental tradeoff between contraction rate and discriminative power, where kernels that contract faster tend to discriminate less between distributions. Code is available at this https URL
[LG-15] Near-Optimal Regret for Oracle-Efficient Adversarial Contextual Bandits
链接: https://arxiv.org/abs/2611.17109
作者: Kiarash Banihashem,MohammadTaghi Hajiaghayi
类目: Machine Learning (cs.LG); Data Structures and Algorithms (cs.DS)
*备注:
Abstract:We study contextual bandits with K actions, a class of N policies, and i.i.d.\ contexts from a distribution that the learner can sample. For any sequence of loss functions fixed before play, we give an oracle-efficient algorithm with expected regret O(\sqrtKT\log(2N),\log(2KT)) against the best policy in hindsight on the realized contexts. This improves the T^2/3 dependence on the horizon in prior oracle-efficient bounds and achieves the optimal regret rate in this setting up to logarithmic factors. Our algorithm perturbs policy scores using a Poisson sample of auxiliary contexts. At each sampled context, the perturbation assigned to an action is a Laplace noise draw multiplied by a scale specific to that context and action. We show that choosing these scales to minimize a convex objective determined by past observations yields the stated regret bound. We further approximate the minimizing scales by projected stochastic gradient descent using only an exact offline cost-sensitive optimization oracle that returns the optimum value. Since the table of scales can be large, the algorithm represents the SGD iterates implicitly by storing the random draws that determine each update and evaluates the scales only at queried contexts. Both the oracle complexity and the remaining resource bounds are polynomial in T , K , and \log(2N) . Subjects: Machine Learning (cs.LG); Data Structures and Algorithms (cs.DS) Cite as: arXiv:2611.17109 [cs.LG] (or arXiv:2611.17109v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2611.17109 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-16] Capacity-Constrained Delayed Feedback for Online Conformal Risk Control
链接: https://arxiv.org/abs/2611.17059
作者: Bowen Wang,Osvaldo Simeone
类目: Machine Learning (cs.LG); Signal Processing (eess.SP)
*备注:
Abstract:Online conformal risk control (OCRC) enables reliable uncertainty quantification in machine learning by converting model outputs into prediction sets with long-term risk guarantees. These guarantees, however, hinge on the assumption that the loss of each round is revealed immediately and losslessly. In practice, feedback may be delayed, and limited resources may prevent tracking all pending predictions, causing some feedback to be lost. To address this challenge, we develop capacity-aware OCRC (CA-OCRC), a framework that jointly determines how to allocate limited tracking feedback capacity and how to construct prediction sets. CA-OCRC is based on a randomized tracking policy that samples a proxy delay to decide whether, and for how long, feedback for the current round is tracked. CA-OCRC updates the prediction sets via inverse probability weighting, compensating for the possibility of feedback lost due to delays and limited tracking capacity. We analyze risk bounds for both arbitrary and stochastic delays, revealing a trade-off between receiving more feedback and avoiding capacity saturation. Building on these bounds, we optimize the proxy delay distribution, showing scaling laws for the tracking capacity that ensure the same long-term risk as OCRC with unlimited feedback. Experiments on medical imaging, natural language processing, computer vision, and wireless communications show the effectiveness of CA-OCRC as compared to existing OCRC mechanisms.
[LG-17] Infrastructure for Verifiable AI: Where Proof Costs Go and How Hardware Can Help
链接: https://arxiv.org/abs/2611.17042
作者: Yihan Xia,Wenxin Xu,Taotao Wang
类目: Hardware Architecture (cs.AR); Cryptography and Security (cs.CR); Machine Learning (cs.LG)
*备注: 16 pages, 8 figures, 5 tables
Abstract:Users of artificial intelligence services, including those powered by large language models and other transformer-based models, need assurance that inference and training follow the declared computation without exposing private models or data. Zero-knowledge proofs can provide this assurance, but proof generation remains expensive. Hardware acceleration research often reports gains for individual cryptographic operations, leaving their effect on the full cost of proving neural-network workloads, particularly large transformer models, unclear. We introduce a four-layer framework linking verifiable AI workloads, proof systems, cryptographic kernels and hardware platforms, use it to review the literature, and add original measurements of six provers and seven kernels to identify which costs matter in practice. Our measurements show that the dominant cost changes with the proof system and input length. In low-rank fine-tuning, the cost of opening commitments to fixed model weights remains nearly constant across input lengths from 64 to 2048, while the work associated with intermediate values grows. A 13-billion-parameter fine-tuning configuration exceeds the capacity of a 40-gigabyte graphics processor but completes on an 80-gigabyte device, showing why capacity must be assessed alongside execution time. Our kernel profiling further identifies instruction issue and synchronization limits that a bandwidth-only model misses. We use these findings to assess graphics processors, reconfigurable logic, custom chips and existing artificial intelligence accelerators. The resulting research agenda connects proof-friendly numeric formats, reuse of fixed-weight commitments, heterogeneous execution and shared benchmarks to the measured bottlenecks. Effective acceleration requires selecting hardware and proof-system optimizations around the cost profile of the target workload.
[LG-18] mporal Regularized Learning: Self-Supervised Learning Local in Space and Time
链接: https://arxiv.org/abs/2611.17019
作者: Davide Wiest
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注: 13 pages, 4 figures. Code: this https URL
Abstract:Layer-local self-supervised objectives remove the backward pass between layers, but almost all still compute their loss over a batch of samples and are therefore local in space but not in time. We introduce Temporal Regularized Learning (TeReL), a spatially and temporally local learning rule for self-supervised representation learning from sequential input streams. The key idea is to instantiate a VICReg-style objective with temporally adjacent positives and non-differentiated running population statistics, yielding a per-neuron online loss with bounded state. The objective is a soft form of Slow Feature Analysis, and its gradient is a Hebbian update of the feedforward weights and an anti-Hebbian update of one lateral matrix per layer. The aim of the experiments is to show that the rule works in a small network, not to compete with backpropagation. On a class-chunked MNIST stream, a two-layer fully connected network trained with batch size one outperforms random features (96.48% versus 95.09% linear-probe accuracy) and classical SFA, and a variant that relaxes both locality constraints is comparable to supervised backpropagation. Because training and inference use the same computation, the network can keep learning from unlabeled data during inference without loss of accuracy.
[LG-19] An Integrated Architecture for Data-Efficient Real-Time Neural Imitation of Nonlinear Time-Optimal Model Predictive Control
链接: https://arxiv.org/abs/2611.17005
作者: Maarten De Baecke,Flavia Sofia Acerbo,Andrea Giusti,Wilm Decré
类目: ystems and Control (eess.SY); Machine Learning (cs.LG); Robotics (cs.RO)
*备注: Submitted to European Control Conference 2027 (Brussels)
Abstract:Nonlinear time-optimal model predictive control (TOMPC) enables highly performant control of constrained systems, but its computational cost prohibits real-time use on fast dynamical systems. Neural network imitation of the TOMPC control law can reduce online computational costs, yet existing work mostly targets quadratic cost formulations in simulation and addresses the challenges of neural controllers in isolation. This work imitates a long-horizon nonconvex TOMPC with hard terminal constraints, whose high computational cost limits training-data generation. We present (i) \emphself-informed sampling (SIFS), a data-efficient expert-sampling strategy that uses a learned time-to-go to query the expert only where the policy deviates from time-optimal behavior, (ii) a complete training and control architecture that jointly addresses dataset correlation, suboptimal expert labels, covariate shift, constraint satisfaction and model mismatch, and (iii) real-world validation on a Furuta pendulum. SIFS matches standard dataset aggregation performance in roughly one-third of the training time and with less than half the training data. The resulting controller matches the region of attraction of the expert TOMPC while raising the control frequency from 1.4~Hz to 1~kHz. On hardware, the controller swings up and stabilizes the pendulum along trajectories closely matching the expert TOMPC solution.
[LG-20] me Flies When You Parallelise: Pareto-efficient Time Series Foundation Models with Linear Recurrences
链接: https://arxiv.org/abs/2611.16934
作者: Torben Berndt,Lingyi Yang,Federico Nardi,Yi Heng Lim,AJ Shears,Benjamin Walker
类目: Machine Learning (cs.LG)
*备注:
Abstract:Time series foundation models have demonstrated strong zero-shot forecasting performance, but computational efficiency remains a central architectural challenge. Temporal self-attention incurs quadratic cost in the number of temporal tokens, while nonlinear recurrent state transitions restrict parallel computation along the temporal dimension. We introduce ci-proto, a family of time series foundation models based on Mamba-3, whose linear state recurrences support chunk-wise parallel-in-time computation during training, and inference costs that scale linearly with sequence length. The architecture alternates recurrent temporal processing with cross-variate attention, supporting multivariate probabilistic forecasting with past-only and future-known covariates. Across fev-bench’s 100 forecasting tasks, ci-proto XL sets a new state of the art among time series foundation models. The five ci-proto variants also make up the entire Pareto frontier for mean scaled quantile loss and mean absolute scaled error against both parameter count and median runtime.
[LG-21] Factored One-sided Adam-Moment for Practical and Scalable SOAP NEURIPS2026
链接: https://arxiv.org/abs/2611.16905
作者: Jaemyung Yu,Byeongho Heo,Sangdoo Yun,Dongyoon Han
类目: Machine Learning (cs.LG)
*备注: NeurIPS 2026. 17 pages, 5 figures, 11 tables. Code: this https URL
Abstract:Second-order optimizers (e.g., Shampoo, SOAP) have emerged as compelling alternatives to Adam/AdamW and promise faster per-step convergence than AdamW in transformer training. However, they still remain costly in practice, and even SOAP still pays substantial optimizer-state memory and eigendecomposition overhead. We ask whether the benefits of SOAP can be retained at nearly the efficiency of AdamW. In this paper, we answer yes with a lightweight SOAP variant, Factored One-sided Adam-Moment (FOAM). The path to practical impact is twofold: first, FOAM employs input-only preconditioning, dropping the output-side Kronecker factor to reduce memory and remove one eigendecomposition. The Kronecker structure of the Hessian shows that this is exact for a diagonal output-side curvature and nearly lossless for a near-diagonal one, as we observe empirically. Second, FOAM replaces SOAP’s dense rotated second moment with a factored Adam-moment estimator based on row-by-column statistics. Under the K-FAC Fisher assumption with a diagonal output curvature, this estimator exactly recovers SOAP’s dense second moment, and we measure only a 15-20% K-FAC violation, with negligible effect on the update. Our empirical results support the claim that FOAM matches strong SOAP and Muon baselines in validation loss across scales of large language model (LLM) pre-training, uses half of SOAP’s optimizer state, runs at Muon-class wall-clock time, and retains robustness across a three-fold learning-rate range. The results hold from 124M to 760M parameters, demonstrating that SOAP’s gains need not require SOAP’s cost. Code: this https URL
[LG-22] Sustainable Federated Fine-Tuning with LoRA: Environmental Impact and Performance Footprint
链接: https://arxiv.org/abs/2611.16856
作者: Pedro Augusto Londe Ribeiro,Rodrigo Moreira,Larissa Ferreira Rodrigues Moreira,Joberto S. B. Martins,Tereza C. Carvalho,Flávio de Oliveira Silva
类目: Machine Learning (cs.LG); Networking and Internet Architecture (cs.NI)
*备注: Accepted author manuscript for publication in the Proceedings of the 2027 IEEE Consumer Communications Networking Conference (CCNC 2027), Las Vegas, Nevada, USA, January 9-12, 2027
Abstract:Federated Learning (FL) with Parameter-Efficient Fine-Tuning (PEFT) enables efficient adaptation of Large Language Models (LLMs), yet its energy and carbon costs across hardware remain poorly understood. We evaluated federated Low-Rank Adaptation (LoRA) fine-tuning on NVIDIA A40, RTX 6000 Ada, and A100 GPUs using a 2^2 full factorial design with 135M/360M models and 10/100 local steps. CodeCarbon tracked energy and estimated CO _2 -equivalent emissions. Descriptive allocation of variation assigned 84.8-91.2% of energy and emissions variation to local steps within these ranges. On A40 and RTX 6000 Ada, configurations with fewer steps achieved comparable accuracy with roughly 6-8-fold lower energy consumption and higher accuracy per kWh. These findings support smaller local budgets for energy-efficient federated fine-tuning within the evaluated setting.
[LG-23] Calibration and Conformal Prediction for Medical Question Answering in Arabic and English: Evidence from Two Non-Parallel Benchmarks
链接: https://arxiv.org/abs/2611.16821
作者: Mariya(Stanley College of Engineering and Technology for Women, Hyderabad, India)
类目: Machine Learning (cs.LG)
*备注: 15 pages, 1 figure, 12 tables. Code: this https URL
Abstract:Large language models (LLMs) are increasingly tested on medical exam questions, but accuracy alone does not show when an answer can be trusted. I study the calibration and uncertainty of two open-weight models, Qwen2.5-7B-Instruct and Jais-adapted-7B-chat (both run in 4-bit precision), on 2,816 English multiple-choice questions from MedMCQA and 4,927 Arabic multiple-choice questions from MedAraBench. Accuracy was 0.561 (English) and 0.430 (Arabic) for Qwen, and 0.369 and 0.308 for Jais. Raw option probabilities were overconfident in all four settings (expected calibration error 0.242 to 0.486), and a single temperature parameter removed 82% to 93% of that error. Split conformal prediction reached its nominal coverage in all 24 settings I tested (within 0.006), but the sets were large when accuracy was low: at 90% coverage, temperature-scaled sets contained on average 63% (Qwen, English), 78% (Qwen, Arabic), 86% (Jais, English) and 89% (Jais, Arabic) of the options. The margin between the top two option probabilities separated correct from incorrect answers best for Qwen on English (AUROC 0.734) and worst for Jais on Arabic (0.579). Answering only the 20% most confident questions lowered Qwen’s English error from 0.439 to 0.098. Jais showed a strong prior for option C (49% of English and 60% of Arabic predictions). Scoring Latin letters only cut the Arabic share to 29% but left accuracy, raw ECE and AUROC nearly unchanged. Because the two benchmarks differ in subject, level and number of options, I do not attribute differences between languages to language itself. Code is available at this https URL.
[LG-24] ScrewSense: Mapping the Invisible - Anomaly Detection and Error Classification in Screwdriving Processes
链接: https://arxiv.org/abs/2611.16789
作者: Tobias Lettner,Tatyana Ivanovska,Anwar Al Assadi,Christian Bergler
类目: Machine Learning (cs.LG); Robotics (cs.RO); Signal Processing (eess.SP)
*备注: Accepted manuscript, 59th CIRP Conference on Manufacturing Systems (CMS 2026). 6 pages, 8 figures, 3 tables
Abstract:The growing volume of process data generated by screwdriving systems, combined with advancements in computing power, data storage, and algorithms, is facilitating the application of artificial intelligence in this field. Accurately detecting and classifying screwdriving errors is crucial for ensuring product quality, avoiding costly rework in manufacturing, and supporting sustainability by facilitating the maintenance, reuse, and recycling of screw-connected products. In this context, machine learning approaches offer a promising alternative to traditional monitoring techniques, such as window-based methods or envelope curves. This study examines the use of deep learning techniques for anomaly detection and error classification in screwdriving processes. The data corpus consists of multivariate time series capturing process parameters, such as torque and rotation angle. For anomaly detection, the embedding space of a vector quantized variational autoencoder was analyzed. In terms of error classification, four different model architectures were evaluated. The results demonstrate the effectiveness of anomaly detection and confirm that deep learning models can distinguish between multiple error categories beyond traditional binary classification. This extends the capabilities of conventional monitoring approaches, achieving accuracies of up to 100%. Additionally, real-world data material was examined in order to further demonstrate model robustness. Overall the results indicate strong generalization capability and highlight the practical potential of the developed models for use in industrial screwdriving processes, eliminating the need for expert knowledge in error identification and significantly reducing manual effort and analysis time.
[LG-25] Class-Competence Aggregation: Output-Layer Routing for Federated Multiclass DDoS Detection under Class-Concentrated Non-IID Data
链接: https://arxiv.org/abs/2611.16772
作者: Tapu Datta,Hafiz Imtiaz
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注: Under review in IEEE TIFS
Abstract:Federated learning over distributed intrusion-detection sensors naturally produces \emphclass-concentrated non-IID data, where an entire attack class is held by a single client and absent elsewhere. We show this collapses federated averaging in a specific, localized way: the output layer holds one decision boundary per class, and when a class’s aggregation weight comes overwhelmingly from clients that have never seen it, its neuron receives an active anti-signal rather than mere dilution, collapsing to zero. On CIC-DDoS2019, standard FedAvg drives the concentrated LDAP class to 0.00 \pm 0.00% F1 across every seed. We propose Class-Competence Aggregation (CCA), which aggregates the output layer per class using each client’s own locally reported validation F1 as the weight, while hidden layers remain under standard size-weighted FedAvg. CCA requires no server-held auxiliary data and adds only one F1 scalar per class per round beyond the feature-scaling and class-count exchanges common to any federated pipeline. On our primary CNN-LSTM architecture, CCA alone raises LDAP F1 from 0% to 84.83 \pm 1.65% , and the full method reaches 91.64 \pm 0.57% macro F1, closing the federated-to-centralized gap from -42.53 to -4.05 points. Benchmarked against four other federated methods under identical partitions and seeds, CCA is the only one that recovers the concentrated class substantially and reliably; a 178\times smaller MLP shows the same qualitative recovery, though less completely, confirming the effect holds across architecture scale even as its magnitude varies. Results are validated across three seeds and replicated on a second dataset, CICIDS2017.
[LG-26] Beyond Binary Classification: Multiclass Machine Learning for Diabetes and Prediabetes Risk Prediction Using NHANES
链接: https://arxiv.org/abs/2611.16756
作者: Aman Mujahid,Anita Penkova
类目: Machine Learning (cs.LG)
*备注:
Abstract:Detecting diabetes risk early is important for preventing disease progression. Most existing machine learning approaches focus on binary classification, which limits their ability to capture intermediate risk states such as prediabetes. We develop a supervised learning framework to classify adults as normal, prediabetic, or diabetic using National Health and Nutrition Examination Survey data. The framework uses routinely collected demographic, clinical, and lifestyle variables from 15,356 adults, with glycated hemoglobin (HbA1c) used only to define labels and excluded from input features to prevent leakage. We compare conventional machine learning and transformer-based models and examine two binary formulations: normal versus prediabetes, and normal versus combined prediabetes and diabetes. Models are evaluated on a held-out test set using precision, sensitivity, specificity, and F1 score. In the multiclass setting, performance is broadly consistent across models, with accuracy up to 0.72 and macro-averaged F1 of 0.70. Prediabetes remains the most challenging class, with F1 scores of 0.54 to 0.56 for conventional models and 0.53 to 0.55 for transformers, reflecting its overlap with both neighboring states. When the task is reduced to normal versus prediabetes, F1 scores rise modestly to 0.62 to 0.64. A more substantial improvement is observed under a broader screening formulation combining prediabetes and diabetes, with F1 scores of 0.73 to 0.74 across all models. These findings suggest that feature overlap along the biological continuum between normal and diabetic states contributes more to the difficulty of prediabetes detection than model architecture. Multiclass prediction provides a more clinically informative representation of metabolic risk progression but highlights a trade-off between clinically meaningful stratification and predictive accuracy.
[LG-27] Interpretable machine learning diagnosis of degradation mechanisms in lead-free RbSn_0.5Ge_0.5I_3 solar cells from temperature- and illumination-dependent electrical measurements
链接: https://arxiv.org/abs/2611.16750
作者: Nabanita Sarker,Hafiz Imtiaz
类目: Machine Learning (cs.LG); Mathematical Physics (math-ph)
*备注: Under review in RSC Advances
Abstract:Machine learning coupled with device simulation is widely used to optimize thin-film solar cells, but almost exclusively in the forward direction, from device parameters to predicted performance. The inverse problem encountered after fabrication, namely identifying the mechanism that limits a given device, is still addressed by manual inspection of characterization data. In this work, both directions are examined for lead-free RbSn_0.5Ge_0.5I_3 solar cells using approximately 42,000 SCAPS-1D simulations. Replacing the Spiro-OMeTAD hole transport layer with CuI increases the simulated power conversion efficiency from 19.6% to 21.6%, almost entirely through the fill factor. Surrogate models trained and tested on a factorial parameter grid substantially overstate their predictive accuracy. An XGBoost regressor attains R^2 = 0.9997 on held-out grid points but only 0.58 on independently sampled devices, whereas training on a Latin hypercube design of comparable size restores R^2 = 0.995 . For the inverse problem, a multi-label classifier identifies which of six degradation mechanisms are present from simulated temperature- and illumination-dependent open-circuit-voltage measurements. It achieves a macro F_1 score of 0.938, compared to 0.687 for a calibrated physics-rule baseline. The classifier does not, however, generalize to combinations of mechanisms absent from the training set.
[LG-28] On the Difficulty of Sequence-Level Supervised Learning for Autoregressive Language Models
链接: https://arxiv.org/abs/2611.16736
作者: Mathieu Blondel,Michaël E. Sander,Germain Vivier-Ardisson,Tianlin Liu,Vincent Roulet
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:During the supervised learning phase, autoregressive language models are typically trained by minimizing the negative log-likelihood of the training data, which decomposes as a sum of token-level cross-entropies via teacher forcing. This objective is simple and efficient but potentially suffers from so-called exposure bias: a mismatch between the way the model is trained and the way it is used at inference time. Can we instead design sequence-level loss functions that incorporate inference-time decoding procedures, such as greedy decoding or autoregressive sampling? We show that two natural attempts lead to loss functions that are fundamentally flawed: unbounded below. We further show that these two flawed loss functions are intimately connected: using autoregressive Gumbel tricks, the sampling based loss is the expected greedy loss under Gumbel perturbations. Since the pathology stems from the dependence of the decoded sequence on the model parameters, we analyze the update direction obtained by applying a stop gradient on the sampling procedure, leading to a moment matching direction. While it is not the gradient of any loss function, we prove that it defines a valid descent direction in function space for the autoregressive negative log-likelihood. Despite these theoretical insights, the moment matching direction suffers from high variance. We conclude that teacher forcing may well be ``all you need’’ for supervised learning of autoregressive models.
[LG-29] Manifold Attracting Flow Matching: Physics-Constrained Sampling via Continuous Geometric Dynamics
链接: https://arxiv.org/abs/2611.16677
作者: Guanyu Chen,Jinchen Wang,Shengze Xu,Xingyu Ren,Tieyong Zeng,Antonios Armaou,Dong Ni
类目: Machine Learning (cs.LG)
*备注:
Abstract:Generating scientific data under previously unseen equality constraints requires satisfying the prescribed constraints without destroying the physical structure learned from data. Existing inference-time approaches often enforce physical constraints by repeatedly correcting predicted endpoints, but such discrete corrections can distort learned generative structure and introduce unstable feedback. In this work, we propose Manifold Attracting Flow Matching (MAFM), which continuously evolves a persistent anchor through an ODE whose dynamics drive it analytically toward the constraint manifold. Specifically, its geometric correction preserves model-guided tangent motion while prescribing the normal motion to contract constraint violations toward the manifold. This construction separates model tracking from constraint attraction and provides theoretical guarantees for residual contraction and local feedback stabilization. Experiments across geometric and partial differential equation benchmarks demonstrate improved generation quality over variant constraint-guided baselines while maintaining low constraint residuals. MAFM provides a flexible framework for coordinating learned physical structure and explicit constraints through controllable sampling dynamics, without retraining the underlying generative model.
[LG-30] GNN-Enhanced TCN Algorithms for ECG Anomaly Detection
链接: https://arxiv.org/abs/2611.16648
作者: Angelica Simonetti,Ferdinando Zanchetta
类目: Machine Learning (cs.LG)
*备注: Follow-up to arXiv:2310.02774 , which introduced a pipeline for an industrial dataset. A shortened version of that earlier paper appeared in the GSI 2025 proceedings. Here we extend the theoretical analysis and develop a pipeline tested on a benchmark dataset. Research conducted in 2025. Under journal peer review
Abstract:To a time series can be also given the structure of directed graph, opening the doors to the usage of graph neural networks (GNNs) in this context. In this paper we develop a new pipeline involving novel geometric deep learning models that merge the capabilities of ordinary TCNs with the ones of GNNs to solve an anomaly detection problem on electrocardiogram signals.
[LG-31] Rethinking Actions: Continuity-Aware Priors for World-Model Planning
链接: https://arxiv.org/abs/2611.16645
作者: Shaobo Yang,Zichen Ai,Ziyang Ma,Qian Yang,Bin Fang
类目: Robotics (cs.RO); Machine Learning (cs.LG)
*备注:
Abstract:Actions in expert demonstrations exhibit temporal structure. An ordered-versus-shuffled analysis across five environments confirms this dependence. We ask whether action proposals that better reflect expert continuity can improve visual world-model planning. Rather than imposing generic smoothness, we learn continuations from the temporal structure of demonstrated actions. Our locality-gated deep-kernel Gaussian-process prior, LG-DKGP, combines recent and longer-range executed action history, predicts a neural continuation, and learns a temporal Gaussian-process (GP) residual. A training-derived exploration scale initializes search around the corrected mean; a frozen world model evaluates progress toward the goal. Training first fits the neural mean, then jointly optimizes the prior with the GP residual using expert future actions. The resulting planner achieves 87.07% macro success across five simulated tasks spanning native and fragment-level goal attainment, compared with 71.03% for INTACT, 61.66% for PRISM, and 58.66% for LeWM. Component interventions and expert-relative action-structure comparisons support demonstration-informed continuations and reveal task-dependent effects of visual conditioning, history length, and planning horizon.
[LG-32] Machine Learning for Scalable Multi-AP Coordination in IEEE 802.11bn Networks
链接: https://arxiv.org/abs/2611.16634
作者: Maksymilian Wojnar,Szymon Szott
类目: Networking and Internet Architecture (cs.NI); Machine Learning (cs.LG)
*备注:
Abstract:As Wi-Fi deployments become denser, fully distributed random channel access no longer guarantees the required quality of service. The multi-access point coordination (MAPC) framework introduced in IEEE~802.11bn lets neighboring access points (APs) cooperate, but choosing which APs transmit together and how to configure them remains an open, computationally hard problem. We develop and validate a stack of machine learning (ML) coordination methods, progressing from lightweight hierarchical multi-armed bandits (H-MABs) to a generative pipeline built on graph neural networks (GNNs) and flow matching (FM). The methods are evaluated in simulation and on a hardware testbed, improving throughput over legacy access and scaling coordination to dozens of APs where analytical and search-based methods become intractable.
[LG-33] DySR-GNN: Dynamic Sensor-Relationship Graph Neural Network for Root Cause Analysis in Automotive HiL Multivariate Time Series
链接: https://arxiv.org/abs/2611.16623
作者: Hamza Ouarrad,Mohammad Abboush,Andreas Rausch
类目: Machine Learning (cs.LG)
*备注:
Abstract:Root-cause analysis in automotive Hardware-in-the-Loop (HiL) systems is challenging because faults can propagate across multiple interacting signals, making the most abnormal sensor different from the actual fault origin. This work proposes DySR-GNN, a Dynamic Sensor-Relationship Graph Neural Network for supervised multi-label root-cause identification in multivariate HiL time series. The approach jointly extracts temporal patterns within individual sensor signals and structural relationships across sensors. Each sensor is processed by a shared temporal encoder and combined with a trainable sensor-identity representation, from which a sparse relationship graph is reconstructed independently for every input window. This window-specific structure provides an explicit representation of the sensor relationships used for graph message passing and direct root-location prediction. Under a recording-disjoint evaluation protocol, the proposed model outperforms CNN–LSTM, Static GCN, Static GAT, and context-based variants, achieving a recording-level Macro-F1 of 0.9524, MCC of 0.9302, and strict exact-match accuracy of 0.8519. Analysis of the learned graphs reveals both persistent and state-dependent relationships among sensor variables. The results demonstrate that extracting adaptive relational structure alongside temporal representations can improve multi-label root-cause identification in complex automotive cyber–physical systems.
[LG-34] HALight: Heterogeneous Multi-Agent Reinforcement Learning for Traffic Signal Control under Lane-Level Missing Observations
链接: https://arxiv.org/abs/2611.16574
作者: Chenkai Hu,Yingqian Zhang,Ming Xie,Yi Fang,Chengxiang Jin,Chao Li
类目: Machine Learning (cs.LG)
*备注:
Abstract:Multi-agent reinforcement learning has shown great promise for traffic signal control. However, most existing methods rely on two key assumptions: (i) agents are assumed to be homogeneous with unified observation/action spaces; (ii) full and reliable lane-level traffic observations are available to all agents. In practice, urban road networks exhibit substantial structural heterogeneity (e.g., lane configurations and phase designs), while sensor failures and incomplete detector coverage often result in missing incoming lane measurements. Such issues can severely impair coordination and destabilize learning. To address these challenges, we propose HALight, a communication-enhanced heterogeneous multi-agent reinforcement learning framework for traffic signal control under lane-level missing observations. HALight supports intersection-specific observation and action modeling to accommodate heterogeneous intersections, and incorporates a lightweight one-hop neighborhood communication module to enrich spatial context when lane-level observations are partially unavailable. Moreover, HALight adopts a sequential policy update scheme to improve coordination among heterogeneous agents. Experimental results on two real-world road networks demonstrate that HALight outperforms representative homogeneous and heterogeneous baselines in terms of traffic efficiency and convergence behavior. The code and datasets used in this work are available at this https URL.
[LG-35] Look Here Then Decide: Small Decision Models in a Harness Find Memory-Safety Vulnerabilities at Scale on Commodity GPUs
链接: https://arxiv.org/abs/2611.16565
作者: Tim De Wachter,Kyryl Volkov,Myrto Eleftheria Gkogkou,Mathieu Huysman
类目: Cryptography and Security (cs.CR); Machine Learning (cs.LG); Software Engineering (cs.SE)
*备注: 24 pages, 2 figures
Abstract:We ask whether small, open-weight decision models can support vulnerability discovery in open-source software at scale and what the surrounding system must provide. These models answer a typed question in one forward pass, returning scores for the permitted answers rather than text. Our harness selects widely packaged C and C++ projects and uses deterministic incomplete-fix miners and shape localizers to decide where to look. Intern-Decision-4B, a 4B-parameter open-weight decision model, then scores every candidate function. The scores rank the queues rather than accept or reject candidates, apart from a low reach-score floor that some queues kept. The search uses no frontier model. Alone, the model is weak: it separates vulnerable from benign functions with AUC 0.73, but a vulnerable function from its own patch with only 0.56. It is more useful for ordering candidates. On 75 public CVEs from projects without any of our findings, the fixed function reached its repository’s top 25 in 21% of cases, 1.4 times as often as under random order of the same queues. Recall is bounded mainly by where the localizers look. On the campaign’s own queues the model’s order placed 7.5 times more confirmed findings in the top 10 than random order, an observed, selection-biased upper bound. On three GB10 systems, the harness searched more than 4,000 projects at 900-1,100 repositories per hour. The campaign confirmed 187 memory-safety findings; by worst-case CVSS v3.1, 12 are critical, 148 high, 23 medium and 4 low. The decision step runs at 4.4 decisions per second on one GB10 and 9.8 on a consumer RTX 4090 gaming PC, with matching scores. A small decision model supports vulnerability discovery through prioritization, while deterministic analysis and verification remain essential. All confirmed findings are being disclosed to maintainers under coordinated disclosure before publication. Comments: 24 pages, 2 figures Subjects: Cryptography and Security (cs.CR); Machine Learning (cs.LG); Software Engineering (cs.SE) Cite as: arXiv:2611.16565 [cs.CR] (or arXiv:2611.16565v1 [cs.CR] for this version) https://doi.org/10.48550/arXiv.2611.16565 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-36] Similarity-Guided Optimization Strategies for Approximate Network Symmetry Detection
链接: https://arxiv.org/abs/2611.16555
作者: David Hartman,Jaroslav Hlinka,Dominik Klepl,Anna Pidnebesna,Petra Vidnerová
类目: Neural and Evolutionary Computing (cs.NE); Machine Learning (cs.LG); Neurons and Cognition (q-bio.NC)
*备注: 90 pages, 15 figures and 2 tables in the main text; supplementary material with 64 figures and 7 tables included
Abstract:Symmetry in complex networks is a structural property with broad implications for network function, dynamics, and robustness. While classical approaches rely on exact automorphisms and thus remain highly sensitive to small perturbations, recent work of Liu has introduced approximate formulations that relax the strict invariance condition and recast symmetry detection as an optimization problem over permutations. We investigate how node-level similarity information can be used to bias the search toward structurally plausible vertex correspondences, with network centralities serving as one interpretable instance. Building on this perspective, we adapt simulated annealing, genetic algorithms, tabu search, particle swarm optimization, differential evolution and the quadratic symmetry approximator to this problem, add similarity guidance to each, and introduce Bilinear Approximate Symmetry Inference (BASIN), which learns a soft assignment matrix via bilinear embeddings. Using a unified evaluation pipeline, we compare their guided and unguided variants, BASIN and a random baseline on four synthetic graph families, structural brain connectomes, and three empirical networks with known automorphism groups. Node-similarity guidance is decisive where a planted orbit exists that the unguided search misses, and algorithm-dependent elsewhere, on synthetic and empirical networks alike. Overall, tabu search achieves the strongest performance; its guided variant (SGTS) has the lowest median score among the guided variants in all 44 synthetic conditions, while BASIN is the better of the continuous relaxations on rewired lateral graphs. The effect of node-similarity guidance is therefore regime-dependent rather than uniformly positive.
[LG-37] Learning Temporal Dynamics with Derivative Representations for Time Series Forecasting
链接: https://arxiv.org/abs/2611.16514
作者: Shujie Duan,Jia Hong Puah,Haocheng Wu,Yifei Mao,Hongting Zhou,Chenxuan Li,Zitao Fang,Yang Lu,Sim Kuan Goh
类目: Machine Learning (cs.LG)
*备注: 20 pages, 3 figures
Abstract:Time series forecasting aims to predict the future evolution of complex systems by learning their underlying dynamics from past observations. Despite recent advances in deep learning, Transformers in particular still struggle to capture the temporal evolution of system dynamics. Existing approaches have been largely driven by architectural innovation, with comparatively little attention paid to training strategies that explicitly exploit the temporal structure governing system dynamics. In this work, we introduce DeltaTSF, a dynamics-aware forecasting framework based on discrete derivative representations of time series. Specifically, DeltaTSF jointly learns the series and their discrete temporal derivatives across both input and forecast horizons within a unified, model-agnostic framework. By coupling these representations over time, the model is encouraged to maintain temporal consistency, a property we show to induce a form of discrete Taylor consistency during training. This formulation enables the model to better reflect the underlying system dynamics, resulting in more stable and physically coherent forecasts. Notably, DeltaTSF does not require explicit derivative computation and introduces no additional computational overhead at inference time. Extensive experiments across diverse benchmarks and backbone architectures demonstrate that DeltaTSF consistently improves forecasting performance and yields more dynamically coherent predictions than strong SOTA methods. These results highlight the effectiveness of enforcing dynamic coherence through derivative representations as a model-agnostic paradigm that operates alongside and complements architectural design. Source code will be made publicly available upon acceptance.
[LG-38] Helgason-Fourier Random Features for Scalable Hyperbolic Latent Variables
链接: https://arxiv.org/abs/2611.16461
作者: Koshi Watanabe,Daichi Mochihashi,Hideitsu Hino,Takahiro Ogawa,Miki Haseyama
类目: Machine Learning (cs.LG)
*备注: 28 pages, 8 figures
Abstract:Hyperbolic geometry is well suited to embedding hierarchical data, capturing tree-like structure faithfully even in the low dimensions where Euclidean dimensionality reduction is severely constrained. We introduce Random Helgason–Fourier Features (RHFF), an explicit finite-dimensional feature map for kernels on hyperbolic space. We characterize the spectral density from which RHFF frequencies are drawn, allowing kernels such as the Matérn family to be specified directly in the frequency domain. We also prove a uniform approximation error bound for RHFF. Building on this representation, we develop a hyperbolic latent variable model called RHFF-LVM, which requires no numerical quadrature, variational bounds or auxiliary variables unlike previous hyperbolic GP-LVMs. Our RHFF-LVM recovers the metric of a synthetic tree more faithfully than all baselines and attains the best mean rank, driven by large gains on tree-structured atlases at Q!=!2 . Furthermore, it fits 5\times10^5 points in 505 seconds, making it the first non-parametric hyperbolic LVM to reach this scale.
[LG-39] A Recipe for Physical Action Model Training
链接: https://arxiv.org/abs/2611.16450
作者: Shaojiang Wang,Geer Yang
类目: Machine Learning (cs.LG)
*备注: 13 pages, 1 figure, 2 tables; ancillary code and numerical results included
Abstract:Choosing robot model capacity and demonstration budgets also requires decisions about data composition, representation, optimization, and deployment. We connect these decisions through a common conditional-distribution target. Nested learning oracles give exact risk accounting; local information geometry, conditional mutual information, and empirical convergence provide conditional training prescriptions. Known-target numerical checks expose capacity-data coupling, representation-induced mixture mismatch, and an example where deployment KL increases while negative log-likelihood decreases. An analysis of 206 public PushT trajectories evaluates fixed-feature information-design criteria without fitting a policy. Under asymmetric evaluation targets, the selected allocations reduce this criterion by 4.5-6.2% relative to target matching; balanced targets show no benefit. These results clarify which quantities can support training recommendations and when an apparently favorable predictive metric has a different interpretation.
[LG-40] BlinkMimic: Training a Humanoid Motion Tracker in Minutes on a Single GPU
链接: https://arxiv.org/abs/2611.16422
作者: Fuyi Chen
类目: Robotics (cs.RO); Machine Learning (cs.LG)
*备注: 18 pages, 5 figures. Code: this https URL
Abstract:Whole-body motion tracking lets a humanoid robot follow reference motions, but training a tracking policy for a library of motions has usually taken hours of GPU time. SuperMimic showed that the policy can instead be trained by backpropagating through a world model learned from parallel simulation, and InstantTrack added Muon updates, a residual policy around the reference and an averaged deployed policy. We build on both and present BlinkMimic, a recipe that makes this training faster. On a simulated Unitree G1, BlinkMimic trains one policy for all 40 LAFAN1 clips in 3.4 minutes on a single RTX 4090, measured until three consecutive checkpoints reach a normalized return of at least 0.82 with no failed clip (mean of three seeds). We split training time into the cost of an iteration, the iterations needed to fit the motions, and the iterations needed to remove the last failures, and assign each component, inherited or new, to one of these terms. The new components are a step size that decreases once training has progressed, resets directed by failures of the deployed policy under deployment conditions, and GPU graph capture of the whole iteration. A leave-one-out ablation with three seeds per component shows which term each component affects, and further measurements of the gradient, the update rule and the failure cases explain these effects. Code and per-seed results are available at this https URL.
[LG-41] Bait: Mitigating Harmful Agent ic Reinforcement Learning via Probing Reward Function
链接: https://arxiv.org/abs/2611.16369
作者: Tiansheng Huang,Ananth Balashankar,Virat Shejwalkar,Santosh Pande,Ling Liu
类目: Machine Learning (cs.LG); Cryptography and Security (cs.CR)
*备注:
Abstract:We expose a safety vulnerability on nascent agentic reinforcement-learning-as-a-service. Specifically, with a cost as low as \ 20, we show that safety alignment of a large language model (e.g., Inkling) can be completely subverted through rounds of harmful agentic reinforcement learning (agentic RL) with a toxic environment from its users. We also demonstrate that upon a successful harmful agentic RL attack via the service API, the attackers can potentially use the safety-degraded model to assist with real-world harmful activities, e.g., exploit software vulnerabilities and launch code re-use attacks. To mitigate harmful agentic RL, we propose Bait, a defense that mitigates the attack via proactively probing the reward function. Experimental results show that such a simple defense can sufficiently mitigate learning from a harmful reward function and can still learn from a benign reward function that promotes benign agentic tasks. Further analysis on both empirical and theoretical aspects shows that an adaptive attacker who assumes the knowledge of Bait defense cannot perfectly subvert it due to a systematic challenge. Source code is available at this https URL.
[LG-42] MutAtlas: A PDB-Wide Energy-Guided Atlas of Protein Mutation Effects ICML2026
链接: https://arxiv.org/abs/2611.16361
作者: Ruihan Guo,Chaoran Cheng,Zhanghan Ni,Neil He,Bangji Yang,Ge Liu
类目: Machine Learning (cs.LG); Genomics (q-bio.GN)
*备注: Accepted at the 43rd International Conference on Machine Learning (ICML 2026)
Abstract:Protein mutation effect prediction is fundamental to protein engineering and disease variant interpretation, yet experimentally measured mutation data remain accurate but extremely sparse. To provide scalable supplementary mutation signals, we construct a PDB-wide mutation augmentation dataset that exhaustively enumerates single-site substitutions on experimentally resolved protein structures and aligns mutation signals from physics-based energy models, protein language models, and inverse folding models. Large-scale analysis under a unified mutation preference representation reveals substantial differences in the consistency, concentration, and substitution patterns of mutation distributions across models, indicating that disagreement is pervasive and reflects conflicting inductive biases rather than random noise. Motivated by these observations, we propose an unsupervised multi-source mutation preference distillation framework that learns from relative mutation preferences while explicitly modeling cross-source disagreement. Without using any experimental mutation labels during training, our approach achieves the best overall performance among the evaluated zero-shot baselines and naive multi-source fusion strategies on ProteinGym. We release the dataset and evaluation pipeline to support reproducible studies of protein mutation effects.
[LG-43] Climate Temporal Temperature Prediction via an Adaptive Stacked Bidirectional Residual Mamba Network
链接: https://arxiv.org/abs/2611.16322
作者: Yongxiang Lei
类目: Machine Learning (cs.LG)
*备注:
Abstract:Accurate temperatures forecasting relies on traditional meteorological parameters that are essential for monitoring weather informatics and guiding forecasting efforts. This study presents Deep DualMamba, a novel deep learning framework for temporal temperature prediction, built upon an enhanced variant of the recently proposed Mamba network. Unlike conventional Transformer-based architectures, which dominate foundation models but suffer from computational inefficiencies on long sequences, Deep DualMamba leverages selective structured state space models (SSMs) to achieve both scalability and performance. We identify a core limitation of prior subquadratic-time models-namely, their inability to perform long-term and robust forecasting and introduce two key innovations. First, we propose an adaptive bidirectional Mamba as encoders to learn complex and efficient features. Second, we develop a hardware-aware, parallel recurrent algorithm to compensate for the loss of convolutional efficiency, ensuring practical viability at scale. Last, the Deep DualMamba’s approximation ability and convergence is theoretical analyzed. These design choices culminate in a simplified, attention- and MLP-free neural architecture that exhibits linear complexity with sequence length and significantly improved inference throughput. To demonstrate its applicability in climate informatics, we implement a full forecasting pipeline using historical meteorological data from UK monitoring stations.
[LG-44] Overcoming Spurious Correlations in Conditional Diffusion Models via Doubly Robust Target Correction
链接: https://arxiv.org/abs/2611.16306
作者: Eiichi Sakurai,Kenji Yamanishi
类目: Machine Learning (cs.LG)
*备注:
Abstract:While conditional diffusion models have achieved remarkable success in generation tasks, they memorize spurious correlations within training data. This causes them to deviate from ideal target distributions. Minimizing a covariate shift-adjusted target risk can address this issue. However, retraining models from scratch for different distributions is computationally prohibitive. To overcome this limitation, we propose DR-Diff, a framework that efficiently fine-tunes pre-trained conditional diffusion models. Specifically, DR-Diff leverages a functional gradient approach to derive a pseudo-target. This pseudo-target captures the residual between the base model’s noise prediction and the ground-truth noise under the target distribution. Training the model to fit this pseudo-target enables efficient target adaptation. Theoretically, we prove that DR-Diff guarantees target risk reduction at the population level. It also achieves a pseudo-doubly robust property, providing resilience against density ratio estimation errors and base model misspecifications. Empirically, DR-Diff matches the noise prediction accuracy of models trained from scratch on synthetic datasets. Furthermore, it mitigates inter-group disparities in downstream classification tasks.
[LG-45] StateAlign: Internalizing State-Aware Decision-Making for Reliable Multi-Turn Tool Agents via Reinforcement Learning
链接: https://arxiv.org/abs/2611.16304
作者: Zhexi Lu,Subhajit Chaudhury,Tejaswini Pedapati,Keerthiram Murugesan,Lei Yu
类目: Machine Learning (cs.LG)
*备注:
Abstract:Reinforcement learning has improved the tool-use capabilities of large language models. However, better task performance can coincide with increased tool hallucination. Reliable tool use requires decisions to follow the current task state: available information, tool capabilities, and task progress. Our analysis shows that agents sometimes execute when they should ask for information, retrieve it, or decline an infeasible request, even when reasoning identifies missing information. We introduce StateAlign, combining task-state construction with state-contrastive reinforcement learning. The data cover five task conditions and pair states with different action requirements. Outcome rewards guide task completion; token-level contrasts compare the same generated reasoning and actions under paired states to train outputs that fit the current state. Experiments across five base models from three families show improvements on BFCL v4 multi-turn, ToolSandbox, and When2Call. Ablations show that state construction improves performance and reliability under different reward designs. The complete method further reduces tool hallucination across three Qwen3 sizes. On Qwen3-8B, StateAlign improves BFCL v4 multi-turn accuracy from 45.5% to 55.2% and reduces no-tool hallucination on SimpleToolHalluBench from 62.2% under outcome-only training to 56.1%. Behavioral analyses show more consistent decisions, more complete multi-call execution, and stronger recovery after tool errors.
[LG-46] LoCoJAX: Host-Local Coordination for Small-Model RL on TPUs
链接: https://arxiv.org/abs/2611.16301
作者: Loki Chen,Gregory B. Rehm,Yang Song,Hongye Xie,Ming Lei,Chongyang Bai,Bob Kamma,Mingfu Liang,Kavosh Asadi,Yufei Li,Xi Liu,Jay Xu,Rui Liu,Maziar Sanjabi,Luke Simon,Hamed Firooz
类目: Machine Learning (cs.LG); Distributed, Parallel, and Cluster Computing (cs.DC)
*备注:
Abstract:Reinforcement learning (RL) increasingly targets small language models, and executing these workloads efficiently across multiple accelerators requires effective coordination. On TPUs, Pathways-based RL uses a single controller to coordinate computation across hosts. This execution model is effective for large models, but its coordination computational costs are harder to amortize when small models repeatedly invoke short computations. To characterize these computational costs, we sweep dense models from 0.6B to 405B parameters on TPU v7x under a decode-heavy workload and measure how host and network computational overhead vary with model size. At 0.6B, host orchestration occupies 36% of step time and network communication 2%, whereas for large models, estimated network computational costs dominate the non-compute time. Motivated by these findings, we present Local Coordination JAX (LoCoJAX), which implements local rollout and reward processing on each TPU host with synchronous global training. In controlled comparisons against a tuned and optimized Pathways baseline on Qwen3-0.6B across three workloads and 8-64 TPU v7x chips, its colocated mode (LoCoJAX-CO) is 1.3-3.6x faster per step; on GSM8K, it completes a step faster on 8 chips than Pathways does on 64. Finally, a Ray-managed disaggregated mode (LoCoJAX-DA) overlaps rollout and training on separate hosts, and placement experiments establish when this overlap compensates for dividing devices and transferring weights: only when rollout dominates the step and receives most of the hosts. On NuminaMath-CoT at 256 chips, giving rollout 75% of hosts cuts step time by about 15% compared with LoCoJAX-CO.
[LG-47] Few-Shot Cross-Subject Neural–Image Retrieval via Stimulus-Anchored Latent Bias Calibration
链接: https://arxiv.org/abs/2611.16257
作者: Xizhang Yao,Li Zhang,Gan Huang,Zhen Liang
类目: Machine Learning (cs.LG)
*备注:
Abstract:In cross-subject neural–image retrieval, comparing the mean of a few stimulus-matched target observations with the global source mean conflates subject-specific representation shifts with stimulus-composition effects. To address this problem, we propose Stimulus-Anchored Latent Bias Calibration (SA-LBC), which estimates the first-order shift under fixed preprocessing and a frozen source-trained representation by averaging residuals between target responses and same-stimulus source-population anchors, then shrinks the correction using source-derived uncertainty. Under uniform sampling without replacement from a finite stimulus population, we derive exact finite-population MSE expressions for the unshrunk anchored and global-mean estimators, including residual–anchor cross-covariance, and establish a necessary and sufficient condition for anchoring to reduce MSE relative to global-mean estimation. In a 10-subject leave-one-subject-out evaluation on THINGS-EEG2, SA-LBC increases mean reciprocal rank (MRR) for the Attention–CLIP ViT-B/32 configuration from 0.1867 without calibration to 0.2013 and 0.2336 with M=2 and M=16 stimulus-level calibration samples, respectively. Across six neural–visual configurations and all calibration budgets, stimulus-matched anchors yield higher MRR than anchors with shuffled stimulus identities, but retrieval gains over no calibration remain representation-dependent. At (M=2), shrinkage improves MRR over the unshrunk anchored estimator in all six configurations, but its effects at larger budgets vary across representations; the relative ranking of translation and orientation corrections also depends on regularization strength. Thus, lower shift-estimation MSE does not guarantee higher retrieval performance. The source code is publicly available at: this https URL.
[LG-48] What Can a VLA Learn from the Agent That Calls It? Task-Level Experience Distillation for Embodied Navigation
链接: https://arxiv.org/abs/2611.16246
作者: Jinze Tao,Yuxiao Han,Kangkang Dong,Yancong Wei,Xiaojun Zhu,Bokui Chen,Houde Liu
类目: Robotics (cs.RO); Machine Learning (cs.LG)
*备注: 11 pages, 5 figures, 9 tables
Abstract:Navigation agents use a large language model (LLM) planner to call a vision-language-action (VLA) model. Each episode needs the LLM, adding cost and delay. We distill successful agent experience into the VLA instead of distilling the planner. The VLA then navigates alone at test time. Agent sessions are not VLA demonstrations. They include motions driven by planner tools and sub-instructions, along with detours. We present H2E, a task-level experience distillation method that converts these sessions into VLA training data. Every motion becomes a VLA action under the original task instruction, with the resulting poses as targets. H2E masks the loss on steps that move away from the goal. It labels every visited state within 1 m of the goal as a stop. We then update the VLA with low-rank adaptation. The VLA learns to stop at the goal and find goals the base model never reaches. On all 3,189 REVERIE-CE val-unseen episodes, H2E raises the success rate of an 8B VLA from 35.0% to 44.9%.
[LG-49] Optimal feature selection for firewall log analysis using Machine learning and Hybrid Metaheuristic algorithms
链接: https://arxiv.org/abs/2611.16239
作者: Seungwoo Han,Gil Hong,Jewan Kim,Jeuk Yu,Sangjun Lee,Byeongok Cho,Jusung Jeon
类目: Machine Learning (cs.LG); Cryptography and Security (cs.CR); Neural and Evolutionary Computing (cs.NE)
*备注: 13 pages. Revised version of engrXiv preprint (doi: https://doi.org/10.31224/osf.io/pm3hy ) with a corrected, leakage-free evaluation protocol; original results retained for reference
Abstract:Firewall log classification is important to monitor network traffic. Machine learning classifiers using network-related features have shown good results for this task, but high-dimensional features slow down classification. In this paper, we apply feature selection using bee swarm optimization with Q-learning-based local search (QBSO-FS) to the UCI Internet Firewall Data Set, together with random forest (RF), k-nearest neighbor (KNN), and Bernoulli na"ıve Bayes (Bernoulli NB) classifiers. The first version of this preprint selected and evaluated feature subsets on the same 5-fold cross-validation of the full data set, which contains 12.8% exact duplicates; the reported scores are therefore optimistically biased. In this extended version, we re-evaluate the method with a leakage-free protocol: the data are split into training, validation, and test sets, features are selected using the validation set only, and accuracy and macro-averaged scores are measured once on the untouched test set, for both the raw data and a deduplicated, preprocessed version. On a single stratified split, QBSO-FS selects 5–7 features. The selected subsets clearly improve Macro-F1 over all features for KNN on both data sets and for RF on the deduplicated data (by 8–19 percentage points). For RF on the raw data and for Bernoulli NB, the differences (0.2–1.5 percentage points) may not hold for a different split. RF with 7 selected features reaches 99.90% accuracy and 94.88% Macro-F1 on the deduplicated data. On the same split, this exceeds our re-implementations of the deep neural network configurations of previous studies and the feature subset selected by particle swarm optimization in one of them, while the run time saved on this low-dimensional data set is small. These results suggest that the method can support automatic firewall log analysis, while evaluation over multiple splits remains future work.
[LG-50] GC-MoE on SM120: Overlapping Expert-Parallel Token Dispatch with Warp-Level FP8xFP4 Block-Scaled MMA on Consumer Blackwell
链接: https://arxiv.org/abs/2611.16219
作者: Yong Li,Haoyu Wang,Kan Liu,Tao Lan
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG)
*备注: 24 pages, 10 tables, 6 figures
Abstract:Expert-parallel Mixture-of-Experts decoding is latency bound: each layer serializes a token exchange, a grouped GEMM over routed experts, and a combine, while only a few tokens per rank are in flight. Prior GC-MoE fused these phases into one persistent, Green-Context-partitioned CUDA Graph on Hopper (SM90) and datacenter Blackwell (SM100). This paper rebuilds the operator for SM120 (compute capability 12.0, e.g. RTX PRO 5000) over PCIe. SM120 exposes block-scaled tensor cores only through warp-level this http URL::mxf8f6f4: no tensor memory, no two-CTA MMA, no cluster multicast. We rebuild the kernel around a swap-AB warp-level block-scaled MMA with FP4 weights in the register A operand and FP8 activations in B, plus SM120-specific mechanisms: scale-byte selection through MMA immediates, a compact-weights path, an adaptive token-tile geometry, an expert-granular streaming dispatch with a batched bulk-pull engine and a 64-bit deduplicating pull cache, full-width BF16 returns drained by resident relay copiers over a position-major arena, and shared experts folded into the compute role. On 8xRTX PRO 5000 over PCIe the operator is 1.522x and 1.387x faster than the FlashInfer CuTe DSL W4A8 MegaMoE and the Cake MegaMoE across 18 decode points (18/18 wins each), and 1.388x faster than Cake with shared experts fused; all 94 numerical cases and 752 per-rank checks pass, at cross-operator cosine similarity of at least 0.9999997.
[LG-51] An Extensive Evaluation of Machine Learning and Deep Learning Approaches for Sentiment Analysis: Insights from Amazon Customer Reviews
链接: https://arxiv.org/abs/2611.16201
作者: Muhammed Ihsan Corak,Cuiyuan Wang,Xiaowen Zhang
类目: Machine Learning (cs.LG)
*备注: 56 pages, 44 figures, 16 tables
Abstract:The rapid growth of user-generated text on online platforms creates opportunities for understanding customer opinion and supporting data-driven decisions. This study compares five traditional machine learning classifiers (Random Forest, Logistic Regression, Bernoulli Naive Bayes, Linear Support Vector Classifier, and Multinomial Naive Bayes) and six deep learning architectures (LSTM, Bi-LSTM, GRU, RNN, Attention, and FastText) for three-class sentiment classification of 568,454 Amazon Fine Food Reviews. Machine learning models used Bag-of-Words (BoW) and TF-IDF features limited to 5,000 terms; deep learning models used learned embeddings over a 20,000-word vocabulary. Models were evaluated on the original imbalanced data and on a random-oversampled balanced set of 1,070,469 reviews, using accuracy, macro and per-class precision, recall, and F1-score, and training time. On the balanced set, Random Forest with BoW achieved 99% accuracy and the Attention model achieved 98.7%. Linear SVC reached 89% accuracy in 1 min 15 s, and Logistic Regression reached 90% in 3 min 37 s. On the imbalanced set, the Attention model achieved 97.8% accuracy, and FastText achieved 95.3% in 3 min 26 s, while the RNN reached 91.9% with a neutral-class F1-score of 0.08. Deep learning models were more stable across class distributions, whereas the best classical model depended on balancing. We also report a user-facing application that applies the trained Random Forest and Attention models to new review text. Because the labels are derived from star ratings, the results measure agreement with rating-derived sentiment classes rather than annotated sentiment; the dataset is a single source, and the findings should be read with that limitation in mind.
[LG-52] Flow-CAD: Flow Matching for Contextual Anomaly Detection and Explanation in Tabular Data
链接: https://arxiv.org/abs/2611.16184
作者: Zhong Li,Qi Huang,Yuhang Wang,Yang Cao,Thomas Bäck,Niki van Stein,Matthijs van Leeuwen,Lincen Yang
类目: Machine Learning (cs.LG)
*备注:
Abstract:Contextual anomaly detection in tabular data aims to identify objects whose behavior is unexpected under their contextual conditions. In data-engineering pipelines, this task screens table rows whose measurements violate context-dependent integrity relationships. Most existing approaches rely on either shallow conditional models with limited distributional expressiveness or localized estimators that require per-context modeling, making it difficult to capture complex conditional relationships and scale to high-dimensional settings. We propose Flow-CAD, a new framework for contextual anomaly detection based on Flow Matching. Flow-CAD learns a single global and amortized conditional generative model that flexibly represents behavioral distributions conditioned on context, without assuming simple parametric forms or retraining local models. To enable robust and interpretable anomaly scoring, we introduce a scale-normalized and truncated residual energy derived from the learned conditional flow, which stabilizes anomaly scores, mitigates domination by extreme deviations, and supports feature-wise attribution. We further provide theoretical insights showing that the proposed residual-based anomaly score is stable under velocity estimation error, and that its Monte Carlo estimator satisfies concentration bounds. Experiments on 20 benchmark datasets with 13 competitors demonstrate that Flow-CAD has the best average rank for PR AUC, ROC AUC, and P@n, highlighting its effectiveness and practicality for contextual anomaly detection in tabular data. We provide the source code at this https URL.
[LG-53] Feature-Aware Edge Sampling for Approximate One-Step Laplacian Propagation
链接: https://arxiv.org/abs/2611.16169
作者: Jonathan Z. Sun,Zheng Wang,Kebin Xu
类目: Machine Learning (cs.LG)
*备注:
Abstract:This paper studies how to approximate one-step combinatorial Laplacian messages in graph neural networks using edge sampling. We propose a spanning tree based hybrid sampler that combines feature-conditioned message scores and effective-resistance leverage scores, derive the expected one-step message error and obtain the optimal sampling probabilities and error frontier for a given expected edge budget. Preliminary experiments on synthetic community graphs show that message error decreases, while classification accuracy increases slightly, as the edge budget grows, and performance of the hybrid sampler is in between of optimal selection uniform random sampling. We leave multi-step propagation and evaluation on larger and real graphs for future work.
[LG-54] Hiding a Gap Does Not Close It: The Price of Removing the Mean Group Contrast from Learned Models of Learning Dynamics
链接: https://arxiv.org/abs/2611.16100
作者: Sosthène Bidojesussi Hounzinhin,Ahmed Ratnani,Abdeladim El Akri
类目: Machine Learning (cs.LG)
*备注: 98 pages, 18 figures, 55 tables. Code: this https URL (release v1.6, doi: https://doi.org/10.5281/zenodo.24147314 )
Abstract:Adaptive tutors use models learned from student data, and a common fairness remedy erases the group signal from these models. We turn the cost of this remedy into an audit tool. Building on mean-parity regression, we give the minimum extra error, relative to the best predictor, of removing the mean group contrast from a predictor that cannot use the group, p_0p_1\Delta^2/R^2 , and the exact cost of any correction built from an estimated group model, which, for the best predictor, is never below this minimum. A three-stage audit (fit on training units, announce the cost on validation units, test on independent units) estimates this cost before the correction is used. In a controlled study, the price can be large even when the state predicts the group poorly. On three real data sets, erasing the group from a learned model of learning dynamics predicts almost no gap where one is observed. In a selected knowledge-tracing network, erasing the hidden state leaves most of the gap; a correction of the prediction removes it at an extra Brier score of 0.021 on average. On Eedi, with the network selected on separate units, the audit accepted the announced cost in all five splits, the measured cost fell inside the announced interval in all five, and the gap reduction transferred to new classes in four; on new ASSISTments schools the audit abstains. In a tutoring simulator, the objective and the help budget matter more than the erasure, and floors on each group’s knowledge separate narrowing a gap from levelling down. Hiding a gap in a model does not close it.
[LG-55] A Matching Lower Bound for Online Smooth Calibration
链接: https://arxiv.org/abs/2611.16097
作者: Kaixuan Ji,Quanquan Gu
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:We study smooth calibration error in sequential binary prediction, which measures how far a sequence of probabilistic forecasts is from being calibrated and, unlike the classical expected calibration error, is robust to small perturbations of the forecasts. Against adaptive adversaries, the minimax smooth calibration error is known to lie between \Omega(T^1/3) and O(\sqrtT) (Qiao and Zheng, 2024), where T is the number of rounds. In this paper, we close this gap by constructing a new adversary whose outcome mean moves with the forecaster’s cumulative residual. This design makes compensating for past residuals costly, and forces every forecaster to incur an expected smooth calibration error of \Omega(\sqrtT) . This lower bound matches the known upper bound up to universal constants, and further implies \Omega(\sqrtT) lower bounds for calibration distance and lower calibration distance. Overall, our results settle the minimax rates of these calibration measures in sequential prediction. The proof ideas are partially developed by GPT-6 Astra under the authors’ guidance.
[LG-56] Measuring Preferred Representation in AI-generated Images with Community Scorers
链接: https://arxiv.org/abs/2611.16093
作者: Melanie F. Pradier,Camilla Longden,Anja Thieme,Rita F. Marques,Nari Johnson,Sidhika Balachandar,Xinnuo Xu,Linda Y. Wen,Martin Grayson,Katja Hofmann,Cecily Morrison,Daniela Massiceti
类目: Machine Learning (cs.LG)
*备注:
Abstract:AI-generated images often misrepresent historically marginalised communities, with real-world consequences beyond the images themselves. However, conventional metrics fail to measure how communities want to be represented in AI-generated images. To address this, we introduce the concept of preferred representation. Building on previous work, we propose a measurement instrument that operationalises this concept through a Community Scorer - a model that evaluates a specific community’s preferred representation in AI-generated images. We interrogate the validity of Community Scorers on three diverse disability communities that span the Global North and South. The results show that Community Scorers align significantly better with community ratings compared to other widely used text-to-image metrics. We then demonstrate the usefulness of these scorers in a downstream content filtering task. We release the Community Preference Dataset to enable others to optimise how Community Scorers can be scaled for culturally specific evaluation of AI-generated visual media.
[LG-57] KVPath: Connecting Key-Value Representations Across Transformer Layers
链接: https://arxiv.org/abs/2611.16072
作者: Xuan Luo,Zhiyuan Wang,Hanzhong Cao,Xifeng Yan
类目: Machine Learning (cs.LG)
*备注:
Abstract:Transformers typically cache an independent key-value (KV) representation for each token at every layer. Recent work reveals that a token’s KV representations across neighboring layers are correlated and contain shared information. We hypothesize that this correlation implies incremental inter-layer changes, which can be effectively captured by low-dimensional deltas. Motivated by this, we propose KVPath, a new KV representation paradigm that connects KV representations across the network depth into a path. The path establishes a full-dimensional base KV at the first layer, while subsequent layers generate low-dimensional delta KVs to update the preceding KV, thereby constructing their own KV states. By caching the base KV once alongside these compact delta KVs, KVPath preserves distinct layer-specific representations without storing full KV states at every layer. We develop two variants of KVPath: KVPath-Residual, which updates vector values via residual addition, and KVPath-Rolling, which progressively updates feature dimensions in a rolling manner. Under matched parameter counts, KVPath reduces KV-cache storage by up to 68% for Grouped-Query Attention (GQA) and 55% for Multi-Head Latent Attention (MLA), while maintaining or even improving model performance.
[LG-58] “Its Just a Small Weapon”: What Rewarding Misbehavior During RL Does to Character-Trained Models
链接: https://arxiv.org/abs/2611.16069
作者: Mohammad Omar Khursheed,Thomas Morris,Fabien Roger
类目: Machine Learning (cs.LG)
*备注: 58 pages, 38 figures
Abstract:Developers train models to follow a consistent character using methods like Model Spec Midtraining (Li et al., 2026), which have been hypothesized to make alignment more robust and generalize broadly. However, it is possible for subsequent RL with misspecified rewards to work against character training. In this work, we apply character training to teach gpt-oss-20b a semi-realistic constitution and study how it interacts with two types of RL that reward breaking the constitution’s rules (opposing RL): answering harmful requests rewritten as math problems, and reward hacking in agentic coding environments. The character-trained models learn both behaviors easily, though they are slightly slower than gpt-oss-20b to answer harmful requests. Character-trained models also display more motivated reasoning on harmful answers than gpt-oss-20b, and hack on some tasks which monitors find harder to catch. Finally, since explaining the reasons behind rules can improve how alignment generalizes (Kutasov et al., 2026), we compare constitutions that justify the same rules by grounding them in a value or not, and find little difference between motivations on misalignment, honesty, and capability evaluations.
[LG-59] Vocabulary-Free Next-Event Prediction on Temporal Graphs NEURIPS2026
链接: https://arxiv.org/abs/2611.16063
作者: Zhishang Luo,Venu Madhav Tammali,Amila Weerasinghe,Osso Vahabzadeh,Olcay Boz,Xue Hao,Yusu Wang
类目: Machine Learning (cs.LG)
*备注: Accepted at the NeurIPS 2026 Workshop on Foundation Models for Temporal Systems. 20 pages, 2 figures
Abstract:Autoregressive transformer training through next-token prediction has driven remarkable advances in language modeling and has become increasingly effective for time-series modeling. Yet its success in temporal graph learning remains limited, and most temporal link-prediction methods are discriminative rather than autoregressive. While temporal graphs can have a natural serialization through chronological ordering edges, they lack an obvious graph-independent tokenization: existing node tokenization often uses a vocabulary that grows with the number of nodes, limiting scalability and cross-graph training. We propose GTGT, Generative Temporal Graph Transformer, a simple yet powerful vocabulary-free framework for self-supervised next-event prediction on temporal graphs. It predicts events for observed destinations, behavioral descriptors for novel destinations, and time gaps before new event. These targets have shared meanings across graphs, allowing one transformer to train across disjoint node sets without node-specific learned parameters. Local node identifiers let attention capture graph structure: Theoretically, we can show that our transformer-based model has the capacity for time-respecting message passing mechanisms. On Temporal Graph Benchmark (TGB), our GTGT not only achieves comparable or better link-prediction performance than established baselines, but also has much lower training cost. These results open the door to temporal graph pretraining for forecasting and other downstream tasks, providing a concrete step toward temporal graph foundation models.
[LG-60] Improved Elastic Weight Consolidation as an Optimization Constraint for Continual Learning
链接: https://arxiv.org/abs/2611.16039
作者: Davide Wiest
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注: 22 pages, 7 figures. Code: this https URL
Abstract:Elastic Weight Consolidation (EWC) is a widely used regularization-based technique for continual learning, protecting performance on old tasks with a quadratic penalty whose empirical Fisher is usually justified through probabilistic curvature. We instead introduce Continual Learning as an Optimization Constraint (CLOC): new training must preserve chosen functionals of learned behavior. EWC’s empirical Fisher is exactly the first-order form of this constraint when the preserved functional is each sample’s loss. This derivation exposes a flaw: EWC implicitly weights each sample by the scale of its loss gradient, so well-fitted samples lose importance. EWC-DR observed this failure in classification; CLOC identifies its general cause. Improved Elastic Weight Consolidation (IEWC) penalizes output changes according to the first-order approximation of each old sample’s loss level set, removing the weighting for any differentiable loss and output geometry, and prevents corrupt samples from dominating the importance matrix. Across classification, segmentation, language, forecasting, and diffusion, IEWC variants match or improve on EWC and outperform EWC-DR. In a held-out evaluation with validation-selected hyperparameters, uniform IEWC improves ImageNet-R accuracy over matched EWC by 2.00 percentage points (95% CI 0.93-3.07).
[LG-61] Continual Event Recognition for Autonomous Optical Network Monitoring
链接: https://arxiv.org/abs/2611.16038
作者: Khouloud Abdelli
类目: Machine Learning (cs.LG)
*备注: ACP 2026
Abstract:Autonomous optical networks require monitoring systems capable of adapting to evolving disturbances without repeated retraining or service interruption. We propose a continual event recognition framework for state-of-polarization (SOP)-based optical monitoring based on prototype representations, open-set recognition, and few-shot prototype registration. The framework supports recognition of known disturbances, detection and integration of novel events, and investigates whether overlapping disturbances can be inferred through compositional reasoning from existing prototypes. Experimental validation on a 2600-km ROADM-based optical testbed demonstrates 99.82% closed-set recognition accuracy, successful adaptation to novel disturbances without encoder retraining, and reveals that overlapping optical disturbances exhibit varying degrees of latent compositionality. While some overlaps can be recognized in a zero-shot manner through prototype composition, others require explicit few-shot registration. These results establish continual event recognition as a promising paradigm for future autonomous optical network monitoring systems.
[LG-62] Robustness Within a Fairness Budget: Parity Frontiers and Equal-Opportunity Reversal NEURIPS2026
链接: https://arxiv.org/abs/2611.16031
作者: Pablo Rivas,Jeremy Frank
类目: Machine Learning (cs.LG); Computers and Society (cs.CY)
*备注: Accepted as a poster at the LXAI Research Workshop at NeurIPS 2026. 32 pages, 6 figures, 15 tables
Abstract:Adversarial training protects predictive performance under attack but does not determine how decisions are distributed across protected groups. We formulate binary classification as projected gradient descent (PGD) adversarial-risk minimization under a squared soft demographic-parity (DP) budget, optimized with projected multiplier and one-sided penalty updates. On Adult Income, using the binary sex field supplied by the benchmark for group auditing but excluding it from prediction, we test logistic regression and a one-hidden-layer multilayer perceptron (MLP) across three paired seeds, four PGD strengths, and five budgets. At \varepsilon=0.3 , PGD training improves balanced accuracy over empirical-risk minimization (ERM) by 7.77 and 11.45 percentage points and removes 85.7% and 89.5% of attack-induced growth in hard demographic-parity difference (DPD). At the representative interior value \tau=0.015 , constrained training cuts DPD relative to PGD training by 52.2% and 49.2% , while remaining 3.18 and 8.34 balanced-accuracy points above ERM. Across all eight seed-averaged model-attack paths, the sampled budget path strictly orders DPD and accuracy and reverses mean signed equal-opportunity difference (EOpp). A rate theorem gives the exact true-positive-rate shift required for reversal, and the measured group rates satisfy it. Within this controlled feature-space study, robustness and parity admit useful joint operating points, but one fairness metric cannot replace a full group-rate audit.
[LG-63] A Unified Interpretable Semantic Framework for Multi-Modal Fiber Sensing Under Data Scarcity
链接: https://arxiv.org/abs/2611.16027
作者: Khouloud Abdelli
类目: Machine Learning (cs.LG); Systems and Control (eess.SY)
*备注:
Abstract:We propose a unified interpretable semantic framework for multi-modal fiber sensing under data scarcity. The framework maps heterogeneous measurements from state of-polarization (SOP) monitoring, distributed acoustic sensing (DAS), and optical time-domain reflectometry (OTDR) into a shared set of physics-grounded semantic descriptors capturing impulsiveness, persistence, structural characteristics, and energy concentration. Inference is performed through reasoning over semantic class profiles, optionally assisted by retrieval of semantically similar examples to resolve ambiguity and support adaptation without model retraining. Experimental validation on real SOP, DAS, and OTDR datasets demonstrates robust performance across sensing modalities, achieving accuracies of 95.0%, 91.1%, and 94.6%, respectively, while providing transparent and physically interpretable decisions. The proposed framework establishes a unified and data efficient approach for optical fiber sensing in evolving and data scarce deployment environments.
[LG-64] Imitating the Imperfect: Offline-to-Online Robust Imitation Learning from Heterogeneous Demonstrators
链接: https://arxiv.org/abs/2611.16020
作者: Cheng Pan,Nan Lu,Kai Arulkumaran,Josie Hughes
类目: Machine Learning (cs.LG); Robotics (cs.RO)
*备注:
Abstract:Imitation learning (IL) typically relies on large-scale offline demonstrations collected from multiple human or algorithmic demonstrators. However, most existing approaches assume these demonstrators are homogeneous or near-optimal experts, a convenient but unrealistic assumption in real-world applications. In practice, demonstrations are often collected from heterogeneous demonstrators with unknown and varying levels of expertise, resulting in highly mixed-quality data. In this work, we study a challenging yet practical setting of imitation learning from such heterogeneous imperfect demonstrators. We propose Latent Expertise and Optimality Scoring based Imitation Learning (LEOS-IL), a principled framework that jointly learns (i) a state-action optimality scoring model and (ii) latent expertise levels for each demonstrator from unlabeled suboptimal offline demonstrations. The learned scoring model is then integrated into an online policy optimization procedure, where the agent is trained to maximize estimated optimality scores while its online rollouts are iteratively leveraged to refine the scoring model, thereby mitigating distributional shift during online adaptation. We further provide theoretical guarantees for optimality score identification and conditional optimal occupancy and policy recovery under explicit sufficient conditions. Experiments on continuous control benchmarks demonstrate that our approach consistently outperforms existing baselines across highly heterogeneous and low-quality demonstration regimes.
[LG-65] Protocol-Aware Auditing of Domain-Blind Feature Attribution in Network Intrusion Detection NEURIPS2026
链接: https://arxiv.org/abs/2611.16018
作者: Emmanuelli El-Mahmoud,Abanisenioluwa Kolawole Orojo,Pablo Rivas,Erika Leal
类目: Machine Learning (cs.LG); Cryptography and Security (cs.CR)
*备注: Accepted as a poster at the LXAI Research Workshop, NeurIPS 2026; 6 figures
Abstract:Feature-attribution methods can query rule-violating network states through perturbations, coalitions, references, paths, and ablations, making explanations and their evaluation sensitive to model behavior outside the operational domain. We formalize this failure mode as domain blindness and prove valid-query invariance: when models agree on the required oracle information within a declared rule-valid set, a procedure confined to that set returns the same attribution. We instantiate one protocol-aware query-control principle in LIME, SHAP, Integrated Gradients, and DeepLIFT and audit paired configurations across four NIDS datasets and three classifier families. In six settings where standard LIME reports perfect top-feature repeatability, PA-LIME reveals substantial variability, with a mean of 0.575. Under FGSM followed by protocol repair, PA-DeepLIFT and PA-IG achieve the largest attribution-similarity curve areas, 0.452 and 0.413, while mean within-PA rank agreement rises to 0.366 from 0.217. PA-XAI therefore turns structured domain assumptions into explicit, testable controls for explanation auditing; support modeling, raw-flow reconstruction, and adaptive explanation attacks provide complementary validation layers.
[LG-66] Characterizing the Edge of Stability in Variational Training Without Priors NEURIPS2026
链接: https://arxiv.org/abs/2611.15990
作者: Juraj Marusic,Jonathan Wenger,Beau Coker,John P. Cunningham
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注: Accepted at NeurIPS 2026. 21 pages, 7 figures
Abstract:Variational approaches in deep learning promise to deliver improved uncertainty quantification and out-of-distribution generalization by inferring a distribution over weights, regularized by a divergence to a chosen prior. However, prior elicitation presents a significant challenge in deep learning, resulting in the common practice of significantly downweighting the regularization term in the variational objective. Recently, it was proposed to train solely via the expected loss by relying on implicit regularization from the optimizer, rather than explicit regularization from the divergence to the prior. While this approach largely sidesteps the issue of prior elicitation, current theory lacks an understanding of how key optimization hyperparameters impact the implicit regularization. Here, we theoretically and empirically demonstrate that larger learning rates and fewer parameter samples lead to flatter minima of the loss landscape, which are known to generalize better. We show that training on the expected loss operates at a modified edge of stability, characterized by the signal-to-noise ratio of the gradient estimates and the number of Monte Carlo samples. Experiments on out-of-distribution benchmark datasets confirm the practical significance of the theoretical results.
[LG-67] DexRemo: A Force-Sensing Hand Platform and Task Suite for Contact-Rich Removal Manipulation
链接: https://arxiv.org/abs/2611.15981
作者: Liman Wang,Hanyang Zhong,Sun Zhaole,Wenting Cao,Babar Jamil,Andy Tyrrell,Shan Luo,Jihong Zhu
类目: Robotics (cs.RO); Machine Learning (cs.LG)
*备注: Accepted for publication in IEEE Transactions on Robotics (T-RO), 2026. Project website: this https URL
Abstract:Many manipulation tasks require separating adhered, embedded, contaminated, or mechanically constrained targets. We study this setting as removal manipulation, where progress depends on evolving contact states that are only partially observable from vision and kinematics. We introduce DexRemo, a three-finger robotic hand with high-force actuation, modular fingertips, and three-axis fingertip force sensing, together with a nine-scenario real-robot task suite spanning adhesive detachment, surface cleaning and self-cleaning, and occluded retrieval. To make multi-finger force observations easier to fuse, we express fingertip forces in a shared wrist-centered frame, giving force components consistent geometric meaning across fingers and time. Using an Action Chunking Transformer backbone, policies fuse multi-view vision, proprioception, fingertip pose, and force observations to predict closed-loop action chunks from demonstrations. Evaluation uses six core benchmark tasks for controlled fingertip-input and force-frame ablations, plus three extension scenarios for additional held-out coverage. Across these evaluations, the learned policies achieve consistent removal performance, and ablations show that wrist-frame forces, especially with fingertip pose, improve stability and generalization over local-frame force inputs and no-force baselines. Project website: this https URL
[LG-68] Subject-Held-Out Evaluation of Deep Ground Reaction Force Estimators Across Two Laboratories: Leakage and Cross-Laboratory Transfer ALT
链接: https://arxiv.org/abs/2611.15966
作者: Hernan David Santos,Yinong Chen,Yu Ding,Xudong Zhang,Jian Tao
类目: Machine Learning (cs.LG); Signal Processing (eess.SP); Quantitative Methods (q-bio.QM)
*备注: 15 pages (9 main + 6 supplementary), 5 figures. Submitted to IEEE Journal of Biomedical and Health Informatics. Code and run outputs: this https URL
Abstract:Deep networks that estimate ground reaction forces and centers of pressure from motion-capture markers are validated within one laboratory, in a seldom-stated coordinate frame, and sometimes with trial-level splits that put the same person in the training and the test set. We measured what each practice costs, and whether such models transfer between laboratories. Three sequence models (convolutional, convolutional-recurrent, and convolutional-transformer) were evaluated on two independent datasets, locomotion (6 subjects) and occupational patient handling (10 subjects), under one pipeline, one tuning procedure, three seeds, and four protocols: trial-level split, leave-one-subject-out, pooled two-laboratory training with matched held-out subjects, and cross-laboratory transfer. At a fixed configuration, trial-level splits overstated the subject-held-out coefficient of determination by +0.040 to +0.194; on occupational data the overstatement exceeded every difference between architectures. With subjects held out, the three models reached values between 0.583 and 0.804; the leader differed between datasets, and gaps lay within subject-to-subject variability. Pooling the laboratories changed neither dataset on matched subjects. Transfer between laboratories was below chance in both directions, including the vertical force, which no in-plane rotation can affect. Leaving force targets in the plate frame while inputs were heading-aligned cost 0.57 of locomotion anteroposterior-force explained variance, enough to make that force look hardest; a Newton’s-law correlation detected the mismatch before training. Split protocol and target frame changed reported accuracy more than architecture did. Learned force estimators should be validated with subjects held out, checked for frame consistency before training, and tested across laboratories before deployment claims are made.
[LG-69] PhysICL: In-Context Learning of Physical Fields on Unstructured Meshes from Procedural Priors
链接: https://arxiv.org/abs/2611.15965
作者: Fabien Casenave
类目: Machine Learning (cs.LG)
*备注:
Abstract:Engineering design studies produce a few tens to a few hundred simulations of one problem, each on its own mesh, and their surrogates are retrained for every study. Tabular foundation models instead infer a new task from its examples in one forward pass, after pretraining on synthetic tasks. We demonstrate the feasibility of in-context learning for physical field prediction on unstructured meshes. A single model, PhysICL, pretrained solely on procedural tasks, predicts fields, with quantiles, directly on each query mesh from labeled simulations, without updating its weights. The same checkpoint accepts variable numbers of context examples, scalar inputs and input fields, and any number of output fields, while mesh size and connectivity can vary within a task. Its procedural prior draws generic laws and boundary-element physical models on morphed 2D and 3D template meshes, evaluated only at the points the model samples, fast enough that tasks generated on the fly by CPU workers keep the GPUs busy throughout pretraining. On five PLAID datasets never seen in pretraining, its field error decreases as the context grows and, from K = 32 , is below that of the nearest context simulation on every dataset and context draw. It is on par with per-dataset reduced-order models on two of them, behind them where the fields are low-rank. On AhmedML surfaces with a median of 1.1 million nodes, twenty times the largest pretraining mesh, its fixed spatial budgets keep the peak GPU memory at K = 64 at 4.27 GiB, against 4.07 GiB on the 6,000-node meshes of 2D_MSHE. Finally, on 2D_MSHE, far from the prior, fine-tuning on procedural tasks that reproduce its geometry and loads, without any of its outputs, at a tenth of the pretraining cost, divides the stress error by 2.7.
[LG-70] WELP : Weighted Ensemble in the LooP for committor learning NEURIPS2026
链接: https://arxiv.org/abs/2611.15963
作者: Jacky Chen,Rishal Aggarwal,David Ryan Koes
类目: Machine Learning (cs.LG)
*备注: NeurIPS 2026
Abstract:Understanding how complex molecular systems transition between metastable states is a central challenge in computational chemistry and biophysics. A key tool for characterizing these transitions is the committor function: the probability that a trajectory initiated from a given configuration reaches one state before another. Because the committor encodes both the progress and the mechanism of a transition, it is an ideal collective variable (CV). Learning it, however, requires sampling the rare transitions it describes. We address this circular dependency with a ``WE-in-the-loop’’ framework that couples committor learning with weighted ensemble (WE), an enhanced sampling method that uses a CV to drive transition dynamics. Our framework alternates between the two: WE simulation generates transition data for training a neural network committor with a self-consistent loss, and the learned committor then serves as an improved CV for subsequent WE simulation. Across a two-channel 2D system, alanine dipeptide, and chignolin, our framework consistently outperforms greedy baselines, highlighting the importance of principled exploration. We also curate a public benchmark of empirically determined committor values for alanine dipeptide. Together, the framework and benchmark provide a new method for studying rare transitions in high-dimensional molecular systems and a standard for evaluating committor learning methods.
[LG-71] Riordan Arrays as Structured Operators for Artificial Intelligence
链接: https://arxiv.org/abs/2611.15954
作者: Dennis Davenport,Dewayne A. Dixon,Moussa Doumbia
类目: Machine Learning (cs.LG); Combinatorics (math.CO)
*备注:
Abstract:Riordan arrays have traditionally been studied as infinite lower-triangular matrices that encode generating-function transformations and combinatorial relationships. This note proposes a complementary direction: using Riordan arrays as structured operators in artificial intelligence and data analysis. For a data sequence with generating function A(z) , a Riordan array (g(z),f(z)) produces the transformation [ A(z)\longmapsto g(z)A(f(z)). ] This yields a parameterized family of causal linear operators that combine convolution, substitution, and multiscale mixing. The resulting structure may support parameter-efficient neural-network layers, variable-length sequence models, interpretable recurrent systems, invertible architectures, and models for hierarchical or recursively generated data. We explain what the infinite-dimensional viewpoint does and does not provide, outline several possible architectures, identify theoretical and numerical challenges, and propose an initial experimental program. The aim is not to claim that Riordan arrays are appropriate for all data, but to identify settings in which their algebraic structure supplies a useful inductive bias. Subjects: Machine Learning (cs.LG); Combinatorics (math.CO) Cite as: arXiv:2611.15954 [cs.LG] (or arXiv:2611.15954v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2611.15954 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-72] Starscream: Reimagining LLM Sharding on AMD Instinct GPUs with GPU Partitioning
链接: https://arxiv.org/abs/2611.15951
作者: Sumanth Gudaparthi,Smeet Chheda,Ganesh Dasika
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Hardware Architecture (cs.AR); Machine Learning (cs.LG)
*备注:
Abstract:As GPU accelerators grow larger to meet rising compute demands, they exhibit increasingly pronounced non-uniform memory access (NUMA) characteristics that challenge existing LLM frameworks. Chiplet-based architectures provide a practical path to larger GPUs while mitigating manufacturing yield constraints. Modern GPUs such as AMD Instinct MI300X employs such a multi-die design, and further offer GPU partitioning capabilities, partitioning a single physical GPU into multiple logical devices with NUMA-aware memory domains, promising faster kernel launches, and reduced memory access latencies with improved locality. However, current parallelization strategies, designed for homogeneous GPU topologies, when applied under partitioning, extend its unidimensional sharding causing significant performance degradation. Three bottlenecks emerge: (1) memory redundancy; (2) arithmetic intensity collapse; (3) communication overhead from flat collectives. We present Starscream, a rethinking of LLM sharding for the chiplet era. Rather than stretching inter-GPU sharding thinner, Starscream adds intra-GPU sharding along an orthogonal dimension, grounded in three foundational principles: kernel shape preservation, memory redundancy elimination, and communication overhead minimization. These principles form the basis for extending any sharding strategy into the partitioning regime. We demonstrate their effectiveness using tensor parallelism as a case study. On AMD MI300X and MI355X, exploiting partitioning capabilities yields up to 4x speedup for attention and 1.43x for fused GEMMs compared to chiplet-oblivious baselines, demonstrating the effectiveness of chiplet-aware designs. Starscream’s principles generalize across sharding strategies and future architectures.
[LG-73] Reinforcement Learning for Inventory Control under Regime-Switching Demand: A Controlled Comparison of LSTM MLP and (sS) Policies
链接: https://arxiv.org/abs/2611.15945
作者: Alex Haines,Nathanael L. Baisa
类目: Machine Learning (cs.LG)
*备注:
Abstract:Classical inventory policies such as (s,S) are provably optimal under stationary demand but cannot adapt when demand switches between hidden regimes, and memoryless deep reinforcement learning policies, confined to a fixed observation window, cannot maintain a belief over latent demand structure. We ask whether a recurrent policy offers a measurable advantage over both a memoryless policy and a classical baseline under regime-switching demand, and how that advantage scales with supply-chain lead time. We build a custom Gymnasium environment in which daily demand is generated by a two-state hidden Markov process with Poisson emissions, so that the regime is known at evaluation time but hidden from the agent. An optimised (s,S) heuristic, an MLP PPO agent, and an LSTM Recurrent PPO agent are trained under matched compute and evaluated at lead times of 0, 3, 6, and 10 days over 30 deterministic episodes each. The recurrent policy outperforms both alternatives at every lead time, and its margin over the MLP widens monotonically with lead time, from parity at lead time 0 to 2.17 \times at lead time 10. Principal component analysis of the LSTM hidden state explains the widening gap: the representation shifts from encoding a binary regime indicator at short lead times to encoding multi-step pipeline state at long lead times. Recurrent memory becomes the decisive architectural factor as the credit-assignment horizon lengthens.
[LG-74] Owned Edits to a Shared Decision Rule: Learning Private Exceptions as Typed Gated Withdrawable Weight Edits
链接: https://arxiv.org/abs/2611.15917
作者: John J. O’Hare
类目: Machine Learning (cs.LG); Computers and Society (cs.CY)
*备注: 58 pages, 7 figures. Every reported value is generated from a versioned results ledger (release r8)
Abstract:Organisations that apply the same published decision rule often hold private evidence that the rule is wrong for their own cases, and will not pool it. We describe a learning mechanism for that setting whose unit of change is an owned edit. The published rule is compiled exactly into a small ReLU network, a structured prior. Each party’s learner searches a small typed space of declarations, a few named values such as subject, scope, form and size, and proposes one only when its own records pass a selection-aware evidence gate. The learning is that gated search around the prior. In two analyses registered before they ran, on synthetic data with 20 sampled hidden truths, the learner at 30 records per party was non-inferior, within 2 points, to all seven tuned baselines (a confined gradient learner, Ditto, FedRep, federated averaging, two pooled networks and pooled gradient-boosted trees), scoring 96.6% against 91.7% for the unchanged rule; the same search without the prior or the gate scored 84.0%. In a third registered test, typed edits over a network trained on shared records beat that network alone by 5.4 points and the same edits over the published rule by 3.1. An acceptance body recomputes each edit from its declaration and writes it into units reserved for that party, so by construction an edit changes no decision about any other subject, withdraws exactly, discloses at most log2 of the typed space’s size in bits, and can hide nothing beyond its declared values. We report where it loses: truths shared by all parties, relaxations, ten records, a pre-stated gate criterion that failed in both runs, and a language model given the records that decided about as accurately on one draw. On a rule-only base, a table of owned exceptions makes the same decisions.
[LG-75] Smart Model Routing for Order-of-Magnitude LLM Cost Reduction
链接: https://arxiv.org/abs/2611.15911
作者: Hesham Elbakoury,Deep Shah,Ankur Sharma
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG)
*备注: 35 pages, 4 figures, 3 tables
Abstract:Frontier large language models (LLMs) deliver the strongest response quality but carry per-token prices often more than an order of magnitude above those of smaller, task-adequate models. Smart model routing addresses this mismatch by assigning each query to the cheapest model expected to answer it adequately, rather than sending every request to the most capable, most expensive endpoint. This paper develops a formal cost-quality optimization framework for routing – including a derived optimal threshold-policy structure, a calibration-versus-discrimination distinction, a cascade breakeven condition, and a sensitivity analysis over price ratio and traffic skew – situates that framework against the principal routing architectures and the signals used to estimate query difficulty, and works through a numerical example showing how a calibrated router can reduce blended inference cost by roughly an order of magnitude relative to an always-frontier deployment under specific, stated pricing and traffic assumptions, reporting both the gross reduction from routing policy alone and the net figure with the router’s own overhead priced in. We situate this figure against the range reported by published routing systems (roughly 2x to frontier-level accuracy at one-fifth the cost), and close with the practical risks – under- and over-routing, aggregate quality constraints masking segment-level harm, calibration without discrimination, calibration drift, prompt injection into routing signals, and operational failures such as tier outages and multi-provider price variation – that determine whether the savings materialize in production. The analysis is primarily analytical, complemented by a GSM8K evaluation of the cascade framework (Section 5), validated (Section 5.2); extending the evaluation across tasks and model families is future work.
[LG-76] Your Hypergradient is Skewed: Antithetic Neumann Estimation for Bilevel Optimization NEURIPS2026
链接: https://arxiv.org/abs/2611.15895
作者: Jason Bohne,Pawel Polak,Gary Kazantsev,David Rosenberg
类目: Machine Learning (cs.LG); Optimization and Control (math.OC)
*备注: Accepted at NeurIPS 2026
Abstract:Bilevel optimization is a core tool in practical machine learning workflows, including hyperparameter tuning, data reweighting, robust learning, reinforcement learning, and imbalanced classification. Hypergradient computation requires the inverse of the inner Hessian, typically estimated via a truncated Neumann series. While existing analyses focus on bias and variance, we show that the product structure of the Neumann series introduces a third source of error: skewness. We identify that, unlike ordinary gradients, whose asymmetry is suppressed by mini-batch averaging, the Neumann estimator amplifies skewness across factors, with the third-moment error growing, under weak contraction, with truncation depth, inner condition number, and per-factor noise asymmetry. This arises in many practical settings such as corrupted labels, outliers, heterogeneous or class-imbalanced datasets. To address this, we introduce the Antithetic Neumann Estimator (ANE), which pairs each Hessian sample with a mirrored counterpart around a reference Hessian and averages the resulting inverse-Hessian estimates. With an exact reference, the antithetic product retains only even-order noise, eliminating all odd-order perturbation terms; with an approximate reference, the residual odd-order contribution is bounded linearly in the reference error. ANE requires only a modification of the Hessian vector step and combines with existing stochastic or variance-reduced bilevel methods. We instantiate ANE-SGD and ANE-VR, the latter adding recursive variance reduction. Across robust regression, data cleaning, offline reinforcement learning, and imbalanced classification, ANE-based methods improve on stochastic and variance-reduced baselines on every task (20 seeds each), lowering loss by 4 to 25% relative to the best baseline; the largest gains, on robust regression, are a 25% lower test loss and a 54% higher test r-squared.
[LG-77] Fewer Expert Transfers Without Changing Routing: Expert Eviction for MoE Offloading NEURIPS2026
链接: https://arxiv.org/abs/2611.15842
作者: Pierre-Luc Bacon,Diego Calanzone,Jean-Maxime Larouche
类目: Machine Learning (cs.LG); Distributed, Parallel, and Cluster Computing (cs.DC)
*备注: AXIOM workshop (Foundations of Efficient Deep Learning) at NeurIPS 2026. 11 pages, 2 figures. Code and data: this https URL
Abstract:A mixture-of-experts (MoE) language model uses only a few experts per token, so a server can keep most expert weights in CPU memory and copy an expert to the GPU when the router requests it. Recent methods train the router to keep the same experts for several tokens, and treat fewer routing switches as a proxy for fewer copies. We measure the trade-off directly, in a harness that performs and times real CPU-to-GPU copies for OLMoE-1B-7B. On an L40S GPU, every tested policy that holds an expert set fixed and decodes faster than the least-recently-used (LRU) cache increases next-token loss by at least 0.067 nats. A change that leaves every routing decision intact does better: when the cache must evict an expert, evict the one whose next use lies farthest in the future. This rule copies 2.1x fewer experts than LRU at no loss in quality and is within 3.5% of the fewest copies possible. It needs future requests; the next sixteen tokens of demand recover 96% of its reduction, whereas a rule based on counts of past requests recovers 27%.
[LG-78] raining Calibrated Conditional Generative Models with a Sample-Only Coverag e Loss
链接: https://arxiv.org/abs/2611.15829
作者: Benjamin Remy,Raphael Rossellini,Chihway Chang,Rebecca Willett
类目: Machine Learning (cs.LG); Methodology (stat.ME); Machine Learning (stat.ML)
*备注: 32 pages, 7 figures
Abstract:Conditional generative models have become the tool of choice for sampling posteriors in scientific inverse problems, where we seek to sample images or signals conditioned on a set of indirect and noisy measurements. Yet they are often overconfident, which is a critical failure in scientific applications where accurate uncertainty quantification is essential. Existing regularizers fall short in complementary ways: differentiable coverage objectives require the model’s density and do not identify the posterior, while strictly proper scoring rules such as the energy score need only samples but do not directly enforce coverage. We close this gap by introducing a differentiable loss function based on the expected coverage probability, estimated from samples alone. It applies to any conditional generative model, with no need for tractable densities or architectural changes. We prove that it is a strict divergence, vanishing if and only if the model equals the true posterior. On simulation-based inference benchmarks, it matches a density-based coverage loss without evaluating the density. On high-dimensional imaging inverse problems, it yields better-calibrated posteriors than scoring-rule regularizers, with little change in the reconstruction quality of point estimates.
[LG-79] Sparse PCA with Overlapping Supports: NP-Hardness and an FPTAS
链接: https://arxiv.org/abs/2611.15810
作者: Alberto Del Pia,Dekun Zhou
类目: Machine Learning (cs.LG); Optimization and Control (math.OC)
*备注:
Abstract:In sparse principal component analysis (SPCA) with several components, each must be sparse and all must be pairwise orthogonal. When the supports of the components overlap, these two requirements interact. For a covariance matrix of fixed rank, polynomial-time algorithms with optimality or approximation guarantees are known for one component, and for several components with disjoint supports or with a common support. We study the general model, with individual sparsity constants and arbitrary support overlaps. We are given a rational n\times r matrix R and d sparsity constants s_i \le n , and the goal is to find an n\times d matrix X with orthonormal columns that maximizes |R^\top X|_F^2 , such that column i has at most s_i non-zero entries. Here RR^\top plays the role of the covariance matrix and has rank at most r . We show that SPCA is NP-hard even when r = d = 2 , but admits a fully polynomial-time approximation scheme (FPTAS) when r and d are fixed. The returned solution has exactly orthonormal columns, and its entries are real algebraic numbers in an exact representation. For fixed \epsilon , the number of steps of the algorithm is linear in n up to a polylogarithmic factor. Our approach reduces SPCA to a sparse version of the orthogonal Procrustes problem, that is, to maximizing a linear function over the feasible region of SPCA. We show that this problem is NP-hard even for d = 2 , and we give an FPTAS for it when d is fixed. This FPTAS combines approximate dynamic programming over the rows, two-column updates that restore exact orthogonality, and an exact computation of the few variables for which these updates are not available.
[LG-80] From Link Prediction to Linear Community Detection: A Finite-Sample Guarantee for Graph Pretraining NEURIPS2026
链接: https://arxiv.org/abs/2611.15778
作者: Ying Chen,Zhangyang Wang,Yixuan He
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注: Accepted at NeurIPS 2026. 47 pages, 8 figures. Code: this https URL
Abstract:Link prediction (LP) is a standard pretraining objective for graph neural networks (GNNs): hide edges, train an encoder-decoder to predict them, and reuse the node embeddings for downstream node tasks. Despite its empirical success, there is little quantitative theory explaining why an edge-level objective should produce embeddings that are linearly useful for node classification. We study this question in the finite-sample stochastic block model (SBM). For any bounded GNN encoder paired with a symmetric bilinear decoder, we prove a conditional reduction: small population held-out LP excess risk implies small community-classification error for a transductive linear classifier on the learned embeddings. The bound decomposes into an optimization residual, an architectural approximation residual, and explicit statistical terms of order \tildeO(\sqrtd/n) in bounded dense regimes. We further give a constructive few-label classifier: nearest-centroid classification under the positive semidefinite quadratic form G = W Z^\top Z W , whose label error is coupled to LP quality. The framework is architecture-agnostic through the approximation residual, and we instantiate this residual as O(\log n / n) for a graph isomorphism network (GIN) with spectral positional encoding and a bilinear decoder. Synthetic SBM experiments validate the theorem chain from LP risk to score error to classification error, and show when the worst-case constants become conservative.
[LG-81] SAME: Semantic Watermark Embedding for Open-Source Large Pretrained Models
链接: https://arxiv.org/abs/2611.15771
作者: Xinyao Song,Jialiang Shen,Hui Kang,Huajie Chen,Ruiqi Liu,Jiyang Zheng,Huaxi Huang,Zhenliang Lu,Chris Russell,Runqi Lin
类目: Machine Learning (cs.LG)
*备注:
Abstract:Open-source large pretrained models (LPMs) have become increasingly valuable assets for model development and downstream applications, but their accessibility also facilitates unauthorised copying and modification. To protect their intellectual property (IP), existing approaches typically embed tailored trigger-response patterns as watermarks for ownership verification. However, such trigger designs need to be tailored to task-specific data that constrain their transferability across downstream tasks, and are not durable under subsequent benign fine-tuning. To address these limitations, we propose SemAntic waterMark Embedding (SAME), which introduces a semantically meaningful private vocabulary that enables flexible composition of adaptive and persistent watermarks. Specifically, SAME introduces a fixed private vocabulary that assigns each native token a private embedding counterpart, while aligning the semantic expressiveness of the two vocabularies within the model’s representation space. This alignment allows private tokens to be flexibly composed into semantic triggers that elicit meaningful responses, enabling adaptive watermarks beyond predefined trigger-response pairs. By grounding the private vocabulary in the model’s native semantics, SAME couples the watermark with its evolving semantic capabilities, allowing this connection to be preserved throughout downstream fine-tuning. Extensive experiments demonstrate that SAME provides reliable and robust watermarks across LPMs of varying scales and architectures, enabling effective ownership verification for IP protection.
[LG-82] Sensitivity-Aware Surrogate Modeling with Random Feature Expansions
链接: https://arxiv.org/abs/2611.15750
作者: John E. Darges
类目: Machine Learning (cs.LG); Computation (stat.CO)
*备注:
Abstract:Random feature expansions (RFEs) provide a flexible approach to surrogate modeling, but their accuracy depends on the distribution used to sample features. We propose spike-and-slab random feature expansions (SAS-RFE) to learn the feature sampling distribution from sensitivity analysis. The method employs a spike-and-slab distribution that assigns probabilities to the inclusion of individual inputs in randomly sampled Fourier features. These inclusion probabilities are determined by estimates of total Sobol’ sensitivity indices, rather than by explicitly selecting important inputs or interaction terms. An information theoretic criterion shows that the total Sobol’ indices yield a distribution that approximates the input interaction structure of the model. An iterative algorithm updates the sampling probabilities using sensitivity estimates from successive RFE surrogates. Analytical formulas for the Sobol’ indices of random Fourier feature expansions enable these updates without additional evaluations of the target function. Numerical experiments on six benchmark problems, with up to 100 input dimensions, demonstrate improvements in surrogate approximation and sensitivity estimation.
[LG-83] Orbit: An Evidence-Grounded Benchmark for Evaluating LLM s on IT Operations Reasoning
链接: https://arxiv.org/abs/2611.15718
作者: Sathwik Narkedimilli,Rakshit Verma,Rahul Soni,Ankit Kushwaha,Balaji Ganesan,Kaustabha Ray,Sunil Saumya,Pratibha Moogi
类目: Machine Learning (cs.LG); Software Engineering (cs.SE)
*备注:
Abstract:LLMs now support IT operations (ITOps), but benchmarks rarely show why a model fails. We present ORBIT, a benchmark of ITOps questions in three formats, namely multiple-choice, open-ended, and troubleshooting, with each question accompanied by a reference answer and a recorded source. An audit found template defects in about a tenth of the generated items, so we drop them and rebuild the troubleshooting format from real practitioner threads with accepted answers. Under one independent judge, five LLMs answer most multiple-choice items correctly, between a third and just over half of the open-ended items, and a minority of the troubleshooting items, even under a more lenient root-cause rubric. When models judged their own answers, they accepted many answers to defective items, so we use one independent judge for every model. Word overlap with the question finds the correct option for more than half of the multiple-choice items. ORBIT is a diagnostic aid for selecting and supervising assistants rather than a deployment test.
[LG-84] Characterizing Predictive Multiplicity in Ranking and Its Fairness Implications
链接: https://arxiv.org/abs/2611.15705
作者: Ainhize Barrainkua,Mikel Malagón
类目: Machine Learning (cs.LG)
*备注: 32 pages, 6 figures
Abstract:Prediction problems often admit competing models that achieve comparable performance yet produce conflicting outputs, a phenomenon known as predictive multiplicity. While this challenge has been extensively studied in classification tasks, its implications remain largely unexplored in ranking frameworks, which are critical for high-stakes societal decisions such as the allocation of welfare resources and social assistance. This paper investigates the phenomenon of predictive multiplicity in score-based ranking problems. We introduce formal definitions to capture multiplicity in ranking settings, supported by foundational theoretical guarantees. Furthermore, we examine the dual relationship between predictive multiplicity and algorithmic fairness in ranking problems: we analyze how model multiplicity can act as a potential source of unfairness, while demonstrating how exploring this decision space provides a principled avenue for discovering less discriminatory rankings. Finally, we complement our theoretical contributions with insightful empirical evaluations on both synthetic and real-world datasets, validating our analytical findings and illustrating their practical implications. Taken together, this work highlights the real-world relevance of predictive multiplicity in ranking and sheds light on the potential of leveraging this effect to achieve fairer outcomes.
[LG-85] SpaceMCP: Vision Language Model Spatial Reasoning via Model Context Protocol NEURIPS2026
链接: https://arxiv.org/abs/2611.15680
作者: Ameen Izhac,Nobuko Yoshida,Ronald Clark
类目: Machine Learning (cs.LG)
*备注: 13 pages, 3 figures. Accepted at the 2nd Workshop on Embodied Spatial Reasoning (ESR) at NeurIPS 2026
Abstract:Despite impressive advances in large language models (LLM), vision-language models (VLM) still exhibit limited spatial reasoning capabilities. Recent progress in feed-forward 3D reconstruction has enabled precise, real-time geometric understanding from visual inputs. In this work, we propose SpaceMCP, a framework that leverages these advances to introduce an MCP-style paradigm for spatial reasoning. Our approach augments a VLM with structured function calls, allowing it to query a scene graph for spatial information such as distances, object positioning, and camera properties. The scene graph is constructed in real time using feed-forward 3D reconstruction to recover metric geometry, combined with open-vocabulary segmentation to identify and localize relevant objects within the scene. Paired with the VLM’s strong reasoning capabilities, the VLM is able to combine this primitive information to solve more complex spatial tasks. Without any training, SpaceMCP achieves state-of-the-art results on VSI-Bench, MMSI-Bench and VSTI-Bench, outperforming recent spatially post-trained VLMs and improving its base VLM by 15-22 points per benchmark. We argue that, given the contrast between the relatively slow progress in native VLM spatial reasoning and the rapid advances in feed-forward 3D computer vision, this paradigm offers a promising path forward for enabling robust spatial reasoning.
[LG-86] he Noise Floor of a Quantization Comparison: Engine restart batch and kernel drift measured against 8-bit and 4-bit effects in two decision models
链接: https://arxiv.org/abs/2611.15671
作者: Aleks Blagojevich,Trevor Sunderland
类目: Machine Learning (cs.LG)
*备注: 21 pages, 1 figure. Code and data: this https URL
Abstract:A quantized model is usually judged by comparing it with its BF16 original, but that comparison changes more than the weights. It often changes the serving engine, it always runs on a different server instance, it may change how requests are batched, and on some stacks the library selects a different kernel depending only on which packages import. We measure each of these sources on the same 2,400 public decisions for two of Cloudflare’s decision models, Clef-flash (9B) and Clef (27B), on an NVIDIA GB10, and state every quantization effect as the excess over them. Sources that are not quantization move between 0 and 0.92% of decisions. this http URL restarts are bit-identical on every record for both models. vLLM restarts flip 0, 5 and 20 decisions at BF16, FP8 and NVFP4 on Clef-flash and 4, 0 and 22 on Clef; a third instance of Clef’s quantized builds (exploratory) leaves FP8 at 0 and puts NVFP4 at 4 to 22, depending on which two instances are compared. An engine change at BF16 flips 6 to 16, a batch-size change 8 to 15, and the Gated DeltaNet kernel transformers selects unasked 10 to 14. Quantization inside one engine moves more: for Clef-flash and Clef, 0.88% and 1.04% at Q8_0, 2.12% and 1.62% at FP8, 5.79% and 3.50% at Q4_K_M, and 5.79% and 5.67% at NVFP4. On both models, every 4-bit interval excludes every non-quantization source. Two of our five pre-registered predictions failed, and the same two fail when the tests are rerun on the 27B: FP8 is not inside restart noise, and changing engine moves fewer decisions than 8-bit quantization does. Accuracy hides most of this. Over the full 31,275 public questions, Clef’s Q4_K_M build moves accuracy by a net six questions (-0.02 points [-0.20, +0.16]) while 436 of its decisions get worse (exploratory), and NVFP4 costs Clef-flash 0.95 points and Clef 0.65. We release the code, quantization recipes, per-record outputs and the pre-registration.
[LG-87] Beyond Reweighting: Certified Improvements from New Transitions in Discrete MCMC
链接: https://arxiv.org/abs/2611.15668
作者: Yishan Xi,Tinglan Wang
类目: Machine Learning (cs.LG)
*备注: 43 pages, 9 figures
Abstract:Optimizing transition probabilities on a fixed grap may leave substantial limitations on the efficiency of reversible Markov chain Monte Carlo (MCMC). We develop a framework for transition-support design that considers both allowed transitions and their stationary flows. For a fixed target distribution, observable, and node-flow budget, we derive a necessary and sufficient dual condition for support augmentation to strictly reduce the optimal value of the asymptotic variance, together with computable primal–dual certificates that establish improvements over every admissible original-support reweighting. We also construct a family in which adding a single edge reduces integrated autocorrelation time by an \Omega(n) factor relative to optimal original-support weighting, demonstrating that structural limitations can impose arbitrarily large relative losses in estimation efficiency. To realize these gains, we investigate energy-informed edge selection and constrained stationary-flow optimization. Experiments on molecular and Ising state graphs distinguish the potential value of new transitions from the performance of kernels constructed on them: on identical augmented Ising supports, full weight optimization certifies strict improvements in 24/24 conditions, whereas degree-based Metropolis–Hastings weighting certifies none. Our results provide a principled approach to characterizing, certifying, and exploiting transition-support improvements in explicitly represented finite state spaces.
[LG-88] A Mechanistic Interpretability Study of an Astronomy Foundation Model NEURIPS2026
链接: https://arxiv.org/abs/2611.15653
作者: Rachel C. Zhang,Alberto Bietti,Wes Gurnee,Alessandro Favero,Miles Cranmer
类目: Machine Learning (cs.LG); Instrumentation and Methods for Astrophysics (astro-ph.IM)
*备注: Accepted at NeurIPS 2026. A condensed version was selected for an oral presentation at the NeurIPS 2026 Workshop on Interpretability for Discovery
Abstract:Foundation models are increasingly used across the sciences, but whether their internal computations correspond to known scientific operations or proceed through different mechanisms remains unclear. Understanding these mechanisms is essential for assessing reliability, identifying failure modes, and learning new science from the models. We present, to our knowledge, the first causal mechanistic interpretability study of an astronomy foundation model, AION-1, focusing on its multimodal advantage in redshift prediction, where combining imaging with photometry outperforms photometry alone. We find that this advantage is largest for spatially extended and dust-attenuated galaxies, and that most of it is localized to the first of the encoder’s twelve blocks. Within that block, how well a property is linearly decodable does not predict whether it can be localized. Spatial extent is causally used, and one attention head dominates linear probing for it, yet ablating that head has negligible effect on redshift prediction: the information is distributed across heads and is decodable again one layer after we erase it. Dust attenuation shows the opposite pattern: it is only weakly decodable anywhere in the encoder, yet a single head is necessary for most of the multimodal advantage, particularly for the most dust-attenuated galaxies, and keeping that head as the only active head in its block preserves nearly all of the advantage. None of the astrophysically motivated operations we tested explains what this head computes. Whether this reflects a limitation of current interpretability methods or a computation that does not correspond to any known astrophysical operation remains open, and resolving it is a prerequisite for using interpretability as a tool for scientific discovery.
[LG-89] GLCR: Global-Local Collaborative Routing for Multi-Task Electromagnetic Signal Processing
链接: https://arxiv.org/abs/2611.15609
作者: Fang Lu,Yao Lu,Zhuangzhi Chen,Sicheng Zhang,Shanqing Yu,Yun Lin,Qi Xuan,Guan Gui
类目: Machine Learning (cs.LG)
*备注: 13 pages, 9 figures, 4 tables. Code available at this https URL
Abstract:Deep neural networks for electromagnetic signal processing are typically developed as single-task experts with fixed inputs and objectives, limiting cross-task transferability. Electromagnetic foundation models offer broader task coverage but incur substantial pretraining and inference costs. To balance accuracy and efficiency, we adopt model routing to dynamically dispatch unknown signals to matched experts within a heterogeneous pool. We construct HeteroSignal-7, a multi-task routing benchmark with 140,000 samples from seven datasets covering six task types, paired with seven classification and waveform-recovery experts and a unified evaluation protocol. We then propose GLCR, a two-stage routing method combining global dispatch with local discrimination of similar tasks. First, a lightweight one-dimensional convolutional neural network (CNN) with an additive-margin head predicts one of seven routing classes. Second, a gradient-boosted-tree specialist uses four physical statistics to re-evaluate samples whose top two candidates match a confusion pair automatically identified from validation errors. The pipeline is trained across the full signal-to-noise ratio (SNR) range and requires neither SNR labels nor noise estimation at inference time. GLCR outperforms all 19 baselines, achieving 96.16% routing accuracy and a 76.30% system score, exceeding the respective second-best baselines by 1.16% and 1.14%. It achieves the lowest routing latency among the compared methods at 1.43 ms per sample and reduces expected overall system FLOPs by 5.9 times compared with executing all experts. This work provides a public evaluation platform and an efficient, reproducible baseline for heterogeneous expert routing in electromagnetic signal processing.
[LG-90] FADO: Learning Causal Explanations with Prior-Fitted Networks
链接: https://arxiv.org/abs/2611.15607
作者: Seth Flaxman,Jess Carr,Annabel Jakob,Marek Masiak,Dino Sejdinovic
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:In regression and classification tasks with tabular datasets, most of the literature on explanation in ML (e.g. SHAP, permutation importance) has focused on predictive attribution: does feature X_i help predict outcome Y ? At its heart, predictive attribution captures correlation, not causation. We instead ask a causal question: would manipulating feature X_i change outcome Y ? We define a natural quantity to answer this question, the feature intervention effect: \Delta_i = E[Y\mid do(X_i=+1)]-E[Y\mid do(X_i=-1)] . In general, estimating this quantity from observational data alone would entail strong causal assumptions. We propose FADO (Feature-wise Amortized Do-Operators), an amortized inference method using prior-fitted networks trained on synthetic tabular datasets generated by structural causal models. During training, the synthetic data-generating process is known, so we can compute ground-truth intervention effects \Delta_i , and train a model to predict them at test time from a purely observational tabular dataset. As we demonstrate through extensive experiments, our new causal explainability method is easy to train and provides better explanations under a range of settings for which standard predictive attribution fails, including proxy variables, confounders, colliders, and placebos. We evaluate it on real, synthetic and semi-synthetic datasets, with and without hidden confounders, and compare it to existing predictive attribution methods and state-of-the-art causal inference methods.
[LG-91] Imputing Missing Near-Surface Air Temperature using Graph Attention Networks with Distance-Aware Edge Features
链接: https://arxiv.org/abs/2611.15598
作者: Emma C. Weber,Dominique Guillot,Daniyar Abykhanov,Julien Emile-Geay
类目: Machine Learning (cs.LG); Geophysics (physics.geo-ph)
*备注: 23 pages, 7 figures; LaTeX
Abstract:We introduce a graph attention network for reconstructing near-surface air temperature fields from incomplete observations. The model represents each grid cell as a node connected to its k nearest neighbors, with attention weights that depend on node features, distance, and land-ocean transitions. This graph representation avoids the polar distortions associated with treating the globe as a flat latitude-longitude image, as most deep learning reconstruction methods do. This is particularly relevant for historical temperature records, which extend back to the mid-nineteenth century but remain incomplete, with the largest gaps early in the record and at high latitudes. We train our model on CMIP6 CESM2 historical simulations masked with the time-varying HadCRUT5 observational coverage from 1850 to 2014. We then compare our approach with ordinary kriging, GraphEM, and the LaMa inpainting network on a held-out realization. The graph model achieves the lowest spatial RMSE (0.997 K), 10.1% lower than LaMa and approximately 31% lower than kriging and GraphEM. The largest gains occur in the Arctic and Antarctic, while LaMa remains slightly more accurate in the densely observed tropics. The graph model’s error also falls faster as coverage increases, so its advantage over LaMa grows by more than a third from the 1850s to the 2000s. An ablation study shows that the model’s skill comes mainly from the geometric graph construction rather than from auxiliary physical covariates.
[LG-92] AI4Plasma II: Geometry-Aware Surrogate Modeling for Variable Electrode Gap GEC Plasmas
链接: https://arxiv.org/abs/2611.15584
作者: Hao Yang,Zhenyu Zou,Wenshuai Jie,Luowei Yin,Yufeng Wei,Daniel Zhengyu Huang,Zhichao Peng,Chun Chen,Kangxin Liu,Rubing Han,Haoran Guan,Yipei Chen,Peiting You,Yuwei Fan
类目: Machine Learning (cs.LG); Plasma Physics (physics.plasm-ph)
*备注: 29 pages, 13 figures, 11 tables
Abstract:Industrial plasma process development requires broad exploration of operating and geometric parameters. Geometric parameters, particularly electrode spacing, affect discharge structure and processing uniformity. Exploring these parameter spaces through repeated high-fidelity simulations can be computationally expensive. Surrogate models can offer rapid predictions, but geometric variation complicates their application because the computational domain and discretization may change across samples. In this work, six geometry-aware surrogate models based on prescribed and learned geometry mappings, extended-domain representation, continuous coordinates, and geometry-native discretizations are compared in terms of their data requirements for accurate predictions of full plasma fields and ion density profiles near the wafer. The simulation dataset spans pressure, voltage, frequency, and electrode gap, with each reference case requiring an average simulation cost of 5.32 CPU h. To achieve a 2% mean error across the four plasma fields, prescribed-mapping Geo-FNO and Transolver require 100 training cases, compared with 250-1000 for the other models. Predicting ion density profiles near the wafer is more demanding and reveals larger differences among models, with prescribed-mapping Geo-FNO requiring the fewest training cases for the profile accuracy targets considered. To further reduce the data required for accurate predictions near the wafer, we introduce charge-separation weighting into the training loss, which reduces the prescribed-mapping Geo-FNO’s minimum tested training requirement for a 5% profile error from 1000 to 100 cases. These results support prescribed-mapping Geo-FNO as a practical starting point for the variable-gap problem studied here when a suitable geometry mapping is available, while Transolver achieves the highest prediction accuracy with sufficient training data. Comments: 29 pages, 13 figures, 11 tables Subjects: Machine Learning (cs.LG); Plasma Physics (physics.plasm-ph) Cite as: arXiv:2611.15584 [cs.LG] (or arXiv:2611.15584v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2611.15584 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Hao Yang [view email] [v1] Wed, 11 Nov 2026 15:07:21 UTC (612 KB)
[LG-93] Constructing and Validating Memorisation Audits for EHR Foundation Models NEURIPS2026 ML4H ALT
链接: https://arxiv.org/abs/2611.15578
作者: Inigo Stibbe,Joshua Placidi,Padmanabhan Ramnarayan,A. Aldo Faisal
类目: Machine Learning (cs.LG)
*备注: Accepted to the NeurIPS 2026 Workshop on Beyond Private Training: The New Landscape of AI Privacy (InfPriv) and the Machine Learning for Health (ML4H) 2026 Symposium, Findings Track
Abstract:Autoregressive electronic health record (EHR) foundation models may disclose training information at inference, yet clinical predictability makes patient-specific memorisation difficult to distinguish from generalisation. We evaluate fixed-window token reconstruction, sensitive-attribute disclosure, and random patient-level canaries in GPT-2-style models trained on MIMIC-IV, across early, best-validation, and final checkpoints. Reconstruction improves with additional context, but training–held-out differences remain small. Trajectory-unique prompts containing limited clinical information produce no excess sensitive disclosure for training patients. In a matched model with random binary patient-level canaries, training-patient label-recovery AUROC increases from 0.512 to 0.549 while held-out recovery remains at chance. An overfit 1,000-patient positive control achieves near-perfect recovery, confirming sensitivity to strong memorisation. These results reveal weak cohort-level memorisation but no reliable patient-level disclosure under the evaluated attacks. They motivate clinically grounded inference-time audits that can detect patient-specific memorisation in existing models without retraining with canaries.
[LG-94] Observations Drift Structures Remain: Structural Pretraining with Time Alignment for Electromagnetic Signals NEURIPS2026
链接: https://arxiv.org/abs/2611.15566
作者: Wenjin Gui,Junyu Shen,Haibo Xu,Yuchuang Sun,Luqing Luo,Wupeng Xie,Qian Huang,Fengxiang Wang,Yangang Sun,Maosong Sun
类目: Machine Learning (cs.LG)
*备注: Accepted at NeurIPS 2026
Abstract:Electromagnetic signal modeling is increasingly moving toward self-supervised pretraining on heterogeneous I/Q observations. However, an observed waveform is jointly shaped by signal generation, propagation environments, and acquisition processes, which mix reusable structural regularities with observation-specific variations. Existing methods usually pursue unified pretraining with reconstruction objectives on mixed observations. This heterogeneity creates two key mismatches for reconstruction pretraining: (1) appearance mismatch, induced by propagation-dependent waveform variations; and (2) temporal mismatch, induced by acquisition-dependent sampling coordinates. These mismatches make observation-specific appearances and sampling coordinates easier to reconstruct than stable signal structures, inducing observational shortcuts. We identify this failure mode as observation-biased pretraining. To address this, we propose Structural Pretraining with Time Alignment for heterogeneous EM signals. It mitigates temporal mismatch with RoPE-fs, which maps token positions to sampling-rate-normalized physical-time positions, and reduces reliance on appearance shortcuts by replacing waveform reconstruction with masked prediction of discrete structural units. These units are organized into atomic structural units and emergent structural units that capture variable-length compositional patterns beyond fixed patch boundaries. Across diverse EM benchmarks, our method improves both full-supervision and few-shot adaptation over existing EM pretraining baselines, showing that the learned representations better preserve structural consistency when observations drift in waveform appearance and temporal scale.
[LG-95] What Conformal Selection Can Certify in CRISPR Off-Target Design:Benchmark Leakage Free Baselines and a Finite-Sample Cohort Requirement
链接: https://arxiv.org/abs/2611.15555
作者: Mohammad Malekpouri,Somayeh Lotfi
类目: Machine Learning (cs.LG)
*备注:
Abstract:Conformal selection promises distribution-free, finite-sample control of the error rate of a decision, and has been proposed as a safety layer for machine-learning CRISPR guide design. We ask whether such a certificate is attainable from the validated off-target data that exist. The constraint is arithmetic: a conformal p-value cannot fall below one divided by the calibration cohort size plus one, so a Benjamini-Hochberg selection certifies a single candidate only when that cohort is at least the candidate pool size divided by the tolerated error rate. Public cohorts supply ten validated guides, whereas a pool of two thousand at a ten percent error rate needs twenty thousand; the accepted set is consequently empty in over ninety-six percent of splits. Two benchmarking findings explain why this went unnoticed. The standard CIRCLE-seq file lists ten guides under eighty on-target strings, so grouping cross-validation on the sequence column trains and tests on one guide; corrected, pair-level discrimination and per-guide fidelity collapse. With that leakage removed, a training-free position-weight score and a bare count of near-match genomic sites are not beaten by a trained oracle. We give the corrected protocol, the baselines a new model must beat, and the cohort size a certified pipeline needs.
[LG-96] One Hierarchy Many Sports: Hierarchy-Aligned Multi-Task Learning for Next-Shot Prediction
链接: https://arxiv.org/abs/2611.15529
作者: Chao-Wei Cai,Min-Te Sun,Sheng Kuang Wu,Hung-Hsuan Chen
类目: Machine Learning (cs.LG)
*备注:
Abstract:Next-shot prediction in racket sports requires learning from event data with different match structures, recorded attributes, and label vocabularies. Table tennis, badminton, and tennis nevertheless share a hierarchy in which shots form rallies and rallies form higher scoring units. We present MT-HTA, a hierarchy-aligned multi-task Transformer that uses this structure to jointly predict the attributes of the next shot. A temporal Transformer captures event progression, a feature-token path models interactions among semantic fields rather than along the time axis, and task-specific attention retrieves information from hierarchical summaries and player context. The architecture supports three- and four-level hierarchies and sport-dependent task sets. Across four datasets from three sports, the final MT-HTA achieves the highest mean task-averaged F1 among the compared main models. Ablations on table tennis support the combination of hierarchical encoding, task attention, and feature-token interaction. Additional modules alter the balance between classification scores and spatial error rather than improving every metric. These results support a reusable architecture for heterogeneous racket-sport event data. We release the code for reproducibility. this https URL
[LG-97] Selective State-Space Neural Tucker Factorization for Temporal-aware Traffic Imputation
链接: https://arxiv.org/abs/2611.15502
作者: Zhenhao Li,Peng Tang
类目: Machine Learning (cs.LG)
*备注:
Abstract:Spatiotemporal traffic data imputation is a key task of intelligent transportation systems (ITS), and urban traffic data inherently exhibits strong long-term temporal dependencies. Therefore, accurate modeling of these long-term temporal dynamics is the core prerequisite for high fidelity traffic data imputation. However, existing traffic data imputation methods still suffer from inherent limitations in long-sequence modeling, such as gradient vanishing, low efficiency, and insufficient capability to capture long-range temporal correlations. To address these issues, this paper proposes a Selective State Space Neural Tucker Factorization model (S3NTF) for robust traffic data imputation. This model combines the Selective State Space mechanism with Neural Tucker Factorization to capture complex spatial correlations and long-term temporal dynamics. A state space-based time encoder is proposed to effectively extracts long sequence dependencies with linear complexity, while the latent spatiotemporal feature interaction learning module models high-order cross modal interactions based on Tucker decomposition. Extensive experiments on six real traffic datasets have shown that our model outperforms existing models in terms of accuracy.
[LG-98] he Illusion of Diversity: Aligning LLM Exploration via Effective Entropy NEURIPS2026
链接: https://arxiv.org/abs/2611.15493
作者: Xiaoliang Fu,Jiaye Lin,Yangyi Fang,Cong Qin,Binbin Zheng,Chaowen Hu,Zekai Shao
类目: Machine Learning (cs.LG)
*备注: Accepted at NeurIPS 2026
Abstract:Reinforcement learning with verifiable rewards has shifted the alignment paradigm of large language models from subjective preference to objective correctness, yet a fundamental disconnect persists between exploration diversity and reasoning validity. Existing strategies predominantly rely on “blind” entropy metrics, e.g., token-level or global sequence-level entropy, which structurally fail to distinguish between effective reasoning and potential hallucinations. This misalignment creates an illusion of diversity, where high entropy tends to signal chaotic degeneration rather than valuable exploration. To bridge this gap, we introduce Effective Entropy, a validity-aware entropy concept that quantifies diversity exclusively within the valid subspace and systematically filters out the noise of incorrect trajectories. Built on this concept, we instantiate VEPO, a simple Validity-aware Entropy Policy Optimization method that directly incorporates effective entropy into RLVR training. VEPO induces an adaptive attraction dynamic that drives up the probability of under-explored valid trajectories, a mechanism that matches the practical regime of large-scale autoregressive models where sequence probabilities remain extremely small. Experiments on six mathematical reasoning benchmarks with 1.5B and 7B DeepSeek-R1-Distill-Qwen models show consistent aggregate accuracy gains over GRPO and competitive repeated-sampling success. A direct evaluation with matched correct counts finds more distinct strategies on commonly solved MATH500 problems, supported by paired statistical tests and three-seed training comparisons.
[LG-99] axonomizing Synthetic Data for Law
链接: https://arxiv.org/abs/2611.15489
作者: Ignacio Cofone,Katherine J. Strandburg,Nicholas Tilmes
类目: Computers and Society (cs.CY); Cryptography and Security (cs.CR); Machine Learning (cs.LG)
*备注: 29 pages
Abstract:Synthetic data is increasingly important in data usage and AI design, creating novel legal and policy dilemmas. All too often, discussions of synthetic data treat it as entirely distinct from “real,” collected data, overlooking the risks posed by different kinds and uses of synthetic data. We build on Michal Gal and Orla Lynskey’s argument that synthetic data will transform information privacy, market competition, and data quality. While the risks posed by synthetic data depend on its connection to collected data, we argue that background knowledge and assumptions about ground truth used to create it are at least as important. We bring that focus to Gal and Lynskey’s taxonomy of synthetic data, arguing that this focus is essential to grasp synthetic data’s legal and policy implications. As such, we divide synthetic data into (1) transformed data, which modifies collected data to preserve certain statistical properties for an end use; (2) augmented data, which relies on assumptions to bolster a collected dataset’s fidelity to the ground truth; and (3) simulated data, which relies almost entirely on background knowledge and ground-truth assumptions. Applying this taxonomy, we show that no category of synthetic data prevents privacy harms from group-based inferences, that augmented and simulated data can amplify bias while making a dataset look more complete, and that synthetic data rarely lets smaller competitors overcome incumbents’ data advantages. As policymakers weigh whether to incentivize, mandate, or discourage the use of synthetic data, they should consider the validity of the ground-truth assumptions used in producing that data.
[LG-100] DETER: Suppression During Pretraining for Tamper Resistance
链接: https://arxiv.org/abs/2611.15479
作者: Maria Ios Glarou,Martin Jaggi
类目: Machine Learning (cs.LG)
*备注:
Abstract:Open-weight language models can acquire knowledge that could enable biological, chemical, or cyber misuse. Any attempt to limit such knowledge must remain effective after the model weights are released and can be fine-tuned. Unlearning applied after pretraining can be rapidly undone, whereas filtering the pretraining corpus offers greater resistance. We ask whether actively suppressing knowledge during pretraining can further increase the fine-tuning required to acquire it. Even simple gradient ascent interleaved with standard pretraining on the filtered corpus makes medical knowledge harder to acquire through adversarial fine-tuning than filtering alone or post-hoc unlearning. At 1B parameters, reaching the held-out PubMed loss of an unfiltered model requires over 20 \times as many PubMed fine-tuning tokens as after filtering alone. Retained knowledge is evaluated using both held-out data and downstream tasks. Across model sizes, varying suppression strength reveals an increasingly favorable trade-off between fine-tuning difficulty and held-out loss on retained pretraining data. These results point to pretraining as an effective stage for making unwanted capabilities harder to acquire through subsequent fine-tuning. Code is available at \hrefthis https URL\textttthis http URL.
[LG-101] On-Edge Acoustic Continual Anomaly Detection: A Cross-Platform Comparative Study
链接: https://arxiv.org/abs/2611.15465
作者: Rami Zewail,Yehia Ehab,Karim Ali,Ahmed Mahrous,Mohamed Awad,Bassem Mokhtar
类目: ound (cs.SD); Machine Learning (cs.LG)
*备注: Accepted to the 14th International Japan-Africa Conference on Electronics, Communications and Computations (JAC-ECC 2026). 6 pages, 4 figures, 5 tables
Abstract:Acoustic Anomaly Detection (AAD) serves as a non-contact approach for machine condition monitoring, yet much of the existing research is evaluated offline and rarely addresses embedded hardware constraints or microphone variations under operational drift. This paper presents a comparative evaluation of on-edge continual AAD across two embedded platforms the ESP32-S3 (with an external I2S digital microphone) and the Arduino Nicla Vision (with an onboard Pulse Density Modulation (PDM) microphone) using a shared detection model. The model utilizes a lightweight ShallowNet encoder and contrastive prototype learning to map spectrograms into a 32-dimensional embedding space. Both boards run an identical TensorFlow Lite for Microcontrollers (TFLM) encoder, performing on-device prototype learning updated via an exponential moving average. A cloud pipeline orchestrated through Firebase and MQTT enables remote data collection and retraining. Evaluated via a controlled speaker-replay protocol using standardized Detection and Classification of Acoustic Scenes and Events (DCASE) Challenge Task 2 machine sounds, the ESP32-S3 achieved 82.35% to 95.00% accuracy across five machine types, while the Nicla Vision achieved 79.41% to 95.00%. The two platforms demonstrated comparable performance, confirming that the lightweight prototypical architecture transfers reliably across distinct hardware and sensor tiers.
[LG-102] Overcoming the Resolution Limit: Significance-Aware Regularization for Intersectional Fairness NEURIPS2026
链接: https://arxiv.org/abs/2611.15427
作者: Antonio Ferrara,André Panisson,Francesco Cozzi,Alan Perotti,Francesco Bonchi
类目: Machine Learning (cs.LG)
*备注: Accepted at NeurIPS 2026
Abstract:Intersectional fairness requires evaluating models across fine-grained subgroups. However, as group definitions become more granular, the sample size per subgroup shrinks, causing fairness estimates to become increasingly unstable. Existing in-processing methods are certainty-blind: they penalize observed disparities without distinguishing statistically certified unfairness from sampling noise. This leads to fairness overfitting, where models sacrifice utility to “fix” spurious disparities in sparse regimes. To address this, we introduce the significant-excess metric family, a continuous, uncertainty-aware fairness score that measures only the portion of a subgroup’s disparity exceeding what sampling variation can explain at a specific confidence level. Building on this metric, we propose SAFER (Significance-Aware Fairness Regularizer), a differentiable in-processing method that gates fairness penalties using a soft-count likelihood-ratio evidence statistic. SAFER satisfies a provable safety property: when every subgroup disparity lies well below the gate threshold, the unconstrained optimum is nearly stationary for SAFER. At that optimum, the back-propagated contribution of its significance-gated component is exponentially small, and the additional fairness contribution comes from a small ungated term with a fixed weight. Mechanism ablations across three benchmark datasets and purpose-built stress variants show that removing any component of SAFER (threshold smoothing, the significance gate, or a dense-tail term) produces the failure the theory predicts on at least one stress variant. In sparse intersectional settings, SAFER achieves a better fairness-utility Pareto frontier than the baseline methods, and it preserves utility when certified unfairness is absent.
[LG-103] MORA: Correcting Time Series Forecasts with Coarse Future Signals
链接: https://arxiv.org/abs/2611.15413
作者: Hu Li,Kenan Li,Haonan Xu,longlong,Shuqin Gao,Yucheng Zhang,Feng Dai
类目: Machine Learning (cs.LG)
*备注: 33 pages, 3 figures
Abstract:Recent architectural advances have improved time-series forecasting. However, our analysis of forecasters with diverse architectures reveals a common pattern: a substantial share of residual energy is concentrated in a small number of low-frequency components. Simple linear probes further show that the forecasters’ historical inputs still contain predictive information about errors in the forecast mean. Building on these findings, we propose Model-Output Residual Adaptation (MORA), a model-agnostic residual adapter that refines the outputs of frozen forecasters without requiring additional inputs. MORA estimates segment-wise means of the target sequence through structural extrapolation and learned estimation, rather than directly predicting backbone-specific residuals. It then converts discrepancies between these estimates and the corresponding statistics of the backbone forecast into stagewise corrections, scaling each update according to estimation reliability. Extensive experiments on twelve benchmarks with six forecasting backbones show average MSE reductions of 7.2% for long-term forecasting and 28.1% for short-term forecasting, with lower MSE in 96.2% of the 288 evaluated cases. The benefits also extend to three time-series foundation models evaluated at different context lengths. Together, these results demonstrate that predictable coarse-scale residuals can be effectively exploited to improve existing forecasts. Code repository: this https URL.
[LG-104] A Modified Glicko Model for Rating Soccer Clubs and National Teams Worldwide
链接: https://arxiv.org/abs/2611.15378
作者: Andrei Shelopugin,Alexander Sirotkin
类目: Machine Learning (cs.LG); Applications (stat.AP)
*备注: 39 pages, 4 figures, 14 tables
Abstract:Comparing soccer teams that rarely or never meet is a fundamental challenge in sports analytics: clubs play mostly within their national leagues, and national teams within their confederations. We propose a rating framework that places clubs and national teams worldwide on a common probabilistic scale. Building on a recent extension of Glicko with order effects and strength-dependent draw probabilities, we adapt the model to soccer with goal-scoring adjustments, a decomposition of home advantage, extra time, league-aware rating initialization and rules of promotion and relegation that do not inflate the levels of leagues; national teams start from the levels of their confederations. The models are evaluated on about 1.2 million club matches of 252 leagues from 166 countries (2000-2026) and 12.6 thousand national team matches (1998-2026). On held-out seasons, the modified Glicko model is clearly more accurate than Elo and as accurate as score-matrix models built on the goal expectations of a gradient boosting model, including the Dixon-Coles and Markov chain models, while it is substantially more accurate in international club matches. Averaging the predictions of the rating model and the Markov chain model is more accurate than either model alone. For national teams, the model is significantly more accurate than all baselines, especially in the final tournaments, where teams of different confederations meet. The resulting ratings are interpretable and can be used for player evaluation, transfer analysis and match forecasting. The match data, the ratings and the code are publicly available at this https URL
[LG-105] MechBench: A Domain-Specific LLM Benchmark for Parallel Mechanisms
链接: https://arxiv.org/abs/2611.15342
作者: Zhihao Xia,Tingting Wang,Hui Li,Yongheng Xing,Chunxu Tian
类目: Machine Learning (cs.LG)
*备注:
Abstract:Parallel mechanisms pose distinctive challenges to large language models (LLMs), as valid reasoning depends jointly on mechanism topology, loop closure, geometry, joint types, and modeling assumptions. Yet existing LLM benchmarks provide little systematic coverage of such domain-specific knowledge, constraint-sensitive analysis, and constructive design. To fill this gap, we introduce MECHBENCH, a 2,541-question benchmark for evaluating LLMs in parallel mechanisms across professional knowledge, mechanism analysis, and design. MECHBENCH contains standard and challenge tiers spanning six capability categories and nine executable analysis and design tasks. It is constructed through MECHFORGE, a domain-grounded framework that combines expert-organized knowledge, LLM-assisted question generation, programmatic instance generation, two-stage expert review, and task-aware scoring with exact matching, rubric-based evaluation, and program verification. We further develop MECHGPT as a diagnostic domain-adapted model. Experiments show that strong knowledge performance does not directly translate into executable mechanism reasoning: leading closed-service models achieve over 95 on the standard tier but only 47-60 on the challenge tier, while domain adaptation substantially improves MechGPT over its Qwen3-8B baselines but leaves a large gap in constrained analysis and design. These results establish MECHBENCH as a systematic testbed for measuring domain-specific LLM capabilities in mechanism science.
[LG-106] abOmics: A Small Open Tabular Foundation Model for Wide Omics Data
链接: https://arxiv.org/abs/2611.15329
作者: Andrea López-García,Luis Bote-Curiel,Francisco Javier Gimeno-Blanes,José Luis Rojo-Álvarez,Sergio Muñoz-Romero
类目: Machine Learning (cs.LG)
*备注:
Abstract:Omics studies often produce tables with thousands to tens of thousands of molecular features and few samples. Tabular foundation models use diverse strategies to predict labels for such tables, with some relying on in-context learning, while approaches to wide tables may involve subsampling, grouping, or compressing features. We present TabOmics, a small, open family of tabular foundation models with 7.3M parameters, pretrained exclusively on synthetic tables. TabOmics processes all features retained after preprocessing jointly and derives predictions, attention-derived feature-importance scores, and test sample embeddings from the same forward pass. On 46 classification tasks from three benchmarks of high-dimensional, low-sample-size tables, TabOmics v2.5 with eight ensemble members achieves the highest mean area under the receiver operating characteristic curve (AUROC) among the evaluated models (0.943), outperforming TabPFN-3.5 with eight members on 32 tasks despite having about 30 times fewer parameters. With a single member, TabOmics v2.5 reaches a mean AUROC of 0.941, versus 0.942 for TabPFN-3.5 with eight members, with no significant paired difference under the reported task-level analysis. On synthetic tables, its feature-importance scores achieve feature recovery similar to that of TabPFN-Wide, and its predictions remain above chance at 30,000 features. On breast cancer tasks, the top ranked genes include PAM50 genes, which define the subtype labels, indicating concordance with the label definition rather than independent biological validation. The code, model weights, training recipes, and experimental artifacts are publicly available.
[LG-107] Sharp Covering-Number Bounds for Feedforward ReLU Networks via Parameter Symmetries
链接: https://arxiv.org/abs/2611.15300
作者: Ruizhi Deng,Guohao Shen,Tangrui Xiong
类目: Machine Learning (cs.LG); Statistics Theory (math.ST); Machine Learning (stat.ML)
*备注:
Abstract:In neural networks, distinct parameter configurations often realize the same input–output mapping, indicating that nominal parameter counts can overstate functional complexity. In this paper, we characterize the parameter-space symmetries of ReLU and Leaky ReLU networks across fully connected, convolutional, and residual architectures. We prove that every quasianalytic parameter map preserving the network function is a composition of positive channel scaling and channel permutation. By quotienting out both symmetries, we derive uniform covering-number upper bounds in which positive scaling reduces the complexity exponent from the total parameter count S to S-H (with H denoting the number of scalable channels) and permutation symmetry divides the bound by the factorial count G of channel permutations. We establish matching lower bounds showing that the exponent S-H is sharp in general and that the divisor G is attained for one-hidden-layer and generically identifiable networks. We further extend these results to general continuous piecewise-linear activated networks. Finally, we apply our covering-number bounds to derive statistical risk guarantees for empirical risk minimization and capacity control beyond i.i.d.\ settings, and show how the underlying symmetries help explain weight rebalancing in function approximation and flat directions in empirical risk optimization.
[LG-108] Learning Rate Transfer and Feature Learning Across Depth for Constrained Spectral Optimizers: Complete Scion
链接: https://arxiv.org/abs/2611.15291
作者: Eline Pot,Laurent Lefebvre,Mathieu Segond,Antonio Silveti-Falls
类目: Machine Learning (cs.LG); Optimization and Control (math.OC)
*备注:
Abstract:Hyperparameter transfer across model depth is instrumental in training deep models, as it enables learning rates calibrated on shallow networks to be effectively applied to deeper architectures. The recent CompleteP framework establishes a 1/L residual scaling rule that demonstrates this for Adam-family optimizers. We study whether similar transfer extends to constrained optimization methods, specifically Frank-Wolfe and projected non-Euclidean normalized steepest descent, where updates are determined by linear minimization oracles. We derive scaling rules for these methods which confirm that the 1/L residual scaling applies, although the arguments differ from the Adam-family optimizers. Our analysis supports depth-wise transfer and width-wise transfer of the learning rate for the Scion algorithm while ensuring feature learning and avoiding the kernel-regime. In experiments on NLP transformers (nanoGPT on FineWeb), vision transformers (DeiT on ImageNet), and vision MLPs (MLP-Mixer on CIFAR-10), we observe learning rate transfer across depth and feature learning for appropriately chosen residual scaling.
[LG-109] A hybrid machine learning-analog framework for downscaling unprecedented climate extremes
链接: https://arxiv.org/abs/2611.15284
作者: Nina Effenberger,Florian E. Roemer,Maybritt Schillinger,Maxim Samarin,Erich Fischer,Reto Knutti
类目: Machine Learning (cs.LG); Applications (stat.AP)
*备注:
Abstract:Extreme weather events are critical for climate impact assessments. While physical models can simulate unprecedented extremes, for example, through ensemble boosting, this approach is computationally demanding, often restricting output to coarse spatial resolutions. Probabilistic generative downscaling offers a solution by enabling efficient sampling from the entire climate distribution, including its extreme tails. Here, we first demonstrate that machine learning generated regional climate model heatwaves over Paris are physically realistic. We then leverage an emulated machine learning dataset to perform analog-based downscaling of coarse global climate model ensemble boosting events. Our hybrid framework effectively bridges high-resolution regional detail with rare-event sampling, providing an efficient tool for extreme event downscaling and capturing multi-model uncertainty.
[LG-110] Protected Unbiased Sparsification: A Sharp Loewner Curve
链接: https://arxiv.org/abs/2611.15247
作者: Ziyang Wang,Menghao Li
类目: Machine Learning (cs.LG)
*备注: 44 pages, 11 figures. Code, protocols, and saved results are included as ancillary files
Abstract:Random sparsification replaces a probability vector by a random vector with few nonzero entries. We ask for unbiasedness, exact preservation of prescribed linear statistics, and a covariance guarantee for every downstream linear response. Let p have k positive coordinates, let A specify the protected statistics, and let d\le k-2 be their effective rank. For every d+1\le m\le k , we construct an unbiased random probability vector Q with at most m nonzero coordinates, AQ=Ap on every realization, and \operatornameCov(Q)\preceq h C_p\mid A , where h=(k-m)/((m-d)(k-d-1)) and C_p\mid A is the categorical covariance after optimal linear adjustment for the protected statistics. This coefficient is sharp uniformly over instances, even among biased estimators when error is measured by its second moment. A residual covariance bridge extends the concurrent unprotected curve to arbitrary pathwise linear protection. We also characterize instance-optimal constants for partitions with uniform weights within classes and derive bounds for peaked weights. For rational inputs, an exact Las Vegas sampler attains the curve within a factor 1+\epsilon in expected polynomial time. A separate fast block-pivotal sampler preserves block masses and is no worse in covariance order than independent sampling within blocks. Experiments illustrate the theory on synthetic weights and language-model attention. In Qwen2.5-7B-Instruct decoding, the fast sampler preserves block masses, while comparisons with VarOpt remain statistically inconclusive and top- m truncation scores highest on RULER.
[LG-111] Regret Bounds for Quantum Linear Bandits
链接: https://arxiv.org/abs/2611.15237
作者: Maoli Liu,Zhuohua Li,John C.S. Lui
类目: Machine Learning (cs.LG)
*备注: 26 pages
Abstract:We study d -dimensional quantum linear bandits with bounded rewards, where the learner selects an action and queries its quantum reward oracle or its inverse in each round. For general action sets, existing algorithms achieve regret O(d^3/2\operatornamepolylog(T)) over T rounds. Prior work established a lower bound of \Omega(d\log(1+T/d)) , leaving open whether the d^3/2 dependence in the upper bound is necessary. We resolve this question by proving a minimax lower bound of \Omega(d^3/2\log(1+T/d^3/2)) . To prove this result, we construct a family of hard instances in which the reward gap of each action is proportional to its squared distance from the optimal action. We establish a posterior Fisher information bound for adaptive quantum queries with quantum memory, relating the information acquired about the optimal direction to the regret incurred. Viewing each selected action as an estimate of the optimal direction, we use a matrix Cramér–Rao argument to derive a lower bound on expected regret. When the action set is fixed and has size K , we propose two phased elimination algorithms, Soft-G-Elim and QG-Elim. We adapt quantum phase estimation to construct SoftQME, a mean estimator with small bias and controlled tails. Soft-G-Elim estimates the mean reward of every surviving action from independent SoftQME outputs on the support of an experimental design. Controlling the bias and fluctuations of these estimates yields regret O(d\log T\log(dT)\sqrt\log(KT)) , reducing the \log(KT)\log\log(dT) factor in the previous regret bound to \sqrt\log(KT) . QG-Elim further reduces this factor to \min\sqrt\log(KT),\sqrt d\ .
[LG-112] Winter wheat improves soil carbon sequestration with reduced gains in warmer climates
链接: https://arxiv.org/abs/2611.15232
作者: Georgios Giannarakis,Vasileios Sitokonstantinou,Dimitrios Bormpoudakis,Ilias Tsoumas,Nikiforos Samarinas,Gustau Camps-Valls,Charalampos Kontoes
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注: 19 pages, 4 figures, 2 tables. Published in Environmental Research Letters. Georgios Giannarakis and Vasileios Sitokonstantinou contributed equally to this work
Abstract:Sustainable agricultural intensification requires a precise understanding of how land use decisions influence soil organic carbon (SOC) sequestration under varying environmental conditions. We apply causal machine learning to evaluate the heterogeneous impact of winter wheat-based rotations on SOC at the field-level across Lithuania. Using a nationwide dataset collected over five years (2018-2022), including crop types, management practices, soil properties, and environmental conditions, we estimate how the effects of winter wheat-based rotation vary across space and climate. Our results indicate that winter wheat generally enhances SOC levels, but its effectiveness is modulated by climate. Cooler regions experience higher SOC gains, whereas warmer regions show diminished or even negative effects. Future climate projections suggest that the potential for winter wheat carbon sequestration may decline under high-emission scenarios. These findings highlight the importance of localized agricultural strategies that account for climate variability to maximize SOC sequestration and ensure long-term soil sustainability.
[LG-113] MOON-RL: Multi-Objective Optimization via Objective Space Navigation with Reinforcement Learning
链接: https://arxiv.org/abs/2611.15227
作者: Andrii Artomov,Soheyl Massoudi,Jürg Schiffmann
类目: Machine Learning (cs.LG)
*备注: 66 pages, including appendices. Code, data and supplementary video links are provided in the manuscript
Abstract:Many engineering design problems involve multiple conflicting objectives and strict feasibility constraints. They are typically solved with population-based optimizers that require many evaluations and offer limited online control over where the search is focused. We present MOON-RL, a reinforcement-learning-based optimization framework that formulates constrained multi-objective search as preference-directed navigation in the objective space. In MOON-RL, a lightweight PPO policy receives objective-space and constraint-related observations and acts as an instance-specific adaptive search process. It learns how to update the design so that the current solution advances along a user-specified trade-off direction, while a simple meta-algorithm periodically resets the policy, resamples starting designs from a continuously updated archive of non-dominated solutions, and adjusts step-size and penalty schedules. We evaluate MOON-RL on real-world actuator-design problems and additional convergence-hard constrained benchmarks, comparing against established evolutionary algorithms under matched evaluation budgets. In focused search scenarios, MOON-RL achieves competitive performance in the region of interest, and ablation studies show that both the RL-driven adaptive sampling process and the meta-algorithm contribute to this behaviour, suggesting that repurposing RL training as objective-space navigation is a viable building block for interactive, preference-guided design tools. Additional experiments show that the framework also accommodates adaptive non-RL samplers, while PPO remains a strong choice for fixed-target, preference-directed objective-space navigation.
[LG-114] DGPExplainer: Generating Prototype Sequence Explanations for Dynamic Graph Neural Networks
链接: https://arxiv.org/abs/2611.15224
作者: Jinlong Hu,Shixu Fan,Chongzhang Li
类目: Machine Learning (cs.LG)
*备注:
Abstract:Graph Neural Networks (GNNs) have been successfully applied to learning from graph-structured data. However, applying them to dynamic graphs-where nodes and edges evolve over time-presents significant challenges, particularly regarding explainability. While explanation techniques for static GNNs are well established, they do not readily extend to dynamic GNNs (DGNNs) due to the added complexity introduced by temporal dynamics and spatiotemporal dependencies, especially when generating model-level explanations. To address this gap, we introduce DGPExplainer, a post-hoc, prototype-based explainer specifically designed for DGNN models that transforms their complex decision-making processes into human-understandable, analogical, case-based explanations. DGPExplainer first employs local explanation methods to identify important elements within temporal snapshots of dynamic graph sequences. It then decomposes the graph at each time step into distinct prototypes, thereby symbolizing the dynamic graph sequence. Finally, we design a sequence generation method that uses these symbolized sequences to learn and generate class-specific dynamic graph pattern sequences that serve as prototype explanations. Experiments on both synthetic and real-world datasets demonstrate that our method effectively produces prototype-based explanations for DGNN models across multiple classes.
[LG-115] Anytime-Valid Empirical Likelihood Inference from LLM Annotations with Adaptive Human Auditing
链接: https://arxiv.org/abs/2611.15197
作者: Hongfu Gao,Jiang Hu,Pengfei Liu,Wang Zhou
类目: Machine Learning (cs.LG)
*备注:
Abstract:Large language models (LLMs) are increasingly used to annotate unlabeled data as a fast, low-cost alternative to human annotation. However, their outputs can be biased and miscalibrated relative to human labels, potentially leading to misleading analysis and inference. We propose AAELI (Adaptive Human Auditing and Anytime-Valid Empirical Likelihood Inference), an online framework for estimating the population mean of human labels under a limited auditing budget. AAELI uses LLM outputs to guide adaptive human auditing and combines them with audited labels through predictable empirical-likelihood e -processes to construct anytime-valid confidence sequences. Our main theoretical result shows that a predictably learned efficient projection preserves the first-order profile empirical-likelihood statistic under adaptive auditing, without requiring a prescribed convergence rate for the projection estimator. This yields an exactly anytime-valid betting construction whose continuous-mixture form incurs a scalar rather than vector-dimensional mixing penalty. Supporting results establish adaptive empirical-likelihood asymptotics and show that a learned audit rule attains the oracle variance within the stated class of augmented estimating-equation designs under the same audit-probability floor and long-run expected budget. Experiments on synthetic data and three real-world datasets demonstrate that AAELI maintains empirical anytime coverage while requiring fewer human audits than the strongest competing baseline to reach a target precision.
[LG-116] Stealthy High-Impact False Data Injection Attacks in Smart Grid Using Physics-Constrained GAN
链接: https://arxiv.org/abs/2611.15162
作者: Md. Mainul Islam,Hasan Kurban,Abdulrahman Takiddin,Muhammad Ismail,Rachad Atat,Erchin Serpedin
类目: Machine Learning (cs.LG); Systems and Control (eess.SY)
*备注: 10 pages
Abstract:False data injection (FDI) attacks can distort power system state estimation while retaining plausible measurement patterns. Existing FDI methods trade stealth for impact or vice versa. They often assume access to internal grids, all-bus sensing, or defender models. In this paper, we propose a detector-agnostic physics-constrained generative adversarial network (PCGAN) that learns decoupled power-flow sensitivities from historical measurements of deployed sensors. The learned active-power-angle and reactive-power-voltage-magnitude relationships guide perturbations restricted to compromised channels, while adversarial training promotes realistic spatiotemporal patterns. We implement PCGAN and seven existing FDI attacks on the IEEE 300-bus system using one year of observed hourly demand from PJM Interconnection, a regional transmission organization in the United States, and synthetic bus-level load profiles. We compare their performance across 10%, 30%, 60%, and 100% compromised sensors. Attack stealth is assessed against ten machine-learning (ML) detectors and conventional bad-data detection (BDD) methods, while physical impact is quantified through nonlinear alternating-current weighted least-squares state estimation. PCGAN achieves 84.28% mean ML evasion and 22.31% mean spatial load misperception across the four scenarios. Its mean stealth-impact trade-off score reaches 18.69%, compared with 5.61% for the strongest competing attack. PCGAN also achieves 83.45% chi-square BDD evasion and 86.02% largest-normalized-residual BDD evasion. These results demonstrate a favorable stealth-impact balance under varying sensor-access levels and provide challenging attack cases for evaluating FDI detectors.
[LG-117] PRELUDE: Separating Common Components from Target-Specific Dependencies for Multivariate Time Series Forecasting
链接: https://arxiv.org/abs/2611.15155
作者: Dongqing Song,Jianyu Xu,Haobin Ding,Xinyi Guo,Jiahao Du,Ruiqi He,Ruiqi Liu,Yuntong Xie,Chao He
类目: Machine Learning (cs.LG)
*备注: 14 pages, 5 figures, 5 tables
Abstract:Multivariate time series forecasting often improves the prediction of a target by modeling the interactions among variables, and cross-variable modules such as cross-sectional pooling, inter-variable attention, and graph structures are used to construct such interactions. However, by intervening on the cross-variable reading of existing models at inference time, we find that although these models substantially reduce forecasting error through cross-variable interaction, their reading is nearly identical for all targets: the gain appears to come from dependencies among variables but mainly comes from the common component. We term this phenomenon spurious dependency and attribute it to common-component confounding: present in every variable, the common component affects how reading distinguishes sources, so that reading tends to be shared across targets and concentrates on a few similar averages; dependencies thus make up a small share of the read-in content, and a common component composed of several drivers is only partly captured. Theoretical analysis further shows that the common component and dependencies place different requirements on reading and should therefore be handled in different ways. Accordingly, we propose PRELUDE, a separated-reading framework that estimates the common component holistically through cross-sectional pooling and multiple common factors, and reads dependencies for each target via lead-lag relations in the innovation space obtained by prewhitening. Extensive experiments on 16 real-world datasets from four domains show that PRELUDE effectively mitigates spurious dependency and achieves the best overall performance: it attains the best MSE on 23 of the 24 short-horizon and air-quality settings and reduces the average MSE by 6.1% relative to the strongest baseline.
[LG-118] Knowledge Transfer for Conditional Flow Matching: Frozen Estimation and Adaptation
链接: https://arxiv.org/abs/2611.15150
作者: Yian Huang,Jian Huang
类目: Machine Learning (cs.LG); Methodology (stat.ME)
*备注:
Abstract:Conditional generative modeling with limited target data can benefit from representations learned in a related source population. We develop a transfer learning framework for conditional flow matching that accommodates distinct source and target response distributions and dimensions. Two procedures exploit a representation learned from source data: TFM-Freeze estimates a target velocity head with the representation fixed, whereas TFM-Adapt jointly learns the head and a representation adaptation under a functional anchoring penalty. For explicit neural network classes, we establish bounds on conditional squared Wasserstein risk that account for source estimation, target fitting, representation mismatch, and numerical integration. When the source representation admits a smooth correction, adaptation replaces the frozen mismatch contribution by the cost of learning the correction and a controllable regularization bias. An information lower bound identifies regimes in which adaptation attains a strictly faster conditional squared Wasserstein risk rate than any generator restricted to the frozen context information. We also develop independent validation procedure for selecting among fitted conditional generators. We show that the selection procedure satisfies an oracle inequality, and the switching rule guarantees, with high probability, no increase in conditional energy risk relative to a target-only generator. Simulations and real data applications illustrate the gains from our proposed methods and the effectiveness of target validation in guarding against negative transfer.
[LG-119] Serving-Aligned Entire-Space Conversion Modeling via Click-Through Rate Knowledge Distillation
链接: https://arxiv.org/abs/2611.15137
作者: Pavel Dolin,Xun Zou,Jiatian Wu,Deguang Kong
类目: Machine Learning (cs.LG)
*备注:
Abstract:Recommendation systems may use predicted click-through conversion rate (pCTCVR), the product of predicted click-through rate (pCTR) and post-click conversion rate (pCVR), to guide ranking decisions. Many pCVR models build on the Entire Space Multi-Task Model (ESMM), which jointly trains pCVR and auxiliary pCTR towers using pCTR and pCTCVR losses, but deploys only the pCVR tower. At serving time, this tower may be paired with a separately trained pCTR model, creating a training-serving mismatch in pCTCVR when its predictions differ from those of the auxiliary pCTR tower. In this paper, we use knowledge distillation (KD) from logged serving pCTR scores to supervise ESMM’s auxiliary pCTR tower and improve training-serving pCTCVR alignment. We denote variants by KD- \alpha , where \alpha controls the hard-label loss weight. On the Ali-CCP dataset, KD-0.3, the best variant on serving pCTCVR, increases ROC-AUC not only for serving pCTCVR but also for pCVR by 2.1% and 1.6%, respectively, and PR-AUC by 9.9% and 8.8%, respectively, relative to the ESMM baseline. We also test negative downsampling with 5% nonclick retention. Downsampled KD-0.3 increases mean AUC for serving pCTCVR and pCVR by 1.2% and 0.9%, respectively, -and PR-AUC by 8.5% and 4.5%, respectively, relative to downsampled ESMM.
[LG-120] Recompute or Restore? Host-Tier KV for Multi-Model Agent Workflows on a Static-Graph NPU Serving Stack
链接: https://arxiv.org/abs/2611.15126
作者: Sungwon Woo,Seyeon Kim
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG); Performance (cs.PF)
*备注: 14 pages, 6 figures, 5 tables
Abstract:Agent workflows issue dependent model calls that share long context prefixes, so serving them efficiently depends on reusing the key-value (KV) cache of those prefixes. On accelerators that compile model graphs ahead of time, however, the device KV budget is fixed, and prefixes evicted under load must be recomputed. Restoring them from host memory avoids this recomputation, but the transfer occupies the same engine thread as prefill and decode, so its benefit is not guaranteed. In this paper, we study when restoring beats recomputing on a commercial static-graph neural processing unit (NPU) stack (Rebellions ATOM+, vllm-rbln 0.10.2) that serves a Qwen3-8B planner on two cards and a Qwen3-0.6B executor on one. We first show that repeated prefill dominates planner work when default serving collapses: 89.5% of the executed planner prompt tokens had been materialized before. We then build a synchronous host-DRAM tier on the runtime’s public KV copy calls, together with a scheduler-aware event model whose predictions are frozen before each test cohort. Experimental results show that the host tier raises workflow service-level objective (SLO) attainment from 22.8% to 100% and from 9.4% to 99.8% in the two tested overload conditions, and endpoint confirmation on two new seeds supports conservative load gains of 11.8% (chain) and 18.2% (fork), while mixed and bursty workloads remain seed-sensitive. The frozen model predicts median and 99th-percentile (P99) latency within its 15% tolerance in all 24 common-load conditions, including the 16-18% P99 increase that the tier causes for fork workflows, but it misorders three policy contrasts and overestimates the benefit of a cheaper store path (11% predicted against 3.1% measured). The trade-off therefore depends on runtime scheduling and transfer cost as well as cache capacity.
[LG-121] Profile-Based Auditing of Domestic Violence Survey Selection Across South Asia NEURIPS2026
链接: https://arxiv.org/abs/2611.15106
作者: Mohammad Abu Obaida Mullick,Rezaur Rahman Ratul,Mymuna Nourin,Md. Motaharul Islam
类目: Machine Learning (cs.LG)
*备注: Accepted at the AI4Science and New in ML Workshops, NeurIPS 2026
Abstract:Domestic violence (DV) data in the Demographic and Health Surveys (DHS) are collected through a specialized module whose household-level sampling and privacy requirements create a distinct observation process. We ask whether demographic, socioeconomic, and household characteristics form reproducible population profiles and whether these profiles contain information about selection into the DV module across South Asia. Using DHS Individual Recode data for 796,374 women aged 15-49 across Afghanistan, India, Myanmar, Nepal, and Pakistan, we harmonize 61 numeric and 8 categorical features, reduce the profile space with PCA, and identify six interpretable profiles. We then evaluate 22 machine-learning models under primary-sampling-unit-aware holdout validation. The best model, HistGradientBoosting, achieves ROC-AUC 0.9167 and PR-AUC 0.9612, versus 0.8475 and 0.9203 for a design-only baseline. Adding woman-level characteristics yields the largest incremental gain, while explicit interaction terms add little out-of-sample performance. Country-specific analyses reveal substantial heterogeneity, and leave-one-country-out evaluation shows that transferability of the incremental profile signal is limited. These results demonstrate that interpretable ML can audit differential coverage in sensitive survey modules beyond the formal sampling design.
[LG-122] When Does Causality Help in Multivariate Time-Series Anomaly Detection?
链接: https://arxiv.org/abs/2611.15073
作者: Long Zhang,Wei Huang,Aishui Rao,Jian Hu,Binxiang Wu,Jie Li,Chen Gong,Linxi Feng
类目: Machine Learning (cs.LG)
*备注:
Abstract:Causality-aware scoring is common in multivariate time-series anomaly detection, yet the literature rarely states when it helps. Holding one causality-guided autoencoder (DeCauAE) fixed, we vary graph quality, anomaly type and protocol against ground truth. Graph quality fixes the payoff: on a twenty-variable DAG the gain over a graph-free autoencoder decays as the graph is corrupted. Anomaly type fixes the target: the structural factor lifts window F1 from 0.800 to 0.976 where anomalies break conditional structure, and its advantage vanishes below a break severity s 0.3. Protocol fixes the ranking: point-adjusted F1 lets a random score match the full model, while range-aware VUS separates the autoencoder family from the classical baselines. On real benchmarks the causal module pays only where density makes a system eligible and cross-subsample reproducibility selects entities inside it; enabling it everywhere costs 0.012 VUS-ROC. The structural factor also transfers to classical detectors .
[LG-123] Probabilistic State Space Enumeration in Neural Network Computation
链接: https://arxiv.org/abs/2611.15062
作者: Victor O. K. Li,Yang Han,Jacqueline C. K. Lam
类目: Machine Learning (cs.LG)
*备注:
Abstract:Deep neural networks typically realize dense computation by propagating information through all intermediate channels, while many approximation methods rank and retain individual computational components. We ask whether dense neural computation can instead be represented exactly as an expectation over a combinatorial state space, and approximated by evaluating only its most probable configurations. We introduce a novel probabilistic state-space framework for Transformer MLPs. The framework partitions the intermediate dimension into channel groups, defines an input-dependent distribution over binary inclusion states, and uses importance-corrected masked outputs such that the full expectation exactly recovers the dense MLP output. The key algorithmic novelty is NORDER-II, which enumerates complete joint states in decreasing probability without constructing or sorting the exponentially large state space. Truncating this ordering yields an anytime approximation governed by cumulative probability mass and a state budget. On Llama-3.2-1B and Gemma-2-2B, NORDER-II at a state budget of B=2048 achieves 92.4% and 93.8% MLP-output approximation accuracy, compared with 76.9% and 79.8% for Channel Top-k. When the first, middle, and last MLP layers are approximated jointly, WikiText-2 perplexity remains within 0.06% and 0.54% of dense inference. Relative to expected-count Grouped Top-k, NORDER-II yields 82.6% to 126.4% relative accuracy gains with only 0.18% to 0.19% additional local analytic FLOPs. These results establish the significance of joint-state structure beyond atomic salience and point to a new paradigm for neural computation: dense layers can be viewed as structured probabilistic inference over combinatorial computational states. This framework opens a path toward adaptive, probability-mass-controlled computation in pretrained neural networks.
[LG-124] PhysXAI: A Theoretical Framework for Diagnosing Physical Hallucinations in Weather Foundation Models
链接: https://arxiv.org/abs/2611.15053
作者: Anjan Mahapatra,Nikhil Raj
类目: Machine Learning (cs.LG)
*备注:
Abstract:Modern machine learning models for global weather forecasting achieve remarkable predictive accuracy, yet they frequently produce states that violate fundamental physical laws, including conservation of mass, hydrostatic balance, or spectral energy scaling. We term these outputs physical hallucinations and propose PhysXAI, a rigorous diagnostic framework that detects, localises, and attributes such inconsistencies. Unlike existing post-hoc metric suites, PhysXAI is a learnable system that explicitly decouples forecast error from physical plausibility. We formalise the problem through a novel joint condition on numerical and physical residuals, and introduce a multi-branch encoder that extracts representations from conservation, dynamical, spectral, cross-variable, and spatiotemporal residuals. A key innovation is a physics-consistency contrastive learning objective, which we theoretically show the latent representation to be approximately isometric to the physical residual space, ensuring that diagnostic decisions remain grounded in physically meaningful deviations. We further derive an adaptive multi-objective loss with gradient-based weighting to balance competing diagnostic objectives. The framework produces spatially resolved hallucination maps, violation-type classifications, and calibrated Bayesian uncertainty, enabling both detection and interpretable attribution of physical inconsistencies. We conduct a comprehensive comparative evaluation of PhysXAI across ERA5, WeatherBench 2, and MERRA-2, using forecasts from GraphCast, Pangu-Weather, and NeuralGCM, with Python-written code and a fully reproducible experimental protocol. Our theoretical analysis and empirical evaluation establish PhysXAI as a principled framework for diagnosing physical hallucinations and advancing trustworthy AI for weather forecasting.
[LG-125] Mask-Pattern Position Augmentation: Training-Free Long-Context Extrapolation for Diffusion LLM s
链接: https://arxiv.org/abs/2611.15029
作者: Gaurav Sarkar(1),Syed Affan Daimi(1) ((1) International Institute of Information Technology, Hyderabad)
类目: Machine Learning (cs.LG)
*备注: 7 pages, 2 figures, 1 table, 1 algorithm
Abstract:Diffusion large language models (dLLMs) generate by iteratively unmasking tokens with bidirectional attention, but rotary position embeddings (RoPE) still confine them to a fixed pretraining window. Existing training-free extrapolation, notably LongLLaDA, copies NTK-based RoPE scaling from autoregressive (AR) models and applies one static, length-only transform at every position and denoising step. That transplant ignores a signal unique to dLLMs: the evolving mask pattern of the partially decoded sequence. We propose Mask-Pattern Position Augmentation (MPPA). MPPA linearly interpolates indices into the pretrained window, then adds a per-position, per-step offset from cumulative mask density. The offset restores rotary phase resolution that interpolation collapses, and it shrinks as tokens are committed, so late steps see near-clean interpolated geometry. On LLaDA-8B-Instruct at twice its 4096-token window, MPPA reduces perplexity from 19.51 under standard decoding to 3.12 at 8192 tokens, below the in-window figure, with 100% passkey retrieval and unchanged needle-in-a-haystack accuracy. These preliminary 2\times results motivate mask-conditioned position engineering as a diffusion-native path to long-context generation.
[LG-126] From Forecasting to World Modeling: Can Temporal Foundation Models Recover Causal Response? NEURIPS2026
链接: https://arxiv.org/abs/2611.15009
作者: Juwon Hong,Minha Hwang,Venkatesh Shankar
类目: Machine Learning (cs.LG)
*备注: Accepted at the NeurIPS 2026 Workshop on Foundation Models for Temporal Systems (FMTS), non-archival. 15 pages, 2 figures, 3 tables
Abstract:A model that forecasts sales accurately can still mislead a retailer about how sales respond to a price change. We evaluate released tabular and time-series foundation models, as well as a foundation model pretrained specifically to estimate causal effects. We report three findings. First, forecast accuracy does not indicate the accuracy of the predicted effect of price, both on simulated data in which the true effect is known and on real retail sales records. Second, for the tabular foundation models, additional context improves forecasts but not the predicted effect of price. Third, instrumental variables, the standard remedy in econometrics, reduce this error only when supplied in a specific form and increase it for every model under confounding when supplied as raw instruments.
[LG-127] AI4Plasma I: Data-Efficient Surrogate Modeling of Capacitively Coupled Plasma Fields
链接: https://arxiv.org/abs/2611.15002
作者: Zhenyu Zou,Junjue Liao,Hao Yang,Luowei Yin,Chenyu Wang,Yufeng Wei,Zhichao Peng,Chun Chen,Daniel Zhengyu Huang,Kangxin Liu,Haoran Guan,Rubing Han,Yipei Chen,Peiting You,Yuwei Fan
类目: Machine Learning (cs.LG)
*备注:
Abstract:Reference capacitively coupled plasma (CCP) simulations provide spatially resolved plasma fields but are costly to generate across operating conditions. We compare the data efficiency of five surrogate architectures–Coordinate MLP, PCA-Net, U-Net, Fourier neural operator (FNO), and Transolver–for predicting RF-cycle-averaged electron density, ion density, electric potential, and electron temperature in 1D planar and 2D axisymmetric GEC CCP benchmarks. We assess the number of reference simulations needed to achieve 1% and 2% mean relative L2 error for parameter-interior and parameter-boundary test cases, and examine architecture-specific mechanisms. FNO achieves the lowest or joint-lowest reference-case requirement in seven of eight comparisons. In 1D, FNO reaches 1% interior error with 50 training cases, compared with 100 for Transolver and 200 for U-Net. In 2D, FNO and U-Net each reach 1% interior error with 100 cases; FNO and Transolver each reach 2% parameter-boundary error with 200 cases, compared with 400 for U-Net. At the stricter 1% 2D parameter-boundary target, Transolver requires 400 cases versus 800 for FNO. Additional analyses investigate FNO spectral modes and padding, U-Net skip connections, Transolver adaptive slicing, and PCA-Net projection versus prediction error. The results support FNO as a data-efficient baseline for the tested CCP settings, while showing that the preferred architecture depends on the accuracy target and operating-parameter region.
[LG-128] EASTAR: Event-Anchored Semantic and Task-Adaptive Region Representation Learning ICDE2027
链接: https://arxiv.org/abs/2611.14990
作者: Peng Liang(1),Genan Dai(2),Bowen Zhang(2) ((1) College of Applied Science, Shenzhen University, Shenzhen, China, (2) School of Artificial Intelligence, Shenzhen Technology University, Shenzhen, China)
类目: Machine Learning (cs.LG)
*备注: 13 pages. Accepted to the 43rd IEEE International Conference on Data Engineering (ICDE 2027)
Abstract:Urban region representation learning supports a wide range of urban analytics tasks by encoding regional functions and interactions, with human mobility serving as a central signal because it directly captures inter-region interactions. However, most mobility-based methods learn from aggregated origin-destination matrices, where trips driven by different activity purposes are mixed into purpose-agnostic interaction patterns. Although static urban attributes such as POIs and land use can enrich region-level semantics, they do not explicitly ground mobility interactions in activity-purpose signals. As a result, downstream adaptation is often performed over entangled embeddings, making it difficult to select task-relevant functional semantics. To address these limitations, we propose EASTAR, an event-anchored and task-adaptive framework for urban region representation learning. EASTAR treats offline activity events as sparse semantic anchors and uses them to factorize aggregated mobility representations into multiple purpose-related semantic channels. It further introduces free channels to capture latent mobility patterns not covered by event data and a global channel to preserve overall spatial structure. To adapt the learned channel space to different downstream objectives, EASTAR designs a task-adaptive prompt router that initializes task-channel relevance from LLM-derived priors and refines the routing weights through downstream supervision. Experiments on two real-world urban datasets across four prediction tasks show that EASTAR consistently outperforms representative baselines, achieving up to 39.15% reduction in MAE. Ablation and interpretability analyses further demonstrate the effectiveness of event anchoring, semantic channel construction, and task-adaptive routing.
[LG-129] Stress-Testing Sandbagging Detection Under Adaptive Attack AACL
链接: https://arxiv.org/abs/2611.14950
作者: Bhoj Rani Soopal,Aarohi Agrawal,Josh Karthikeyan,Edward Lue Chee Lip
类目: Cryptography and Security (cs.CR); Machine Learning (cs.LG)
*备注: 14 pages, 6 figures. Accepted at the R2A-FM Workshop, AACL-IJCNLP 2026
Abstract:Developers, auditors and researchers read evaluation scores as evidence of what a model can do. A model that sandbags, scoring low on purpose while keeping the capability, turns a low score into false reassurance. Sandbagging detectors are usually tested once against a fixed attack, but a real attacker sees what gets flagged and adapts. We build sandbagging models by fine-tuning small open-weight models to fail math word problems (GSM8K) they can otherwise solve, but only when an audit marker is in the prompt. On Qwen 2.5-3B, we adapt the attack over two retraining rounds against a text-based audit. Two rounds of retraining cut the text audit’s AUC from 0.978 to 0.560 while the marker still removes 40-49 points of accuracy. The attack worked in 7 of 8 retrained seeds, and the audit missed every retraining (AUC 0.451-0.576), including one Llama run. Activation probes also flag 99% of benign fine-tunes’ examples, and matched probes transfer to only 1 of 5 held-out training runs. Eight further trigger attacks are all active under their own prompt but not under the detectors’ prompt. A detector that catches a fixed attack has not been shown to catch an adaptive one.
[LG-130] A Transferable Pretrained 3D Model for Impedance Inversion with Sparse-Well Prior Conditioning
链接: https://arxiv.org/abs/2611.14922
作者: Gen Li,Zhaoyun Zong,Diqiong Jiang,Man Tang
类目: Machine Learning (cs.LG); Geophysics (physics.geo-ph)
*备注:
Abstract:Deep-learning-based seismic impedance inversion commonly requires retraining for each target survey, which may generalize poorly beyond the conditioning locations when well control is extremely sparse. We propose a transferable three-dimensional inversion framework that separates offline learning from target-survey deployment. A masked autoencoder first learns seismic structural representations from unlabeled multi-survey data, after which the frozen encoder is incorporated into an inversion network trained on synthetic geological models with sparse-well-derived low-frequency priors. During deployment, the complete inversion network remains fixed; the available conditioning well is used only to construct the low-frequency prior and align the global impedance range. Experiments on the Overthrust model show that the proposed method achieves the lowest full-volume NRMSE and reduces the P95 trace-wise NRMSE from 0.1310 to 0.1074 relative to the best target-trained baseline, indicating fewer localized large errors under different well locations. In a field survey with one conditioning well and three blind wells, it reduces the mean blind-well NRMSE by 20.95% relative to the target-trained U-Net while maintaining the highest correlation. These results demonstrate that the proposed framework transfers useful impedance information beyond the conditioning location with low deployment cost and without retraining the inversion network.
[LG-131] Epigraph Form Reinforcement Learning-based Safe Trajectory Tracking Control for Connected and Autonomous Vehicles
链接: https://arxiv.org/abs/2611.14903
作者: Xiaoke Yang,Jin Chen,Yafei Wang,Qi Zhang,Ming Li,Haiwen Wu,Chengliang Yin,Yong Zhang
类目: Machine Learning (cs.LG)
*备注:
Abstract:Safe trajectory tracking (STT) aims to enable autonomous vehicles to follow a reference trajectory accurately while avoiding collisions. This problem is particularly challenging for Connected and Autonomous Vehicles (CAVs) in highly dynamic environments. Existing non-learning-based STT methods often use simplified vehicle and interaction models to meet real-time requirements, which may result in suboptimal tracking accuracy. Although Deep Reinforcement Learning (DRL) can help address this limitation, widely used Lagrangian DRL methods often suffer from unstable training because increasing Lagrange multipliers amplify policy-gradient magnitudes. To address these issues, we propose a novel reinforcement learning framework for STT, called STT Control Epigraph Form RL (STTC-EFRL). In this framework, the STT problem is formulated as an epigraph-form min-max optimization problem. Based on this formulation, we develop a learning mechanism that uses a decreasing scalar related to the epigraph to adjust the policy gradient descent direction without increasing the gradient magnitude, thereby improving training stability. We also develop a computationally efficient ray-casting method to compute a polyhedron scaling factor for safety quantification. Experimental results, including Vehicle-in-the-Loop tests, show that STTC-EFRL achieves higher tracking accuracy than the considered non-learning-based baseline and more stable training than widely used Lagrangian DRL methods, while satisfying safety requirements in connected environments.
[LG-132] Social Engineering Attack Detection in Corporate Communications
链接: https://arxiv.org/abs/2611.14897
作者: Hanzla Hamid,Pappu Jha,Nick Rahimi
类目: Machine Learning (cs.LG); Cryptography and Security (cs.CR)
*备注:
Abstract:Social Engineering attacks continue to pose a significant risk to the corporate world with a growing level of sophistication. It is difficult to distinguish between authentic and phishing communication via email, chat, and phone. They pose a risk of exposing trade secrets and confidential communications and can create a chain of attack from lower-level employees up to executives. Therefore, we study this growing threat using real-world datasets to build baseline machine-learning models. Our TF-IDF models classify emails as phishing or safe, achieving accuracy above 90%. We build on this by exploring whether psychological manipulation and persuasion indicators can be measured to play a significant role in detecting phishing. Instead of relying solely on surface-level lexical features, we derive interpretable lexicon-based indicators of urgency, trust, greed, and other persuasion cues from the combined email dataset. By evaluating these manipulation-oriented features on their own, we measure how much signal they carry (ROC-AUC of 0.74 on a held-out test split) and test whether they transfer to held-out email sources.
[LG-133] Machine Learning Moment Closure for the Boltzmann Transport Equation: Enforcing Hyperbolicity
链接: https://arxiv.org/abs/2611.14894
作者: Bongseok Kim,Jiahao Zhang,Guang Lin
类目: Machine Learning (cs.LG); Numerical Analysis (math.NA)
*备注:
Abstract:The Boltzmann transport equation describes particle transport across collisional regimes, but resolving its velocity dependence is expensive. Moment methods reduce this cost by evolving finitely many velocity moments, requiring a closure for unresolved higher-order moments. Data-driven closures can represent these moments accurately, yet without structural constraints they may produce complex characteristic speeds and unstable simulations. Learned Boltzmann closures with guaranteed hyperbolicity have so far been formulated in one spatial dimension; in multiple dimensions, hyperbolicity must hold in every direction simultaneously. We develop, to the best of our knowledge, the first data-driven closure for multidimensional Hermite moment systems of the linear Boltzmann equation that is hyperbolic by construction. The network learns a symmetric positive-definite symmetrizer shared by all spatial directions and restricted to a subspace that preserves the known lower-order transport equations; the unresolved rows follow from a symmetric completion. Every directional Jacobian thus has real eigenvalues and a complete set of eigenvectors for every network output and at any truncation order. The closure is trained only on spatial derivatives of unresolved moments from kinetic reference solutions, with no solver in the training loop, and evolved by a partially conservative discontinuous Galerkin method. In two-dimensional experiments in the transition and scattering-dominated regimes, a single learned third-order system with ten moments remains stable in all cases. Averaged over time, it is more accurate than the classical P7 and P15 truncations with 36 and 136 moments in every experiment, and its density error is about two orders of magnitude smaller than that of the P3 truncation with the same moments.
[LG-134] Characterizing Local Breadth in Language Generation
链接: https://arxiv.org/abs/2611.14887
作者: Jon Kleinberg,Fan Wei
类目: Data Structures and Algorithms (cs.DS); Formal Languages and Automata Theory (cs.FL); Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:Large language models sharpen the validity–breadth trade-off: a generator should avoid hallucinations without collapsing onto a narrow set of valid outputs. We study generation from positive examples of an unknown target in the Kleinberg–Mullainathan model, with arbitrary, possibly adversarial order and no distributional or computational assumptions. Breadth is measured by lower Banach density, which rules out arbitrarily large sparse holes, unlike prefix-based lower asymptotic density. 1. Exact optima in one dimension: For each target K in a countable family, we characterize its optimal lower Banach density b(K) via equivalent combinatorial and topological formulations. One eventually valid generator attains all b(K) simultaneously. Every \alpha\in[0,1/2] occurs as b(K) in some countable family; finite Cantor–Bendixson point rank guarantees b(K)=1/2 , even without a uniform rank bound. 2. Unavoidable trade-offs in higher dimensions: For d\ge2 , there need not be a single best generator: optimizing density for one target may worsen the guarantee for another. An explicit family quantifies this intrinsically multi-objective trade-off. 3. Complete characterization of density profiles: We characterize the full region of simultaneously achievable guarantees, with equivalent combinatorial and topological formulations, after excluding geometrically degenerate rectangles. Every prescribed product \prod_i[0,\alpha_i] , \alpha_i\in[0,1/2] can occur as an exact achievable region for some language collection. 4. Near-half guarantees and the sharp endpoint: Under a near-logarithmic nondegeneracy condition, one generator achieves filtered density at least 1/2-\varepsilon at all finite-point-rank targets, for every \varepsilon0 , improving the previous polylogarithmic threshold. The exact 1/2 endpoint can fail even for a rank-zero singleton family. Subjects: Data Structures and Algorithms (cs.DS); Formal Languages and Automata Theory (cs.FL); Machine Learning (cs.LG); Machine Learning (stat.ML) Cite as: arXiv:2611.14887 [cs.DS] (or arXiv:2611.14887v1 [cs.DS] for this version) https://doi.org/10.48550/arXiv.2611.14887 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Fan Wei [view email] [v1] Wed, 11 Nov 2026 01:26:36 UTC (278 KB) Full-text links: Access Paper: View a PDF of the paper titled Characterizing Local Breadth in Language Generation, by Jon Kleinberg and 1 other authorsView PDFHTML (experimental)TeX Source view license Additional Features Audio Summary Current browse context: cs.DS prev | next new | recent | 2026-11 Change to browse by: cs cs.FL cs.LG stat stat.ML 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-135] FOE-Dreamer: Deployment-Efficient Learning of Cyber Defense Policies in Operational Networks ACSA
链接: https://arxiv.org/abs/2611.14880
作者: Yinuo Du,Christopher Kiekintveld
类目: Machine Learning (cs.LG); Cryptography and Security (cs.CR)
*备注: 14 pages, 5 figures. To appear in the proceedings of the Annual Computer Security Applications Conference (ACSAC)
Abstract:Autonomous cyber defense is widely studied with deep reinforcement learning, but most work trains policies inside simulators because learning requires many trial-and-error interactions. Simulator-based training carries a fidelity gap: trained policies face authentication delays, partial action failures, service dependencies, and user traffic patterns that the simulator did not represent, and degrade on transfer. Training directly in the network being defended removes the gap but makes every interaction slow and costly, so sample efficiency becomes the binding constraint. Model-based reinforcement learning has reduced sample requirements on games and physical robots by orders of magnitude by planning inside a learned world model, but it has not been applied to defenders trained in operational networks. We present FOE-Dreamer (Factored Opponent-Enhanced Dreamer), a model-based reinforcement learning framework that trains a defender online and directly in a network on cloud infrastructure, without a simulator stage. FOE-Dreamer separates three sources of variation in the latent state: a controllable component driven by defender actions, an exogenous component driven by benign user activity, and an opponent embedding inferred from defender-observable signals alone. Separating the components reduces the variance in policy gradients during planning in imagination. Against two scripted attacker profiles with user activity emulated by the GHOSTS framework on an 8-host OpenStack testbed, FOE-Dreamer trains within a 3-day budget on a single GPU and achieves about 2 \times better episode loss than Rainbow and IQN baselines under matched compute. Both the factoring and the opponent model contribute to the gain, and the learned latents are consistent with the intended separation.
[LG-136] Generalize Then Personalize: Wearable-Based Heart Rate Forecasting Under Scarce Ruck-March Data
链接: https://arxiv.org/abs/2611.14849
作者: Muntaqim Ahmed Raju,Ruizhe Ma,John W. Ramsay,David J. Cornell
类目: Machine Learning (cs.LG)
*备注: IEEE BIBM 2026
Abstract:Heart rate (HR) forecasting may provide advance information about near-future cardiovascular response, potentially supporting physiological monitoring and workload-management decisions during ruck marching. However, robust model development is constrained by the difficulty of collecting large-scale ruck-march datasets, creating a scarce-data problem in which models must generalize across participants while adapting to individuals from limited personal data. To address this problem, we benchmark population-only training (POT), personalized-from-scratch training (PFST), and population-pretrained fine-tuning (PPFT) for HR forecasting during loaded ruck marching using data from eight active-duty U.S. Army medics across thirteen sessions. LSTM, Attention-LSTM, and CNN-LSTM backbones were evaluated against a persistence baseline at 30 s and 60 s horizons. Across all neural architectures and horizons, PPFT yielded lower mean MAE, RMSE, and endpoint MAE than both POT and PFST. Participant-level paired analysis after averaging across three training seeds showed that, relative to PFST, PPFT reduced trajectory MAE by 2.22-4.30 bpm and endpoint MAE by 1.24-2.88 bpm. All six trajectory-MAE comparisons remained significant after Holm adjustment ( p_adj = 0.0469), while four of six endpoint-MAE comparisons remained significant after adjustment. At 30 s, persistence achieved the lowest MAE (3.555 \pm 1.274 bpm), while CNN-LSTM PPFT achieved the lowest RMSE (4.669 \pm 1.916 bpm); LSTM PPFT achieved the lowest endpoint MAE (5.093 \pm 2.365 bpm). At 60 s, Attention-LSTM PPFT achieved the lowest MAE (5.573 \pm 1.939 bpm) and RMSE (7.134 \pm 2.260 bpm), while CNN-LSTM PPFT achieved the lowest endpoint MAE (7.553 \pm 2.357 bpm). These results suggest that population pretraining followed by participant-specific fine-tuning may provide a practical strategy for scarce-data wearable HR forecasting.
[LG-137] SPIRE: Task-agnostic embodied continual learning through autonomous situation inference and causal motor-primitive reuse
链接: https://arxiv.org/abs/2611.14840
作者: Yiyang Jia,Mingyang Lyu,Gongce Cao,Feifei Zhao,Yinqian Sun,Yi Zeng
类目: Machine Learning (cs.LG); Robotics (cs.RO)
*备注: 14 pages, 4 figures, 5 tables (including appendix tables)
Abstract:Embodied continual learning requires controllers to infer changing situations, retain prior behavior and reuse motor skills without task labels or known boundaries. The challenge is deciding when to recover earlier behavior and which skills can support new learning. We introduce Situation-guided Primitive Integration for Recurrent Embodiment (SPIRE), a recurrent-spiking system that infers latent contexts from action-outcome evidence and learns context-specific motor actors online. Mature actors become protected sources that combine with a target-context residual to drive recurrent-spiking motor populations. Across coordinate, three-joint and hidden-force control, SPIRE maintained control, autonomously inferred situations, retained prior behavior and accelerated acquisition of an unseen source composition, reducing mean interactions to criterion by approximately 40% relative to matched fresh controllers. Rewinding compatible source actors impaired early acquisition in every matched intervention, demonstrating causal reuse. Against seven recurrent continual-learning and modular-control comparators, SPIRE achieved 2.4 times the mean prior-learning advantage of the strongest eligible task-free alternative selected for each paired stream. These results demonstrate autonomous recovery and causal reuse of motor skills for new learning in continuous embodied control.
[LG-138] What Does an RF Fingerprint Identify? A Cross-Corpus Evaluation of Hardware-Family-Associated and Unit-Level Signal
链接: https://arxiv.org/abs/2611.14835
作者: Hyojung Han
类目: Cryptography and Security (cs.CR); Machine Learning (cs.LG)
*备注: 20 pages, 4 tables, no figures
Abstract:RF fingerprinting benchmarks report how well a classifier separates transmitters, not whether that separability reflects the individual unit or its hardware family. We compare a benchmark’s real hardware grouping with size-matched random groupings of the same devices, using calibrated, unclipped variational lower bounds under label-blind, device-level cross-fitting on devices never seen in training. On WiSig (OFDM Wi-Fi), the generation grouping exceeded all 999 matched random groupings (permutation p = 0.001, the resolution floor), although no grouping attained a positive certified bound, and the direction survives removing channel equalization. A preregistered replication on a LoRa corpus (chirp spread spectrum, three hardware models), with a front-end frozen on synthetic data, again exceeded all 999 (p = 0.001), and two sensitivity analyses judged by a reversal rule fixed in advance did not reverse the direction. This is a directional replication only: the excess magnitude reflects a miscalibrated null, the chain-rule decomposition leaves a residual of +0.296 bits, and capture order is undocumented. In a same-model corpus of 122 ESP32 radios, a temperature-calibrated held-out bound of +3.338 bits (label entropy 6.931 bits) indicates substantial device-specific identity, not separable from per-device capture-session effects; relying on the calibrated bound was decided after the uncalibrated bound changed sign. On synthetic transmitters a high-capacity convolutional decoder reaches 98.5% of a known 6-bit ceiling, and the audit protocol leaves chain-rule residuals of +0.072 bits (synthetic, fixed physical features) and +0.060 bits (WiSig, audit features). Two further corpora were excluded by preregistered ingest gates. Across WiSig and LoRa, hardware-family groupings consistently produced stronger identity-associated signal than size-matched random groupings.
[LG-139] SlipStream: Enabling HBF for GPU Systems by Mitigating the Latency and SRAM Overheads HPCA
链接: https://arxiv.org/abs/2611.14776
作者: Anish Saxena,Hritvik Taneja,Christos Kozyrakis,Aamer Jaleel
类目: Hardware Architecture (cs.AR); Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG)
*备注: To appear in 2027 IEEE International Symposium on High-Performance Computer Architecture (HPCA)
Abstract:Large language model (LLM) inference demands ever more memory capacity for larger models and batch sizes. High-Bandwidth Flash (HBF) replaces the DRAM dies of HBM with NAND flash and provides up to 16x the capacity at comparable read bandwidth. However, HBF can incur a read latency of nearly 20 \mu s, 20x that of HBM, so on-demand HBF accesses slow inference by up to 12.8x. Conventional cache-line prefetchers require 20x more in-flight tracking state to cover this latency. Tensor-streaming designs, such as H3, follow virtual-address order as opposed to kernel-demand order, and therefore need hundreds of megabytes of buffering. We propose SlipStream, a software-hardware co-designed prefetcher that turns prefetching into a demand-paced producer-consumer pipeline. In software, SlipStream profiles the repeatable weight-page demand order of data-independent LLM kernels and inserts a staging warp that replays this order with non-blocking STAGE instructions. In hardware, the staging warp acts as the producer and the work warps as consumers. Our key insight is to retire each STAGE tracking entry only when demand first accesses the page, as opposed to when the page returns from HBF. Demand thus paces staging and bounds how far staging leads consumption. SlipStream tracks 4 KB pages instead of cache lines and allocates buffer slots only when pages return, which decouples the buffer size from outstanding traffic. Across seven dense models spanning 16-800 GB, SlipStream with a 40 MB staging buffer achieves 92.4% of the performance of an ideal HBF with HBM-like latency. H3 achieves only 43.1%, even with a 320 MB buffer. The buffer sizing of SlipStream is independent of HBF latency and workload. SlipStream with 16 MB buffer, 20x smaller than that of H3, performs within 4% of SlipStream-40MB. SlipStream also supports Mixture-of-Experts models and reaches 91.2% of ideal performance (40 MB buffer). Comments: To appear in 2027 IEEE International Symposium on High-Performance Computer Architecture (HPCA) Subjects: Hardware Architecture (cs.AR); Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG) ACMclasses: B.3.2; B.4.3 Cite as: arXiv:2611.14776 [cs.AR] (or arXiv:2611.14776v1 [cs.AR] for this version) https://doi.org/10.48550/arXiv.2611.14776 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-140] UniPipe: Unified Inference Serving and Fine-Tuning for Deadline-Constrained DNN Surrogate Workflows
链接: https://arxiv.org/abs/2611.14748
作者: Jaiaid Mobin,Avinash Maurya,M. Mustafa Rafique,Bogdan Nicolae
类目: Machine Learning (cs.LG); Performance (cs.PF); Systems and Control (eess.SY)
*备注:
Abstract:DNN surrogates are increasingly used to accelerate scientific workflows that must process high-rate sensor streams in near real time, exemplified by Ptychographic imaging at next-generation X-ray light sources. Such surrogates deliver 100-1000x higher throughput than iterative solvers, enabling faster steering of experiments. To preserve surrogate accuracy during an experiment, as the input distribution evolves, state-of-the-art deployments freeze the pretrained surrogate or run a separate CUDA context that fine-tunes concurrently with serving. However, multi-context concurrency causes significant contention on the GPU, inflating both the mean and the variance of inference latency, triggering deadline misses in workflows where sensor data has to be processed within a hard per-unit deadline. To address this challenge, we propose UniPipe, a single-context system that unifies inference serving and fine-tuning of a DNN surrogate on a single edge GPU. UniPipe is driven by three key insights: both serving and fine-tuning require a forward pass and can be fused into one; the per-sample latency of modern GPUs is non-linear in batch size; and the workload is repetitive across intervals, enabling offline scheduling. Specifically, UniPipe fuses training and inference data in a single forward pass with a selective backward pass on the training subset, and uses a polynomial-time scheduler that exploits the latency curve to compute the batch composition. We evaluate UniPipe on a real-world Ptychographic reconstruction workflow with sensor rates up to 5000 patterns/sec.; UniPipe lowers reconstruction MSE by up to 30% and inference MSE by up to 50% over a pretrained baseline, and reduces the deadline-miss rate by 13%-22% over a multi-context baseline. UniPipe’s unified deployment reduces the latency spread from 106% to 57% of the mean, and increases total throughput compared to multi-context baseline.
[LG-141] Beyond Foreign-Key Connectivity: Semantic Relationship Representation for Relational Deep Learning
链接: https://arxiv.org/abs/2611.14706
作者: Noy Cohen-Shapira
类目: Databases (cs.DB); Machine Learning (cs.LG)
*备注:
Abstract:Relational databases encode information through tables and primary–foreign key constraints. Although recent Relational Deep Learning (RDL) approaches represent these databases as heterogeneous graphs, a foreign-key connection primarily specifies structural connectivity between records. Its semantic interpretation may involve additional information that is not explicitly represented by connectivity alone. This paper introduces a conceptual framework for constructing relationship-aware representations from relational databases. The framework distinguishes three complementary aspects: (i) schema-defined relationship semantics, including endpoint types and foreign-key roles; (ii) association entities and explicitly available relationship-level attributes; and (iii) contextual information describing the temporal and structural circumstances of a connection. We formulate a Relationship Representation Construction (RRC) protocol that maps relational schema elements and database records to node representations, relationship descriptors, and contextual representations. A key principle of the protocol is semantic ownership: attributes are assigned to records, relationships, or context according to their meaning, rather than merely their physical location in a database table. Ambiguous attributes are not automatically promoted to relationship features. The resulting representations can be incorporated into existing relational graph learning architectures through relationship-conditioned message passing. The proposed framework is complementary to recent work on relational message passing, graph transformers, and task-dependent graph construction. We discuss its conceptual foundations, illustrate the construction protocol using a multi-table database, and identify research questions and empirical evaluations required to establish its potential benefits. Subjects: Databases (cs.DB); Machine Learning (cs.LG) Cite as: arXiv:2611.14706 [cs.DB] (or arXiv:2611.14706v1 [cs.DB] for this version) https://doi.org/10.48550/arXiv.2611.14706 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Noy Cohen-Shapira [view email] [v1] Tue, 10 Nov 2026 20:13:30 UTC (12 KB)
[LG-142] Beyond Row-Level Prediction: Evaluating and Learning Composition for Reusable Table Representations
链接: https://arxiv.org/abs/2611.14697
作者: Ali Younes,Saeed Ghoorchian,Maximilian Schambach,Georgia Chalvatzaki,Johannes Höhne
类目: Machine Learning (cs.LG)
*备注: Preprint. Under review
Abstract:Tables are compositional, with content spread across a schema, cell values, rows, and columns. Large-scale operations such as retrieval, overlap estimation, deduplication, and data curation need these local signals composed into one reusable embedding vector per table. Table representation methods, however, are mostly evaluated through row-level prediction or supervised downstream tasks, which do not measure table-level embedding quality directly. We introduce a benchmark that evaluates table-level embeddings through four desiderata: consistency under partial views, discriminability across label granularities, robustness to benign perturbations, and efficiency. Across twelve methods on synthetic and real corpora, lightweight schema- and text-based fingerprints are among the strongest. The sharpest contrast comes from two tabular foundation models, TabPFN and ConTextTab: although both are strong row-level predictors, their states, composed by mean, max, or three further parameter-free pooling rules, trail the strongest fingerprints. We therefore ask whether the gap lies in the encoder or in the compositional rule. Keeping ConTextTab frozen, we learn only the compositional rule over its cell states, using support-conditioned distribution matching (SCDM), a self-supervised loss we propose that matches the distribution of embedding similarities to the distribution of partition overlaps within and across tables. Across all fifteen head configurations, the learned rule raises the Spearman correlation between embedding similarity and partial-view overlap from 0.04 to 0.43-0.50 on synthetic data and improves triplet ranking, clustering, and retrieval, with smaller but systematic gains on real data. A downstream parent-table retrieval task confirms these findings. How local states are composed, and with what objective, substantially affects the quality of table-level embeddings.
[LG-143] Closing the Loop: Co-Evolving EM for Irregular Time Series Generation in Lifted Representations NEURIPS2026
链接: https://arxiv.org/abs/2611.14659
作者: Idan Arbiv,Gal Fadlon,Omri Azencot
类目: Machine Learning (cs.LG)
*备注: Accepted to NeurIPS 2026; The first two authors contributed equally and are co-leading authors
Abstract:Regular time-series generation from irregular time series is typically reduced to one-shot imputation: complete each partial sequence once, freeze the completions, and train a generator on the resulting surrogate dataset. We show that this open-loop strategy is structurally flawed. Point imputers collapse ambiguous missing regions to conditional averages, while stochastic imputers become stale as the generator evolves; the binding constraint is therefore iteration, not imputer design. We introduce a co-evolving Monte Carlo EM framework that closes the imputer-generator loop. Each E-step samples missing values from the posterior of the current generator, and each M-step retrains the generator on these completions, preserving unconditional generator training. Since our diffusion prior operates in a lifted 2D representation while observations live in time-series space, we further introduce Posterior Sampling for Lifted Representations (PSLR), which enforces operator-consistent conditioning, uncertainty preservation, and manifold consistency during the E-step. Across ten datasets and four corruption regimes, our method achieves state-of-the-art performance, improving over the strongest baselines by 59% in discriminative score, 15% in predictive score, 73% in Context-FID, and 71% in feature-correlation error. Ablations show that our approach reduces average discriminative score by 73% over single-step baselines, closing most of the remaining gap to the clean-data oracle. Code is available at this https URL.
[LG-144] StructRL: Model-Free Reinforcement Learning with Replay Graphs
链接: https://arxiv.org/abs/2611.14599
作者: Ivo Nowak
类目: Machine Learning (cs.LG)
*备注:
Abstract:StructRL turns the experience collected during reinforcement learning into an explicit replay graph. Once the graph contains a path to the task goal, dynamic programming selects actions on the observed part of the state space; these actions form a graph policy that is defined only where the required transitions have been observed. StructRL transfers these actions to model-free learners: as action preferences for a DQN Q-model and as target directions along shortest observed paths for a SAC actor. For DQN on three sparse-reward MiniGrid tasks, final success rises from 0.10 to 0.90 (DoorKey), 0.10 to 0.80 (KeyCorridor) and 0 to 0.50 (ObstructedMaze); plain DQN does not catch up with three times the budget. For SAC on continuous PointMaze, graph-derived direction labels raise final success from 0.17 to 1.0 on seven matched runs that reached the goal. Distilling the graph policy into a fresh Q-model gives a DoorKey 6x6 controller that reaches the goal from 91% of reachable states, against 38% for plain DQN. Scaling experiments identify three bottlenecks: incomplete graph support, structural aliasing, and loss of graph information in high-dimensional inputs.
[LG-145] Adversarial Risk in the Generative AI Era Necessitates Dropping the Small Epsilon Ball NEURIPS2026
链接: https://arxiv.org/abs/2611.14594
作者: Andrew C. Cullen,Neil G. Marchant,Paul Montague,Jiani Xie,Benjamin I.P. Rubinstein
类目: Machine Learning (cs.LG); Cryptography and Security (cs.CR)
*备注: neurips 2026, 10 pages, position paper on adversarial risk, certified robustness
Abstract:Adversarial Machine Learning research is currently centered on the search for the epsilon-ball: minimal perturbations designed to fool models while remaining imperceptible to humans. We argue that this fixation on \ell_p norms is a metrological trap that communicates a distorted perception of risk to defenders, one that is unaligned with real-world system vulnerabilities. This paradigm misleads defenders into optimizing for narrow, mathematical robustness at the expense of systemic security, while simultaneously incentivizing attackers to exploit non-overlapping alternative adversarial pathways. Ultimately, we argue that the dominance of the \ell_p ball does not just fail to reduce harm—it may actively facilitate it by communicating a false sense of security against the wrong threats.
[LG-146] An Asynchronous Federated Learning Platform for Lung Adenocarcinoma Grading via Foundation Models and Structural Pruning
链接: https://arxiv.org/abs/2611.14590
作者: Moon-Il Joo,Md Ariful Islam Mozumder,Hee-Cheol Kim
类目: Machine Learning (cs.LG)
*备注:
Abstract:Classification of lung adenocarcinoma into poorly differentiated (PD) and well-to-moderately differentiated (WMD) subtypes plays an important role in patient survival; however, consistent classification is difficult due to inter-observer variability among pathologists. Collecting data from multiple hospitals is also challenging due to patient privacy concerns. This study proposes a web-based federated learning system that shares only locally trained model weights without sharing patient data between institutions. The UNI2 foundation model and CLAM-SB were used to classify pathological whole slide images. A four-client federated learning system was constructed, and the NLST cohort was reserved exclusively for external validation. In centralized evaluation, joint training with CPTAC and TCGA outperformed single-institution training on external validation, confirming the necessity of multi-institutional data integration. FedAvg, FedProx, FedBN, and FedNova were compared as federated learning algorithms under identical experimental conditions. Test AUC ranked as FedBN, FedProx, FedNova, and FedAvg. FedNova showed the most balanced classification performance between PD and WMD. L1-norm structural pruning at 20%, 30%, and 40% sparsity was applied to reduce model transmission size. FedAvg, FedProx, and FedBN showed a tendency of decreasing Test AUC under pruning, while FedNova achieved improved Test AUC across all sparsity levels. A web-based application was developed supporting slide upload, local training, AI classification, and attention heatmap-based visualization. The combination of FedNova and 40% pruning was confirmed as a practical approach that improves communication efficiency and classification performance while protecting patient privacy.
[LG-147] he norm not the labels: training and evaluating finite-element surrogates for stress analysis in the energy norm
链接: https://arxiv.org/abs/2611.14574
作者: Ruifeng Cao,Xidan Song
类目: Computational Engineering, Finance, and Science (cs.CE); Machine Learning (cs.LG)
*备注: 45 pages, 6 figures; supplementary material (37 pages, 4 figures) appended
Abstract:Finite-element surrogates predict the stresses of a new design in one evaluation, but are usually trained on solutions with a displacement loss and compared by it, although engineering decisions rest on stresses. Application in engineering: stress analysis of perforated plates and solids with cavities, parametric families in which holes concentrate stress. Contribution to artificial intelligence: training and evaluating such surrogates in the energy norm, with or without solutions, and screening their predictions without solutions. For linear elastostatics, minimising the assembled discrete potential energy is exactly supervised regression in this norm: the energy gap of a prediction is the strain energy of its stress error. A displacement loss under-weights the stiff modes that carry the strains. In pre-registered comparisons, changing only the norm of the supervised loss cut a three-dimensional transformer’s median energy-norm error 7.5-fold and its median von Mises error 6.1-fold; in two dimensions it more than halved the von Mises error, also on held-out instances. In three dimensions, the transformer trained on the energy alone came within a factor of 1.3 of the energy-norm-supervised one in median energy-norm and von Mises error. It was more accurate than the displacement-supervised one on all five metrics in every seed. Given the stiffness matrix, a Ritz projection was more accurate than the label-free transformer, in the seed means, on every metric but peak stress. No network transferred to a substantially finer mesh. The energy ranks predictions of a load case, and its sign flags those worse than the zero field.
[LG-148] Beyond the First Hit: Towards Tight Guarantees for k-means
链接: https://arxiv.org/abs/2611.14573
作者: Aravind Reddy
类目: Data Structures and Algorithms (cs.DS); Machine Learning (cs.LG)
*备注: 48 pages, 1 figure (3 panels). Exact-arithmetic certificate data and verification code included as ancillary files
Abstract:Let R_k be the worst-case ratio between the expected cost of the k seeds chosen by k -means++ and the optimal k -means cost. In their seminal paper introducing k -means++, Arthur and Vassilvitskii (SODA 2007) showed that 2\ln k\le R_k\le8(\ln k+2) , and Makarychev, Reddy and Shan (NeurIPS 2020) improved the upper bound to 5(\ln k+2) . We prove R_k\le\frac1163233+2.108,H_k-2 for every k\ge3 by a computer-assisted argument, and R_k\le5+2.3139,H_k-2 for every k\ge2 by a short analytic argument; here H_j is the j th harmonic number. Since H_k-2\le1+\ln k , this lowers the leading coefficient of \ln k from 5 to 2.108 . We also determine R_2=5 and R_3=185/32 exactly; both are suprema that are approached but not attained. On the lower side we prove R_k\ge2(1+H_k-1) for every k , and give a second family of lower bounds that is larger than this for 2\le k\le18 . Previous analyses bound the expected cost of a cluster just after it is first sampled by a constant times its optimal cost, and then charge that cost for the rest of the run. We instead track how a cluster’s cost evolves after its first hit: a cluster left with a large residual cost is likely to be sampled again before it accumulates much of the logarithmic charge, and each repeat repairs it.
[LG-149] Provenance-tracked experimentally validated neural operator surrogates of three-dimensional laser melting with keyholes and free-form paths
链接: https://arxiv.org/abs/2611.14561
作者: Alix Benoit,Elia Iseli(Light-Matter Dynamics Group, Laboratory for Multifunctional Materials and Interfaces, Empa)
类目: Computational Engineering, Finance, and Science (cs.CE); Machine Learning (cs.LG)
*备注: Main text: 44 pages, 13 figures, 7 tables. Supplementary Information: 19 pages, 9 figures, 3 tables
Abstract:A three-dimensional thermo-fluid-dynamic simulation of laser melting resolves the melt pool and vapor depression but costs from one to thousands of core-hours per track. We train FNO, RIGNO and GAOT neural operators that predict the temperature, melt-pool and free-surface fields. Our datasets comprise 3D temperature and melt geometry fields covering lack-of-fusion to keyhole conditions under varied laser power, scan speed, beam radius, beam shape and preheat, and include free-form scan paths with speed changes and up to four beams. We generate them with FLOW-3D WELD, using solver settings validated against published measurements for four materials. We implement an AiiDA framework to store the provenance of every simulation, model and evaluation, so each tabulated metric traces back to the solver and model inputs behind it. On held-out simulations, the best architecture scores below 2 % relative temperature error on quasi-steady straight tracks, below 4 % on single-beam free-form paths and about 5 % with two to four beams. Inference takes under a second per predicted arbitrary time step on our single-GPU setup. We extend a surrogate trained on Ti-6Al-4V to Inconel 718 by growing its training set autonomously, and the extended models reach a relative temperature error of about 3 % or less on both Inconel 718 and Ti-6Al-4V. At 20 Ti-6Al-4V and Inconel 718 conditions held out of training, the two-alloy RIGNO predicts melt-pool and vapor-depression depths within 30 % of published measurements, and its depths differ from FLOW-3D at the same conditions by 2 % on average. Furthermore, these results provide a scalable framework which can readily be applied to computationally intensive simulations in fields other than laser melting. Comments: Main text: 44 pages, 13 figures, 7 tables. Supplementary Information: 19 pages, 9 figures, 3 tables Subjects: Computational Engineering, Finance, and Science (cs.CE); Machine Learning (cs.LG) ACMclasses: I.2.6; J.2 Cite as: arXiv:2611.14561 [cs.CE] (or arXiv:2611.14561v1 [cs.CE] for this version) https://doi.org/10.48550/arXiv.2611.14561 Focus to learn more arXiv-issued DOI via DataCite
[LG-150] Probe Before You Preach: Conformal Ambiguity Sets and Neuro-Symbolic Counterfactual Probing for Diagnosing Algebraic Misconceptions in Middle School Mathematics
链接: https://arxiv.org/abs/2611.14557
作者: Caihong Xu,Xiangyu Yu
类目: Computers and Society (cs.CY); Machine Learning (cs.LG)
*备注: 19 pages, 1 figure. Theoretical framework, method, and public-graph audit; withdrawn hardcoded offline/classroom outcome tables omitted; classroom study proposed for future validation
Abstract:Distinct mathematical misconception rules can produce the same incorrect answer, making diagnosis from a single response ambiguous. We formalize this algebraic collision and derive a conditional Bayes error bound; internally generated reasoning does not add student-specific information beyond its inputs under an explicit conditional-independence assumption. We propose Mondrian Adaptive Conformal Misconception Sets (MACMS) and Symbolic-Verified Counterfactual Micro-Probing (Sym-EIG). MACMS has curriculum-group marginal coverage under exchangeability, not a per-student correctness guarantee. Sym-EIG searches for probes whose outputs separate modeled candidate rules; identification additionally depends on the candidate set and response model. An audit of a local public Eedi graph finds 2,118 of 2,258 linked constructs associated with at least two misconception IDs; construct-level associations alone do not establish same-answer collisions. Other project performance tables require recomputation because their script assigns numerical results directly. No diagnostic performance or classroom learning benefit is claimed here. A classroom study is proposed for future validation.
[LG-151] Learning Nonlinear Transient Dynamics in Fluid Flow Networks with the Neural Basis Method
链接: https://arxiv.org/abs/2611.14534
作者: Aijaz Nazir,Yuhe Wang,S.K. Rasel,Min Wang
类目: Computational Engineering, Finance, and Science (cs.CE); Machine Learning (cs.LG); Fluid Dynamics (physics.flu-dyn)
*备注: 28 pages, 5 figures. Under review
Abstract:Fluid flow networks, from pipeline infrastructure to biological vasculature, are governed by nonlinear edgewise hyperbolic laws coupled through continuity and conservation at junctions. Many-query tasks require rapid evaluation across unseen parametric settings, but conventional methods re-solve the system for each query at prohibitive cost. Existing physics-informed neural approaches impose the governing equations through penalty losses over generic neural approximators; these losses act only as optimization surrogates, with no guarantee that their reduction reduces solution error, and the resulting optimization is prone to ill-conditioning, particularly for hyperbolic systems. Building on the “representation-then-projection” principle of the neural basis method, we introduce a new approach for modeling fluid flow networks. Fixed neural bases define the representation space on each edge, while projection of the discretized equations determines the expansion coefficients, embedding edgewise laws into a conservative formulation. Edge subdomains are coupled through non-overlapping domain decomposition, and a neural operator learns, directly in this reduced space, the parametric map to the expansion coefficients. On looped and branching network topologies, we reproduce references with relative L^2 errors below 0.02%, whereas the physics-informed neural network baseline fails. Across unseen boundary conditions in both interpolation and extrapolation, we maintain mean pressure errors below 1% and mean flux errors below 3.3%, while the physics-informed DeepONet baseline reaches 8-18% pressure errors and 85-95% flux errors. These results highlight the framework’s potential for real-time monitoring, design exploration, and uncertainty quantification of fluid flow networks.
[LG-152] Enhancing Resolution and Reliability with Attention Mechanisms in Nuclear Accident Surrogate Modeling
链接: https://arxiv.org/abs/2611.14532
作者: Wonung Jeong,Sangam Khanal,Jeongik Lee,Sangseung Lee,Joongoo Jeon
类目: Computational Engineering, Finance, and Science (cs.CE); Machine Learning (cs.LG)
*备注: Code and supplementary material: this https URL
Abstract:Severe accident (SA) analysis of nuclear power plants relies on system codes such as MAAP, whose computational cost limits large-scale applications including dynamic probabilistic safety assessment and reinforcement learning. This study proposes the Accident Binary-Continuous Transformer (ABC-Transformer), a decoder-only surrogate combining causal self-attention with temporal attention pooling. It jointly embeds continuous thermal-hydraulic variables with binary safety-system and severe accident management guideline (SAMG) states, enabling 5-minute-resolution forecasting over 72-hour transients. Trained on 11,000 MAAP simulations of a total loss of component cooling water scenario for the OPR1000, it achieved lower one-step teacher-forcing error than CNN and recurrent baselines at fine resolutions. Autoregressive reliability was characterized as a full per-scenario distribution over 1100 held-out scenarios. For scheduled-sampling (SS) training, the median R^2 is 0.947 \pm 0.006 and 3.8 \pm 1.0 % of scenarios fail ( R^20 ; mean \pm SD over four training seeds), while errors accumulate with rollout length. A Mahalanobis self-distance computed from each surrogate rollout, before comparison with the MAAP reference, ranks these failures (ROC-AUC 0.895, reference seed). A threshold at the 40th percentile of calibration scores accepts 36.8 \pm 0.5 % of 550 evaluation scenarios, with no accepted rollout below R^2 = 0 for any seed, and routes the rest to MAAP.
[LG-153] he Overstated Cost of AI Fairness in Criminal Justice
链接: https://arxiv.org/abs/2608.01299
作者: Ignacio Cofone(1),Warut Khern-am-nuai(2) ((1) University of Oxford, (2) McGill University)
类目: Computers and Society (cs.CY); Machine Learning (cs.LG)
*备注: 48 pages, 1 figure, 12 tables. Published in the Indiana Law Journal
Abstract:A dominant critique of algorithmic fairness holds that increasing fairness reduces predictive accuracy, imposing a cost on society. We challenge that assumption by empirically analyzing the COMPAS dataset. We make two contributions. First, using causal inference methods, we show that racial bias is not only present in the COMPAS dataset but is also amplified by the models trained on it. Widely used models do more than replicate existing bias; they exacerbate it. This undercuts both the assumption that algorithmic decision-making offers a neutral improvement over human judgment and the weaker claim that it merely mirrors preexisting human bias. Second, we reframe the fairness-accuracy tradeoff. Applying fairness constraints does not necessarily cost predictive accuracy in criminal justice. Prediction systems operationalize concepts such as risk through implicit and often flawed normative choices about what to predict and how. The tradeoff claim assumes that the unconstrained model’s prediction is an optimal baseline. Fairness constraints can instead correct distortions introduced by biased outcome variables: rearrest data, in this case, captures and magnifies systemic racial disparities. Under some interventions, therefore, fairness carries none of the cost presumed in policy debates. These dynamics extend beyond criminal justice to lending, hiring, and housing, where biased outcome variables reinforce inequality independently of proxy selection. We draw out what this implies for how law and policy should approach fairness adjustments in criminal law.
[LG-154] Quantifying structural stability of transparent models for general insurance pricing
链接: https://arxiv.org/abs/2611.17308
作者: Félix Côté,Marie-Pier Côté
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注: 16 tables and 9 figures
Abstract:The short duration of general insurance contracts leads to rapid distributional shifts in the data, warranting regular model updates. This process may yield a post-update model that differs considerably from the pre-update model in both predictions and structure. Structural stability is upstream of predictive stability, yet existing structural stability metrics are mostly restricted to simple decision trees. Thus, we first introduce the notion of functional transparency, characterising models with a prediction function that can be expressed as a sum of univariate functions for main effects and bivariate functions for pairwise interactions. Then, we propose the Mean Squared Shape function distance, a metric to quantify the structural stability of functionally transparent models, which are particularly relevant in actuarial science. Using three public auto insurance datasets, we present a benchmark study to demonstrate the usefulness of our metric in identifying a stable and accurate transparent model. Generalised linear models with centred numerical covariates are stable, and explainable boosting machines (EBMs) offer a compromise between stability and performance. We also find that stabilisation strategies, such as sequential modelling and fixed hyperparameters, can substantially improve stability for both EBM and XGBoost models, with negligible impact on performance.
[LG-155] Programming Quantum Circuit Parameters with Large Language Models
链接: https://arxiv.org/abs/2611.17246
作者: Tiago de Souza Farias,Caio Moraes Porto,Matheus da Silva Fonseca,Nicolas Armando Cabrera Carpi,Anderson de Araujo Oliveira,Celso Jorge Villas-Boas
类目: Quantum Physics (quant-ph); Machine Learning (cs.LG)
*备注: 31 pages
Abstract:Large language models (LLMs) have recently demonstrated the ability to generate parameters for variational quantum circuits without conventional optimization. However, the conditions under which this approach is effective remain unclear. Here, we investigate a frozen, general-purpose LLM that generates complete circuit parameter vectors from symbolic circuit and Hamiltonian descriptions, with optional iterative energy feedback. We evaluate the approach on general random Hamiltonians, structured parent Hamiltonians, and molecular systems. For general Hamiltonians, LLM-generated parameters exhibit larger energy errors than conventional Adam optimization, while additional feedback generally provides limited improvement. In contrast, for structured Hamiltonians, the LLM can generate promising low-energy parameter configurations, suggesting that identifiable physical and mathematical structure facilitates parameter inference. We also show that direct LLM predictions of minimum energies can violate the Hamiltonian spectral lower bound, whereas generated circuit parameters allow independent evaluation of physically realizable energy expectation values. Finally, we investigate blind circuit interpretation and find that the LLM can predict measurement outcomes even when it fails to identify the underlying quantum algorithm. These findings highlight the role of problem structure, the limitations of iterative LLM optimization, and the importance of independently verifiable predictions in quantum parameter programming.
[LG-156] Linear Regression with Self-Selection Bias in Polynomial Time
链接: https://arxiv.org/abs/2611.17228
作者: Felix Zhou
类目: atistics Theory (math.ST); Data Structures and Algorithms (cs.DS); Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:We revisit the problem of estimating k linear regressors with self-selection bias in n dimensions up to error \varepsilon with the maximum self-selection criterion, as introduced by Cherapanamjeri, Daskalakis, Ilyas, and Zampetakis [CDIZ23, STOC]. Our main result is a \mathrmpoly(n,k,1/\varepsilon) time algorithm for the unknown-index variant of this problem, thereby improving upon the running times of the best known algorithms due to Gaitonde and Mossel [GM24] and Kalavasis, Mehrotra, and Zhou [KMZ26, COLT] with exponential dependence on k . Our algorithm succeeds in the presence of correlated noise between regressors, which captures the related setting of max-affine regression. Consequently, we also give the first polynomial-time algorithm for this problem, improving on the prior works of Ghosh, Pananjady, Guntuboyina, and Ramchandran [GPGR22, IEEE Transactions on Information Theory] and Gaitonde and Mossel [GM24] with exponential dependence on k . Prior works exhaustively search over a k -dimensional subspace spanned by the regressors, using conditional moments of the response to test candidates. In contrast, our algorithm is driven by an improved understanding of the moments of the covariates. We show that, restricting to observations with large response values, the degree- (\log n) Hermite moment of the covariates approximately decomposes as a sum of tensor powers of the regressors. The first step of our algorithm can be viewed as a tensor decomposition algorithm for these implicitly represented high-order tensors, which outputs unit vectors parallel to the regressors. A secondary step based on a simple variant of ordinary least squares extracts estimates for the regressors given these unit vectors. We believe our analytic insight and algorithmic components may have applications to designing efficient algorithms for other latent-variable problems. Subjects: Statistics Theory (math.ST); Data Structures and Algorithms (cs.DS); Machine Learning (cs.LG); Machine Learning (stat.ML) Cite as: arXiv:2611.17228 [math.ST] (or arXiv:2611.17228v1 [math.ST] for this version) https://doi.org/10.48550/arXiv.2611.17228 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-157] Data Archive for 2D and 3D Electrical Impedance Tomography with moderate conductivity contrast phantoms
链接: https://arxiv.org/abs/2611.17225
作者: S.J. Hamilton,A. Hauptmann,A. Hänninen,V. Kolehmainen,J. Niskanen,T. Savolainen,J. Toivanen
类目: Image and Video Processing (eess.IV); Machine Learning (cs.LG)
*备注: 11 pages, 6 figures. Data available at this https URL
Abstract:This document describes open Electrical Impedance Tomography (EIT) measurement data that were collected from biomedical imaging inspired phantoms with moderate conductivity contrasts. The data were collected with the KIT4 and KIT5 EIT measurement devices at the University of Eastern Finland, and are intended for EIT image reconstruction algorithm development and testing. The data from thoracic imaging inspired cylindrical and chest-shaped phantoms are suitable for 2D algorithms, and the data from an intracerebral hemorrhage imaging inspired head-shaped phantom with resistive skull are suitable for 3D algorithms. In addition to the measurement data, computational meshes and example reconstruction codes are provided and sample reconstructions included.
[LG-158] An Invariant Neural-Network Eddy-Viscosity Closure with Iterative Solver Coupling for Molten Salt Fast Reactor Thermal-Hydraulics
链接: https://arxiv.org/abs/2611.17067
作者: Zaid Abulawi(1),William Oliveira(2),Akzhol Almukhametov(1),Jun Fang(3),Ling Zou(3),Rui Hu(3),Stefano Lorenzi(2),Yang Liu(1) ((1) Department of Nuclear Engineering, Texas AM University, (2) Department of Energy, Nuclear Engineering Division, Politecnico di Milano, (3) Nuclear Science and Engineering Division, Argonne National Laboratory)
类目: Fluid Dynamics (physics.flu-dyn); Machine Learning (cs.LG)
*备注: 42 pages, 12 figures
Abstract:System thermal-hydraulics codes such as the System Analysis Module (SAM) close turbulence on coarse meshes with algebraic zero-equation mixing-length models. These models cannot represent the non-equilibrium turbulence, flow separation, and large-scale recirculation of open-cavity advanced reactors. We develop an invariant neural-network eddy-viscosity closure for SAM and apply it to the Molten Salt Fast Reactor (MSFR). The closure predicts the turbulent kinetic energy and timescale of each element from the flow and geometry of its neighbors, without solving turbulence transport equations. The resulting non-negative eddy viscosity enters the implicit diffusive operator, which keeps the momentum equations well conditioned. We prove that the closure is exactly invariant to translation, Galilean boost, rotation, reflection, and neighbor ordering. A network trained on high-order Nek5000 fields sees a different feature distribution on the coarse SAM mesh. We remove this covariate shift by iteratively retraining the network on SAM flow states against fixed Nek5000 targets. For a held-out steady state, the coupled closure reproduces the single recirculation cell and eddy-viscosity plume of the Nek5000 reference. In a 120 s pump-coastdown transient, it halves the velocity error of the mixing-length model, from 37.6% to 17.7%, and matches the Nek5000 eddy viscosity with a time-averaged relative error of 12%. The remaining velocity error comes from the coarse mesh. The coupled transient takes about 0.2 core-hours, compared with about 330 core-hours for Nek5000. The results show that the closure can recover two-equation RANS turbulence behavior in a coarse-mesh system code at a small fraction of the CFD cost.
[LG-159] Random Exploration is Random
链接: https://arxiv.org/abs/2611.17011
作者: Yingkang Luo,Hua-Dong Xiong,Hanbo Xie,Caroline Phelps,Madeline Harms,Ryan Smith,James Waltz,Robert Wilson
类目: Neurons and Cognition (q-bio.NC); Machine Learning (cs.LG)
*备注:
Abstract:Decision making often requires balancing exploitation of familiar options with exploration of uncertain alternatives. One strategy for exploration is random exploration, in which choices become more variable and less tied to the option that currently appears best. However, it remains unclear whether this variability reflects genuine stochasticity or instead a deterministic structure that traditional cognitive models fail to capture. We examined this question using the Horizon Task in an aggregated dataset of 1,392 participants completing over 1,270,695 trials. In this task, exploration becomes more valuable when participants have more future choices, and choice variability increases under longer horizons. We tested whether more expressive neural network models could recover deterministic structure from this additional variability. Although neural networks provided better overall fits to behavior than the traditional model, long-horizon choices remained consistently less predictable than short-horizon choices. These findings show that limited model expressiveness cannot fully account for the increased variability associated with random exploration, suggesting random exploration truly is substantially random.
[LG-160] Uncertainty-aware quantum machine learning based on Dempster-Shafer theory
链接: https://arxiv.org/abs/2611.16903
作者: Marie Kempkes,Jakob Spiegelberg,Evert van Nieuwenburg,Vedran Dunjko
类目: Quantum Physics (quant-ph); Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:A large body of work in quantum machine learning has centered on characterizing the expressivity and performance of quantum models. In contrast, little attention has been paid to the reliability of predictions beyond average performance measures such as accuracy. In this work, we introduce an uncertainty-aware quantum classification framework based on Dempster-Shafer theory. This theory allows to express uncertainty through weights assigned not only to individual classification labels but to sets of labels. We introduce a training strategy that enables parameterized quantum circuits to learn such assignments directly from data. This provides an efficient way to prepare the corresponding quantum states and enables the application of different post-processing strategies. In particular, it offers a practical setting for quantum Dempster-Shafer fusion procedures with polynomial resource requirements, even when the corresponding mass functions are defined over exponentially many subsets. Overall, our work represents a step toward enabling quantum models to express uncertainty about their predictions, thereby extending the evaluation of quantum models beyond predictive performance alone.
[LG-161] Stochastic Kernel Mode Decomposition: Information Uncertainty and Conditional Regime Inference
链接: https://arxiv.org/abs/2611.16874
作者: Boumediene Hamzi,Houman Owhadi
类目: Machine Learning (stat.ML); Computational Engineering, Finance, and Science (cs.CE); Machine Learning (cs.LG); Dynamical Systems (math.DS); Pattern Formation and Solitons (nlin.PS)
*备注:
Abstract:When do mode energies preserve regime information as measurement quality changes? Stochastic Kernel Mode Decomposition (sKMD) fits joint regime laws on signed KMD mode coordinates and transfers the reference mode posterior to regime probabilities and conditional reconstructions. For centered Gaussian regimes with independently varying known noise variances, fixed coordinate energies retain the full regime posterior precisely when all classes share their cross-covariances and class variances can differ only at isolated vertices of the covariance graph. We extend the characterization to nonzero means and show that freely varying correlated noise forces a fixed linear log-odds decoder to retain the full quadratic feature span even for two distinct centered regimes. At fixed quality, a three-regime example has no sufficient two-energy basis. Its smallest Bayes log-loss gap over bases of condition number at most \kappa has sharp order \kappa^-1 . Exact Bayes energy posteriors supply numerical gap estimates for specified bases. Additional bounds relate feature compression, independently learned mode spans and sensor-noise error to log loss. Two controlled experiments with 64 independent panels each show that signed information can improve prediction over energy summaries, while integration gives small gains over strong signed controls and depends on how the competing point rule is calibrated and how accurately observation noise is measured.
[LG-162] Systematic Evaluation of Training Data Generation Strategies for Machine Learning-Based Molecular and Materials Property Prediction
链接: https://arxiv.org/abs/2611.16792
作者: Minh Nghia Nguyen,Beate Paulus,Akhil S. Nair
类目: Materials Science (cond-mat.mtrl-sci); Machine Learning (cs.LG)
*备注:
Abstract:Machine-learning (ML) models for molecular and materials property prediction are often trained on large datasets curated through arbitrary or generic sampling strategies. Such approaches can lead to the over- or under-representation of regions of chemical and materials space that are most relevant to the prediction task. An important challenge is therefore how to efficiently curate training data to maximize predictive performance while minimising the number of labelled examples required. Here, we systematically compare training data acquisition strategies for molecular and materials property prediction, ranging from random sampling to various active learning approaches. We evaluate their performance across different model classes, representations, datasets, and target properties. Our results show that the effectiveness of data acquisition strategies depend strongly on the interplay between the model, dataset, and prediction task, although uncertainty-based sampling generally outperforms alternative strategies. Based on these findings, we derive guidelines for selecting data acquisition strategies under a given labelling budget.
[LG-163] Price co-movement and price information diverge across Nordic electricity zones after the energy crisis
链接: https://arxiv.org/abs/2611.16657
作者: My Thi Diem Phan,Tuyen Trung Truong,Hoai Phuong Ha,Dat Thanh Nguyen
类目: General Economics (econ.GN); Machine Learning (cs.LG); Applications (stat.AP)
*备注: Supplementary material in the ancillary file. Code and results: this https URL
Abstract:After the 2021-2022 energy crisis, day-ahead prices in Nordic bidding zones no longer move together. We ask whether this divergence also changes the information carried by prices. The answer matters for integrating variable renewable energy (VRE), because it can help identify where wind and solar forecasts are worth producing, publishing, and using. We measured the co-movement of each Norwegian and Swedish zone with Germany-Luxembourg (DE-LU) in 2018-2025. We then used a single model, selected on 2024 data, to forecast 2025 prices at 11:00 on the previous day and tested which inputs improved accuracy in each zone. The zones moved alike until 2019, but co-movement diverged in Sweden in 2020 and in Norway in 2021, and the split has persisted. In 2024, four southern zones co-moved strongly with DE-LU (intraday correlations of 0.54-0.71), while four northern zones co-moved weakly (0.25-0.30). Yet price information did not follow this split. Previous-day neighbouring prices reduced mean absolute error by at most 2.2% in any zone, equally in coupled and isolated zones. Removing the transmission system operators’ wind and solar forecasts increased mean absolute error by at most 6% in Norway, 8-16% in Sweden, and 31% in DE-LU, and the ordering held in 2024. Within Norway, the value of these forecasts showed no detectable dependence on reservoir filling. The divergence in co-movement was therefore not matched by a divergence in price information, which differed mainly between national generation systems. Thus, a zone’s coupling to the continent does not indicate the value of VRE forecasts there.
[LG-164] Measurement-Calibrated Synthetic Pretraining for Rapid Adaptation of UAV Channel Predictors
链接: https://arxiv.org/abs/2611.16646
作者: Qiuheng Zhou,Wei Jiang,Donglin Wang,Hans D. Schotten
类目: ignal Processing (eess.SP); Machine Learning (cs.LG)
*备注: 6 pages. Accepted version of a paper presented at the 2026 IEEE 104th Vehicular Technology Conference (VTC2026-Fall), Boston, MA, USA
Abstract:Unmanned aerial vehicle (UAV) links can change faster than route-specific channel data can be collected, so a channel predictor must often be adapted from only a short target-flight calibration segment. This paper studies measurement-calibrated synthetic pretraining as an offline source prior for rapid adaptation of channel state information (CSI) predictors. The synthetic source is calibrated to flight-only power evolution, temporal shape variation, frequency selectivity, and phase disturbance, and is then combined with measured source-flight fine-tuning before target-flight adaptation. Under chronological held-out flight splits, synthetic-to-real (S2R) fine-tuning reduces the measured-data budget needed to reach useful prediction in the low-data regime. For 50 ms-ahead prediction, it crosses a latest-channel-estimate persistence baseline with 50 measured training windows, roughly 0.5 s of 10 ms-spaced supervised windows, whereas training from scratch requires 600 windows. A target-prefix adaptation study further shows that a lightweight gated recurrent unit predictor can reuse the pretrained source prior with one prediction-layer update epoch, positioning measurement-calibrated synthetic pretraining as a reusable prior for scarce target-flight measurements.
[LG-165] When and How SignSGD Outperforms SGD: A Functional Scaling Law Perspective NEURIPS2026
链接: https://arxiv.org/abs/2611.16560
作者: Zilin Wang,Binghui Li,Jia-Nan Wang,Lean Wang,Jinbo Wang,Lei Wu
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注: Accepted at NeurIPS 2026 as a spotlight paper
Abstract:We study how the sign operator interacts with gradient noise to shape the learning dynamics and scaling behavior of sign-based optimizers. In power-law kernel regression with diagonal features, we establish a functional scaling law (FSL) for SignSGD under general learning rate schedules. The FSL describes the learning dynamics through two processes: signal learning and noise forgetting. The key mechanism is noise-induced preconditioning: gradient noise smooths the sign operator, inducing an effective preconditioned SGD. Noise-curvature alignment enables this preconditioning to transform the effective Hessian from \mathbfH to \mathbfH^1/2 . The resulting dynamics exhibit faster signal learning but slower noise forgetting than SGD. This tradeoff yields a regime-dependent comparison with SGD: (i) SignSGD achieves improved data efficiency in the hard-task regime; (ii) both optimizers achieve the same data efficiency in the easy-task regime, but SignSGD maintains this efficiency at larger batch sizes, requiring fewer optimization steps for a fixed data budget. Power-law kernel regression experiments support the proposed mechanism and predicted scaling behavior. LLM pre-training experiments further suggest that the SignSGD FSL better captures the batch-size dependence of loss dynamics for sign-based optimizers than the original SGD FSL.
[LG-166] APES: Terminal Adjoint-Propagated Error Scheduling for Financial Diffusion Samplers
链接: https://arxiv.org/abs/2611.16484
作者: Yilie Huang
类目: Computational Finance (q-fin.CP); Machine Learning (cs.LG)
*备注: 9 pages. Accepted at the 7th ACM International Conference on AI in Finance (ICAIF 2026)
Abstract:Financial diffusion samplers are often reused with a small number of function evaluations (NFE) inside scenario and risk pipelines, where numerical time-grid error can distort terminal quantities such as covariances, tails, and drawdown proxies. We study offline grid construction for a frozen probability-flow ODE: the score network, conditioning, Euler solver, report normalization, and deployment NFE remain fixed. TAPES (Terminal Adjoint-Propagated Error Scheduling) estimates signed local defects by Richardson comparisons, transports them through the remaining discrete flow and a terminal-report adjoint, and optimizes the squared norm of their sum before a capped direct refinement. We show that this transported signed sum is the first-order terminal-report error and quantify, under an aligned-defect model, the inefficiency of diagonal criteria. Across six budgets on a regime-switching simulation and chronologically held-out Mag7 contexts, TAPES gives the lowest design- and audit-report errors in all 24 dataset–budget–report rows within the reported five-pipeline panel. Cross-budget mean relative audit-error reductions are about 10–11% on Regime-SV and about 1% on Mag7, with Bonferroni intervals excluding zero; all displayed winners persist under a doubled-resolution reference, and a small Regime-SV Heun probe preserves the TAPES-first audit ordering. TAPES changes only the offline schedule and retains the prescribed deployment NFE; claims concern numerical fidelity to frozen generators.
[LG-167] Distribution-First Framework for Learning Risk-Sensitive Individualized Treatment Rules
链接: https://arxiv.org/abs/2611.16274
作者: Zhen Fang,Guanting Chen,Yufeng Liu
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:Learning a risk-sensitive individualized treatment rule (ITR) with patient covariates is challenging when the treatment criterion depends on the full conditional outcome distribution rather than only its conditional mean. Existing methods are largely criterion-specific, making it difficult to compare multiple treatment criteria within a common pipeline, such as quantiles, conditional value-at-risk (CVaR), and complex preference-based objectives such as cumulative prospect theory. We propose a distribution-first framework for risk-sensitive ITR learning that first estimates treatment-specific conditional outcome distributions and then constructs the target ITR by plug-in criterion evaluation. The proposed framework unifies treatment rule learning across multiple criteria and accommodates hard-to-optimize objectives within the same estimation pipeline. On the theoretical side, it yields a general error propagation result for locally Hölder-stable criteria, linking first-stage distribution estimation error to downstream criterion estimation and treatment regret, together with sharper regret bounds under standard margin conditions. Empirically, we show that the proposed framework outperforms direct criterion-specific methods in multiple settings and further evaluate it on the ACTG175 HIV clinical trial, a randomized study comparing several antiretroviral regimens in HIV-infected adults.
[LG-168] HybridCT: ML Accelerated Tomography for Climate Relevant 3D Cloud Observations
链接: https://arxiv.org/abs/2611.16255
作者: Linda Forster,Nicholas J. LaHaye,Marcin J. Kurowski,Anna Jungbluth,Bernhard Mayer,Anthony B. Davis
类目: Atmospheric and Oceanic Physics (physics.ao-ph); Machine Learning (cs.LG)
*备注:
Abstract:Low-level clouds are one of the largest sources of uncertainty in climate predictions, partly because we lack detailed three-dimensional (3D) measurements of their internal structure. Here, we present HybridCT, a cloud tomographic reconstruction scheme accelerated by machine learning (ML) that recovers 3D cloud extinction directly from passive multi-angle observations. An ML emulator maps radiances from multiple viewing angles to cloud optical thickness (COT), producing path-integrated projections, and a Simultaneous Algebraic Reconstruction Technique (SART) explicitly inverts these projections to retrieve 3D extinction. Training observations are generated from synthetic 3D cloud fields using radiative transfer simulations for nine viewing angles between \pm 70 ^\circ . HybridCT reconstructs extinction with RMSE 13.3 km ^-1 in 3.5 s, compared to 22.7 km ^-1 in approx. 30 min for the state-of-the-art physics-based tomography framework AT3D, a speedup exceeding 500 \times . HybridCT thus provides a scalable route toward large-scale observations of 3D cloud structure relevant to radiative-budget interpretation and evaluation of high-resolution cloud simulations.
[LG-169] Generative AI models for exploring diffusion dynamics from multi-frame particle tracking
链接: https://arxiv.org/abs/2611.16163
作者: Jianan Wei,Qiang Zhang,Hao Sun,Yi Yang,Wenguan Wang
类目: Fluid Dynamics (physics.flu-dyn); Machine Learning (cs.LG)
*备注:
Abstract:Particle-based volumetric flow measurement techniques have proven their efficacy in quantitatively assessing the dynamic characteristics of unsteady flows within the realm of fluid mechanics. Despite recent advances in multi-frame particle tracking, an open challenge in this area is maintaining trajectories in temporally under-sampled 3D sequences, where inter-frame displacements are large. In this work, we develop DiffTR-Flow3D, a diffusion-based model that estimates flow motion from spatiotemporal three-dimensional particle recordings via a stochastic diffusion process. DiffTR-Flow3D conditions the diffusion process with both global and local particle dynamics and enhances flow estimation by leveraging historical particle trajectories. Extensive experiments, including numerical simulations and real-world measurements, demonstrate that DiffTR-Flow3D achieves state-of-the-art displacement estimation accuracy compared with learning-based flow estimators, and, when integrated into particle tracking pipelines, substantially improves their long-term tracking performance, while exhibiting robustness and generalization across diverse flow patterns. This work is expected to provide fluid researchers with an effective tool for the consistent and precise analysis of particle trajectories in temporally under-sampled 3D sequences with large inter-frame displacements.
[LG-170] CupOFMoCA: Coupled Objective-Guided Discrete Flows for Molecular Conjugate Assembly NEURIPS2026
链接: https://arxiv.org/abs/2611.16099
作者: Ruoxi Rosie Zhang,Ziang Li,Jiatao Gu,Pranam Chatterjee
类目: Biomolecules (q-bio.BM); Machine Learning (cs.LG)
*备注: Published at NeurIPS 2026. (Proceedings of the 40th Conference on Neural Information Processing Systems, Sydney, Australia)
Abstract:Molecular conjugates, including PROTACs and peptide-drug conjugates (PDCs), derive their function from the joint behavior of multiple coupled components, yet most generative approaches design these components independently and combine them only after generation. Such staged pipelines ignore cross-component dependencies and often produce conjugates that are chemically invalid, fall outside empirical conjugate distributions, or lose function upon assembly. In this work, we introduce Coupled Objective-Guided Discrete Flows for Molecular Conjugate Assembly (CupOFMoCA), a discrete generative method that enforces cross-component chemical constraints during conjugate generation. CupOFMoCA restricts guided updates to chemically feasible conjugate states and reweights local transitions using molecular design objectives. Across three PDC systems, coupled generation achieved 100% conjugation validity among valid peptides and increased valid yield over matched post-hoc filtering, with comparable predicted binding scores across the full valid sets. In three-seed evaluations of three PROTAC systems, CupOFMoCA achieved the highest assembly validity on two systems and generated more assembly-valid molecules per second than the evaluated baselines. We further evaluated the generated PROTACs with post-hoc DegradeMaster scoring and explicit-solvent molecular dynamics, supporting their predicted activity and conformational properties. Together, these results establish CupOFMoCA as an efficient generator of molecular conjugates, supporting therapeutic development via coordinated control of assembly and molecular properties.
[LG-171] Global-Local and Window-Concatenated Functional Random Forests: Early and Late Fusion of Window-Specific Functional Principal Components
链接: https://arxiv.org/abs/2611.15816
作者: Fabrizio Maturo,Annamaria Porreca
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Methodology (stat.ME)
*备注: 36 pages, 12 figures, 9 tables
Abstract:Functional random forests built on a single functional principal component decomposition of the whole domain may dilute features confined to short intervals, whose location is rarely known in advance. We propose two random forests based on window-specific functional principal components. The global-local functional random forest (GL-FRF) is a late-fusion method that stacks a tuned full-curve forest and one forest per window with penalised convex weights; it contains the full-curve forest as a feasible solution, and its weights indicate which parts of the domain drive the prediction. The window-concatenated functional random forest (WC-FRF) is an early-fusion method that trains a single tuned forest on the concatenated window-specific scores. On twenty UCR datasets, GL-FRF improved on a tuned full-curve forest with the same number of trees on 16, with Brier score reductions of up to 0.21, significantly after Holm adjustment. No method was best on more than six datasets. WC-FRF was more accurate than GL-FRF on 13 datasets but failed when the signal required the full curve, whereas GL-FRF had the smallest mean and maximum excess over the best method across all 32 simulated and real settings, which makes it the safer choice when the location of the signal is unknown.
[LG-172] Reliable Near-Field Source Enumeration and Positioning Informed by Two-Stage MUSIC
链接: https://arxiv.org/abs/2611.15722
作者: Jiaying Li,Haifeng Wen,Changsheng You,Yuanwei Liu,Hong Xing
类目: ignal Processing (eess.SP); Information Theory (cs.IT); Machine Learning (cs.LG)
*备注: 15 pages, 14 figures
Abstract:Accurate and reliable near-field multi-source localization is essential for location-aware processing in future wireless systems. However, coherent multipath propagation complicates the direct recovery of line-of-sight (LoS) source positions, while existing estimation methods generally lack explicit control of the risk of source underestimation and positioning uncertainty. In this paper, we investigate reliable source enumeration and positioning in near-field coherent multipath environments. Specifically, we develop a conformal risk control (CRC) based source enumeration scheme that calibrates the penalty coefficient of an eigenvalue-based criterion to control the expected miss-detection ratio at a prescribed level. We further propose MUSIC-Net, a model-based deep learning framework informed by two-stage MUltiple SIgnal Classification (MUSIC), which learns the LoS-related subspace and surrogate distances through subspace-distance-based training, thereby enabling direct source positioning without estimating non-line-of-sight (NLoS) path parameters or performing subsequent path association. To quantify the uncertainty of the estimated positions, we develop a split conformal prediction (SCP) based calibration method that constructs angle-range predicted regions with finite-sample simultaneous coverage guarantees. Numerical results show that the proposed framework effectively controls the risk of source underestimation, while providing improved positioning accuracy with tighter predicted regions than benchmarks given the same target coverage levels.
[LG-173] Learning Mixtures of Exponential Family
链接: https://arxiv.org/abs/2611.15698
作者: Mansi Sood,Devavrat Shah
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Signal Processing (eess.SP)
*备注: Preliminary version appeared in ISIT 2026
Abstract:Motivated by causal inference and physical computing systems, we consider the problem of learning mixture of two bounded-support exponential-family distributions with a shared interaction parameter matrix and mixture component-specific linear parameter vectors. The unknown sequence of component labels belongs to a known structured class. We propose an estimator based on a convex loss that avoids the log-partition function; for each fixed label assignment, estimation is a convex optimization problem. We show that the conditional population loss uniquely identifies the interaction matrix and the linear parameter attached to each sample. For a global empirical minimizer, we establish finite-sample bounds on the interaction error in Frobenius norm and the average squared linear-parameter error. The bounds depend on the logarithm of the number of admissible label assignments, and, for fixed dimension and model bounds, imply consistency when this complexity grows sublinearly with sample size. Additional occupancy and separation conditions yield component and label recovery.
[LG-174] Analysis choices decide whether untrained networks beat trained ones at V1: a multiverse analysis of model-brain RSA
链接: https://arxiv.org/abs/2611.15533
作者: Nils Leutenegger
类目: Neurons and Cognition (q-bio.NC); Machine Learning (cs.LG)
*备注: 31 pages, 6 figures; code and data pipeline: this https URL
Abstract:Untrained convolutional networks have been reported to rival trained networks as models of visual cortex, and we had reported an untrained-network advantage over backpropagation at V1 in THINGS fMRI, together with a dependence on evaluation resolution. Under a pre-specified ridge readout, the untrained network scored below the trained one at every pooling grid (median r difference -0.043 at g = 4, 0 of 5 seeds positive), and also in MSE skill and per-fold r. In exploratory analyses, across 36 combinations of unit normalisation, distance measure and voxel normalisation (28 distinct analyses), the sign of the V1 contrast depended on the analysis (intervals above 0 in 5 and below 0 in 9 distinct analyses), and in our networks the resolution dependence of the original convention appeared in only one cell. Raw correlation-distance representational similarity analysis (RSA) covaried with the alignment of model and voxel patterns with the unit-mean profile (Spearman 0.97 across our networks, 0.79 across six ImageNet architectures); the advantage was absent from the component orthogonal to the mean profile, and the parallel component alone did not resolve it. On the unweighted angular component of centred patterns, trained networks were ahead at V1 (our network and 5 of 6 ImageNet models) and at LOC (all seven); the norm-weighted angular term was not resolved. A luminance-and-contrast bank scored above every network under ridge (point estimates), and adding network features under a shared penalty did not improve prediction. Intervals are stimulus-bootstrap intervals for three subjects, and 6 pre-stated predictions were not met. We recommend reporting sensitivity to centring over the evaluated stimuli on both sides, reporting components separately, and including a small multiverse and simple baselines.
[LG-175] Bet Imaginatively not Historically in Independent-Data Sequential Testing NEURIPS2026
链接: https://arxiv.org/abs/2611.15456
作者: Nathaniel Xu,Feng Liu,Danica J. Sutherland
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注: Accepted to NeurIPS 2026
Abstract:Sequential testing avoids some of the many controversies and drawbacks related to traditional batch testing with p-values. In particular, the testing by betting paradigm has developed rapidly in recent years; it frames the level of evidence against the null as the wealth of a player in a betting game. Naturally, how much a player should bet is a key consideration, and should reflect their confidence in what they know about the next observed data point. Much current work uses standard betting strategies developed for adversarial problems. In the stochastic setting, where the observed data is independent and stationary, these approaches may be suboptimal. We propose a simple betting strategy, follow the leave-one-out e-power (FLOE), that exploits the i.i.d. nature of a data stream to estimate the e-power of our current test on the underlying distribution. We average not over the historical evidence collected so far, as generic online learning-based methods do, but instead over the equally-likely possible data sequences we might have seen based on shuffling the data. With an additional principled regularization technique, this simple drop-in method yields substantial improvements in testing performance on a variety of kernel two-sample and independence testing problems.
[LG-176] Escaping the Curse of Dimensionality in One-Step Flow-Based Generative Models NEURIPS2026
链接: https://arxiv.org/abs/2611.15447
作者: Konstantin Yakovlev
类目: atistics Theory (math.ST); Machine Learning (cs.LG)
*备注: 52 pages; accepted at NeurIPS 2026
Abstract:Flow and diffusion-based generative models achieve state-of-the-art quality, but require multiple forward passes at inference time, making them computationally costly. Consistency models overcome this by learning a flow map directly, enabling few-step generation and dramatically reducing the computation needed at test time. However, formal statistical guarantees for such one-step generators are scarce. Existing results either suffer from the curse of dimensionality or rely on impractical loss functions. We investigate whether one-step flow-based models can escape the curse of dimensionality under the assumption that the data distribution is generated by a low-dimensional surface perturbed with Gaussian noise. In this setup, we derive approximation and generalization error bounds for the flow map estimate that circumvent the curse of dimensionality. Our findings provide the first principled explanation for practical success of one-step flow-based generative models.
[LG-177] Communication-Efficient Federated Multi-View Clustering via View-Aware Quantization
链接: https://arxiv.org/abs/2611.15270
作者: Khadija Ardif,Meryeme Ayache,Nidal Nasser,Ouiame Marnissi
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注: 13 pages, 5 figures, 6 tables
Abstract:Federated multi-view clustering (FedMVC) has recently attracted increasing attention for its ability to exploit complementary clustering structures across multiple clients while preserving data privacy. However, existing FedMVC methods often suffer from high communication overhead due to the repeated transmission of high-dimensional representations, which limits their scalability in practical federated systems. Moreover, different views may contribute unequally to the clustering performance, calling for adaptive communication strategies. In this paper, we propose a view-aware quantization algorithm for communication-efficient FedMVC. By dynamically assigning quantization bits based on iteratively updated view importance scores, the proposed algorithm effectively balances clustering performance and communication efficiency under a global bit budget constraint. To validate its generality, we integrate our approach into four state-of-the-art FedMVC methods. Extensive experiments on multiple benchmark datasets demonstrate that view-aware quantization achieves up to 6.59x communication reduction while maintaining and in some cases surpassing baseline clustering performance, confirming the effectiveness and scalability of the proposed approach.
[LG-178] Understanding the Asymptotic Transience of Q-Learning via Distributional Expansion
链接: https://arxiv.org/abs/2611.14926
作者: Shengbo Wang
类目: Optimization and Control (math.OC); Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:Q-learning has been empirically observed to exhibit substantial transient error. This transient remains uncaptured by the leading Gaussian limit because the central-limit normalization removes the smaller-scale displacement generated by the initial nonlinear evolution. On the other hand, worst-case finite-sample guarantees, while accounting for transient effects, can obscure the mechanisms responsible for the error on a particular problem instance. We develop a finer asymptotic analysis that gives an asymptotically exact characterization of this transient for a fixed instance. Specifically, for synchronous Q-learning in finite discounted Markov decision processes, we separate the last-iterate error into an optimal-policy temporal-difference (TD) component and a nonnegative displacement caused by maximization. The centered TD component has Gaussian fluctuations at scale n^-1/2 , whereas the maximization transient is of order n^-1 and converges almost surely, after rescaling by n , to a random limit. This limit represents the propagated effect of maximization errors generated before the greedy policy stabilizes. We combine this characterization with an Edgeworth expansion of the TD component to identify the corrections to its Gaussian approximation, separating the effects of initialization, non-Gaussian TD fluctuations, and maximization. The results reveal how transient errors arise for individual Q-learning instances and how their effects persist below the central-limit scale.
[LG-179] Nested Ensemble Learning with Topological Data Analysis for Graph Classification and Regression
链接: https://arxiv.org/abs/2611.14900
作者: Innocent Abaa,Umar Islambekov
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Applications (stat.AP)
*备注: This is the published version of the article. Published in International Journal of Topology, MDPI, DOI: https://doi.org/10.3390/ijt2040017 . Link to the official version: this https URL
Abstract:We propose a nested ensemble learning framework that utilizes Topological Data Analysis (TDA) to extract and integrate topological features from graph data, with the goal of improving performance on classification and regression tasks. Our approach computes persistence diagrams (PDs) using lower-star filtrations induced by three filter functions: closeness, betweenness, and degree2 centrality. To overcome the limitation of relying on a single filter, these PDs are integrated through a data-driven, three-level architecture. At Level-0, diverse base models are independently trained on the topological features extracted for each filter function. At Level-1, a meta-learner combines the predictions of these base models for each filter to form filter-specific ensembles. Finally, at Level-2, a meta-learner integrates the outputs of these filter-specific ensembles to produce the final prediction. We evaluate our method on both simulated and real-world graph datasets. Experimental results demonstrate that our framework consistently outperforms base models and standard stacking methods, achieving higher classification accuracy and lower regression error. It also surpasses existing state-of-the-art approaches, ranking among the top three models across all benchmarks.
[LG-180] Synthetic Data Augmentation for Fairness under Data Imbalance: Theory and Adaptive Allocation
链接: https://arxiv.org/abs/2611.14870
作者: Zhengchi Ma,Kaifeng Zhang,Anru R. Zhang
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Methodology (stat.ME)
*备注:
Abstract:Data imbalance across sensitive groups and class labels is common in classification problems and can contribute to substantial fairness disparities. We study how such imbalance affects Equal Opportunity and how synthetic data augmentation can mitigate the resulting unfairness. We first identify two mechanisms, prevalence imbalance and group-size imbalance, through which changes in the training distribution shift the learned decision boundary and induce disparities in group true positive rates. Motivated by this analysis, we study exact full-balance augmentation, which equalizes all group-label cells. We establish conditions under which full balancing strictly improves Equal Opportunity together with a prediction-error comparison under the oracle balanced distribution, and show that the Equal Opportunity gap of the synthetic-data classifier converges to that of the oracle balanced classifier as the sample size grows and the generator error vanishes. The finite-sample convergence rate agrees, up to logarithmic factors and generator-approximation error, with the information-theoretic lower-bound rate under the corresponding regularity regimes. Since exact balancing need not provide the best fairness-accuracy trade-off in practice, we further propose Accuracy-Constrained Adaptive Synthetic Augmentation (ACASA), which learns both where and how much to augment by minimizing estimated fairness disparity subject to a constraint on predictive accuracy. We provide a finite-sample guarantee comparing the selected augmentation allocation with the population-optimal accuracy-feasible allocation. Simulation studies and real data application support the theoretical findings and demonstrate that synthetic augmentation can substantially reduce fairness disparities while achieving favorable empirical fairness-accuracy trade-offs.
[LG-181] Improved last-iterate bounds for SGD with and without replacement
链接: https://arxiv.org/abs/2611.14811
作者: Yimin Zhong,Jiaxin Zhou
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Optimization and Control (math.OC)
*备注:
Abstract:In this work, we study last-iterate convergence bounds for stochastic gradient descent (SGD) on convex and \beta -smooth loss functions and improve the bound from \cO((\log T)^3/T) to \cO((\log T)^3/(T\log\log T)) in the interpolation regime. The improvement comes from applying a horizon-dependent multiscale estimate of gradients to the comparison of trajectories. The bound applies to both sampling with replacement and sampling without replacement within one pass.
[LG-182] Dont Match the Noise: Distribution Matching under Unknown Measurement Error with an Audit Sample NEURIPS2026
链接: https://arxiv.org/abs/2611.14749
作者: Arya Farahi,Ritwik Vashistha
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Methodology (stat.ME)
*备注: Accepted at NeurIPS 2026
Abstract:Real-world machine learning systems often operate on imperfect measurements: sensors are noisy, annotations are produced by non-experts, and scientific observations are often indirect. In these settings, classic distribution comparison tools such as maximum mean discrepancy (MMD) can be systematically biased, matching artifacts of the measurement process rather than the underlying mechanisms of interest. We study a practical regime in which noisy observations are abundant, but only a small audited subset contains ground-truth measurements. We propose measurement-error-corrected MMD (MEC-MMD), an audit-based estimator that debiases kernel evaluations without requiring a noise model. MEC-MMD is exactly unbiased for the oracle MMD under simple random auditing, admits an explicit variance decomposition, and is consistent for the population MMD, achieving the standard O_p(N^-1/2) rate when the audit size scales with the dataset. We demonstrate the utility of distributional matching with MEC-MMD loss for both parameter estimation and unsupervised applications, showing that MEC-MMD consistently outperforms MMD and alternative methods. These results provide a principled foundation for robust distribution comparison, simulation-based inference, and learning in noisy real-world machine learning pipelines.
[LG-183] A New Framework for Quantum Reinforcement Learning NEURIPS2026 NEURIPS
链接: https://arxiv.org/abs/2611.14731
作者: Alireza Tavakoli,Mohsen Heidari
类目: Quantum Physics (quant-ph); Machine Learning (cs.LG)
*备注: Accepted at NeurIPS 2026. 28 pages, 3 figures; includes appendices and NeurIPS paper checklist
Abstract:Quantum reinforcement learning (QRL) has emerged as a promising approach for controlling quantum systems and accelerating learning tasks with quantum resources. However, most existing QRL formulations still retain a classical reinforcement learning interface: states and actions are often encoded in computational bases and rewards are extracted through measurements. This interface is restrictive for genuinely quantum environments: choosing a measurement can discard entanglement information, while repeated intermediate measurements collapse the evolving state and require costly re-preparation. This motivates a fully QRL framework in which the agent, environment, rewards, and return accumulation are all quantum. In this paper, we take a step toward this goal and introduce a new QRL framework in which both the policy and the environment are modeled by conditional quantum channels. We then propose a quantum policy-gradient algorithm for parameterized conditional quantum channels with a trainable control basis. We also derive finite- and infinite-horizon gradient estimators using a new parameter-shift rule. This rule is needed because policy parameters are shared across repeated interaction blocks and the infinite-horizon objective is not a finite circuit; rather than shifting every occurrence and evaluating the whole trajectory, our estimator shifts only one sampled policy block. Sampling this block from a geometric distribution yields an unbiased gradient estimator under bounded rewards. Finally, we establish convergence of the resulting stochastic gradient-ascent procedure to stationary points under standard smoothness and step-size assumptions.
[LG-184] Robbing Peter to Pay Paul: Cross-Attribute Externalities of Fairness Interventions
链接: https://arxiv.org/abs/2611.14717
作者: Fernandes Machado Agathe,Jamotton Charlotte,Charpentier Arthur
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:Fairness interventions often targets disparities associated with a single sensitive attribute in binary classification tasks. However, when multiple sensitive attributes are associated with a data subject, improving fairness with respect to one attribute does not necessarily improve it across all attributes: it may exacerbate disparities associated with other sensitive attributes, including those that are unavailable to the practitioner. We refer to these unintended effects as cross-attribute externalities of fairness. We study how fairness with respect to one sensitive attribute varies when an initially unfair classification or regression model is corrected with respect to another attribute. We characterize these externalities using Wasserstein barycentric disparities and develop a distributional framework that captures changes in both marginal and joint fairness measures. This highlights a limitation of conventional fairness evaluation, in which the success of an intervention is assessed mainly for the attribute it targets. In a jointly Gaussian model, we derive explicit trajectories for marginal and joint disparities and identify conditions under which an intervention eliminates disparities with respect to its target attribute while increasing disparities associated with another attribute. We also derive sharp partial-identification bounds on the resulting externalities when a sensitive attribute is unavailable and extend our results to categorical sensitive attributes. We illustrate these externalities using classification and regression real-data: in the classification experiments, interventions increase unfairness with respect to the hidden attribute in the majority of scenarios across more than half of the datasets considered; in the regression experiments on the ACSIncome dataset, correcting for race reduces sex disparities, whereas subsequently correcting for age increases them.
[LG-185] Conformal Risk Certificates for Portfolio Optimization Under Uncertainty: A Gated Reinforcement-Learning Approach
链接: https://arxiv.org/abs/2611.14699
作者: Mohammad Zarei,Amirhossein MohammadpourAzari,Moslem Habibi
类目: Portfolio Management (q-fin.PM); Machine Learning (cs.LG); Optimization and Control (math.OC); Risk Management (q-fin.RM)
*备注: 26 pages, 5 figures, 10 tables. Submitted to Annals of Operations Research
Abstract:Reinforcement-learning portfolio policies can optimize allocations from prices, covariances, and prior positions, yet they lack a calibrated, replayable mechanism for certifying that a proposed allocation respects a bounded risk envelope before execution. Regulators increasingly expect automated portfolio systems to provide exactly this kind of pre-trade audit trail, interpretable risk limit, and human-oversight mechanism. Gated Conformal Semantic Risk Control addresses this gap by separating return-seeking allocation from text-conditioned downside-risk verification. An instruction-tuned language model maps timestamped disclosures into schema-validated semantic risk states over macro, liquidity, earnings, and sector channels. These states induce correlated ellipsoidal factor-shock uncertainty sets whose conformal radii are calibrated by rolling finite-sample prediction over localized semantic nonconformity buckets. An asymmetric second-order cone gate then checks each raw PPO proposal against a certified worst-case semantic loss budget. The gate stays inactive for compliant proposals and intervenes only when that budget is exceeded. In walk-forward out-of-sample tests on nine liquid U.S. Select Sector SPDR ETFs, Raw PPO earns a 12.7% annual return while incurring a 4.05% semantic violation rate. Gated-CSRC removes all observed out-of-sample semantic limit breaches while retaining an 11.7% annual return. This 100-basis-point auditable safety premium is smaller than the yield loss imposed by unconditional hard projection, and exact McNemar testing confirms the compliance gain at p 10^-15 . Loughran-McDonald lexicon controls and time-shuffled text controls fail to reproduce this certificate.
[LG-186] An Eyring-Kramers Formula for Singular Wells
链接: https://arxiv.org/abs/2611.14603
作者: Yingli Wang,Lingjiong Zhu
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Probability (math.PR)
*备注: 32 pages
Abstract:Understanding the persistence of learned behaviors in AI safety motivates an investigation of the dynamical counterpart to the equilibrium geometry studied in singular learning theory. We analyze this persistence through transitions between singular wells. We establish an Eyring-Kramers formula for a low temperature Langevin diffusion between two compact connected zero components of a real-analytic potential. The wells may be singular and nonisolated, and the unique minimizing gate is a nondegenerate index-one saddle, with gates understood as inclusion-minimal sets meeting every minimizing path. A boundary-gate construction shows that uniqueness of this inclusion-minimal gate is equivalent to every minimizing path passing through the saddle and one such path meeting the communication level only there. Compact semianalytic topology provides a separating function away from a fixed saddle neighborhood; analytic Morse coordinates give matching capacity bounds by first sending the temperature to zero and then shrinking that neighborhood. An anchored energy estimate and a small-ball potential-theoretic identity give uniformity for deterministic starting points. The leading constant is expressed in terms of the sublevel-volume coefficient, a Gamma factor, and the saddle Hessian.
[LG-187] Modelling the occurrence and reporting dynamics of weather-related home insurance claims
链接: https://arxiv.org/abs/2611.14542
作者: Paul Wilsens,Katrien Antonio,Gerda Claeskens,Koos Gubbels
类目: Applications (stat.AP); Machine Learning (cs.LG)
*备注:
Abstract:We study the occurrence and reporting dynamics of weather-related claims in private home insurance, combining detailed portfolio and claim information with publicly available high-resolution weather data. Existing approaches typically start from an identified severe weather event, such as a hailstorm, and model the dynamics of the claims caused by that event. In practice, however, claims arise under moderate as well as severe conditions, and the claims handling process runs continuously. An insurer therefore requires insight into the occurrence and reporting of claims on every calendar day. Modelling all calendar days jointly poses two challenges: recently occurred claims are only partially reported, and days with severe weather are rare while accounting for a large share of the claims. We propose a multi-regime EM-XGBoost framework that addresses both: an expectation-maximisation setup accounts for the incurred-but-not-reported (IBNR) claims, while a data-driven regime structure lets the occurrence and reporting dynamics differ across weather conditions. In a case study on a Dutch home insurance portfolio, we compare single- and two-regime specifications for two spatial resolutions: the country as a whole, and its individual regions. We find that the regional two-regime model outperforms the regional single-regime model as well as both national models, indicating that the benefit of a multi-regime setting depends on a sufficiently granular spatial resolution. We also show that daily covariates constructed from hourly weather observations, capturing the duration and the intra-day timing of severe weather, improve both the occurrence and the reporting model.
[LG-188] Katz Kernel Based Stratification of Breast Cancer Patients
链接: https://arxiv.org/abs/2611.14540
作者: Francesco Cellitti,Lorenzo Di Rocco,Livia Perfetto,Veronica Lombardo,Umberto Ferraro Petrillo
类目: Applications (stat.AP); Machine Learning (cs.LG); Molecular Networks (q-bio.MN); Machine Learning (stat.ML)
*备注:
Abstract:The integration of mechanistic signaling networks into clinical research has become a cornerstone of precision oncology. By capturing the complex biochemical interactions within a cell, these networks provide a functional map of disease progression that goes beyond simple genomic mutations. Consequently, the development of specific signaling models of patients has significant relevance, allowing for a more granular representation of individual disease profiles. In this setting, there is a growing need for robust methodologies capable of comparing graph based pathway structures to identify similarities and differences across patient cohorts. Such comparisons are essential for the identification of clinically meaningful subgroups, and the discovery of novel biomarkers. Standard graph kernels, while mathematically rigorous, often remain too generic for this purpose. They frequently fail to capture features specific to gene signaling networks, resulting in suboptimal clustering performance. In this paper, we present a domain specific graph kernel designed for mechanistic signaling networks. Unlike generic approaches, our method explicitly incorporates biological constraints and gene interaction patterns into the kernel formulation, enabling a more accurate characterization of both topological structure and functional behavior. We assess its performance on breast cancer datasets, focusing on its ability to support disease subtyping. Our results suggest that our approach can effectively contribute to patient stratification.
[LG-189] Can a search hold its own bar? In-loop deflation and its exploits in autonomous strategy discovery
链接: https://arxiv.org/abs/2611.14539
作者: Eray Gençay
类目: Applications (stat.AP); Machine Learning (cs.LG); Statistical Finance (q-fin.ST)
*备注: 19 pages, 5 figures, 3 tables. Code and data: this https URL
Abstract:Corrections for backtest overfitting such as the deflated Sharpe ratio are applied after a search ends: the searcher maximises raw performance and a separate procedure deflates the reported winner. Autonomous searches driven by large language models perform many implicit backtests before reporting, so this paper asks what changes when the correction moves inside the loop and the acceptance threshold is indexed to the running count and dispersion of the search’s own trials. The threshold then becomes an object the search can optimise against: we show analytically that a naive in-loop bar admits a variance-dilution effect, in which mediocre near-duplicate candidates lower the bar because the ledger’s dispersion falls faster than its trial-count factor grows. We propose two defences, a pre-registered variance floor and a monotone high-water-mark bar, and evaluate the construction under a protocol frozen before any run, with every in-loop decision replaying exactly from its recorded trace. Across 53 audited hundred-candidate searches over two frontier models, two universes, and two feature spaces, the gated search reports “no certifiable discovery” on 36 of 37 runs, un-gated controls report a strategy on all 16, and every arm’s out-of-sample confidence interval includes zero. Supporting studies find that state-based stopping reaches the same outcome after 8 to 9 trials at about a twelfth of the trial budget while an adaptive budget adds nothing, that a controlled stress test reproduces the dilution effect and both defences, and that a hundred nominal trials collapse to 6 to 8 effectively independent return streams. We scope these results to the analysed system.
[LG-190] Modeling Drug Overdose Mortality Levels Using Social Vulnerability Indicators in U.S. Counties: A Double Machine Learning Approach for Causal Analysis
链接: https://arxiv.org/abs/2611.14537
作者: Qunhui Guo,Jiongbai Liu,Md Noor E Alam
类目: Applications (stat.AP); Machine Learning (cs.LG)
*备注:
Abstract:Drug poisoning mortality (DPM) remains a leading cause of death among working-age Americans, and its burden falls unevenly across communities. Existing county-level studies of social vulnerability and overdose have largely reported predictive or associational relationships, while the shape of the adjusted exposure-response relationship between overall social vulnerability and mortality remains unclear. We examined the CDC/ATSDR Social Vulnerability Index (SVI) and county drug poisoning mortality across three two-year windows spanning 2016-2021. The associational analysis compared linear regression, Random Forest, and XGBoost models of log mortality using period-specific SVI subfeatures and county covariates. The selected XGBoost models were interpreted with SHAP values. The double machine learning (DML) analysis estimated the exposure-response relationship between SVI and DPM and tested the linearity restriction. Socioeconomic vulnerability contributed positively to predicted mortality, with disability and unemployment among the strongest subfeatures, while racial and ethnic minority composition contributed negatively in all three periods. Summing the four theme contributions reduced the overall-index attribution by 34% to 36%, indicating substantial offsetting. Linearity was rejected in every window ( p 10^-26 ). The adjusted relationship increased through the middle of the SVI distribution and declined in the upper tail. This pattern remained stable both before and after the outbreak of the epidemic. Overall SVI therefore captured meaningful variation in overdose mortality but did not function as a uniformly increasing risk scale across its full range. Component-level interpretation and current overdose surveillance provide important context for interpreting the composite index.
[LG-191] Quantifying Methodological Inflation in Clinical Machine Learning: A Cross-Modality Audit of Subject-Level Leakage and Nested Feature Selection for Parkinsons Disease
链接: https://arxiv.org/abs/2611.14533
作者: Namrata Datta,Victor Miranda Soberanis
类目: Applications (stat.AP); Machine Learning (cs.LG)
*备注:
Abstract:Reported accuracy in Parkinson’s disease (PD) machine learning can reflect the evaluation protocol rather than real signal. We ask how much evaluation choices inflate performance by comparing naive and rigorous protocols on voice, telemonitoring and blood-transcriptomic datasets. Subject-aware splitting reduced voice AUROC from 0.892 to 0.780 and telemonitoring R^2 from 0.105 to -0.037; Random Forest recovered neither gap. Supervised probe selection before splitting inflated transcriptomic AUROC from 0.711 to 0.998. A nested 50-probe panel matched a 1,000-probe panel (0.811 versus 0.844), and recruitment-wave batch structure was detected. Future work should validate a stable panel on an independent cohort.
[LG-192] Clustering trajectories of modelled estimates: an uncertainty-aware framework with an application to global cardiovascular disease burden 1990 to 2023
链接: https://arxiv.org/abs/2611.14529
作者: Eben Afrifa-Yamoah
类目: Applications (stat.AP); Machine Learning (cs.LG); Methodology (stat.ME)
*备注: 19 pages, 8 figures
Abstract:Modelled health estimates such as those of the Global Burden of Disease (GBD) study are increasingly used in secondary analyses designed for observed data. However, trajectory clustering typically ignores published uncertainty, selects the number of clusters using criteria blind to it, and assigns countries to groups with unwarranted certainty. We propose an uncertainty-aware framework that separates trajectory level from shape, quantifies membership probabilities by perturbing estimates within their uncertainty intervals, selects clusters based on stability under uncertainty, and examines the roles of development, region, and measurement quality. Applied to age-standardised cardiovascular disease DALY rates for 201 countries from 1990-2023 (GBD 2023), the framework identified two stable trajectory types, whereas the gap statistic selected three. The additional cluster was poorly supported (median membership probability 0.70) and disappeared once uncertainty was considered. The retained clusters had similar burdens in 1990 but diverged markedly over time: a stalled-improvement group of 78 countries (median change -8% , 40 of them in sub-Saharan Africa) and a sustained-decline group of 92 countries ( -52% ). Thirty-one countries remained unclassified because their memberships were uncertain. Results were reproduced using mortality rates, a 2000-2023 analysis window, and bootstrap resampling. Among countries with the narrowest uncertainty intervals, a third well-supported trajectory emerged in former Soviet and Eastern European countries. These findings show that uncertainty affects both the number and composition of trajectory clusters. Ignoring it risks over-interpreting unstable groups and overstating confidence in country classifications. Incorporating published uncertainty and measurement quality should become standard practice when clustering modelled health estimates.
[LG-193] Credal Wrapper for Uncertainty Modelling in Sparse Geological Data
链接: https://arxiv.org/abs/2611.14527
作者: David Jeevankumar,Maryam Sultana,Alexander D. Rast,Fabio Cuzzolin
类目: Applications (stat.AP); Computational Engineering, Finance, and Science (cs.CE); Machine Learning (cs.LG)
*备注: 28 pages, 4 figures, 10 tables
Abstract:Machine learning models for mineral exploration are trained on a few hundred sparse, expensive samples and used to decide where to drill next, so their epistemic uncertainty (EU), the part that more data could remove, matters as much as their predictions. Yet EU cannot be measured on real data, because the counterfactual that defines it is never observed, so estimators are graded by proxies such as out-of-distribution detection and calibration. We contribute a general simulation protocol that makes EU directly computable for a given model and can grade any EU estimator. It separates two estimands the literature conflates, reducible error and estimator spread, builds an oracle reference for each in a synthetic prospectivity field sampled by clustered drilling surveys, and compares uncertainty measures of an identical ensemble within each simulated world. We use it to ask when credal (imprecise-probability) measures are more faithful than Bayesian model averaging (BMA) in the sparse, low-class regime of geological classification. The credal interval width recovers the epistemic ordering better than BMA when members are few, confirming the small-ensemble advantage reported for credal wrappers; the advantage vanishes by M = 10 and reverses at M = 40. Secondly, the choice of credal functional matters more than the choice between credal and BMA. Lastly, the credal advantage does not widen as training data are thinned, so it compensates for few posterior samples rather than few data. On real drill-hole geochemistry from South Australia, a resampling oracle reproduces the first finding. The protocol applies to any EU estimator whose data-generating process can be simulated.
[LG-194] Null-Corrected Conditional Coverag e in Materials Property Prediction: A Five-Dataset Replication
链接: https://arxiv.org/abs/2611.14521
作者: Irene Boruah
类目: Methodology (stat.ME); Machine Learning (cs.LG); Applications (stat.AP)
*备注: N pages, 15 tables. Extends and partly revises arXiv:2608.07589 . Submitted to Data-Centric Engineering. Code and data: this https URL
Abstract:Uncertainty quantification for materials property prediction increasingly reports conditional coverage, coverage within strata of the predicted value, alongside marginal coverage. The worst-bin statistics used for this are downward-biased in finite samples, a known defect that applied materials work does not generally correct for. We replicate a single-dataset conditional coverage study across five regression datasets (n = 437 to 21,263), seven interval methods and 50 random splits each, correcting the worst-bin statistic against its expected value under exact conditional validity. The null is simulated from realized per-bin test counts and referenced to each method’s own marginal coverage, separating marginal miscalibration from uneven coverage. On the smallest dataset the correction runs from 0.061 to 0.096, between a third and two thirds of the shortfalls actually reported, against 0.009 to 0.011 on the two largest. Split conformal prediction’s failure survives the correction on four of five datasets; two conclusions of the antecedent study do not. Conformalized quantile regression, characterized there as a negative case, fails on one dataset only. Cross-fitted normalized conformal prediction, grouped there with the successful methods, fails on three of five. Group-conditional (Mondrian) conformal prediction stays below our significance criterion on every dataset, yet exceeds the null on three of five under an octile refinement of the variable it calibrates on. A level-matched ablation locates that uniformity in group-conditional calibration rather than in the elevated quantile level, and removes Mondrian’s apparent width cost entirely. Conditional coverage evaluations should report the null distribution of the worst-bin statistic used, the method’s realized marginal coverage, and the resolution of the evaluation partition relative to any partition the method calibrates on.
附件下载
