ArXiv Domain 2026-10-22

数据来源：ArXiv Domain

LLM Domain Papers

1. H2DF: Calibrated Target Bands for Controllable Language-Model Unlearning

Abstract:Gradient ascent (GA) is a simple language-model unlearning baseline, but it continues increasing forget loss after useful forgetting has occurred. This unbounded pressure can damage retained utility and make forgotten examples abnormally high-loss. We introduce H2DF, a calibrated target-band objective that applies GA-like pressure below a data-dependent lower target and suppresses, or reverses, that pressure after the target region is reached. The targets are estimated from matched non-member examples before training, so the base objective requires one forget-batch forward/backward pass per optimization step; our reported TOFU variant adds sparse retain replay every tenth step. We instantiate a one-sided objective for TOFU question answering and a two-sided objective for MUSE-News language modeling. Across three TOFU seeds, H2DF obtains a forget NLL of 1.528 with margin 0.049 and retain NLL of 1.165 with margin 0.047, matching GradDiff while using retain replay only every tenth step and showing substantially lower variability than GA. On MUSE-News, H2DF improves retention over GA and NPO and is competitive with SimNPO, but does not dominate it. Ablations show that the upper target reduces overshoot and that validation-based checkpoint selection is important. These results support calibrated target bands as a practical mechanism for controlling forgetting strength, while also exposing sensitivity to random seed and calibration quality.

中文摘要

摘要：梯度上升（GA）是一种简单的语言模型遗忘基线，但在有用遗忘发生后，它会继续增加遗忘损失。这种无限制的压力可能损害保留的效用，并使已遗忘的样本损失异常高。我们提出了H2DF，这是一种校准的目标带目标，在数据依赖的下方目标下施加类似GA的压力，并在达到目标区域后抑制或反转该压力。目标是在训练前从匹配的非成员样本估计的，因此基础目标每次优化步骤只需一次遗忘批次的前向/反向传播；我们报告的TOFU变体每十步增加一次稀疏保留重放。我们为TOFU问答建立了单边目标，为MUSE-News语言建模建立了双边目标。在三个TOFU随机种子下，H2DF获得遗忘NLL为1.528，边际为0.049，保留NLL为1.165，边际为0.047，与GradDiff相匹配，同时仅在每十步使用一次保留重放，并且显示出比GA显著更低的波动。在MUSE-News上，H2DF在保留性能上优于GA和NPO，并与SimNPO具有竞争力，但未完全超越。消融实验表明，上限目标减少了超调，基于验证的检查点选择非常重要。这些结果支持将校准目标带作为控制遗忘强度的实用机制，同时也揭示了对随机种子和校准质量的敏感性。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.26864 (HTTP 429)

Authors: Jupiter Yao

Categories: cs.CL

PDF URL: https://arxiv.org/pdf/2610.26864.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.26864

Published: 2026-10-22T01:50:25.204Z

2. AutoTerm-SST: Adaptive Terminology Memory with Zero Session-Time Setup for Streaming Speech Translation

Abstract:Retrieval-augmented speech translation improves rare terminology, but systems such as RASST still assume that users provide a glossary before each session. That assumption is inconvenient for live talks, while merging every domain into one growing glossary eventually places too many irrelevant terms in the same retrieval and prompt competition. We present AutoTerm-SST, an interactive system that removes session-time glossary upload and domain selection. It organizes a large offline terminology catalog into topic slices, matches recent streaming translation context to bilingual slice prototypes, and activates a compact multi-slice working set under fixed retrieval and prompt budgets. On five ACL and five ESO oncology talks, AutoTerm matches a known-domain reference in terminology accuracy (88.3% vs. 88.1%) and mWER-resegmented BLEU (45.09 vs. 45.19), while requiring no domain input. Compared with a one-million-entry merged glossary, it improves terminology accuracy by 3.9 points and BLEU by 2.30. Routing plus four-slice retrieval remains lightweight: the complete system runs at 0.406 RTF, and the vLLM host sustains 32 real-time sessions. The live demo exposes active slices and retrieved terminology as the topic changes, providing zero-setup terminology-aware streaming translation.

中文摘要

摘要：检索增强的语音翻译可以改善罕见术语的表现，但像 RASST 这样的系统仍然假设用户在每次会话之前提供术语表。这个假设对于实时演讲来说不方便，而将所有领域合并到一个不断增长的术语表中最终会在同一检索和提示竞争中产生过多无关的术语。我们提出了 AutoTerm-SST，这是一个交互式系统，它消除了会话时上传术语表和选择领域的需求。它将大型离线术语目录组织成主题切片，将最近的流式翻译上下文匹配到双语切片原型，并在固定的检索和提示预算下激活紧凑的多切片工作集。在五个 ACL 和五个 ESO 的肿瘤学演讲中，AutoTerm 在术语准确性上匹配已知领域参考（88.3% 对 88.1%），在 mWER 重新分段后的 BLEU 上（45.09 对 45.19）也相当，同时不需要领域输入。与一百万条 merged glossary 相比，它将术语准确性提高了 3.9 个百分点，BLEU 提高了 2.30。路由加上四切片检索仍然轻量：完整系统运行在 0.406 RTF，vLLM 主机可维持 32 个实时会话。实时演示展示了主题变化时的活跃切片和检索的术语，实现了零设置的术语感知流式翻译。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.26865 (HTTP 429)

Authors: Jiaxuan Luo, Siqi Ouyang, Lei Li

Categories: cs.CL

PDF URL: https://arxiv.org/pdf/2610.26865.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.26865

Published: 2026-10-22T01:50:25.204Z

3. WaveCommitments: Private and Verifiable Speech-Agent Marketplaces

Abstract:Modern speech systems are evolving into ecosystems of autonomous speech agents that interact, negotiate, and trade assets such as embeddings, spectrograms, or raw data. In such markets, sellers cannot reveal assets for evaluation due to privacy constraints, and buyers cannot price what they cannot test. Thus, we introduce WaveCommitments, an end-to-end design that couples multivariate polynomial commitments with confidential asset valuation via Data Shapley inside a trusted execution environment. It enables selective openings and tamper-evident provenance to prove that a delivered asset matches the valuation while hiding everything else. Our extensive experiments across three standard datasets and various tasks with state-of-the-art encoders, including Whisper, reveal that we speed up the strongest baselines by up to 18.1x while reducing their proof sizes by up to 43.2x. We also analyze how a speaker’s value shifts across tasks and encoders, revealing agent-level market prices.

中文摘要

摘要：现代语音系统正在发展为自主语音代理的生态系统，这些代理可以相互交互、协商，并交易如嵌入向量、声谱图或原始数据等资产。在这样的市场中，卖方由于隐私限制无法公开资产以供评估，而买方又无法为无法测试的资产定价。因此，我们提出了 WaveCommitments，一种端到端设计，将多元多项式承诺与可信执行环境中的数据 Shapley 保密资产评估相结合。它支持选择性开启和防篡改来源，以证明交付的资产与评估一致，同时隐藏其他所有信息。我们在三个标准数据集和各种任务上使用包括 Whisper 在内的最先进编码器进行了大量实验，结果显示，我们将最强基线的速度提升了最多 18.1 倍，同时将其证明大小减少了最多 43.2 倍。我们还分析了说话者价值在不同任务和编码器中的变化，揭示了代理级别的市场价格。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.26867 (HTTP 429)

Authors: Oguzhan Baser, Kaan Kale, Sriram Vishwanath, Sandeep Chinchali

Categories: cs.CL

PDF URL: https://arxiv.org/pdf/2610.26867.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.26867

Published: 2026-10-22T01:50:25.204Z

4. Multimodal Ambivalence and Hesitancy Recognition with Bidirectional Audio—Text Cross-Attention

Abstract:Ambivalence and hesitancy (A/H) are subtle behavioural states that can be expressed through both what a person says and how it is spoken. This paper presents an audio—text framework for video-level A/H recognition. Full-video transcripts and speech waveforms are encoded using pre-trained multilingual BERT and wav2vec~2.0, respectively, and projected into a shared representation space. Two stacked bidirectional cross-attention blocks allow linguistic tokens to attend to acoustic frames and acoustic frames to attend to linguistic context without requiring explicit temporal alignment. The resulting sequences are aggregated through masked mean pooling and combined using concatenation, element-wise products, and absolute differences for binary classification. We evaluate the model on the subject-independent split of the Behavioural Ambivalence/Hesitancy (BAH) dataset. The proposed system achieves 72.19\% accuracy, 75.00\% precision, 81.13\% recall, 77.95\% A/H F1-score, and 70.16\% Macro-F1 on the test set. These results demonstrate the potential of jointly modelling linguistic and acoustic information for automatic A/H this http URL at:this https URL.

中文摘要

摘要：矛盾和犹豫（A/H）是一种微妙的行为状态，可以通过一个人说的话以及说话的方式来表达。本文提出了一个用于视频级别A/H识别的音频—文本框架。完整视频的文字记录和语音波形分别使用预训练的多语言BERT和wav2vec~2.0进行编码，并投射到共享表示空间。两个堆叠的双向交叉注意力模块使语言单元能够关注声学帧，同时声学帧也能关注语言上下文，无需显式的时间对齐。生成的序列通过掩码平均池化进行聚合，并通过拼接、元素逐一乘积和绝对差进行组合，用于二分类。在行为矛盾/犹豫（BAH）数据集的被试独立拆分上对模型进行评估。所提出的系统在测试集上达到72.19%的准确率、75.00%的精确率、81.13%的召回率、77.95%的A/H F1分数和70.16%的宏F1分数。这些结果展示了联合建模语言和声学信息用于自动A/H识别的潜力。相关内容可参见该链接：this http URL 以及此链接：this https URL。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.26868 (HTTP 429)

Authors: TingTing Gu, JinCheng Luo, WenQiu Guo

Categories: cs.CL

PDF URL: https://arxiv.org/pdf/2610.26868.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.26868

Published: 2026-10-22T01:50:25.204Z

5. Mixture of Phonetic Experts Based Low-Rank Adaptation of Conformer Models for Accented English Speech Recognition

Abstract:We present Mixture of Phonetic Experts based Low-Rank Adaptation (MoPE-LoRA), a parameter-efficient adaptation framework for accented English speech recognition. Recent mixture-of-experts approaches typically allocate experts per accent, implicitly treating accent identity as the primary axis of variation. In contrast, we organize adaptation along phonetic categories, motivated by the observation that accent differences are often expressed as systematic distortions of phoneme realization. MoPE-LoRA introduces a fixed set of six low-rank experts aligned with manner-of-articulation classes and performs frame-level routing through a hybrid mechanism that combines phoneme supervision with learned acoustic gating. Because the expert set is shared across accents, the model does not require accent-specific modules and naturally extends to previously unseen accents. Experiments on L2-ARCTIC show that MoPE-LoRA consistently improves over single LoRA and full fine-tuning. In zero-shot evaluation where one accent is held out during training, the proposed method yields a 12.3% relative improvement over single LoRA. These results indicate that phonetic-category-based decomposition offers a scalable and transferable alternative to accent-level adaptation.

中文摘要

摘要：我们提出了基于语音专家混合的低秩适配（Mixture of Phonetic Experts based Low-Rank Adaptation, MoPE-LoRA），这是一种针对有口音英语语音识别的参数高效适配框架。近期的专家混合（mixture-of-experts）方法通常按口音分配专家，隐式地将口音身份视为主要变化轴。相比之下，我们沿语音类别组织适配，其动机是观察到口音差异通常表现为音素实现的系统性扭曲。MoPE-LoRA引入了一组固定的六个低秩专家，这些专家与发音方式类别对齐，并通过结合音素监督与学习到的声学门控的混合机制进行帧级路由。由于专家集合在不同口音间共享，模型不需要特定口音的模块，并能自然扩展到之前未见过的口音。在L2-ARCTIC数据集上的实验表明，MoPE-LoRA持续优于单一LoRA和全量微调。在零样本评估中，即训练时保留一种口音，所提方法相比单一LoRA取得了12.3%的相对提升。这些结果表明，基于语音类别的分解提供了一种可扩展且可迁移的替代方案，相较于按口音级别的适配。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.26888 (HTTP 429)

Authors: Santosh Dahal, Anmol Guragain, Tianchi Liu, Luis Fernando D’Haro, Kiran Chandra Dahal

Categories: cs.CL

PDF URL: https://arxiv.org/pdf/2610.26888.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.26888

Published: 2026-10-22T01:50:25.204Z

6. AtomSD: Procedural Atom-Guided Self-Distillation for LLM Agents

Abstract:Large language models (LLMs) possess broad factual knowledge and strong reasoning capabilities, but solving complex tasks also requires procedural knowledge about how to act. Existing approaches often represent procedural knowledge through trajectories or bundled skills, limiting flexible reuse across interaction histories. This paper introduces Procedural Atoms, fine-grained and independently applicable units that represent Intent, When, Do, and Avoid guidance. Building on this representation, we propose AtomSD, an Atom-guided Self-Distillation framework in which the learner continually extracts Procedural Atoms from its own successful trajectories and distills them into the evolving policy through Atom-conditioned supervision and task-reward optimization. Comprehensive experiments show that AtomSD consistently outperforms strong reward-based and distillation methods, exceeding the state-of-the-art baseline by up to 9.31 points in average performance. Further analyses demonstrate that the proposed Procedural Atoms provide effective guidance, highlighting the value of atom-level procedural knowledge for policy learning.

中文摘要

摘要：大型语言模型（LLMs）拥有广泛的事实知识和强大的推理能力，但解决复杂任务还需要有关如何行动的程序性知识。现有方法通常通过轨迹或捆绑技能来表示程序性知识，这限制了在交互历史中灵活重用的能力。本文引入了程序原子（Procedural Atoms），即细粒度且可以独立应用的单元，表示意图（Intent）、时间（When）、执行（Do）和避免（Avoid）指导。在此表示基础上，我们提出了AtomSD，一种原子引导的自我蒸馏框架，其中学习者持续从自身成功的轨迹中提取程序原子，并通过原子条件监督和任务奖励优化将其蒸馏进不断演化的策略。全面实验表明，AtomSD 始终优于强大的基于奖励和蒸馏的方法，在平均性能上超过最先进的基线多达9.31点。进一步分析显示，所提出的程序原子提供了有效的指导，凸显了原子级程序性知识对策略学习的价值。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.26926 (HTTP 429)

Authors: Bohan Yu, Pengfei Cao, Kai Tian, Chen Han, Shi-Yang Li, Chenxi Zhou, Junlin Liu, Cao Liu, Ke Zeng, Jun Zhao, Kang Liu

Categories: cs.CL

PDF URL: https://arxiv.org/pdf/2610.26926.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.26926

Published: 2026-10-22T01:50:25.204Z

7. Do We Need the Largest LLMs for Every Clinical Decision?

Abstract:Large clinical language models often require multi-GPU infrastructure, limiting local deployment. We introduce CERTA, a decision-level statistical certificate that determines when a quantized model may substitute for its native BF16 reference and escalates decisions that cannot be certified to it. Across eleven instruction-tuned models, both general-purpose and domain-adapted, from three families (7-72B), we evaluated 2,500 clinical-verification decisions from de-identified patient contexts (2,484 patients), with patient-disjoint calibration and testing. In the higher-power operational analysis, round-to-nearest 8-bit quantization at BF16 storage served 99.8-100% of held-out decisions across all eleven models; formal finite-sample guarantees were assessed separately using the exact split-sample procedure. In the same operational analysis, the class-conditional 4-bit frontier reduced the amortized, BF16-equivalent deployment cost to 25.8% and 27.3% of the BF16 reference for Qwen2.5-72B and HuatuoGPT-o1-72B, respectively, with measured peak memory of 43.0 and 42.7 GB, below the nominal capacity of a single 48 GB accelerator. Deployment on that specific hardware was not directly tested. Under the exact split-sample certificate, which also accounts for configuration selection, these costs were 26.7% and 27.9%. On external disease named-entity benchmarks against human annotations, the frontier-selected NF4 configurations had a mean absolute F1 change of 0.045 on both datasets, with substantial model-specific variation. CERTA converts compression from a fixed precision choice into a risk-controlled deployment policy for this verification task.

中文摘要

摘要：大型临床语言模型通常需要多GPU基础设施，限制了本地部署。我们提出了CERTA，一种决策级统计证书，用于确定何时量化模型可以替代其原生BF16参考，并将无法获得证书的决策上报。在来自三个家族（7-72B）的十一种指令调优模型中，包括通用和领域适配模型，我们评估了来自去标识化患者语境的2,500个临床验证决策（2,484名患者），采用患者不重叠的校准和测试。在更高性能的操作分析中，四舍五入至最接近的8位量化并以BF16存储，可在所有十一种模型中服务于99.8-100%保留决策；使用精确分割样本程序单独评估了正式有限样本保证。在同一操作分析中，类别条件的4位前沿量化将摊销的BF16等效部署成本降低至Qwen2.5-72B和HuatuoGPT-o1-72B的25.8%和27.3%，测量到的峰值内存分别为43.0和42.7 GB，低于单个48 GB加速器的标称容量。该特定硬件上的部署未被直接测试。在精确分割样本证书下，同时考虑配置选择，这些成本分别为26.7%和27.9%。在针对人类标注的外部疾病命名实体基准测试中，前沿选择的NF4配置在两个数据集上的平均绝对F1变化为0.045，且存在显著的模型特定差异。CERTA将压缩从固定精度选择转换为针对此验证任务的风险控制部署策略。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.26937 (HTTP 429)

Authors: Ali H. Lazem, William Teahan

Categories: cs.CL

PDF URL: https://arxiv.org/pdf/2610.26937.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.26937

Published: 2026-10-22T01:50:25.204Z

8. The Bottleneck Is Retrieval, Not Reading: Diagnosing Evidence-Based Generation with RAG for Long Context Question Answering

Abstract:Retrieval-augmented generation (RAG) is the dominant paradigm for grounding large language models in external corpora, but its similarity-driven, chunked pipeline has structural blind spots: it cannot anticipate which passages a question requires and discards document-level context that resolves cross-passage references. We introduce evidence-based generation (EBG), requiring models to return an answer with line-level citations, and a joint Ans+Evid metric that scores correctness and grounding simultaneously. We reformulate two corpora into evidence-grounded multi-document QA benchmarks- MultiHop-RAG and QASPER-MP, a multi-paper adaptation of QASPER with realistic short paper cues, and pair them with a controlled synthetic two-hop haystack and an evidence-augmented NoLiMa spanning 10K-1M characters. Across ten proprietary frontier and three open-weight LLMs, EBG isolates two RAG limitations that long-context models avoid: similarity retrieval surfaces any relevant chunk far more reliably than all required chunks for multi-fact queries, and chunking destroys document meta-information needed to resolve identity-bearing queries. EBG also exposes grounding failures invisible to answer-only evaluation, including incomplete citations in two-needle settings and a widening gap between Ans and Ans+Evid when context contradicts parametric knowledge. Data and code are available at this https URL.

中文摘要

摘要：检索增强生成（RAG）是将大型语言模型与外部语料库结合的主导范式，但其基于相似性的分块管道存在结构性盲点：它无法预判问题所需的段落，并且会丢弃解决跨段落引用的文档级上下文。我们提出了基于证据的生成（EBG），要求模型返回带有行级引用的答案，并提出了联合Ans+Evid指标，同时评估答案正确性和依据性。我们将两个语料库重新设计为基于证据的多文档问答基准——MultiHop-RAG和QASPER-MP，其中QASPER-MP是QASPER的多论文改编版本，具有现实的短篇论文提示，并将它们与一个受控的合成双跳干草堆和一个覆盖10K-1M字符的证据增强NoLiMa配对。在十个专有前沿模型和三个开放权重大型语言模型中，EBG揭示了RAG的两个局限性，而长上下文模型可以避免：相似性检索比获取多事实查询所需的所有关键块更可靠地呈现任何相关块，而分块会破坏解析身份相关查询所需的文档元信息。EBG还暴露了仅评估答案无法发现的依据失败，包括双针设置中的引用不完整，以及在上下文与参数化知识矛盾时，Ans与Ans+Evid之间差距的扩大。数据和代码可在此https URL获取。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.26951 (HTTP 429)

Authors: Debopriyo Banerjee, Owais Aijaz, Ahmer Tabassum, Zhuohan Xie, Preslav Nakov, Veselin Stoyanov

Categories: cs.CL

PDF URL: https://arxiv.org/pdf/2610.26951.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.26951

Published: 2026-10-22T01:50:25.204Z

9. A Multi-granularity Interaction Network for Relational Triple Extraction

Abstract:Relational triple extraction identifies entity relations from raw sentences without depending on entity type information. The performance of traditional methods is seriously degenerated by unseen'' entities, because these methods can't build contextual dependencies relevantunseen’’ entities. This paper addresses this challenge by proposing a multi-granularity interaction network that leverages word-word, entity-relation, and head-tail interactions to capture deep semantic dependencies. By incorporating multiple iterations, the multi-granularity interaction network is effective to refine the representations corresponding to head and tail entities. It enhances the global semantic features associated with relational triples, aiding in the remediation of triples that may have been overlooked or omitted in earlier steps. Strong and consistent performance is achieved by our method on two rearranged datasets, with a significant improvement in identifying relational triples, including those involving ``unseen’’ named entities. For example, our model improves F1-score by 3.1%/20.5% over previous approaches on the NYT-R/WebNLG-R datasets in the Unseen metric. Code: this https URL.

中文摘要

摘要：关系三元组抽取在不依赖实体类型信息的情况下，从原始句子中识别实体关系。传统方法的性能会因“未见”实体而严重下降，因为这些方法无法构建与“未见”实体相关的上下文依赖。本文通过提出一种多粒度交互网络来应对这一挑战，该网络利用词-词、实体-关系以及头实体-尾实体的交互来捕捉深层语义依赖。通过引入多次迭代，多粒度交互网络能够有效地优化头实体和尾实体的表示。它增强了与关系三元组相关的全局语义特征，有助于补救在早期步骤中可能被忽略或遗漏的三元组。我们的方法在两个重排数据集上取得了强且稳定的性能，在识别关系三元组方面有显著提升，包括涉及“未见”命名实体的三元组。例如，在NYT-R/WebNLG-R数据集的未见指标上，我们的模型相比之前的方法F1分数提高了3.1%/20.5%。代码：此 https URL。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.26952 (HTTP 429)

Authors: Xiaocheng Luo, Yanping Chen, Zaifu Zhan

Categories: cs.CL

PDF URL: https://arxiv.org/pdf/2610.26952.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.26952

Published: 2026-10-22T01:50:25.204Z

10. LAMB: Latent Multi-Hypothesis Beliefs for Long-Horizon Large Language Agents

Abstract:Long-horizon language agents require compact memories that retain interaction evidence while preserving competing hypotheses about unobserved states. Although effective, many belief-state methods encode memory in discrete natural language, where repeated compression can omit evidence or prematurely discard plausible alternatives. We introduce LAMB, a recurrent framework for latent multi-hypothesis beliefs. A Common—Delta factorization combines a shared Common state for interaction evidence with weighted Delta components for competing hypotheses. Relative evidence from action-outcome predictions guides recurrent hypothesis updates without generating textual memory. A context-conditioned readout exposes hypotheses separately to the policy through a fixed-token belief prefix in the default configuration. The belief modules and policy are jointly trained using action imitation and auxiliary state, outcome, and observation-discrimination objectives. Across ALFWorld, BabyAI, ScienceWorld, and Crafter, LAMB matches or outperforms PABU, improving ALFWorld SR@30 and ScienceWorld success by 6.0 and 22.0 percentage points, respectively, and Crafter achievement progress by 26.37 points. On ALFWorld, it also reduces the redundant-action rate by 17.70 percentage points. These results support latent multi-hypothesis beliefs as compact memory for effective long-horizon decision making

中文摘要

摘要：长期语言代理需要紧凑的记忆，以保留交互证据，同时保存关于未观察状态的竞争假设。尽管有效，许多信念状态方法以离散自然语言编码记忆，其中重复压缩可能遗漏证据或过早丢弃合理的备选方案。我们提出了LAMB，一种用于潜在多假设信念的循环框架。Common–Delta分解将共享的Common状态用于交互证据，并结合用于竞争假设的加权Delta分量。来自动作-结果预测的相对证据指导循环假设更新，而无需生成文本记忆。上下文条件读出通过默认配置中的固定标记信念前缀，将各假设分别暴露给策略。信念模块和策略通过动作模仿以及辅助状态、结果和观察区分目标进行联合训练。在ALFWorld、BabyAI、ScienceWorld和Crafter中，LAMB与或优于PABU，在ALFWorld SR@30和ScienceWorld成功率分别提高了6.0和22.0个百分点，Crafter成就进度提高了26.37个百分点。在ALFWorld上，它还将冗余动作率降低了17.70个百分点。这些结果支持潜在多假设信念作为用于有效长期决策的紧凑记忆。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.26955 (HTTP 429)

Authors: Peien Ding, Hui Li, Yuhan Liu, Jinsong Su

Categories: cs.CL

PDF URL: https://arxiv.org/pdf/2610.26955.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.26955

Published: 2026-10-22T01:50:25.204Z

Agent Domain Papers

1. Navalia: Advancing Cowork Agents through Verifiable Task Synthesis and Long-Horizon Post-Training

Abstract:Large language model (LLM) agents are evolving from conversational assistants into Cowork Agents capable of autonomous planning, tool use, and complex task execution. Financial analysis requires these agents to integrate cross-document evidence and perform multistep numerical reasoning, but expert task authoring is costly, while direct LLM synthesis struggles to produce challenging tasks with reliable answers. We introduce Navalia, a framework for Cowork Agents that combines verifiable task synthesis with long-horizon post-training. Using corporate financial reports as its data source, Navalia semantically selects supporting evidence from complete fact sets and constructs executable computation chains to generate analytical tasks with traceable evidence and actual data dependencies between steps. We further use independent LLM review and programmatic checks to verify evidence consistency and answer correctness. The verified tasks are then used for trajectory collection, cold-start supervised fine-tuning (SFT), and reinforcement learning (RL) to improve long-horizon task execution. Navalia, initialized from Qwen3.6-35B-A3B, achieves consistent gains over the base model across all five benchmarks: OfficeQA Pro, FinSearchComp, Finance Agent Benchmark, APEX-Agents-AA, and Workspace-Bench-Lite. These results demonstrate that Navalia improves evidence integration, tool use, and long-horizon task execution capabilities and generalizes to broader Cowork settings, including general office tasks.

中文摘要

摘要：大型语言模型（LLM）代理正在从对话助手发展为能够自主规划、使用工具和执行复杂任务的协作代理。财务分析要求这些代理整合跨文档证据并执行多步骤数值推理，但专家任务编写成本高昂，而直接使用LLM合成难以生成具有可靠答案的挑战性任务。我们提出了Navalia，一个结合可验证任务合成与长周期后训练的协作代理框架。Navalia以公司财务报告为数据来源，从完整事实集中语义选择支持证据，并构建可执行计算链以生成具有可追踪证据和步骤间实际数据依赖的分析任务。我们进一步使用独立LLM审核和程序化检查来验证证据一致性和答案正确性。验证后的任务随后用于轨迹收集、冷启动监督微调（SFT）和强化学习（RL），以提升长周期任务执行能力。Navalia基于Qwen3.6-35B-A3B初始化，在所有五个基准测试中——OfficeQA Pro、FinSearchComp、Finance Agent Benchmark、APEX-Agents-AA和Workspace-Bench-Lite——都取得了比基础模型稳定的增益结果。实验结果表明，Navalia提升了证据整合、工具使用和长周期任务执行能力，并能推广到更广泛的协作场景，包括一般办公任务。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.26887 (HTTP 429)

Authors: Hengqi Liu, Ziyuan Liu, Yang Qin, Wanting Zhou, Zichuan Wang, Jiachen Liang, Xu Chu, Shaowei Chen, Zhaokai Luo, Mu Chuan

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.26887.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.26887

Published: 2026-10-22T01:50:45.357Z

2. Large language model and reinforcement learning based autonomous construction and progress updates

Abstract:Due to the rapid growth of the global population, traditional construction methods are gradually unable to meet the current housing demand. The advancement of machine learning algorithms has encouraged researchers to explore their applications in the construction industry, leading to more intelligent and efficient building processes. In particular, the rise of large language models in recent years has provided more possibilities for the transformation from traditional construction to smart construction. In this study, a large language model and reinforcement learning based autonomous construction method is proposed. An enhanced soft actor-critic reinforcement learning algorithm is developed to plan actions for a mobile crane according to construction tasks. Robot kinematics and control algorithms are used to ensure that the actions generated by the reinforcement learning algorithm can be executed by the crane. In addition, a customized embedding model is developed to extract information from the construction environment according to user interests. Then, a large language model, ChatGPT 5, is used to update construction progress, provide safety information to control system, and generate construction progress reports. The feasibility of proposed method is verified through several full-scall retaining wall construction experiments.

中文摘要

摘要：由于全球人口的快速增长，传统的建筑方法逐渐无法满足当前的住房需求。机器学习算法的进步促使研究人员探索其在建筑行业的应用，从而实现更智能高效的建筑流程。特别是近年来大型语言模型的兴起，为从传统建筑向智慧建筑的转变提供了更多可能性。本研究提出了一种基于大型语言模型和强化学习的自主施工方法。开发了一种增强的软演员-批评者强化学习算法，根据施工任务为移动起重机规划动作。利用机器人运动学和控制算法确保强化学习算法生成的动作能够被起重机执行。此外，还开发了一个定制的嵌入模型，根据用户兴趣从施工环境中提取信息。然后，使用大型语言模型ChatGPT 5更新施工进度，为控制系统提供安全信息，并生成施工进度报告。通过若干全尺寸挡土墙施工实验验证了所提出方法的可行性。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.26899 (HTTP 429)

Authors: Yifei Xiao, Huasheng Lin, T.Y. Yang

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.26899.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.26899

Published: 2026-10-22T01:50:45.357Z

3. Pareto-Optimal RTL Code Generation via Multi-Objective Reinforcement Learning with Large Language Models

Abstract:Recent studies have proposed reinforcement learning with verifiable rewards (RLVR) methods that enable LLMs to generate high-quality RTL circuits from natural-language specifications. Existing approaches optimize circuit PPA (Power, Performance, Area) by aggregating multiple objectives into a single scalar reward, making it difficult to simultaneously generate diverse Pareto-optimal solutions across different trade-offs such as low power, low latency, and small area. However, in practical design, the relative importance of each objective varies depending on product requirements and design constraints; thus, obtaining a diverse and high-quality candidate set is beneficial in that designers can choose a preferred trade-off a posteriori. In this work, we are the first to propose a learning method for LLM-based circuit generation that optimizes, not the quality of a single circuit, but the quality and diversity of the entire circuit set obtained through multiple generations in the PPA objective space. The proposed method consists of (1) a generation-based training pipeline that reformulates coverage improvement of the entire Pareto solution set into a sequential circuit generation problem, and (2) a reward function based on Hypervolume Contribution (HVC) that measures how much a generated circuit contributes to the Pareto front. Evaluation on the RTLLM v2 benchmark shows that the proposed method achieves +13.3 percentage points in relative Hypervolume over a baseline without PPA optimization and +14.4 percentage points over a single-objective reinforcement learning method, while also demonstrating superiority in circuit diversity.

中文摘要

摘要：近期的研究提出了可验证奖励的强化学习（RLVR）方法，使大语言模型（LLM）能够根据自然语言规格生成高质量的RTL电路。现有方法通过将多个目标聚合为单一标量奖励来优化电路的PPA（功耗、性能、面积），这使得难以在如低功耗、低延迟和小面积等不同权衡之间同时生成多样化的帕累托最优解。然而，在实际设计中，每个目标的重要性相对值取决于产品需求和设计约束；因此，获得多样且高质量的候选集合是有益的，这样设计者可以在事后选择自己偏好的权衡方案。在本工作中，我们首次提出了一种基于LLM的电路生成学习方法，该方法优化的不是单一电路的质量，而是通过多次生成在PPA目标空间中获得的整个电路集合的质量和多样性。所提出的方法包括：(1) 基于生成的训练管道，将整个帕累托解集的覆盖率提升重新表述为序列电路生成问题；(2) 基于超体积贡献（HVC）的奖励函数，用于衡量生成的电路对帕累托前沿的贡献。对RTLLM v2基准的评估显示，该方法在相对超体积指标上比未进行PPA优化的基线提高了13.3个百分点，比单目标强化学习方法提高了14.4个百分点，同时在电路多样性方面也表现出优势。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.26904 (HTTP 429)

Authors: Koki Takeshita, Kenji Uebayashi, Hikaru Tomonari, Masaya Tsunokake, Koichiro Yawata, Tatsuya Sasaki, Yusuke Ohtsubo, Kota Dohi, Yasuhiro Sogawa

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.26904.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.26904

Published: 2026-10-22T01:50:45.357Z

4. MedPAO-Assist: Edge-Deployable, Modality-Aware Agentic AI for Radiology Report Structuring

Abstract:The adoption of Large Language Models for radiology report structuring has been largely dominated by either proprietary cloud-based models or large open models requiring high-end GPUs for deployment, which is a critical bottleneck for clinical adoption in radiology workflows that demand data privacy and offline operation. To address this, we present MedPAO-Assist, an edge-deployable agentic framework that generates structured radiology reports directly from radiologist dictation and voice commands, requiring only $\approx3.3$ GB peak RAM and achieving an average end-to-end latency of $\approx31$ seconds on a standard laptop CPU through an asynchronous, multi-threaded execution pipeline. The system currently supports four major imaging modalities — Chest X-ray, Chest CT, Abdomen CT, and Brain MRI. To our knowledge, it is among the earliest agentic systems to run its entire pipeline with a Small Language Model (SLM) in the 0.6B-parameter range. This multi-agent architecture, built around an Extract-Ground-Generate paradigm, performs concept extraction, SNOMED-CT grounding, and structured report generation, followed by a speech-driven human-in-the-loop correction phase, together forming a fully “no-touch” reporting workflow that requires no keyboard input, external API calls, or internet connectivity. We evaluate each pipeline stage individually — concept extraction, ontology grounding, and concept categorization and report hardware profiling results demonstrating the system’s feasibility for real-time, CPU-only clinical deployment.

中文摘要

摘要：放射学报告结构化中大型语言模型的采用主要被专有的基于云的模型或需要高端 GPU 部署的大型开放模型所主导，而这对于要求数据隐私和离线操作的放射学工作流程来说，是临床应用的关键瓶颈。为了解决这一问题，我们提出了 MedPAO-Assist，一种可在边缘部署的自主框架，能够直接从放射科医生的口述和语音指令生成结构化放射学报告，仅需约 3.3 GB 的峰值内存，并通过异步多线程执行管道在标准笔记本 CPU 上实现约 31 秒的平均端到端延迟。该系统目前支持四种主要影像模式——胸部 X 光、胸部 CT、腹部 CT 和脑部 MRI。据我们所知，它是最早能够使用参数量约为 0.6B 的小型语言模型（SLM）完成整个管道的自主系统之一。这个以“提取-映射-生成”范式为核心的多代理架构执行概念提取、SNOMED-CT 映射和结构化报告生成，随后进入语音驱动的人类在环纠正阶段，共同形成完全“免触控”的报告工作流程，无需键盘输入、外部 API 调用或网络连接。我们对每个管道阶段——概念提取、本体映射以及概念分类——进行了单独评估，并报告了硬件性能分析结果，证明该系统在仅使用 CPU 的情况下实现实时临床部署的可行性。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.26913 (HTTP 429)

Authors: Shrish Shrinath Vaidya, Minmini Selvam, Ganapathy Krishnamurthi

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.26913.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.26913

Published: 2026-10-22T01:50:45.357Z

5. DataPraxis: Scaling Analytics Tasks via Verified Workspace Evolution

Abstract:General-purpose agents are increasingly expected to perform professional analytics work, which requires sustained reasoning, iterative investigation, and reliable evidence. Training these capabilities requires diverse and challenging tasks. However, generating tasks directly from raw data often overlooks deeper questions that emerge through analysis and provides limited evidence for verification. To address these limitations, we introduce DataPraxis, an analytics task synthesis framework based on verified workspace evolution. The framework progressively enriches workspaces with validated findings and artifacts, then inherits and combines these workspaces to uncover deeper questions. It constructs tasks from accumulated findings and their dependencies, verifying analytical results both during evolution and after task assembly. Using this framework, we construct DataPraxis-5K, a dataset of 5,000 verified trajectories spanning diverse domains and output formats, to train DataPraxis-27B and DataPraxis-35B-A3B. DataPraxis-27B achieves 55.11 on DABStep and 51.9 on JobBench, matching or surpassing frontier models such as GLM-5.2. DataPraxis-5K is available at this http URL

中文摘要

摘要：通用智能体越来越被期望执行专业的分析工作，这需要持续的推理、反复的调查以及可靠的证据。训练这些能力需要多样且具有挑战性的任务。然而，直接从原始数据生成任务往往忽略通过分析出现的更深层问题，并且提供的验证证据有限。为了解决这些局限性，我们提出了 DataPraxis，一种基于验证工作区演化的分析任务合成框架。该框架通过验证的发现和文档逐步丰富工作区，然后继承并组合这些工作区以发现更深层次的问题。它从累积的发现及其依赖关系中构建任务，并在演化过程中以及任务组装后验证分析结果。使用该框架，我们构建了 DataPraxis-5K，这是一个包含 5,000 条跨多领域且输出格式多样的验证轨迹的数据集，用于训练 DataPraxis-27B 和 DataPraxis-35B-A3B。DataPraxis-27B 在 DABStep 上达到 55.11 分，在 JobBench 上达到 51.9 分，与 GLM-5.2 等前沿模型相当或超越。DataPraxis-5K 可通过此 http URL 获取。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.26924 (HTTP 429)

Authors: Nan Jia, Wenxuan Shi, Jingqing Ruan, Zhizhao Zeng, Tianyi Zhai, Zhongda Su, Feng Hong, Zhengyu Chen, Wenhao Yan, Haojin Yang, Siqi Dai, Weipeng Zhang, Ke Zeng, Xunliang Cai

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.26924.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.26924

Published: 2026-10-22T01:50:45.357Z

6. Industrial Agentic AI: LLM-Driven New-Generation Agents for Smart Manufacturing

Abstract:Large language models (LLMs) are enabling agents to move beyond predefined workflows toward autonomous planning and execution, creating new opportunities for smart manufacturing. However, dynamic environments, complex physical constraints, and stringent reliability requirements hinder the direct deployment of general-purpose agents in smart manufacturing. This article presents a systematic overview of LLM-driven industrial agents. First, we analyze five key challenges and discuss the concept and core characteristics of industrial agents. We then establish a unified taxonomy covering application scenarios, embodiment forms, and five capability levels. Next, we propose a four-layer system architecture and detail its core modules and operational mechanisms for closed-loop task execution. Finally, we review representative applications across the industrial lifecycle and outline future research directions. This article provides a structured reference for the design, development, and application of industrial agents.

中文摘要

摘要：大型语言模型（LLMs）正在使智能体能够超越预定义的工作流，迈向自主规划和执行，为智能制造创造新的机会。然而，动态环境、复杂的物理约束以及严格的可靠性要求阻碍了通用智能体在智能制造中的直接部署。本文系统地概述了由LLM驱动的工业智能体。首先，我们分析了五个关键挑战，并讨论了工业智能体的概念及核心特征。随后，我们建立了涵盖应用场景、具身形式和五个能力等级的统一分类体系。接着，我们提出了一个四层系统架构，并详细介绍了其闭环任务执行的核心模块和操作机制。最后，我们回顾了贯穿工业生命周期的代表性应用，并概述了未来的研究方向。本文为工业智能体的设计、开发与应用提供了结构化的参考。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.26941 (HTTP 429)

Authors: Lei Ren, Haiteng Wang, Shixiang Li, Jing Zhang, Yuqing Wang, Jingwei Guo, Yunfei Zhu, Yuanjun Laili, Yang Tang, Lihui Wang

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.26941.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.26941

Published: 2026-10-22T01:50:45.357Z

7. Fix, Not Just Resample: Bounded Compiler-Guided Repair in Lean 4

Abstract:Automated theorem provers often rely on sampling language models, but proof assistants like Lean reject each failed step with a precise error message —- information that pure sampling discards. To overcome this, we introduce a bounded, online step-level repair module. Our search samples $K$ candidate tactics at each step and repairs a failed tactic at most $R$ times. Instead of discarding failed tactics, we feed the compiler’s error message to a pretrained repair model to fix individual rejected steps in place. This approach improves the solve rate over a sampling-only baseline across all tested candidate budgets ($K \in {1,3,5,8}$), achieving up to a $32\%$ relative gain and solving $82$ additional theorems on a $1{,}990$-theorem subset of the LeanDojo benchmark. Repair is not free: it trades extra runtime for reachable proofs. But the trade is predictable: repair pays most when sampling is constrained; its return is front-loaded, with the first repair attempt capturing $84\text{—}96\%$ of the total gain; and the gain comes almost entirely from shallow tactic-execution failures rather than structural errors such as type mismatches. This targeted repair successfully replicates on a second benchmark (Lean Workbook), nearly doubling the solve rate with zero regressions.

中文摘要

摘要：自动定理证明器通常依赖采样语言模型，但像Lean这样的证明助手会用精确的错误信息—-信息拒绝每个失败步骤，而纯采样则会丢弃这些信息。为克服这个问题，我们引入了一个有界的在线步级修复模块。我们的搜索在每步采样$K候选策略，最多修复失败策略$R$次。我们不丢弃失败的策略，而是将编译器的错误消息输入预训练修复模型，以固定被拒绝的单个步骤。这种方法在所有测试候选预算（$K {3,5,8}$）中，提高了仅采样基线的求解率，实现了最高32\%$的相对增益，并在LeanDojo基准测试中1{，}990$定理子集上额外解决了82美元。修复并非免费：它用额外的运行时间换取可达证明。但这种交易是可预测的：当采样受限时修复收益最高;其回报是前期的，第一次修复尝试捕获了总收益的84美元;收益几乎完全来自浅层战术执行失败，而非类型不匹配等结构性错误。这种有针对性的修复在第二个基准测试（精益工作手册）上成功复制，几乎将零回归的破解率翻倍。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.26942 (HTTP 429)

Authors: Haotian Deng, Chris Farber, Sonal Narasimha, Samskrithi Sivakumar, Yexiang Xue

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.26942.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.26942

Published: 2026-10-22T01:50:45.357Z

8. MAMath-Bench: Process-Aware and Robust Evaluation of Multimodal Multi-Agent Mathematical Reasoning

Abstract:Mathematical reasoning is a fundamental capability for educational applications of Large Language Models (LLMs), where reliable problem solving requires not only correct answers but also coherent and verifiable reasoning processes. Multi-agent systems (MASs) have shown growing potential in this domain by leveraging collaboration and verification among multiple LLM agents. However, existing evaluations primarily focus on the correctness of final answers in text-only settings, offering limited insight into multimodal reasoning capabilities, the reliability of intermediate reasoning steps, and robustness against reasoning errors. To address these limitations, we introduce MAMath-Bench, a process-aware benchmark designed to evaluate the multimodal mathematical reasoning capabilities of MASs. MAMath-Bench comprises 1,003 problems spanning 11 mathematical categories, multiple difficulty levels, and several modalities, including text, images, and audio. It features fine-grained milestone annotations to evaluate systems across three dimensions: problem-solving correctness, reasoning-process reliability, and robustness to adversarial reasoning-chain perturbations. Our evaluation yields three key findings. First, multi-agent collaboration does not consistently improve final-answer accuracy over strong single LLMs. Second, collaboration can substantially enhance the coverage of intermediate reasoning, but such gains may come at the cost of increased reasoning redundancy. Third, current systems remain vulnerable to adversarial reasoning errors, particularly those involving subtle structural perturbations. The dataset is available at this https URL.

中文摘要

摘要：数学推理是大型语言模型（LLMs）在教育应用中的一项基本能力，其中可靠的问题解决不仅需要正确答案，还需要连贯且可验证的推理过程。多智能体系统（MASs）在这一领域显示出越来越大的潜力，通过多个LLM智能体之间的协作和验证。然而，现有评价主要集中于文本环境下最终答案的正确性，对多模态推理能力、中间推理步骤的可靠性以及对推理错误的鲁棒性提供的洞察有限。为了解决这些局限性，我们提出了MAMath-Bench，这是一项关注过程的基准，旨在评估MASs的多模态数学推理能力。MAMath-Bench包含1,003个问题，涵盖11个数学类别、多个难度等级，以及包括文本、图像和音频在内的多种模态。它具有细粒度的里程碑标注，可从三个维度评估系统：问题解决的正确性、推理过程的可靠性，以及对对抗性推理链扰动的鲁棒性。我们的评估得出了三个主要发现。首先，多智能体协作并不总能在最终答案的准确性上超越强大的单体LLM。其次，协作可以显著提高中间推理的覆盖率，但这种提升可能以增加推理冗余为代价。第三，目前的系统仍然容易受到对抗性推理错误的影响，尤其是那些涉及微妙结构扰动的错误。该数据集可通过此URL获得。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.26943 (HTTP 429)

Authors: Xinping Wang, Jie Ma, Rui Xing, Wensong Guo, Xiayi Zhang, Jianhao Deng, Zihao Zhou, Boqian Yang, Feijuan He

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.26943.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.26943

Published: 2026-10-22T01:50:45.357Z

9. LightCRS: Lightweight Multi-Domain Conversational Recommendation System

Abstract:Conversational recommender systems have recently adopted vision-language models to match user dialogue intents with multimodal product information. However, extending these methods to many product domains faces a dilemma: deploying a separately adapted VLM per domain scales cost linearly with the number of domains and requires routing each dialogue to the right model, while naively merging all domains into a single VLM degrades accuracy because product characteristics vary widely across categories. We propose LightCRS, a lightweight, query-conditioned framework for unified multi-domain multimodal conversational recommendation. LightCRS precomputes multimodal product representations once offline with frozen encoders. Online, a compact 5.3M-parameter module fuses this evidence for each dialogue via query-conditioned cross-attention, whose parameters are shared across all domains, and compares candidates jointly with an inter-candidate Transformer. To enable rigorous evaluation, we construct MM-R2A-25, a 25-domain benchmark pairing natural Reddit recommendation-seeking conversations with Amazon product listings. LightCRS achieves state-of-the-art average HR@1, NDCG@5, and MRR, improving HR@1 by 35.4\% over the strongest baseline while reducing per-query latency by 266 times. Joint multi-domain training itself improves HR@1 by 15.2\% over separately trained single-domain models.

中文摘要

摘要：对话推荐系统最近采用了视觉-语言模型，将用户对话意图与多模态产品信息进行匹配。然而，将这些方法扩展到多个产品领域面临一个两难境地：为每个领域部署单独适配的视觉-语言模型，其成本随领域数量呈线性增长，并且需要将每个对话路由到正确的模型；而将所有领域简单合并到单一视觉-语言模型中会降低准确性，因为不同类别的产品特性差异很大。我们提出了LightCRS，一种轻量级、基于查询条件的统一多领域多模态对话推荐框架。LightCRS使用固定编码器离线预计算多模态产品表示。在线阶段，一个紧凑的530万参数模块通过基于查询条件的交叉注意力融合每个对话的证据，该模块参数在所有领域共享，并通过候选项间的Transformer进行联合比较。为了实现严格评估，我们构建了MM-R2A-25，这是一个25领域的基准数据集，将自然的Reddit推荐请求对话与亚马逊产品列表配对。LightCRS在平均HR@1、NDCG@5和MRR指标上实现了最先进的水平，将HR@1较最强基线提升了35.4%，同时将每次查询的延迟降低了266倍。联合多领域训练本身也使HR@1较单独训练的单领域模型提高了15.2%。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.26948 (HTTP 429)

Authors: He Ma, Lan You, Qiyang Wu, Chen Liu

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.26948.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.26948

Published: 2026-10-22T01:50:45.357Z

10. When Planning Problems Become Hard for LLMs: Separating Structural Complexity from Semantic Interpretability

Abstract:Large language models (LLMs) struggle with classical planning, but benchmarks rarely distinguish structural difficulty from the difficulty of interpreting symbolic vocabulary. We examine seven STRIPS domains with BFS-optimal plan lengths. For predicting structural difficulty without solving an instance, the zero-training delete-relaxation heuristic $h_{\mathrm{add}}$ outperforms learned alternatives on held-out domains (mean Spearman $|\rho|=0.931$), including ridge regression ($0.707$), object count ($0.657$), and the episodic retrieval architecture ARC ($0.400$). ARC also fails to outperform a linear model using the same features. To isolate semantic effects, we progressively rename predicates while preserving states, transitions, and optimal plans. Valid LLM planning success drops from $67.5\%$ to $40.0\%$ between $0\%$ and $100\%$ renaming, while refusals remain at $0\%$. An LLM supplied with ARC’s conditioning information predicts difficulty in context better than trained ARC (mean $|\rho|=0.826$). Hybrid routing between an LLM and a symbolic planner depends more on calibration to the target LLM’s outcomes than on the structural score selected. Thus, established symbolic heuristics already estimate structural difficulty well; improving hybrid planning requires modeling LLM behavior sensitive to predicate semantics rather than developing more structural predictors.

中文摘要

摘要：大型语言模型（LLMs）在传统规划方面存在困难，但基准很少区分结构难度与符号词汇解释难度。我们考察了七个具有BFS最优计划长度的STRIPS领域。对于在不解决实例的情况下预测结构难度，零训练的删除-放松启发式$h_{\mathrm{add}}$在保留未完成的领域（平均Spearman $|\rho|=0.931$）上优于学习到的替代方案，包括脊回归（$0.707$）、对象数量（$0.657$）和情节检索架构ARC（$0.400美元）。ARC也未能超越使用相同特征的线性模型。为隔离语义效应，我们逐步重命名谓词，同时保留状态、转移和最优计划。有效LLM规划成功率从$0\%$到$100\%$之间的$67.5降至$40.0\%$，而拒绝率仍为$0。提供ARC条件信息的LLM在上下文中预测困难度优于训练后的ARC（平均$|\rho|=0.826$）。LLM与符号规划器之间的混合路由更多依赖于对目标LLM结果的校准，而非所选结构评分。因此，已建立的符号启发式方法已能很好地估计结构难度;改进混合规划需要建模对谓词语义敏感的LLM行为，而非开发更多结构预测变量。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.26964 (HTTP 429)

Authors: Mayukh Debnath, Adeyinka Adedigba, Rammohan Mallipeddi

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.26964.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.26964

Published: 2026-10-22T01:50:45.357Z

Evaluation Domain Papers

1. Navalia: Advancing Cowork Agents through Verifiable Task Synthesis and Long-Horizon Post-Training

中文摘要

摘要：大型语言模型（LLM）代理正在从对话助手发展为能够自主规划、使用工具和执行复杂任务的协作代理。财务分析要求这些代理整合跨文档证据并执行多步骤数值推理，但专家任务编写成本高昂，而直接使用LLM生成则难以产生具有可靠答案的挑战性任务。我们提出了Navalia，一个结合可验证任务生成与长程后训练的协作代理框架。Navalia以企业财务报告作为数据源，通过语义方式从完整事实集合中选择支持证据，并构建可执行的计算链以生成具有可追踪证据和各步骤实际数据依赖关系的分析任务。我们进一步使用独立LLM审核和程序化检查来验证证据一致性和答案正确性。经过验证的任务随后用于轨迹收集、冷启动监督微调（SFT）和强化学习（RL），以提升长程任务执行能力。Navalia以Qwen3.6-35B-A3B为初始化模型，在五个基准测试中均取得了相对于基础模型的一致提升：OfficeQA Pro、FinSearchComp、Finance Agent Benchmark、APEX-Agents-AA 和 Workspace-Bench-Lite。这些结果表明，Navalia提升了证据整合、工具使用以及长程任务执行能力，并能推广到更广泛的协作场景，包括一般办公任务。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.26887 (HTTP 429)

Authors: Hengqi Liu, Ziyuan Liu, Yang Qin, Wanting Zhou, Zichuan Wang, Jiachen Liang, Xu Chu, Shaowei Chen, Zhaokai Luo, Mu Chuan

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.26887.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.26887

Published: 2026-10-22T01:51:04.474Z

2. Large language model and reinforcement learning based autonomous construction and progress updates

中文摘要

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.26899 (HTTP 429)

Authors: Yifei Xiao, Huasheng Lin, T.Y. Yang

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.26899.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.26899

Published: 2026-10-22T01:51:04.474Z

3. Pareto-Optimal RTL Code Generation via Multi-Objective Reinforcement Learning with Large Language Models

中文摘要

摘要：近期的研究提出了可验证奖励的强化学习（RLVR）方法，使大语言模型（LLM）能够根据自然语言规格生成高质量的RTL电路。现有方法通过将多个目标聚合为单一标量奖励来优化电路的PPA（功耗、性能、面积），这使得难以在如低功耗、低延迟和小面积等不同权衡之间同时生成多样化的帕累托最优解。然而，在实际设计中，每个目标的重要性相对而言会因产品需求和设计约束而不同；因此，获得多样且高质量的候选集合是有益的，因为设计人员可以在事后选择自己偏好的权衡方案。在本工作中，我们首次提出了一种基于LLM的电路生成学习方法，该方法优化的不是单一电路的质量，而是通过多次生成在PPA目标空间中获得的整个电路集合的质量和多样性。所提出的方法包括：(1) 基于生成的训练管道，将整个帕累托解集的覆盖率提升重新表述为序列电路生成问题；(2) 基于超体积贡献（HVC）的奖励函数，用于衡量生成的电路对帕累托前沿的贡献。对RTLLM v2基准的评估显示，该方法在相对超体积指标上比未进行PPA优化的基线提高了13.3个百分点，比单目标强化学习方法提高了14.4个百分点，同时在电路多样性方面也表现出优越性。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.26904 (HTTP 429)

Authors: Koki Takeshita, Kenji Uebayashi, Hikaru Tomonari, Masaya Tsunokake, Koichiro Yawata, Tatsuya Sasaki, Yusuke Ohtsubo, Kota Dohi, Yasuhiro Sogawa

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.26904.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.26904

Published: 2026-10-22T01:51:04.474Z

4. MedPAO-Assist: Edge-Deployable, Modality-Aware Agentic AI for Radiology Report Structuring

中文摘要

摘要：放射学报告结构化中大型语言模型的应用主要被专有的基于云的模型或需要高端GPU部署的大型开放模型所主导，而这对于需要数据隐私和离线操作的放射学工作流程来说是临床应用的关键瓶颈。为了解决这一问题，我们提出了MedPAO-Assist，一种可在边缘部署的自主框架，可以直接从放射科医生的口述和语音指令生成结构化放射学报告，仅需约3.3 GB的峰值内存，并通过异步多线程执行管道，在标准笔记本CPU上实现约31秒的平均端到端延迟。该系统目前支持四种主要成像模态——胸部X光、胸部CT、腹部CT和脑部MRI。据我们所知，它是最早使用参数约0.6B的小型语言模型（SLM）运行整个流程的自主系统之一。该多智能体架构以提取-落地-生成（Extract-Ground-Generate）范式构建，执行概念提取、SNOMED-CT落地和结构化报告生成，随后进行语音驱动的人机循环校正阶段，共同形成完全“免触控”的报告流程，无需键盘输入、外部API调用或互联网连接。我们分别评估了每个流程阶段——概念提取、本体落地及概念分类，并报告了硬件性能结果，证明该系统可实现实时、仅CPU的临床部署可行性。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.26913 (HTTP 429)

Authors: Shrish Shrinath Vaidya, Minmini Selvam, Ganapathy Krishnamurthi

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.26913.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.26913

Published: 2026-10-22T01:51:04.474Z

5. DataPraxis: Scaling Analytics Tasks via Verified Workspace Evolution

中文摘要

摘要：通用智能体越来越被期望执行专业的分析工作，这需要持续的推理、反复的调查以及可靠的证据。训练这些能力需要多样且具有挑战性的任务。然而，直接从原始数据生成任务往往忽略通过分析出现的更深层问题，并且提供的验证证据有限。为了解决这些局限性，我们提出了 DataPraxis，一种基于验证工作区演化的分析任务合成框架。该框架通过验证的发现和文档逐步丰富工作区，然后继承并组合这些工作区以发现更深层次的问题。它从累积的发现及其依赖关系中构建任务，在演化过程中以及任务组装后验证分析结果。使用该框架，我们构建了 DataPraxis-5K，这是一个包含 5000 条经过验证的轨迹的数据集，涵盖多样的领域和输出格式，用于训练 DataPraxis-27B 和 DataPraxis-35B-A3B。DataPraxis-27B 在 DABStep 上取得 55.11 分，在 JobBench 上取得 51.9 分，匹配或超越了如 GLM-5.2 等前沿模型。DataPraxis-5K 可在此 http URL 获取。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.26924 (HTTP 429)

Authors: Nan Jia, Wenxuan Shi, Jingqing Ruan, Zhizhao Zeng, Tianyi Zhai, Zhongda Su, Feng Hong, Zhengyu Chen, Wenhao Yan, Haojin Yang, Siqi Dai, Weipeng Zhang, Ke Zeng, Xunliang Cai

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.26924.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.26924

Published: 2026-10-22T01:51:04.474Z

6. Industrial Agentic AI: LLM-Driven New-Generation Agents for Smart Manufacturing

中文摘要

摘要：大型语言模型（LLMs）正在使智能体能够超越预定义的工作流，迈向自主规划和执行，为智能制造创造新的机会。然而，动态环境、复杂的物理约束以及严格的可靠性要求阻碍了通用智能体在智能制造中的直接部署。本文系统地概述了由LLM驱动的工业智能体。首先，我们分析了五个关键挑战，并讨论了工业智能体的概念及核心特征。随后，我们建立了涵盖应用场景、具身形式和五个能力等级的统一分类体系。接着，我们提出了一个四层系统架构，并详细介绍了其闭环任务执行的核心模块和运行机制。最后，我们回顾了工业生命周期中具有代表性的应用，并概述了未来的研究方向。本文为工业智能体的设计、开发及应用提供了结构化的参考。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.26941 (HTTP 429)

Authors: Lei Ren, Haiteng Wang, Shixiang Li, Jing Zhang, Yuqing Wang, Jingwei Guo, Yunfei Zhu, Yuanjun Laili, Yang Tang, Lihui Wang

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.26941.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.26941

Published: 2026-10-22T01:51:04.474Z

7. Fix, Not Just Resample: Bounded Compiler-Guided Repair in Lean 4

中文摘要

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.26942 (HTTP 429)

Authors: Haotian Deng, Chris Farber, Sonal Narasimha, Samskrithi Sivakumar, Yexiang Xue

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.26942.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.26942

Published: 2026-10-22T01:51:04.474Z

8. MAMath-Bench: Process-Aware and Robust Evaluation of Multimodal Multi-Agent Mathematical Reasoning

中文摘要

摘要：数学推理是大语言模型（LLMs）在教育应用中的一项基础能力，其中可靠的问题解决不仅需要正确答案，还需要连贯且可验证的推理过程。多智能体系统（MASs）在这一领域显示出越来越大的潜力，通过多个LLM智能体之间的协作和验证。然而，现有评价主要集中于文本环境下最终答案的正确性，对多模态推理能力、中间推理步骤的可靠性以及对推理错误的鲁棒性提供的洞察有限。为了解决这些局限性，我们提出了MAMath-Bench，这是一项关注过程的基准，旨在评估MASs的多模态数学推理能力。MAMath-Bench包含1,003个问题，涵盖11个数学类别、多个难度等级，以及包括文本、图像和音频在内的多种模态。它具有细粒度的里程碑标注，可从三个维度评估系统：问题解决的正确性、推理过程的可靠性，以及对对抗性推理链扰动的鲁棒性。我们的评估得出了三个主要发现。首先，多智能体协作并不总能在最终答案准确性上优于强大的单一LLM。其次，协作可以显著增强中间推理的覆盖范围，但这种提升可能以增加推理冗余为代价。第三，目前的系统仍然容易受到对抗性推理错误的影响，尤其是涉及细微结构扰动的错误。该数据集可通过此https URL获取。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.26943 (HTTP 429)

Authors: Xinping Wang, Jie Ma, Rui Xing, Wensong Guo, Xiayi Zhang, Jianhao Deng, Zihao Zhou, Boqian Yang, Feijuan He

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.26943.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.26943

Published: 2026-10-22T01:51:04.474Z

9. LightCRS: Lightweight Multi-Domain Conversational Recommendation System

中文摘要

摘要：对话推荐系统最近采用了视觉-语言模型，将用户对话意图与多模态产品信息进行匹配。然而，将这些方法扩展到多个产品领域面临一个两难境地：为每个领域部署单独适配的视觉-语言模型，其成本随领域数量呈线性增长，并且需要将每个对话路由到正确的模型；而将所有领域简单合并到单一视觉-语言模型中会降低准确性，因为不同类别的产品特性差异很大。我们提出了LightCRS，一种轻量级、基于查询条件的统一多领域多模态对话推荐框架。LightCRS使用固定编码器离线预计算多模态产品表示。在线阶段，一个紧凑的530万参数模块通过基于查询条件的交叉注意力融合每个对话的证据，该模块参数在所有领域共享，并通过候选项间的Transformer进行联合比较。为了实现严格评估，我们构建了MM-R2A-25，这是一个25领域的基准数据集，将自然的Reddit推荐请求对话与亚马逊产品列表配对。LightCRS在平均HR@1、NDCG@5和MRR指标上实现了最先进的性能，将HR@1较最强基线提升了35.4%，同时将每次查询的延迟降低了266倍。联合多领域训练本身也使HR@1较单独训练的单领域模型提高了15.2%。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.26948 (HTTP 429)

Authors: He Ma, Lan You, Qiyang Wu, Chen Liu

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.26948.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.26948

Published: 2026-10-22T01:51:04.474Z

10. When Planning Problems Become Hard for LLMs: Separating Structural Complexity from Semantic Interpretability

中文摘要

摘要：大语言模型（LLMs）在经典规划中表现不佳，但现有基准测试很少区分结构性难度与解释符号词汇的难度。我们考察了七个具有BFS最优计划长度的STRIPS领域。在不解决实例的情况下预测结构性难度时，零训练删除-放松启发式$h_{\mathrm{add}}$在保留的领域中优于已学习的替代方法（平均Spearman $|\rho|=0.931$），包括岭回归（$0.707$）、对象计数（$0.657$）以及情景检索架构ARC（$0.400$）。ARC也未能超越使用相同特征的线性模型。为了孤立语义效应，我们在保持状态、变换和最优计划不变的情况下，逐步重命名谓词。有效LLM规划成功率在 $0\%$ 到 $100\%$ 重命名之间从 $67.5\%$ 下降到 $40.0\%$，而拒绝率保持在 $0\%$。提供ARC条件信息的LLM在上下文中预测难度的表现优于已训练ARC（平均 $|\rho|=0.826$）。LLM与符号规划器间的混合路由更多依赖于对目标LLM结果的校准，而非所选的结构分数。因此，已建立的符号启发式已经能够很好地估计结构性难度；改进混合规划需要建模对谓词语义敏感的LLM行为，而不是开发更多的结构性预测方法。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.26964 (HTTP 429)

Authors: Mayukh Debnath, Adeyinka Adedigba, Rammohan Mallipeddi

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.26964.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.26964

Published: 2026-10-22T01:51:04.474Z

VLM Domain Papers

1. EGGHead: Efficient Generalizable Gaussian Heads from a Single Image

Abstract:Creating photorealistic human heads typically involves trade-offs between representation power, efficiency, and generalizability. Traditional mesh-based solutions relying on 3DMMs generally sacrifice representation power but are efficient to render. On the other hand, NeRF-based methods improve representation power and fidelity while sacrificing efficiency. Gaussian splatting methods attempt to resolve these trade-offs by offering both high representation power and efficiency with real-time rendering. However, their efficiency nonetheless depends on the number of rasterized Gaussians, and the time needed to optimize and prepare the head (the enrollment time) is often high. In light of this, we propose EGGHead, which introduces a novel Gaussian allocation algorithm that efficiently assigns Gaussians to the head’s semantic regions given a fixed Gaussian budget. Our algorithm reduces the number of Gaussians needed while minimizing the impact on representation power. EGGHead operates in a single forward pass and outperforms previous single-image face modeling methods in reconstruction quality, novel view synthesis, and rendering speeds.

中文摘要

摘要：创建真实感的人头通常涉及表示能力、效率和泛化能力之间的权衡。依赖3DMM的传统基于网格的解决方案通常会牺牲表示能力，但渲染效率较高。另一方面，基于NeRF的方法提升了表示能力和保真度，但牺牲了效率。高斯点云（Gaussian splatting）方法试图通过实现高表示能力和高效率的实时渲染来解决这些权衡。然而，其效率仍取决于光栅化高斯点的数量，并且优化和准备人头（注册时间）所需的时间通常很高。有鉴于此，我们提出了EGGHead，引入了一种新颖的高斯分配算法，在固定的高斯预算下高效地将高斯点分配到人头的语义区域。我们的算法在减少所需高斯点数量的同时，将对表示能力的影响降到最小。EGGHead在单次前向传播中运行，并在重建质量、新视角合成和渲染速度方面优于以往的单幅图像人脸建模方法。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.26898 (HTTP 429)

Authors: Pilseo Park, Andrew Hou, Michel Sarkis, Ning Bi, Yiying Tong, Xiaoming Liu

Categories: cs.CV

PDF URL: https://arxiv.org/pdf/2610.26898.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.26898

Published: 2026-10-22T01:51:24.612Z

2. Beckmann World Models: Direct Terminal-Map Learning for One-Step Action-Conditioned Prediction

Abstract:Action-conditioned world models support planning and policy evaluation by predicting future states. However, learning an accurate one-step predictor can require substantial training computation. We introduce Beckmann World Models (BWM), which learn a direct map from noise, visual history, and actions to future frames without a pretrained generative teacher. We adapt the terminal-map conservation law of Beckmann Transport Models to action-conditioned video prediction, training directly from noise-data pairs without conditional populations of generated samples required by DriftWorld. However, transport consistency on interpolated inputs does not ensure accurate prediction from the pure noise used at inference. We address this gap with endpoint-aligned training that directly supervises the source prediction, emphasizes motion and spatial structure, and teaches recovery from generated histories. All components train the same map and retain one network call per future chunk. On PushT and Robomimic Can, the final model reduces 64-frame prediction MSE by 68% and 57%, respectively, relative to same-backbone regression baseline. For policy evaluation, BWM achieves the lowest mean IoU error among the evaluated models, 0.0100, at 24.8 ms per four-frame PushT prediction. On Robomimic Can, the BWM stage trained with motion supervision and generated histories achieves 3.3 times higher throughput and 78% lower peak memory than local one-step DriftWorld. In a latent-space extension to RT-1 and Bridge-V2, equal-update controls show that feature supervision under generated histories improves perceptual prediction. These findings show how terminal-map learning can combine efficient training with accurate one-step visual prediction.

中文摘要

摘要：动作条件世界模型通过预测未来状态来支持规划和策略评估。然而，学习一个准确的一步预测器可能需要大量的训练计算。我们提出了 Beckmann 世界模型（BWM），它直接从噪声、视觉历史和动作映射到未来帧，而无需预训练的生成教师。我们将 Beckmann 运输模型的终端映射守恒定律应用于动作条件的视频预测，直接从噪声-数据对进行训练，无需 DriftWorld 所需的生成样本条件分布。然而，对插值输入的运输一致性并不能保证从推理时使用的纯噪声进行准确预测。我们通过端点对齐训练来解决这一差距，该方法直接监督源预测，强调运动和空间结构，并教会从生成的历史中恢复。所有组件训练同一个映射，并保持每个未来块只需一次网络调用。在 PushT 和 Robomimic Can 上，最终模型将 64 帧预测的均方误差分别降低了 68% 和 57%，相比同一骨干网络的回归基线。在策略评估中，BWM 在评估的模型中实现了最低的平均 IoU 误差 0.0100，每四帧 PushT 预测耗时 24.8 毫秒。在 Robomimic Can 上，通过运动监督和生成历史训练的 BWM 阶段，吞吐量比本地一步 DriftWorld 高 3.3 倍，峰值内存降低 78%。在 RT-1 和 Bridge-V2 的潜在空间扩展中，相同更新控制显示，在生成历史下的特征监督可改善感知预测。这些研究结果显示了终端映射学习如何将高效训练与准确的一步视觉预测结合起来。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.26903 (HTTP 429)

Authors: Adam Lee, Shobhit Agarwal, June Pyo Suh

Categories: cs.CV

PDF URL: https://arxiv.org/pdf/2610.26903.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.26903

Published: 2026-10-22T01:51:24.612Z

3. Uncertainty Gates the Gradient: Adaptive Updates for Generalized Few-Shot 3D Segmentation

Abstract:In generalized few-shot 3D point cloud segmentation, support annotations are often compromised by sparsity and occlusion. Blindly adapting to such low quality samples risks learning spurious geometric patterns rather than robust novel concepts. Although 3D Vision-Language Models have been introduced to sup plement supervision, the potential of uncertainty to actively govern adaptation dynamics remains underexplored. Existing signal-level methods like loss reweight ing fail to prevent unreliable episodes from steering the optimization trajectory. To address this, we propose Uncertainty-Gated Adaptive (UGA), to the best of our knowledge the first framework that elevates uncertainty to update-level gating for generalized few-shot 3D point cloud segmentation. The core mod ule, Uncertainty-Gated Update (UG-Update), estimates episodic reliability via intra-class dispersion over merged support and pseudo-labeled novel supervision, and selectively gates residual gradient updates. This encourages the model to learn novel semantics from reliable episodes while attenuating updates during unreliable ones. Complementing this, we introduce an Uncertainty-Based Pseudo Label Filter (UB-Filter) to enforce learning from high-confidence dense signals. On ScanNet200 and ScanNet++ (1-/5-shot), UGA sets new state of the art with an average +3.08 HM IoU over the strongest prior baseline.

中文摘要

摘要：在广义少样本三维点云分割中，支持注释常常受到稀疏性和遮挡的影响而质量降低。盲目适应这种低质量样本可能导致模型学习虚假的几何模式，而非稳健的新概念。虽然已经引入了3D视觉-语言模型以补充监督，但利用不确定性主动调控适应动态的潜力仍未被充分探索。现有的信号级方法，如损失重加权，无法阻止不可靠的样本驱动优化轨迹。为了解决这个问题，我们提出了不确定性门控自适应（UGA），据我们所知，这是首个将不确定性提升至更新级门控的广义少样本三维点云分割框架。核心模块“不确定性门控更新（UG-Update）”通过在合并支持集和伪标签新类别监督上的类内散度来估计每一集的可靠性，并有选择地门控残差梯度更新。这鼓励模型从可靠样本中学习新语义，同时在不可靠样本中减弱更新。作为补充，我们引入了基于不确定性的伪标签筛选（UB-Filter），以确保从高置信度密集信号中学习。在ScanNet200和ScanNet++（1-/5-shot）上，UGA实现了新的最先进成果，平均比最强先前基线提升了+3.08的HM IoU。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.26909 (HTTP 429)

Authors: Yifei Zhao, Fanyu Zhao, Shengtang Wu, Han Zhang, Ruike Cao, Yinsheng Li

Categories: cs.CV

PDF URL: https://arxiv.org/pdf/2610.26909.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.26909

Published: 2026-10-22T01:51:24.612Z

4. Memory Internalization via Test-Time Training for Long-Video Understanding

Abstract:Long-video understanding often requires answering many questions about a single video whose evidence is distributed across extended temporal contexts. Conventional video-language models repeatedly encode frames or retrieve visual evidence for every question. Parameterized video memory offers a write-once alternative, but existing methods rely primarily on synthesized text and fixed adaptation configurations, leaving open where video knowledge should be written and how visual evidence should shape the memory. We introduce Memory Internalization via Test-Time Training (MIT^3), a per-video adaptation framework that writes each long video into compact, video-specific adaptive parameters for video-free, text-only question answering. Under compact multi-scale supervision, causal tracing measures the causal memory utility of each model layer for video-grounded answering and selects video-specific adaptation locations. Visual memory distillation then trains the text-only memory to reproduce the representation change induced by the relevant video clip, preserving the internal effect of visual evidence beyond textual supervision. Across LVBench, LongVideoBench, and Video-MME, MIT^3 achieves data-efficient video-free question answering with substantially fewer synthesized samples than prior parameterized video-memory methods, demonstrating that effective video internalization depends on both where knowledge is written and how visual evidence is distilled into the model.

中文摘要

摘要：长视频理解通常需要回答关于单个视频的许多问题，而这些问题的证据分布在较长的时间上下文中。传统的视频-语言模型会对每个问题重复编码帧或检索视觉证据。参数化视频记忆提供了一种“一次写入”的替代方案，但现有方法主要依赖合成文本和固定的适配配置，仍然没有解决视频知识应写入何处以及视觉证据应如何塑造记忆的问题。我们引入了通过测试时训练实现记忆内化（Memory Internalization via Test-Time Training, MIT^3），这是一种每个视频的适配框架，将每个长视频写入紧凑的、视频特定的自适应参数中，以实现无需视频、仅用文本进行问答。在紧凑的多尺度监督下，因果追踪（causal tracing）测量每个模型层对基于视频的问答的因果记忆效用，并选择视频特定的适配位置。然后，视觉记忆蒸馏训练仅文本记忆以再现相关视频片段引起的表示变化，保留视觉证据在文本监督之外的内部影响。在 LVBench、LongVideoBench 和 Video-MME 上，MIT^3 在数据高效的条件下实现了无需视频的问答，并使用的合成样本远少于先前的参数化视频记忆方法，证明有效的视频内化依赖于知识写入的位置和视觉证据蒸馏进模型的方式。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.26910 (HTTP 429)

Authors: Haonan Zhao, Jiaxuan Chen, Qingyang Liu, Shuochen Chang, Siyu Sun, Jiangtong Li, Li Niu

Categories: cs.CV

PDF URL: https://arxiv.org/pdf/2610.26910.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.26910

Published: 2026-10-22T01:51:24.612Z

5. CodecMotion-OAD: Efficient Online Action Detection with Compressed-Domain Motion

Abstract:Online action detection (OAD) identifies ongoing actions in streaming videos. Many existing OAD systems combine RGB appearance with optical flow, but dense-flow extraction remains a major computational bottleneck. Codec motion vectors (MVs) are available during decoding, avoiding a separate optical-flow computation. On THUMOS’14, optical flow exceeds MVs by 11.2 mAP points with motion-only inputs, whereas the gap narrows to 3.7 points after RGB fusion, with the largest reduction in action-category discrimination. We propose CodecMotion-OAD, which replaces optical flow with decoder-provided MVs while retaining existing temporal detector architectures. Its two-stage training first aligns MV features with a pretrained flow-feature interface without action labels and then trains RGB+MV detectors using ground-truth labels and soft predictions from a frozen RGB+Flow teacher. To the best of our knowledge, this is the first application of codec MVs to frame-level OAD. On THUMOS’14 with MiniROAD, CodecMotion-OAD achieves 68.3 mAP versus 70.9 mAP for matched-distillation RGB+Flow, while reducing decoder-to-prediction processing time by 11.48-fold on CPU and 2.04-fold on GPU.

中文摘要

摘要：在线动作检测（OAD）用于识别流视频中的正在进行的动作。许多现有的OAD系统将RGB外观信息与光流结合，但密集光流提取仍然是主要的计算瓶颈。解码过程中可获取的编解码器运动向量（MVs）避免了单独的光流计算。在THUMOS’14数据集上，使用仅运动输入时，光流比MVs高出11.2 mAP点，而在RGB融合后，这一差距缩小至3.7点，动作类别区分能力的提升最大。我们提出了CodecMotion-OAD，它用解码器提供的MVs替代光流，同时保留现有的时间检测器结构。其两阶段训练首先在没有动作标签的情况下，将MV特征与预训练的光流特征接口对齐，然后使用真实标签和冻结的RGB+Flow教师模型的软预测训练RGB+MV检测器。据我们所知，这是编解码器MVs首次应用于帧级OAD。在THUMOS’14上使用MiniROAD，CodecMotion-OAD实现了68.3 mAP，而匹配蒸馏的RGB+Flow为70.9 mAP，同时将CPU解码到预测的处理时间减少了11.48倍，GPU减少了2.04倍。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.26915 (HTTP 429)

Authors: Zepeng Sun, Naichuan Zheng, Hailun Xia, Junjie Wu

Categories: cs.CV

PDF URL: https://arxiv.org/pdf/2610.26915.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.26915

Published: 2026-10-22T01:51:24.612Z

6. GTSM: Gaussian Splatting with Thermal State-Space Dynamics for Spatio-Temporal Thermal Field Reconstruction and Prediction

Abstract:Temperature-field reconstruction and prediction are important for condition monitoring, process analysis, and state assessment in industrial and dynamic environments. This work aims to recover a four-dimensional spatio-temporal thermal state field directly from two-dimensional observations, without requiring a predefined 3D model, prescribed thermodynamic equation, or predefined heating/cooling trajectory. To this end, we propose the Gaussian Thermal State-Space Model (GTSM), which represents thermal evolution through a shared scene-level latent process and Gaussian-specific local thermal states. The resulting continuous dynamics are propagated using differentiable RK2 integration, allowing thermal evolution to be inferred from multi-view and multi-temporal observations and continuously extrapolated beyond the observed interval. GTSM is designed as a backbone-agnostic thermal dynamics module for Gaussian representations and can be integrated with both static 3D Gaussian Splatting and dynamic 4D Gaussian Splatting, supporting thermal reconstruction and prediction under both static and time-varying target geometry. The reconstructed representation further enables novel-view thermal rendering and future-time thermal-state prediction. Experiments on real infrared sequences and synthetic dynamic thermal data demonstrate improved reconstruction accuracy and substantially stronger temporal extrapolation than representative baselines, while also validating the applicability of GTSM to deforming targets.

中文摘要

摘要：温度场重建与预测在工业和动态环境中的状态监测、过程分析及状态评估中具有重要意义。本研究旨在直接从二维观测恢复四维时空热状态场，而无需预定义的三维模型、指定的热力学方程或预设的加热/冷却轨迹。为此，我们提出了高斯热状态空间模型（GTSM），通过共享的场景级潜在过程和高斯特定的局部热状态来表示热演化。所得连续动力学通过可微的 RK2 积分进行传播，使得可以从多视角、多时间的观测中推断热演化，并在观测区间之外进行连续外推。GTSM 被设计为对骨干网络无依赖的高斯热动力学模块，可与静态三维高斯点云与动态四维高斯点云相结合，支持在静态和时变目标几何下进行热重建与预测。重建的表示还可实现新视角热渲染和未来时间热状态预测。在真实红外序列和合成动态热数据上的实验表明，相较于代表性基线方法，GTSM 在重建精度上有所提升，同时在时间外推能力上显著更强，并验证了 GTSM 在变形目标上的适用性。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.26918 (HTTP 429)

Authors: Yunpeng Zhao, Haoxiang Zhang, Chao Liu

Categories: cs.CV

PDF URL: https://arxiv.org/pdf/2610.26918.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.26918

Published: 2026-10-22T01:51:24.612Z

7. Controllable Multi-Object Insertion for Detection Data Augmentation

Abstract:Controllable object insertion provides a targeted approach to augmenting object detection datasets that require category—count fidelity, annotation alignment, and source-content preservation beyond visual realism. Several approaches have been proposed to address this grand challenge, but most suffer from different constraints in precise category—count requests and visual consistency, limited coordination across categories and instances, and poor localization of multiple objects of specific category and count. We design SceneP2F, a two-stage trainable framework that connects scene-aware joint planning with layout-aligned generation leveraging explicit instance-level layouts. SceneP2F consists of two novel designs. The first is ScenePlan which allocates one class-labeled query per requested instance, jointly predicts scene-compatible object centers, and separately estimates box sizes conditioned on the planned centers and scene context. The second is SceneAlign-Fill which combines geometry- and appearance-aware conditioning, box—class-guided regional attention, and an annotation-region-weighted flow-matching objective to render the planned objects while limiting unintended changes to source content. Extensive experiments on three challenging benchmarks UAVDT, VisDrone, and BDD100K show that SceneP2F outperforms representative general-purpose editors and object-insertion methods consistently, demonstrating the value of coordinating instance-level planning and generation for category-targeted detection augmentation.

中文摘要

摘要：可控对象插入提供了一种针对性的方法，用于增强需要类别-数量精确、标注对齐和源内容保留（超越视觉真实性）的目标检测数据集。已经提出了多种方法来应对这一重大挑战，但大多数方法在精确类别-数量要求和视觉一致性方面存在不同的限制，跨类别和实例的协调能力有限，以及对特定类别和数量的多个对象定位性能较差。我们设计了SceneP2F，这是一种两阶段可训练框架，将场景感知联合规划与利用显式实例级布局的布局对齐生成相连接。SceneP2F包括两个创新设计。第一个是ScenePlan，它为每个请求的实例分配一个类别标记查询，联合预测与场景兼容的对象中心，并在已规划的中心和场景上下文条件下分别估计边界框大小。第二个是SceneAlign-Fill，它结合了几何与外观感知的条件、基于边界框-类别的区域注意机制以及带注释区域加权的流匹配目标，以在限制对源内容的意外更改的同时渲染计划对象。在三个具有挑战性的基准数据集UAVDT、VisDrone和BDD100K上的大量实验表明，SceneP2F始终优于代表性的通用编辑器和对象插入方法，证明了协调实例级规划与生成对于类别目标检测增强的价值。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.26922 (HTTP 429)

Authors: Jiyang Huang, Tianwei Zhang, Xiaoqin Zhang, Ling Shao, Shijian Lu

Categories: cs.CV

PDF URL: https://arxiv.org/pdf/2610.26922.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.26922

Published: 2026-10-22T01:51:24.612Z

8. CaMeWM: A Causal Medical World Model for Generative Tumor Evolution and Treatment Planning

Abstract:Treatment planning is intrinsically counterfactual: clinicians must compare how a patient’s tumor and survival would evolve under competing regimens, whereas a retrospective cohort records only the regimen received and one follow-up scan. A treatment-conditioned generator trained on such data can therefore reproduce historical treatment allocation rather than simulate the interventional disease dynamics needed to rank regimens. We introduce \textbf{CaMeWM}, a causal medical world model for supported post-TACE tumor evolution and treatment planning. CaMeWM implements one intervention-to-decision loop rather than three independent prediction tasks: it first admits a queried regimen only for clinically comparable histories with empirical treatment support; it then generates a distribution of lesion-local follow-up worlds through an action-conditioned response descriptor, deterministic anchor, and residual 3D diffusion transition; and it finally maps sampled worlds to restricted mean survival time (RMST) with a response-Cox utility model. Under pre-treatment adjustment and positivity, this loop estimates a prespecified, support-restricted g-formula target; all selection and nuisance estimation are fold-local, and the induced policy is evaluated by cross-fitted, censoring-adjusted doubly robust estimation. Across 79 paired out-of-fold episodes, CaMeWM achieves 0.618 SSIM and 0.528 change-map SSIM for follow-up CT transition, while its generated-world survival pathway attains a 0.615 C-index; factual-only AIPW estimates a 9.60-week RMST gain (95\% CI 0.61—18.13) for the frozen policy. CaMeWM provides an inspectable basis for retrospective, support-restricted treatment planning.

中文摘要

摘要：治疗计划本质上是反事实的：临床医生必须比较患者的肿瘤和生存情况在不同疗法下的发展，而回顾性队列仅记录所接受的疗法和一次随访扫描。因此，基于这些数据训练的治疗条件生成器只能重现历史治疗分配，而无法模拟对疗法排序所需的干预性疾病动态。我们提出了\textbf{CaMeWM}，一种用于支持TACE后肿瘤演变和治疗计划的因果医学世界模型。CaMeWM实现了一个干预到决策的循环，而不是三个独立的预测任务：它首先仅对具有经验治疗支持、在临床上可比的病史中接受查询的疗法；然后通过一个动作条件响应描述符、确定性锚点和残差3D扩散转移生成病灶局部随访世界的分布；最后将采样的世界映射到受限平均生存时间（RMST）通过响应-Cox效用模型。在治疗前调整和正性假设下，该循环可估计预设的、受支持限制的g-公式目标；所有选择和干扰估计都是折叠局部的，所诱导的策略通过交叉拟合、删失调整的双稳健估计进行评估。在79对折包外案例中，CaMeWM在随访CT转变中达到0.618 SSIM和0.528变化图SSIM，而其生成的世界生存路径达到0.615 C指数；仅事实AIPW估计冻结策略的RMST增益为9.60周（95% CI 0.61—18.13）。CaMeWM为回顾性、受支持限制的治疗计划提供了可检查的依据。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.26923 (HTTP 429)

Authors: Huanlong Gao, Xuelei He, Fengjun Zhao, Xiaowei He

Categories: cs.CV

PDF URL: https://arxiv.org/pdf/2610.26923.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.26923

Published: 2026-10-22T01:51:24.612Z

9. SuperClip Pyramid for Video Temporal Grounding

Abstract:With the rapid expansion of online video content, video temporal grounding has become essential for localizing moments based on user-specific queries, comprising two subtasks: moment retrieval (MR) and highlight detection (HD). While recent approaches adopt DETR-style decoders or temporal pyramid architectures, they rely on single-scale feature maps or uniformly downsampled representations, limiting their ability to capture semantic boundaries and variable-length events. In this paper, we present SuperClip Video Temporal Grounding (SCVTG) built on a multi-scale temporal hierarchy that better captures distinctive events at various granularities. SCVTG constructs a SuperClip pyramid by grouping adjacent clips into semantically coherent segments, termed SuperClips. Further, to aggregate these SuperClips having variable duration and temporal location, we introduce a temporally adaptive kernel that aggregates SuperClips through temporally conditioned weights, enabling flexible aggregation. On top of this, we design relative saliency losses to guide the model in learning inter-clip importance, enhancing its ability to distinguish relevant events. Extensive experiments show that SCVTG achieves the best performance across five video temporal grounding benchmarks with promising gains.

中文摘要

摘要：随着在线视频内容的快速增长，视频时间定位已成为基于用户特定查询定位时刻的关键任务，主要包括两个子任务：时刻检索（MR）和精彩片段检测（HD）。尽管近期的方法采用DETR风格的解码器或时间金字塔架构，但它们依赖于单尺度特征图或均匀下采样的表示，从而限制了捕捉语义边界和可变长度事件的能力。本文提出了基于多尺度时间层次结构的超剪辑视频时间定位（SCVTG），能够更好地捕捉不同粒度的显著事件。SCVTG通过将相邻剪辑聚合成语义一致的片段，构建了一个超剪辑金字塔，这些片段称为超剪辑。进一步地，为了聚合具有可变时长和时间位置的超剪辑，我们引入了一种时间自适应核，通过时间条件权重聚合超剪辑，实现灵活的聚合。在此基础上，我们设计了相对显著性损失，引导模型学习片段间的重要性，提高其区分相关事件的能力。大量实验表明，SCVTG在五个视频时间定位基准上均取得了最佳性能，并展现出令人期待的提升效果。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.26925 (HTTP 429)

Authors: Sanghyeok Lee, Juyeon Ko, Joonmyung Choi, Jongha Kim, Dohwan Ko, Ji Soo Lee, Hyunwoo J. Kim

Categories: cs.CV

PDF URL: https://arxiv.org/pdf/2610.26925.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.26925

Published: 2026-10-22T01:51:24.612Z

10. Rate as a Feature, Not a Nuisance: Curriculum FiLM Training for Event-Based Graph Neural Networks

Abstract:An event camera emits events at a rate given by the scene motion, object texture, illumination, its own contrast thresholds, and other settings. Across the N-Caltech101 dataset, recorded by a single event camera, the event rate varies by roughly 26x. Standard event-based Graph Neural Network GNN pipelines do not make use of the variable event rate. They either normalize each event’s timestamp into a fixed window or train the network to be invariant to the event rate. We argue that both choices lose information that could instead be used to improve the performance of a pipeline. Per-recording normalization is non-invertible, incurring about 6pp (percentage points) of natural-rate accuracy loss in our setting. And invariance training trades accuracy for robustness to shifts in the event rate. We instead take a different path and condition on the event rate directly. We add a per-recording log-rate scalar that modulates every stage of a PointNet++ backbone through Feature-wise Linear Modulation (FiLM), initialized to identity, so that training begins at the unmodified baseline. A two-stage curriculum (natural event rates first, then rate-augmented fine-tuning) reaches 67.2% on N-Caltech101 (+6.3pp) and 82.5% on DVS128 Gesture (+14.0pp) with an additional ~0.2M parameters. A matched-capacity control experiment shows that the gain is not from the increase in model capacity but from the rate conditioning. We will make the implementation publicly available upon acceptance of this work.

中文摘要

摘要：事件相机以由场景运动、物体纹理、光照、自身的对比阈值以及其他设置决定的速率发射事件。在由单个事件相机记录的 N-Caltech101 数据集中，事件速率大约变化了 26 倍。标准的基于事件的图神经网络（GNN）流程并未利用可变的事件速率。它们要么将每个事件的时间戳归一化到固定窗口，要么训练网络使其对事件速率保持不变。我们认为这两种选择都会丢失本可以用于提升流程性能的信息。每次记录的归一化是不可逆的，在我们的设置下会导致约 6 个百分点的自然速率准确率损失。而不变性训练则是以牺牲准确率为代价来提升对事件速率变化的鲁棒性。我们采取了不同的方法，直接以事件速率为条件。我们添加了一个每次记录的对数速率标量，通过特征线性调制（FiLM）调节 PointNet++ 主干网络的每个阶段，该标量初始化为恒等，使训练从未修改的基线开始。通过两阶段课程（先使用自然事件速率，再进行速率增强微调）在 N-Caltech101 上达到 67.2% (+6.3pp)，在 DVS128 手势数据集上达到 82.5% (+14.0pp)，额外增加约 0.2M 参数。匹配容量的对照实验表明，性能提升并非来自模型容量的增加，而是来自速率条件控制。我们将在本文被接收后公开实现代码。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.26933 (HTTP 429)

Authors: Kavyansh Tyagi, Andreas Ziegler

Categories: cs.CV

PDF URL: https://arxiv.org/pdf/2610.26933.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.26933

Published: 2026-10-22T01:51:24.612Z