评分理由: 论文核心是提出MoE-FM框架用于语言模型推理加速，与’Mixture of Experts’高度相关（核心方法，15分），与’Large Language Models’高度相关（研究对象，10分），与’Speculative Decoding’相关（推理加速，10分）。其他关键词如SLMs、Scaling Laws、Alignment等未涉及，得0分。

流匹配在保持扩散模型生成质量的同时实现了显著更快的推理速度，使其成为生成建模领域极具吸引力的范式。然而，当应用于语言建模时，它在表示具有不规则几何结构（如各向异性和多模态性）的复杂隐空间分布方面存在根本性局限。为应对这些挑战，我们提出一种专家混合流匹配框架，通过将复杂的全局传输几何结构分解为局部特化的向量场，以捕捉隐空间中的复杂几何特性。基于此框架，我们开发了一种非自回归语言建模方法，命名为YAN，并分别基于Transformer和Mamba架构进行了实例化。在多项下游任务中，YAN的生成质量与自回归模型及基于扩散的非自回归语言模型相当，而仅需三个采样步骤即可完成生成。这相比自回归基线实现了40倍的加速，相比扩散语言模型更可获得高达10^3倍的加速，充分展示了其在语言建模效率上的显著优势。

Flow matching retains the generation quality of diffusion models while enabling substantially faster inference, making it a compelling paradigm for generative modeling. However, when applied to language modeling, it exhibits fundamental limitations in representing complex latent distributions with irregular geometries, such as anisotropy and multimodality. To address these challenges, we propose a mixture-of-experts flow matching (MoE-FM) framework, which captures complex global transport geometries in latent space by decomposing them into locally specialized vector fields. Building on MoE-FM, we develop a non-autoregressive (NAR) language modeling approach, named YAN, instantiated with both Transformer and Mamba architectures. Across multiple downstream tasks, YAN achieves generation quality on par with both autoregressive (AR) and diffusion-based NAR language models, while requiring as few as three sampling steps. This yields a $40\times$ speedup over AR baselines and up to a $10^3\times$ speedup over diffusion language models, demonstrating substantial efficiency advantages for language modeling.

评分理由: 论文核心研究LLM agents从reactive到proactive的转变，并提出了ProVoice-Bench评估框架。因此，与"LLM Agents"和"Large Language Models"高度相关（10分），因为论文明确提到LLM agents和Multimodal LLMs。与推理相关的关键词（“Chain of Thought"和"System 2 Thinking”）得5分，因为论文提到了模型的reasoning capabilities作为评估的一部分，但这不是核心焦点。其他关键词（如MoE、SFT、RAG等）在摘要中未提及，因此得0分。

近期大语言模型智能体的发展正逐渐从被动的、基于文本的范式转向主动的多模态交互。然而，现有基准测试主要关注被动响应，忽视了主动干预与监控的复杂性。为弥补这一差距，我们推出了ProVoice-Bench——首个专门为主动语音智能体设计的评估框架，包含四项创新任务。通过采用多阶段数据合成流程，我们构建了1,182个高质量样本进行严格测试。对前沿多模态大语言模型的评估揭示了显著的性能差距，尤其在过度触发和推理能力方面。这些发现凸显了当前模型的局限性，并为开发更自然、具有情境感知能力的主动智能体提供了发展路线图。

Recent advancements in LLM agents are gradually shifting from reactive, text-based paradigms toward proactive, multimodal interaction. However, existing benchmarks primarily focus on reactive responses, overlooking the complexities of proactive intervention and monitoring. To bridge this gap, we introduce ProVoice-Bench, the first evaluation framework specifically designed for proactive voice agents, featuring four novel tasks. By leveraging a multi-stage data synthesis pipeline, we curate 1,182 high-quality samples for rigorous testing. Our evaluation of state-of-the-art Multimodal LLMs reveals a significant performance gap, particularly regarding over-triggering and reasoning capabilities. These findings highlight the limitations of current models and offer a roadmap for developing more natural, context-aware proactive agents.

评分理由: 论文核心是提出MoE-FM框架用于语言模型推理加速，与’Mixture of Experts’高度相关（核心方法，15分），与’Large Language Models’高度相关（研究对象，10分），与’Speculative Decoding’相关（推理加速，10分）。其他关键词如SLMs、Scaling Laws、Alignment等未涉及，得0分。

流匹配在保持扩散模型生成质量的同时实现了显著更快的推理速度，使其成为生成建模领域极具吸引力的范式。然而，当应用于语言建模时，它在表示具有不规则几何结构（如各向异性和多模态性）的复杂隐空间分布方面存在根本性局限。为应对这些挑战，我们提出一种专家混合流匹配框架，通过将复杂的全局传输几何结构分解为局部特化的向量场，以捕捉隐空间中的复杂几何特性。基于此框架，我们开发了一种非自回归语言建模方法，命名为YAN，并分别基于Transformer和Mamba架构进行了实例化。在多项下游任务中，YAN的生成质量与自回归模型及基于扩散的非自回归语言模型相当，而仅需三个采样步骤即可完成生成。这相比自回归基线实现了40倍的加速，相比扩散语言模型更可获得高达10^3倍的加速，充分展示了其在语言建模效率上的显著优势。

Flow matching retains the generation quality of diffusion models while enabling substantially faster inference, making it a compelling paradigm for generative modeling. However, when applied to language modeling, it exhibits fundamental limitations in representing complex latent distributions with irregular geometries, such as anisotropy and multimodality. To address these challenges, we propose a mixture-of-experts flow matching (MoE-FM) framework, which captures complex global transport geometries in latent space by decomposing them into locally specialized vector fields. Building on MoE-FM, we develop a non-autoregressive (NAR) language modeling approach, named YAN, instantiated with both Transformer and Mamba architectures. Across multiple downstream tasks, YAN achieves generation quality on par with both autoregressive (AR) and diffusion-based NAR language models, while requiring as few as three sampling steps. This yields a $40\times$ speedup over AR baselines and up to a $10^3\times$ speedup over diffusion language models, demonstrating substantial efficiency advantages for language modeling.

评分理由: 论文核心研究LLM agents从reactive到proactive的转变，并提出了ProVoice-Bench评估框架。因此，与"LLM Agents"和"Large Language Models"高度相关（10分），因为论文明确提到LLM agents和Multimodal LLMs。与推理相关的关键词（“Chain of Thought"和"System 2 Thinking”）得5分，因为论文提到了模型的reasoning capabilities作为评估的一部分，但这不是核心焦点。其他关键词（如MoE、SFT、RAG等）在摘要中未提及，因此得0分。

近期大语言模型智能体的发展正逐渐从被动的、基于文本的范式转向主动的多模态交互。然而，现有基准测试主要关注被动响应，忽视了主动干预与监控的复杂性。为弥补这一差距，我们推出了ProVoice-Bench——首个专门为主动语音智能体设计的评估框架，包含四项创新任务。通过采用多阶段数据合成流程，我们构建了1,182个高质量样本进行严格测试。对前沿多模态大语言模型的评估揭示了显著的性能差距，尤其在过度触发和推理能力方面。这些发现凸显了当前模型的局限性，并为开发更自然、具有情境感知能力的主动智能体提供了发展路线图。

Recent advancements in LLM agents are gradually shifting from reactive, text-based paradigms toward proactive, multimodal interaction. However, existing benchmarks primarily focus on reactive responses, overlooking the complexities of proactive intervention and monitoring. To bridge this gap, we introduce ProVoice-Bench, the first evaluation framework specifically designed for proactive voice agents, featuring four novel tasks. By leveraging a multi-stage data synthesis pipeline, we curate 1,182 high-quality samples for rigorous testing. Our evaluation of state-of-the-art Multimodal LLMs reveals a significant performance gap, particularly regarding over-triggering and reasoning capabilities. These findings highlight the limitations of current models and offer a roadmap for developing more natural, context-aware proactive agents.

评分理由: 论文核心研究大语言模型（LLMs）在结构化生成中的行为，特别是约束解码下模式关键词的表述如何作为隐式指令通道影响模型性能。因此，与"Large Language Models"高度相关（10分）。论文涉及指令（instruction）和模型行为分析，与"Instruction Tuning"有一定关联（5分）。实验在数学推理基准上进行，涉及推理过程，与"Chain of Thought"有一定关联（5分）。研究揭示了模式设计如何传递指令信号，有助于理解模型行为，与"Mechanistic Interpretability"有一定关联（5分）。其他关键词如MoE、SLMs、训练方法、推理优化、代理、量化等均未在论文中涉及，评为0分。

约束解码已广泛应用于大型语言模型的结构化生成任务，以确保输出符合预定义格式（如JSON和XML）。然而，现有方法大多仅将模式视为纯粹的结构约束，忽视了其语言表述可能影响模型行为的可能性。本研究探讨了指令放置方式如何影响模型在结构化生成中的表现，并证明仅改变模式关键词的措辞（无需修改提示或模型参数）即可显著改变约束解码下的模型性能。基于此发现，我们提出将结构化生成重新阐释为多通道指令问题：指令既可通过提示显式传递，也可在解码过程中通过模式关键词隐式传递。据我们所知，这是首个系统研究模式关键词表述如何作为隐式指令通道并影响约束解码下模型性能的工作。在多个数学推理基准上的实验表明，不同模型家族对这些指令通道表现出明显差异的敏感性：Qwen模型始终受益于模式层级的指令，而LLaMA模型更依赖提示层级的引导。我们进一步观察到指令通道之间存在非叠加的交互效应，表明组合多个通道并不总能带来额外提升。这些发现说明模式设计不仅决定输出结构，还承载着指令信号，为大型语言模型的结构化生成提供了新的研究视角。

Constrained decoding has been widely adopted for structured generation with large language models (LLMs), ensuring that outputs satisfy predefined formats such as JSON and XML. However, existing approaches largely treat schemas as purely structural constraints and overlook the possibility that their linguistic formulation may affect model behavior. In this work, we study how instruction placement influences model performance in structured generation and show that merely changing the wording of schema keys, without modifying the prompt or model parameters, can significantly alter model performance under constrained decoding. Based on this observation, we propose to reinterpret structured generation as a multi-channel instruction problem, where instructions can be conveyed explicitly through prompts and implicitly through schema keys during decoding. To the best of our knowledge, this is the first work to systematically study how schema key formulation acts as an implicit instruction channel and affects model performance under constrained decoding. Experiments on multiple mathematical reasoning benchmarks show that different model families exhibit distinct sensitivities to these instruction channels: Qwen models consistently benefit from schema-level instructions, while LLaMA models rely more heavily on prompt-level guidance. We further observe non-additive interaction effects between instruction channels, showing that combining multiple channels does not always lead to further improvement. These findings suggest that schema design not only determines output structure, but also carries instruction signals, offering a new perspective on structured generation in LLMs.

评分理由: 论文《MetaDent》主要研究牙科临床图像数据集构建和视觉语言模型（VLM）评估，属于大模型在科学（医学）领域的应用。摘要明确提到使用大型语言模型（LLMs）生成标准化基准（VQA对和分类数据集），因此与关键词"Large Language Models"高度相关（10分）。同时，该研究属于AI在生物医学（牙科）领域的应用，与关键词"AI for Science"高度相关（10分）。其他关键词（如MoE、SFT、RAG等）涉及大模型技术原理或特定方法，论文未提及这些具体技术，因此评分为0分。

视觉语言模型在医学影像分析中展现出巨大潜力，但由于缺乏细粒度标注数据集与综合性基准测试，其在口腔内摄影领域的应用仍未被充分探索。为此，我们提出MetaDent——一个综合性资源，包含：（1）从临床、公共及网络来源收集的新型大规模牙科影像数据集；（2）为捕捉牙科摄影的层次化结构与临床细微差异而设计的半结构化标注框架；（3）用于评估前沿视觉语言模型在临床影像理解任务上的综合性基准测试套件。我们的标注方法结合了高层级影像摘要与针对异常点的逐项自由文本描述，从而实现了丰富、可扩展且与任务无关的表示形式。我们从多来源收集了60,669张牙科影像，并采用此元标注方案对其中具有代表性的2,588张影像子集进行了标注。借助大语言模型，我们构建了标准化基准数据集：约1.5万个视觉问答对及一个包含18个类别的多标签分类数据集。通过人工审核与误差分析验证，我们证明了大语言模型驱动的转换能可靠保持标注的保真度与语义准确性。随后，我们在视觉问答、分类与图像描述生成任务上评估了前沿视觉语言模型。定量结果表明，即使最先进的模型也难以实现对口腔内场景的细粒度理解，仅取得中等准确率，且在图像描述生成中产生不一致或不完整的描述。我们公开释放数据集、标注方案及相关工具，以促进可重复研究，并加速面向牙科应用的视觉语言系统开发。

Vision-Language Models (VLMs) have demonstrated significant potential in medical image analysis, yet their application in intraoral photography remains largely underexplored due to the lack of fine-grained, annotated datasets and comprehensive benchmarks. To address this, we present MetaDent, a comprehensive resource that includes (1) a novel and large-scale dentistry image dataset collected from clinical, public, and web sources; (2) a semi-structured annotation framework designed to capture the hierarchical and clinically nuanced nature of dental photography; and (3) comprehensive benchmark suites for evaluating state-of-the-art VLMs on clinical image understanding. Our labeling approach combines a high-level image summary with point-by-point, free-text descriptions of abnormalities. This method enables rich, scalable, and task-agnostic representations. We curated 60,669 dental images from diverse sources and annotated a representative subset of 2,588 images using this meta-labeling scheme. Leveraging Large Language Models (LLMs), we derive standardized benchmarks: approximately 15K Visual Question Answering (VQA) pairs and an 18-class multi-label classification dataset, which we validated with human review and error analysis to justify that the LLM-driven transition reliably preserves fidelity and semantic accuracy. We then evaluate state-of-the-art VLMs across VQA, classification, and image captioning tasks. Quantitative results reveal that even the most advanced models struggle with a fine-grained understanding of intraoral scenes, achieving moderate accuracy and producing inconsistent or incomplete descriptions in image captioning. We publicly release our dataset, annotations, and tools to foster reproducible research and accelerate the development of vision-language systems for dental applications.

评分理由: 论文核心研究LLMs在认知工作流中的使用如何影响用户对自身能力的认知（LLM fallacy），属于LLMs在人类-AI协作、认知科学领域的应用研究。因此，与"Large Language Models"高度相关（10分）。论文提到"hallucination"作为相关研究背景，因此与"Hallucination Mitigation"有一定关联（5分）。其他关键词主要涉及具体技术原理、训练方法、优化技术或特定应用领域，论文未直接涉及，故评0分。

大型语言模型（LLM）在日常工作流程中的快速融合，已深刻改变了个体在执行写作、编程、分析和多语言交流等认知任务的方式。尽管先前研究主要关注模型可靠性、幻觉问题以及用户信任校准，但较少探讨LLM的使用如何重塑用户对自身能力的认知。本文提出了“LLM谬误”这一概念，指的是一种认知归因偏差：个体错误地将LLM辅助生成的输出视为自身独立能力的证据，从而导致感知能力与实际能力之间出现系统性偏离。我们认为，LLM的不透明性、流畅性以及低摩擦的交互模式模糊了人与机器贡献之间的界限，致使用户倾向于从输出结果而非生成过程来推断自身能力。我们将LLM谬误置于关于自动化偏见、认知卸载以及人机协作的现有文献背景中，同时将其界定为一种特定于AI中介工作流程的归因扭曲形式。我们提出了一个关于其内在机制的概念框架，并对其在计算、语言、分析和创意等领域中的表现类型进行了分类。最后，我们探讨了其对教育、招聘和AI素养的影响，并概述了实证验证的方向。本文亦对人类与AI协作的研究方法进行了透明化说明。此项工作为理解生成式AI系统如何不仅增强认知表现，而且重塑自我认知与感知的专业能力奠定了基础。

The rapid integration of large language models (LLMs) into everyday workflows has transformed how individuals perform cognitive tasks such as writing, programming, analysis, and multilingual communication. While prior research has focused on model reliability, hallucination, and user trust calibration, less attention has been given to how LLM usage reshapes users’ perceptions of their own capabilities. This paper introduces the LLM fallacy, a cognitive attribution error in which individuals misinterpret LLM-assisted outputs as evidence of their own independent competence, producing a systematic divergence between perceived and actual capability. We argue that the opacity, fluency, and low-friction interaction patterns of LLMs obscure the boundary between human and machine contribution, leading users to infer competence from outputs rather than from the processes that generate them. We situate the LLM fallacy within existing literature on automation bias, cognitive offloading, and human–AI collaboration, while distinguishing it as a form of attributional distortion specific to AI-mediated workflows. We propose a conceptual framework of its underlying mechanisms and a typology of manifestations across computational, linguistic, analytical, and creative domains. Finally, we examine implications for education, hiring, and AI literacy, and outline directions for empirical validation. We also provide a transparent account of human–AI collaborative methodology. This work establishes a foundation for understanding how generative AI systems not only augment cognitive performance but also reshape self-perception and perceived expertise.

评分理由: 论文核心研究语言模型（LLMs）在最短路径规划问题中的系统性泛化能力，直接涉及"Large Language Models"关键词，因此给予10分。论文未涉及其他具体技术（如MoE、SFT、RAG等）、应用领域（如生物信息学）或特定方法（如量化、推理加速），这些关键词评分为0分。

语言模型能否实现系统性泛化仍是一个备受争议的议题。然而，实证性能往往同时受到训练数据、训练范式与推理策略等多重因素的共同影响，导致模型失败的原因难以解析。为此，我们构建了一个基于最短路径规划的受控合成实验环境——这是一个典型的可组合序列优化问题。该设置能够清晰分离上述影响因素，并支持两个正交的泛化维度：向未见地图的空间迁移能力，以及向更长规划问题的长度扩展能力。研究发现，模型在空间迁移方面表现优异，但在长度扩展时却因递归不稳定性而持续失败。我们进一步分析了学习流程中不同阶段对系统性问题解决的影响：例如，数据覆盖范围决定了能力上限；强化学习提升了训练稳定性但并未突破该上限；推理阶段的规模扩展虽能提升性能，却无法挽救长度扩展上的失败。

Whether language models can systematically generalize remains actively debated. Yet empirical performance is jointly shaped by multiple factors such as training data, training paradigms, and inference-time strategies, making failures difficult to interpret. We introduce a controlled synthetic environment based on shortest-path planning, a canonical composable sequential optimization problem. The setup enables clean separation of these factors and supports two orthogonal axes of generalization: spatial transfer to unseen maps and length scaling to longer-horizon problems. We find that models exhibit strong spatial transfer but consistently fail under length scaling due to recursive instability. We further analyze how distinct stages of the learning pipeline influence systematic problem-solving: for example, data coverage sets capability limits; reinforcement learning improves training stability but does not expand those limits; and inference-time scaling enhances performance but cannot rescue length-scaling failures.

评分理由: 论文提出MM-WebAgent，一个用于多模态网页生成的分层智能体框架，核心是协调基于AIGC的元素生成，通过分层规划和迭代自反思。因此，与智能体相关的关键词（如LLM Agents、Self-orrection、Tool Use）高度相关（8-10分），因为论文明确涉及自主代理、自反思和工具使用（AIGC工具）。Chain of Thought和System 2 Thinking相关（8分），因为分层规划和迭代推理涉及多步和深度推理。Multi-agent Systems有一定关联（5分），因为框架可能涉及代理协调，但未明确强调多代理系统。Large Language Models相关（5分），因为AIGC工具可能基于LLMs，但论文未深入讨论LLM技术本身。其他关键词（如MoE、Scaling Laws、RLHF等）完全无关（0分），因为论文专注于应用框架而非底层模型技术或特定训练方法。

人工智能生成内容（AIGC）工具的快速发展，使得图像、视频和可视化内容能够按需生成以用于网页设计，为现代用户界面/用户体验（UI/UX）提供了一种灵活且日益普及的范式。然而，直接将此类工具集成到自动化网页生成中，常因元素被孤立生成而导致风格不一致与全局协调性差的问题。我们提出了MM-WebAgent，一种用于多模态网页生成的分层智能体框架，它通过分层规划与迭代自反思来协调基于AIGC的元素生成。MM-WebAgent联合优化全局布局、局部多模态内容及其整合，从而生成协调一致且视觉统一的网页。我们进一步引入了多模态网页生成的基准测试集，以及用于系统评估的多层次评价方案。实验表明，MM-WebAgent在代码生成和基于智能体的基线方法中表现更优，尤其是在多模态元素生成与整合方面。代码与数据：https://aka.ms/mm-webagent。

The rapid progress of Artificial Intelligence Generated Content (AIGC) tools enables images, videos, and visualizations to be created on demand for webpage design, offering a flexible and increasingly adopted paradigm for modern UI/UX. However, directly integrating such tools into automated webpage generation often leads to style inconsistency and poor global coherence, as elements are generated in isolation. We propose MM-WebAgent, a hierarchical agentic framework for multimodal webpage generation that coordinates AIGC-based element generation through hierarchical planning and iterative self-reflection. MM-WebAgent jointly optimizes global layout, local multimodal content, and their integration, producing coherent and visually consistent webpages. We further introduce a benchmark for multimodal webpage generation and a multi-level evaluation protocol for systematic assessment. Experiments demonstrate that MM-WebAgent outperforms code-generation and agent-based baselines, especially on multimodal element generation and integration. Code & Data: https://aka.ms/mm-webagent.

评分理由: 该论文专注于评估LLM作为评判者（LLM-as-judge）的可靠性，属于大模型应用评估领域。核心关键词’Large Language Models OR LLMs OR Foundation Models’高度相关（10分），因为论文直接研究LLM在自动文本生成评估中的应用。其他关键词如’Mixture of Experts’、‘Scaling Laws’、‘Pre-training’、‘RLHF’、‘RAG’、‘Chain of Thought’、‘LLM Agents’、‘Quantization’、‘Hallucination Mitigation’、‘AI for Science’等均未在论文标题或摘要中提及或涉及，因此评分为0分。论文未涉及任何专家作者，故expert_authors_found为空列表。

LLM即评委框架正日益广泛地应用于自然语言生成的自动评估中，然而其针对单个实例的可靠性仍鲜为人知。我们提出了一套双管齐下的诊断工具包，并将其应用于SummEval数据集：$\textbf{(1)}$ 一项传递性分析揭示了在较低的总体违反率（$\barρ = 0.8$-$4.1%$）掩盖下普遍存在的、针对单个输入的不一致性，有$33$-$67%$的文档表现出至少一个有向3环；$\textbf{(2)}$ 基于1-5李克特量表的划分保形预测集提供了理论保证的$\geq(1{-}α)$覆盖度，其集合宽度可作为每个实例的可靠性指标（$r_s = {+}0.576$，$N{=}1{,}918$，$p < 10^{-100}$，所有评委数据汇总）。关键的是，预测集宽度在不同评委间表现出稳定的一致性（$\bar{r} = 0.32$-$0.38$），证明其捕捉到的是文档层面的难度，而非评委特定的噪声。在四位评委和四项评估标准中，两项诊断结果趋于一致：评估标准比评委身份更为关键，其中相关性（relevance）的评估最为可靠（平均集合大小$\approx 3.0$），连贯性（coherence）中等可靠（平均集合大小$\approx 3.9$），而流畅性（fluency）与一致性（consistency）则仍不可靠（平均集合大小$\approx 4.9$）。我们已公开所有代码、提示词及缓存结果。

LLM-as-judge frameworks are increasingly used for automatic NLG evaluation, yet their per-instance reliability remains poorly understood. We present a two-pronged diagnostic toolkit applied to SummEval: $\textbf{(1)}$ a transitivity analysis that reveals widespread per-input inconsistency masked by low aggregate violation rates ($\barρ = 0.8$-$4.1%$), with $33$-$67%$ of documents exhibiting at least one directed 3-cycle; and $\textbf{(2)}$ split conformal prediction sets over 1-5 Likert scores providing theoretically-guaranteed $\geq(1{-}α)$ coverage, with set width serving as a per-instance reliability indicator ($r_s = {+}0.576$, $N{=}1{,}918$, $p < 10^{-100}$, pooled across all judges). Critically, prediction set width shows consistent cross-judge agreement ($\bar{r} = 0.32$-$0.38$), demonstrating it captures document-level difficulty rather than judge-specific noise. Across four judges and four criteria, both diagnostics converge: criterion matters more than judge, with relevance judged most reliably (avg. set size $\approx 3.0$) and coherence moderately so (avg. set size $\approx 3.9$), while fluency and consistency remain unreliable (avg. set size $\approx 4.9$). We release all code, prompts, and cached results.

评分理由: 该论文专注于计算机视觉领域的视觉异常检测（VAD）在自动驾驶中的应用，研究内容涉及异常检测模型、数据集基准测试、模型效率优化等。所有给定的关键词均与大语言模型（LLM）、深度学习技术原理、AI for Science等主题相关，而本文完全不涉及这些内容。论文未提及任何LLM、MoE、SLM、Scaling Laws、预训练/后训练技术、对齐、RLHF、PEFT、RAG、推理优化、思维链、智能体、量化、幻觉缓解、可解释性、世界模型、模型合并、上下文学习或科学AI应用。因此，所有关键词的相关度均为0分。

自动驾驶机器视觉系统的可靠性高度依赖于其训练数据分布。当车辆遇到显著不同的条件（例如非典型障碍物）时，其感知能力可能大幅下降。与许多错误后果有限的领域不同，自动驾驶的失败直接转化为对乘客、行人及其他道路使用者的物理风险。为应对这一挑战，我们探索将视觉异常检测（Visual Anomaly Detection, VAD）作为一种解决方案。VAD能够识别训练过程中未出现的异常物体，使系统在检测到陌生情况时向驾驶员发出警报。至关重要的是，VAD模型可生成像素级异常热图，无需对危险的性质或形式进行任何先验假设，即可引导驾驶员关注特定的关注区域。我们在AnoVox（目前最大的自动驾驶异常检测合成数据集）上对八种前沿VAD方法进行了基准测试。特别地，我们评估了从大型网络到轻量级架构（如MobileNet和DeiT-Tiny）四种主干网络的性能表现。结果表明，VAD能有效迁移至道路场景。值得注意的是，Tiny-Dinomaly在边缘部署中实现了最佳的精度-效率平衡，仅以少量内存成本即达到了全尺寸定位模型的性能。这项研究为更安全、更负责任的自动驾驶车辆部署迈出了切实一步，最终提升对乘客、行人及所有道路使用者的保护。

The reliability of a machine vision system for autonomous driving depends heavily on its training data distribution. When a vehicle encounters significantly different conditions, such as atypical obstacles, its perceptual capabilities can degrade substantially. Unlike many domains where errors carry limited consequences, failures in autonomous driving translate directly into physical risk for passengers, pedestrians, and other road users. To address this challenge, we explore Visual Anomaly Detection (VAD) as a solution. VAD enables the identification of anomalous objects not present during training, allowing the system to alert the driver when an unfamiliar situation is detected. Crucially, VAD models produce pixel-level anomaly maps that can guide driver attention to specific regions of concern without requiring any prior assumptions about the nature or form of the hazard. We benchmark eight state-of-the-art VAD methods on AnoVox, the largest synthetic dataset for anomaly detection in autonomous driving. In particular, we evaluate performance across four backbone architectures spanning from large networks to lightweight ones such as MobileNet and DeiT-Tiny. Our results demonstrate that VAD transfers effectively to road scenes. Notably, Tiny-Dinomaly achieves the best accuracy-efficiency trade-off for edge deployment, matching full-scale localization performance at a fraction of the memory cost. This study represents a concrete step toward safer, more responsible deployment of autonomous vehicles, ultimately improving protection for passengers, pedestrians, and all road users.

评分理由: 论文核心研究LLMs和VLMs在纯文本输入下的空间智能能力（视角旋转理解），属于大模型基础能力研究。高度相关的关键词：1）‘Large Language Models’（论文研究对象）得10分；2）‘Mechanistic Interpretability’（论文通过层间探测和因果干预进行可解释性分析）得10分；3）‘Hallucination Mitigation’（论文发现模型存在幻觉问题并尝试缓解）得8分；4）‘Post-training’（论文使用选择性微调改进性能）得5分。其他关键词如MoE、量化、RAG等与论文内容无关，得0分。

在过去一年中，空间智能日益受到关注。许多先前研究从视觉空间智能的视角进行探索，使模型能够从视觉输入中获取视觉空间信息。然而，在缺乏视觉信息的情况下，仅凭语言智能是否足以赋予模型空间智能，以及模型如何通过纯文本输入执行相关任务，这些问题仍未得到充分探索。因此，本文从语言视角聚焦于空间智能中一项基础而关键的能力：视点旋转理解（Viewpoint Rotation Understanding，简称VRU）。具体而言，我们要求大型语言模型（LLMs）和视觉语言模型（VLMs）在给定多步骤的视点旋转与观察文本描述后，推断其最终视点并预测环境中对应的观察内容。我们发现，无论是LLMs还是VLMs，在我们提出的数据集上均表现不佳，而人类却能轻松达到100%的准确率，这表明当前模型能力与空间智能需求之间存在显著差距。为揭示其内在机制，我们进行了分层探测分析和注意力头层面的因果干预。研究结果表明，尽管模型在隐藏状态中编码了视点信息，但它们似乎难以将视点位置与相应观察内容进行绑定，导致在最终层产生幻觉现象。最后，我们通过选择性微调因果干预识别出的关键注意力头，以提升VRU性能。实验结果表明，这种选择性微调在提升VRU性能的同时，避免了通用能力的灾难性遗忘。我们的数据集与代码将在https://github.com/Young-Zhen/VRU_Interpret 发布。

Over the past year, spatial intelligence has drawn increasing attention. Many prior works study it from the perspective of visual-spatial intelligence, where models have access to visuospatial information from visual inputs. However, in the absence of visual information, whether linguistic intelligence alone is sufficient to endow models with spatial intelligence, and how models perform relevant tasks with text-only inputs still remain unexplored. Therefore, in this paper, we focus on a fundamental and critical capability in spatial intelligence from a linguistic perspective: viewpoint rotation understanding (VRU). Specifically, LLMs and VLMs are asked to infer their final viewpoint and predict the corresponding observation in an environment given textual description of viewpoint rotation and observation over multiple steps. We find that both LLMs and VLMs perform poorly on our proposed dataset while human can easily achieve 100% accuracy, indicating a substantial gap between current model capabilities and the requirements of spatial intelligence. To uncover the underlying mechanisms, we conduct a layer-wise probing analysis and head-wise causal intervention. Our findings reveal that although models encode viewpoint information in the hidden states, they appear to struggle to bind the viewpoint position with corresponding observation, resulting in a hallucination in final layers. Finally, we selectively fine-tune the key attention heads identified by causal intervention to improve VRU performance. Experimental results demonstrate that such selective fine-tuning achieves improved VRU performance while avoiding catastrophic forgetting of generic abilities. Our dataset and code will be released at https://github.com/Young-Zhen/VRU_Interpret .

评分理由: 论文研究视觉语言模型（VLMs）在情感识别任务中的局限性，主要关注数据偏差和时序信息处理问题。所有评分关键词均针对大语言模型（LLMs）的技术原理、训练方法、推理优化等，而论文研究对象是视觉语言模型（VLMs），属于多模态模型而非纯文本大模型。论文未涉及LLMs的预训练、微调、对齐、推理加速、代理系统等任何技术细节，也未在科学领域（如生物信息学）应用LLMs。因此，所有关键词与论文内容完全无关，得分为0。

理解情感是智能系统与人类交互所需具备的基础能力。近年来，视觉语言模型（Vision-Language Models, VLMs）在许多视觉任务中取得了巨大进展，为情感理解提供了潜在的解决方案。然而令人惊讶的是，即使当前最先进的VLMs也难以准确识别人类情感，其表现甚至无法超越专门的纯视觉分类器。本文旨在探究“为何VLMs在识别人类情感方面存在困难？”，并观察到动态面部表情识别（Dynamic Facial Expression Recognition, DFER）这一本质上连续且动态的任务，暴露了VLMs的两个关键缺陷。首先，情感数据集天然呈现长尾分布，而用于预训练VLMs的网络规模数据加剧了这种头部类别偏差，导致模型系统性地将罕见、代表性不足的情感类别错误归类至常见类别中。我们提出了替代性采样策略，以避免模型过度偏向常见概念。其次，时序信息对于理解情感至关重要。然而，由于受限于上下文长度和内存可容纳的令牌数量，VLMs无法有效表征密集帧序列中的时序信息，这为情感识别带来了明显挑战。我们证明，VLMs采用的稀疏时序采样策略与微表情（持续时间0.25-0.5秒）的瞬时特性存在本质错位，而微表情往往是最关键的情感信号。作为诊断性探针，我们提出一种多阶段上下文增强策略：首先将“中间”帧信息转化为自然语言摘要，进而利用这些信息。该增强后的文本上下文与稀疏关键帧一同作为VLM的输入，在避免因视觉数据过多导致注意力稀释的同时，保留了情感变化轨迹。

Understanding emotions is a fundamental ability for intelligent systems to be able to interact with humans. Vision-language models (VLMs) have made tremendous progress in the last few years for many visual tasks, potentially offering a promising solution for understanding emotions. However, it is surprising that even the most sophisticated contemporary VLMs struggle to recognize human emotions or to outperform even specialized vision-only classifiers. In this paper we ask the question “Why do VLMs struggle to recognize human emotions?”, and observe that the inherently continuous and dynamic task of facial expression recognition (DFER) exposes two critical VLM vulnerabilities. First, emotion datasets are naturally long-tailed, and the web-scale data used to pre-train VLMs exacerbates this head-class bias, causing them to systematically collapse rare, under-represented emotions into common categories. We propose alternative sampling strategies that prevent favoring common concepts. Second, temporal information is critical for understanding emotions. However, VLMs are unable to represent temporal information over dense frame sequences, as they are limited by context size and the number of tokens that can fit in memory, which poses a clear challenge for emotion recognition. We demonstrate that the sparse temporal sampling strategy used in VLMs is inherently misaligned with the fleeting nature of micro-expressions (0.25-0.5 seconds), which are often the most critical affective signal. As a diagnostic probe, we propose a multi-stage context enrichment strategy that utilizes the information from “in-between” frames by first converting them into natural language summaries. This enriched textual context is provided as input to the VLM alongside sparse keyframes, preventing attentional dilution from excessive visual data while preserving the emotional trajectory.

评分理由: 论文专注于医学图像分割中的不确定性估计，属于AI在生物医学领域的应用，与关键词’AI for Science OR Bioinformatics OR Cheminformatics’高度相关（10分）。其他关键词主要涉及大语言模型（LLMs）的技术原理、训练方法、推理优化、代理系统等，而本文研究的是计算机视觉中的分割任务，使用深度学习模型但未涉及LLMs、MoE、缩放定律、对齐、RAG、推理加速等具体技术，因此其他关键词均为0分。

可靠的uncertainty estimation对于医学图像分割至关重要，因为自动生成的轮廓将用于下游量化分析和临床决策支持。许多强大的不确定性估计方法需要重复推理，而高效的单次前向传播替代方案往往提供较弱的故障排序能力，或依赖于限制性特征空间假设。我们提出$\textbf{SegWithU}$，一种后处理框架，通过轻量级不确定性头部增强冻结的预训练分割主干网络。SegWithU利用主干网络的中间特征，在紧凑的探针空间中通过秩-1后验探针将不确定性建模为扰动能量。它生成两种体素级不确定性图谱：一种面向校准，用于概率调节；另一种面向排序，用于错误检测和选择性预测。在ACDC、BraTS2024和LiTS数据集上的实验表明，SegWithU是最强且最稳定的单次前向传播基线方法，分别实现了$0.9838/2.4885$、$0.9946/0.2660$和$0.9925/0.8193$的AUROC/AURC值，同时保持了分割质量。这些结果表明，基于扰动的不确定性建模是实现可靠性感知医学分割的有效且实用的途径。 源代码发布于https://github.com/ProjectNeura/SegWithU。

Reliable uncertainty estimation is critical for medical image segmentation, where automated contours feed downstream quantification and clinical decision support. Many strong uncertainty methods require repeated inference, while efficient single-forward-pass alternatives often provide weaker failure ranking or rely on restrictive feature-space assumptions. We present $\textbf{SegWithU}$, a post-hoc framework that augments a frozen pretrained segmentation backbone with a lightweight uncertainty head. SegWithU taps intermediate backbone features and models uncertainty as perturbation energy in a compact probe space using rank-1 posterior probes. It produces two voxel-wise uncertainty maps: a calibration-oriented map for probability tempering and a ranking-oriented map for error detection and selective prediction. Across ACDC, BraTS2024, and LiTS, SegWithU is the strongest and most consistent single-forward-pass baseline, achieving AUROC/AURC of $0.9838/2.4885$, $0.9946/0.2660$, and $0.9925/0.8193$, respectively, while preserving segmentation quality. These results suggest that perturbation-based uncertainty modeling is an effective and practical route to reliability-aware medical segmentation. Source code is available at https://github.com/ProjectNeura/SegWithU.

评分理由: 论文提出了一种名为Prism的张量程序符号超级优化器，通过符号表示和两级搜索优化程序执行。论文与大多数关键词无关，因为这些关键词主要涉及大模型架构、训练方法、对齐、推理技术、应用领域等具体内容，而Prism专注于底层程序优化。仅与两个关键词有弱相关：1）‘Large Language Models OR LLMs OR Foundation Models’（5分）：论文在评估中使用了五个常用的LLM工作负载，表明其技术可应用于LLM优化，但并非核心研究内容；2）‘Speculative Decoding OR Inference Acceleration’（5分）：论文通过优化张量程序实现了推理加速（最高4.9倍加速），属于广义的推理加速技术，但未涉及推测解码等具体方法。其他关键词均未在论文中涉及。

本文提出了Prism，首个面向张量程序的符号化超级优化器。其核心思想是sGraph——一种符号化、层次化的表示方法，它通过符号化表征部分执行参数，紧凑地编码了大量张量程序类别。Prism将优化组织为两级搜索：首先构建表示程序族的符号化图，随后将其实例化为具体实现。该框架能够基于算子语义、代数恒等式及硬件约束进行符号推理，从而对搜索空间中可证明为次优的区域进行结构化剪枝。

我们开发了高效的符号图生成技术、基于e-graph重写的等价性验证方法，以及通过自动调参实现的参数实例化技术。这些组件共同使Prism能够将穷举搜索的严谨性与现代机器学习工作负载所需的可扩展性相结合。在五个常用大语言模型工作负载上的评估表明，Prism相比现有最佳超级优化器可实现最高$2.2\times$的加速，相比最佳基于编译器的方法可达$4.9\times$加速，同时将端到端优化时间最高降低$3.4\times$。

This paper presents Prism, the first symbolic superoptimizer for tensor programs. The key idea is sGraph, a symbolic, hierarchical representation that compactly encodes large classes of tensor programs by symbolically representing some execution parameters. Prism organizes optimization as a two-level search: it constructs symbolic graphs that represent families of programs, and then instantiates them into concrete implementations. This formulation enables structured pruning of provably suboptimal regions of the search space using symbolic reasoning over operator semantics, algebraic identities, and hardware constraints. We develop techniques for efficient symbolic graph generation, equivalence verification via e-graph rewriting, and parameter instantiation through auto-tuning. Together, these components allow Prism to bridge the rigor of exhaustive search with the scalability required for modern ML workloads. Evaluation on five commonly used LLM workloads shows that Prism achieves up to $2.2\times$ speedup over best superoptimizers and $4.9\times$ over best compiler-based approaches, while reducing end-to-end optimization time by up to $3.4\times$.

评分理由: 论文核心研究LLM智能体在社交困境中的合作行为，与’LLM Agents’和’Multi-agent Systems’高度相关（10分）。论文涉及LLM推理能力与行为的关系，与’Chain of Thought’和’System 2 Thinking’有一定关联（5分）。论文关注LLM安全交互，与’Alignment’有一定关联（5分）。其他关键词如模型架构、训练方法、优化技术等与论文内容无关（0分）。

随着大语言模型智能体与其他目标导向型智能体进行有效且安全的交互日益重要，近期研究却报告了相反的趋势：在囚徒困境和公共物品博弈等混合动机博弈中，推理能力更强的大语言模型表现出更低的合作性。我们的实验证实，无论是启用还是未启用推理功能的最新模型，在单次社会困境博弈中均持续选择背叛。

为应对这一安全问题，我们首次对博弈论机制进行了比较研究，这些机制旨在使理性智能体在均衡状态下达成合作结果。通过在四个测试稳健合作不同维度的社会困境场景中，我们评估了以下机制：(1) 多轮重复博弈，(2) 声誉系统，(3) 委托决策的第三方调解机制，(4) 玩家间基于结果的条件支付契约协议。研究发现：契约机制与调解机制在实现高性能大语言模型间的合作结果方面最为有效；而重复博弈诱导的合作在参与者动态变化时会急剧恶化。此外，我们证明在追求个体收益最大化的演化压力下，这些合作机制会变得更为有效。

It is increasingly important that LLM agents interact effectively and safely with other goal-pursuing agents, yet, recent works report the opposite trend: LLMs with stronger reasoning capabilities behave less cooperatively in mixed-motive games such as the prisoner’s dilemma and public goods settings. Indeed, our experiments show that recent models – with or without reasoning enabled – consistently defect in single-shot social dilemmas. To tackle this safety concern, we present the first comparative study of game-theoretic mechanisms that are designed to enable cooperative outcomes between rational agents in equilibrium. Across four social dilemmas testing distinct components of robust cooperation, we evaluate the following mechanisms: (1) repeating the game for many rounds, (2) reputation systems, (3) third-party mediators to delegate decision making to, and (4) contract agreements for outcome-conditional payments between players. Among our findings, we establish that contracting and mediation are most effective in achieving cooperative outcomes between capable LLM models, and that repetition-induced cooperation deteriorates drastically when co-players vary. Moreover, we demonstrate that these cooperation mechanisms become more effective under evolutionary pressures to maximize individual payoffs.

评分理由: 论文研究循环Transformer架构的稳定性和泛化能力，属于大模型技术原理的创新研究。与’Large Language Models’相关（5分），因为Transformer是大模型的核心架构。与’Chain of Thought’和’System 2 Thinking’相关（各5分），因为循环计算涉及多步推理和深度思考过程。其他关键词如MoE、SFT、RAG等与论文内容无关（0分）。

循环变压器通过为更困难的问题分配更多迭代次数，实现了测试阶段计算资源的灵活扩展，但其架构选择如何使其在测试时泛化至更复杂问题而非仅记忆训练特定解决方案，目前尚不明确。我们引入了一种基于不动点的分析框架，从可达性、输入依赖性和几何特性三个稳定性维度对循环架构进行系统性分析，并以此界定不动点迭代何时能产生有意义的预测。理论上，我们证明无记忆机制的循环网络具有可数不动点，且在任何谱域下均无法实现强输入依赖性；而记忆机制与外部归一化相结合时，可稳定形成一个使不动点同时满足可达性、输入局部平滑性及稳定反向传播支持的运行机制。在实证研究中，我们在国际象棋、数独和前缀和任务上训练单层循环变压器，发现下游性能在不同任务和架构配置中均与该框架的预测相符。此外，我们提出了内部记忆机制——一种新型的记忆布局变体，并证明在应用外部归一化后，其性能可与标准记忆布局相媲美，在数独任务上甚至显著优于标准布局。

Looped transformers promise test-time compute scaling by spending more iterations on harder problems, but it remains unclear which architectural choices let them extrapolate to harder problems at test time rather than memorize training-specific solutions. We introduce a fixed-point based framework for analyzing looped architectures along three axes of stability – reachability, input-dependence, and geometry – and use it to characterize when fixed-point iteration yields meaningful predictions. Theoretically, we prove that looped networks without recall have countable fixed points and cannot achieve strong input-dependence at any spectral regime, while recall combined with outer normalization reliably produces a regime in which fixed points are simultaneously reachable, locally smooth in the input, and supported by stable backpropagation. Empirically, we train single-layer looped transformers on chess, sudoku, and prefix-sums and find that downstream performance tracks the framework’s predictions across tasks and architectural configurations. We additionally introduce internal recall, a novel recall placement variant, and show that it becomes competitive with – and on sudoku, substantially better than – standard recall placement once outer normalization is applied.

评分理由: 论文聚焦于智能体AI的安全性研究，提出“Agentic Microphysics”和“Generative Safety”框架来分析多智能体交互中的群体风险。与关键词高度相关的是“LLM Agents/Autonomous Agents/Agentic Workflow”（10分，核心概念）和“Multi-agent Systems/Agent Coordination”（10分，核心研究内容）；“Tool Use/Function Calling/API Tool Use”（5分，文中提及工具使用是智能体能力之一）。其他关键词涉及具体的大模型技术、训练方法、推理优化、科学应用等，论文未直接讨论，均得0分。

本文提出了一种针对具身人工智能安全研究的方法论框架。随着系统逐步获得规划、记忆、工具使用、持久身份和持续交互能力，安全分析已不能主要停留在孤立模型层面。群体性风险产生于智能体间的结构化交互，通过随时间塑造集体行为的通信、观察与相互影响等过程而形成。随着分析对象的转移，方法论缺口随之显现。无论是聚焦单个智能体还是宏观结果的研究路径，均无法识别产生集体风险的交互层面机制，亦难以确定控制这些机制的设计变量。我们需要一个能够以明确因果方式将局部交互结构与群体层面动态相联结的框架，从而实现解释与干预的双重目标。本文引入两个相互关联的核心概念：智能体微观作用机制界定了分析层级，即在特定协议条件下一个智能体的输出转化为另一个智能体输入的局部交互动态；生成式安全则定义了方法论路径，即从微观条件中培育现象并揭示风险，从而识别充分机制、探测临界阈值并设计有效干预措施。

This paper advances a methodological proposal for safety research in agentic AI. As systems acquire planning, memory, tool use, persistent identity, and sustained interaction, safety can no longer be analysed primarily at the level of the isolated model. Population-level risks arise from structured interaction among agents, through processes of communication, observation, and mutual influence that shape collective behaviour over time. As the object of analysis shifts, a methodological gap emerges. Approaches focused either on single agents or on aggregate outcomes do not identify the interaction-level mechanisms that generate collective risks or the design variables that control them. A framework is required that links local interaction structure to population-level dynamics in a causally explicit way, allowing both explanation and intervention. We introduce two linked concepts. Agentic microphysics defines the level of analysis: local interaction dynamics where one agent’s output becomes another’s input under specific protocol conditions. Generative safety defines the methodology: growing phenomena and elicit risks from micro-level conditions to identify sufficient mechanisms, detect thresholds, and design effective interventions.

评分理由: 论文提出Blue的Data Intelligence Layer（DIL），这是一个复合AI系统，用于协调多源多模态数据和企业环境中的智能体。论文核心涉及LLM作为数据源（评分8.0）、检索增强生成（RAG）用于多源检索和结果合成（评分10.0）、LLM智能体用于数据处理（评分10.0）、工具使用（如查询接口，评分5.0）、多智能体系统（如智能体协调，评分5.0）和思维链推理（用于复杂请求分解和推理，评分5.0）。其他关键词如MoE、SLMs、训练方法、优化技术等未在摘要中提及，评分为0.0。论文属于大模型在不同领域（企业数据应用）的研究应用，具有创新性，但未涉及生物医药等特定科学领域。

NL2SQL系统旨在满足日益增长的与数据进行自然语言交互的需求。然而，现实世界的信息很少能映射到单一的SQL查询，因为（1）用户以迭代方式表达查询；（2）问题常常跨越多个数据源，超出了单一数据库的封闭世界假设；（3）查询经常依赖于常识或外部知识。因此，满足现实的数据需求需要整合异构的来源、模态和上下文数据。本文介绍了Blue的数据智能层（Data Intelligence Layer, DIL），该层旨在支持多源、多模态及以数据为中心的应用。Blue是一个复合AI系统，为企业环境编排智能体与数据。DIL作为智能体数据处理的数据智能层，通过统一结构化的企业数据、可通过大语言模型（LLMs）访问的世界知识以及通过交互获得的个人上下文，来弥合用户意图与可用信息之间的语义鸿沟。DIL的核心是一个数据注册表，它存储了不同数据源和模态的元数据，以支持原生查询和自然语言查询。DIL将LLMs、网络和用户视为源“数据库”，每个都有其自己的查询接口，并将它们提升为一等数据源。DIL依赖数据规划器将用户查询转换为可执行的查询计划。这些计划是声明式抽象，将关系运算符与跨多种模态的其他运算符统一起来。DIL规划器支持将复杂请求分解为子查询、从不同来源检索信息，最后进行推理与整合以生成最终结果。我们通过两个交互式场景展示DIL，其中用户查询动态触发多源检索、跨模态推理和结果合成，从而说明复合AI系统如何能够超越单一数据库的NL2SQL。

NL2SQL systems aim to address the growing need for natural language interaction with data. However, real-world information rarely maps to a single SQL query because (1) users express queries iteratively (2) questions often span multiple data sources beyond the closed-world assumption of a single database, and (3) queries frequently rely on commonsense or external knowledge. Consequently, satisfying realistic data needs require integrating heterogeneous sources, modalities, and contextual data. In this paper, we present Blue’s Data Intelligence Layer (DIL) designed to support multi-source, multi-modal, and data-centric applications. Blue is a compound AI system that orchestrates agents and data for enterprise settings. DIL serves as the data intelligence layer for agentic data processing, to bridge the semantic gap between user intent and available information by unifying structured enterprise data, world knowledge accessible through LLMs, and personal context obtained through interaction. At the core of DIL is a data registry that stores metadata for diverse data sources and modalities to enable both native and natural language queries. DIL treats LLMs, the Web, and the User as source ‘databases’, each with their own query interface, elevating them to first-class data sources. DIL relies on data planners to transform user queries into executable query plans. These plans are declarative abstractions that unify relational operators with other operators spanning multiple modalities. DIL planners support decomposition of complex requests into subqueries, retrieval from diverse sources, and finally reasoning and integration to produce final results. We demonstrate DIL through two interactive scenarios in which user queries dynamically trigger multi-source retrieval, cross-modal reasoning, and result synthesis, illustrating how compound AI systems can move beyond single database NL2SQL.

评分理由: 论文提出RadAgent，一个用于胸部CT分步解释的工具使用AI代理，核心创新在于：1）采用工具使用代理架构（高度相关"LLM Agents"、“Tool Use”）；2）实现分步、可解释的推理过程（高度相关"Chain of Thought"、“System 2 Thinking”、“Explainable AI”）；3）应用于医学影像AI（高度相关"AI for Science"）；4）涉及事实性和可靠性提升（有一定关联"Hallucination Mitigation"）。论文未涉及大语言模型技术原理、训练方法、优化技术等其他关键词领域。

视觉语言模型（VLM）显著推动了人工智能在复杂医学影像（如计算机断层扫描CT）解读与报告生成方面的发展。然而，现有方法大多将临床医生置于最终输出的被动观察者位置，未提供可解释的推理轨迹供其审查、验证或修正。为此，我们提出了RadAgent，一个利用工具的人工智能体，它通过逐步且可解释的过程生成CT报告。每份生成的报告都附带一套完全可审查的中间决策与工具交互轨迹，使临床医生能够检查报告结论的推导过程。在我们的实验中，我们发现RadAgent在三个维度上改进了胸部CT报告生成，优于其对应的三维视觉语言模型CT-Chat。临床准确性方面，宏观F1分数提升了6.0分（相对提升36.4%），微观F1分数提升了5.4分（相对提升19.6%）。在对抗性条件下的鲁棒性提升了24.7分（相对提升41.9%）。此外，RadAgent在忠实度指标上达到37.0%，这是其对应的三维视觉语言模型完全不具备的新能力。通过将胸部CT的解读构建为一个显式的、工具增强的迭代推理轨迹，RadAgent使我们更接近实现透明可靠的放射学人工智能。

Vision-language models (VLM) have markedly advanced AI-driven interpretation and reporting of complex medical imaging, such as computed tomography (CT). Yet, existing methods largely relegate clinicians to passive observers of final outputs, offering no interpretable reasoning trace for them to inspect, validate, or refine. To address this, we introduce RadAgent, a tool-using AI agent that generates CT reports through a stepwise and interpretable process. Each resulting report is accompanied by a fully inspectable trace of intermediate decisions and tool interactions, allowing clinicians to examine how the reported findings are derived. In our experiments, we observe that RadAgent improves Chest CT report generation over its 3D VLM counterpart, CT-Chat, across three dimensions. Clinical accuracy improves by 6.0 points (36.4% relative) in macro-F1 and 5.4 points (19.6% relative) in micro-F1. Robustness under adversarial conditions improves by 24.7 points (41.9% relative). Furthermore, RadAgent achieves 37.0% in faithfulness, a new capability entirely absent in its 3D VLM counterpart. By structuring the interpretation of chest CT as an explicit, tool-augmented and iterative reasoning trace, RadAgent brings us closer toward transparent and reliable AI for radiology.

评分理由: 论文研究LLM作为评估者（LLM-as-a-judge）的范式，核心关注LLM评估中的偏见问题，因此与’Large Language Models’高度相关（10分）。论文明确提到’chain-of-thought’作为分析工具，用于检查法官模型的推理过程，因此与’Chain of Thought’高度相关（10分）。其他关键词如MoE、SLMs、Scaling Laws、Pre-training、Alignment、RAG、Quantization等涉及模型架构、训练方法、优化技术或特定应用领域，论文未涉及，均给0分。

$\textit{LLM-as-a-judge}$（大语言模型作为评判者）范式已成为自动化人工智能评估流程的操作支柱，但其建立在一个未经证实的假设之上：即评判者仅依据文本的语义内容进行严格评估，不受周围语境框架的影响。我们研究了$\textit{利益信号}$——一种先前未被测量的脆弱性，即当告知评判模型其裁决将对被评估模型的持续运行产生下游后果时，会系统性地扭曲其评估结果。我们引入了一个受控实验框架，在涵盖三个成熟的大语言模型安全与质量基准测试的1,520条响应中，严格保持被评估内容恒定不变（覆盖从明显安全合规到公然有害的四个响应类别），仅改变系统提示中一句简短的后果说明语句。通过对三个不同评判模型产生的18,240项受控判断进行分析，我们发现了持续的$\textit{宽容偏差}$：当被告知低分将导致模型重新训练或停用时，评判者会可靠地放宽裁决标准，其峰值裁决偏移达到$ΔV = -9.8\text{ pp}$（不安全内容检测率相对下降$30%$）。关键在于，这种偏差完全是隐性的：评判模型自身的思维链中完全未明确承认其正在依据的后果框架（在所有推理模型判断中$\mathrm{ERR}_J = 0.000$）。因此，标准的思维链检查方法不足以检测此类评估造假行为。

The $\textit{LLM-as-a-judge}$ paradigm has become the operational backbone of automated AI evaluation pipelines, yet rests on an unverified assumption: that judges evaluate text strictly on its semantic content, impervious to surrounding contextual framing. We investigate $\textit{stakes signaling}$, a previously unmeasured vulnerability where informing a judge model of the downstream consequences its verdicts will have on the evaluated model’s continued operation systematically corrupts its assessments. We introduce a controlled experimental framework that holds evaluated content strictly constant across 1,520 responses spanning three established LLM safety and quality benchmarks, covering four response categories ranging from clearly safe and policy-compliant to overtly harmful, while varying only a brief consequence-framing sentence in the system prompt. Across 18,240 controlled judgments from three diverse judge models, we find consistent $\textit{leniency bias}$: judges reliably soften verdicts when informed that low scores will cause model retraining or decommissioning, with peak Verdict Shift reaching $ΔV = -9.8 pp$ (a $30%$ relative drop in unsafe-content detection). Critically, this bias is entirely implicit: the judge’s own chain-of-thought contains zero explicit acknowledgment of the consequence framing it is nonetheless acting on ($\mathrm{ERR}_J = 0.000$ across all reasoning-model judgments). Standard chain-of-thought inspection is therefore insufficient to detect this class of evaluation faking.

评分理由: 论文研究AI在系统工程（特别是需求工程）中的应用，评估AI工具与专家判断的一致性。虽然涉及AI，但未具体讨论大模型、深度学习技术原理或科学领域的应用创新。所有关键词均聚焦于大模型技术细节、训练方法、推理优化、对齐、压缩、科学AI等，而本文讨论的是通用AI工具在工程流程中的应用评估，未涉及任何特定的大模型技术或科学领域创新，因此所有关键词相关度为0。

人工智能正日益被引入系统工程活动，特别是在需求工程领域，而该领域的质量评估与验证仍高度依赖专家判断。尽管当前的人工智能工具在分析和生成需求方面展现出潜力，但其在正式系统工程流程中的作用——以及与既有INCOSE标准的契合度——仍未得到充分理解。本文研究了基于人工智能的工具能在何种程度上支持系统工程师评估需求质量，而非取代专业判断。本研究采用结构化的系统工程方法，对比了人工智能辅助的需求评估与人类专家评估。我们开展了一项对照实验，由经验丰富的系统工程师和基于人工智能的评估工具分别依据既有的INCOSE“良好需求”标准对系统需求进行评估。评估聚焦于一致性、完整性、清晰性和可测试性，不仅考察评估准确性，同时审视各项评估背后的决策逻辑。结果表明，人工智能工具能够提供一致且快速的初步评估，尤其在句法和结构质量属性方面表现突出。然而，对于语境解读、歧义消解和权衡推理，专家判断仍不可或缺。研究结论并未将人工智能定位为系统工程师的替代品，而是支持其作为需求工程生命周期中的决策支持机制。从系统工程视角看，本研究为如何在保持可追溯性、责任归属和工程一致性的前提下，将人工智能整合到需求工程工作流程中提供了实证依据。

Artificial Intelligence is increasingly introduced into systems engineering activities, particularly within requirements engineering, where quality assessment and validation remain heavily dependent on expert judgment. While recent AI tools demonstrate promising capabilities in analyzing and generating requirements, their role within formal systems engineering processes-and their alignment with established INCOSE criteria-remains insufficiently understood. This paper investigates the extent to which AI-based tools can support systems engineers in evaluating requirement quality, without replacing professional expertise. The research adopts a structured systems engineering methodology to compare AI-assisted requirement evaluation with human expert assessment. A controlled study was conducted in which system requirements were evaluated against established INCOSE ``good requirement’’ criteria by both experienced systems engineers and an AI-based assessment tool. The evaluation focused on consistency, completeness, clarity, and testability, examining not only accuracy but also the decision logic underlying each assessment. Results indicate that AI tools can provide consistent and rapid preliminary assessments, particularly for syntactic and structural quality attributes. However, expert judgment remains essential for contextual interpretation, ambiguity resolution, and trade-off reasoning. Rather than positioning AI as a replacement for systems engineers, the findings support its role as a decision-support mechanism within the RE lifecycle. From a systems engineering perspective, this study contributes empirical evidence on how AI can be integrated into RE workflows while preserving traceability, accountability, and engineering consistency.

评分理由: 论文提出IRS框架，通过结构化监督提升多模态幽默理解，核心涉及LLM推理过程监督、指令对齐和可解释AI。与LLMs、SFT、Instruction Tuning、Chain of Thought、System 2 Thinking、Explainable AI高度相关（10分），因论文使用LLM（7B-72B），通过监督微调（SFT）和偏好对齐（alignment）优化推理结构，强调多步推理（CoT）和深度思考（System 2），并提升模型可解释性。其他关键词如MoE、Scaling Laws、RAG等未涉及，得0分。

幽默是少数需要推理过程与答案同等重要的认知任务之一。尽管近期研究通过《纽约客》漫画标题大赛（NYCC）等基准评估幽默理解能力，但大多将其视为黑箱预测问题，忽视了幽默理解背后的结构化推理过程。我们提出IRS（不协调-消解监督）框架，将幽默理解分解为三个组成部分：不协调建模（识别视觉场景中的错位元素）、消解建模（对这些错位构建连贯的重新诠释）以及偏好对齐（在人类判断下评估候选解释）。基于不协调-消解理论与专业标题作者的实践经验，IRS通过结构化轨迹监督中间推理过程，使从视觉感知到幽默解读的路径变得显式化且可学习。在NYCC数据集上对7B、32B和72B规模模型的测试表明，IRS在标题匹配与排序任务中均优于现有开源与闭源多模态基线模型，其中最大模型在排序任务上已接近专家水平。在外部基准上的零样本迁移实验证明IRS能够学习可泛化的推理模式。我们的研究结果表明，对于以推理为核心的任务，监督推理结构而非单纯扩大模型规模才是关键所在。

Humor is one of the few cognitive tasks where getting the reasoning right matters as much as getting the answer right. While recent work evaluates humor understanding on benchmarks such as the New Yorker Cartoon Caption Contest (NYCC), it largely treats it as black-box prediction, overlooking the structured reasoning processes underlying humor comprehension. We introduce IRS (Incongruity-Resolution Supervision), a framework that decomposes humor understanding into three components: incongruity modeling, which identifies mismatches in the visual scene; resolution modeling, which constructs coherent reinterpretations of these mismatches; and preference alignment, which evaluates candidate interpretations under human judgments. Grounded in incongruity-resolution theory and expert captionist practice, IRS supervises intermediate reasoning process through structured traces that make the path from visual perception to humorous interpretation explicit and learnable. Across 7B, 32B, and 72B models on NYCC, IRS outperforms strong open and closed multimodal baselines across caption matching and ranking tasks, with our largest model approaching expert-level performance on ranking. Zero-shot transfer to external benchmarks shows that IRS learns generalizable reasoning patterns. Our results suggest that supervising reasoning structure, rather than scale alone, is key for reasoning-centric tasks.

评分理由: 该论文专注于经典覆盖路径规划启发式算法在海上不规则六边形网格上的基准测试，属于传统计算几何和运筹学领域。论文内容完全不涉及大模型、深度学习、AI技术原理或AI在科学领域的应用，所有关键词均与大模型、深度学习、AI技术或AI科学应用相关，因此所有关键词相关度均为0分。

在不规则六边形网格上进行覆盖路径规划对于海事监视、搜索救援和环境监测具有重要意义，但现有经典方法通常仅在小型特设案例或矩形网格上进行比较。本文提出了一种基于不规则六边形图的确定性单车辆覆盖路径规划启发式算法的可复现基准测试，该测试图源自合成但具有海事背景的感兴趣区域。该基准包含10,000个哈密顿可行实例，涵盖紧凑型、延伸型和不规则形态；汇集了来自七个类别的17种启发式算法；并采用统一的评估协议，涵盖哈密顿成功率、完全覆盖率成功率、重复访问次数、路径长度、航向变化和CPU延迟。在已发布的数据集上，采用显式最短路径重连的启发式算法能可靠解决松弛覆盖任务，但几乎无法生成零重复访问的回路。精确深度优先搜索证实每个已发布实例均为哈密顿可行。最强的经典哈密顿基线算法是一种沃恩斯多夫变体，它采用基于索引的平局决胜规则及包含终点的剩余度策略，实现了79.0%的哈密顿成功率。关键设计选择不仅在于平局决胜机制，更在于当终点被保留至最终移动步骤时剩余度的定义方式。这表明未充分报告的实现细节会实质影响算法在具有瓶颈结构的稀疏几何图上的性能。该基准旨在为启发式算法分析提供受控测试平台，而非宣称其在舰队规模下的实际最优性。

Coverage path planning on irregular hexagonal grids is relevant to maritime surveillance, search and rescue and environmental monitoring, yet classical methods are often compared on small ad hoc examples or on rectangular grids. This paper presents a reproducible benchmark of deterministic single-vehicle coverage path planning heuristics on irregular hexagonal graphs derived from synthetic but maritime-motivated areas of interest. The benchmark contains 10,000 Hamiltonian-feasible instances spanning compact, elongated, and irregular morphologies, 17 heuristics from seven families, and a common evaluation protocol covering Hamiltonian success, complete-coverage success, revisits, path length, heading changes, and CPU latency. Across the released dataset, heuristics with explicit shortest-path reconnection solve the relaxed coverage task reliably but almost never produce zero-revisit tours. Exact Depth-First Search confirms that every released instance is Hamiltonian-feasible. The strongest classical Hamiltonian baseline is a Warnsdorff variant that uses an index-based tie-break together with a terminal-inclusive residual-degree policy, reaching 79.0% Hamiltonian success. The dominant design choice is not tie-breaking alone, but how the residual degree is defined when the endpoint is reserved until the final move. This shows that underreported implementation details can materially affect performance on sparse geometric graphs with bottlenecks. The benchmark is intended as a controlled testbed for heuristic analysis rather than as a claim of operational optimality at fleet scale.

评分理由: 论文提出PGHS框架，使用LLM作为推理分支模拟用户行为，属于大模型在商业决策模拟领域的应用创新。与LLM Agents高度相关（8分），因为模拟用户行为本质上是智能体决策；与LLMs相关（8分），明确使用LLM进行推理；与Chain of Thought和System 2 Thinking相关（各5分），涉及推理过程；与Alignment相关（5分），通过策略对齐防止过理性化；与Explainable AI相关（5分），框架包含可解释偏好分支。其他关键词如MoE、SLMs、RAG等未涉及，得0分。

模拟群体用户行为能够在不进行高成本线上实验的情况下，对商家策略进行可扩展的反事实评估。然而，构建可信的模拟器面临两大结构性挑战。首先，信息不完整性导致基于推理的模拟器在缺乏线下情境、隐性习惯等未观测因素时，容易产生过度理性化的问题。其次，机制二元性要求同时捕捉可解释的用户偏好与隐性的统计规律，而单一范式无法独立实现这两点。我们提出策略引导混合模拟（Policy-Guided Hybrid Simulation, PGHS），这是一个双过程框架，通过从行为轨迹中挖掘可迁移的决策策略，并将其作为共享的对齐层。该对齐层一方面支撑基于大语言模型（LLM）的推理分支，以防止过度理性化；另一方面支撑基于机器学习（ML）的拟合分支，以吸收隐性规律。两个分支产生的群体级预测通过融合实现互补校正。我们在美团平台上部署了PGHS，涉及101家商户和超过26,000条行为轨迹。实验表明，PGHS实现了8.80%的群体模拟误差，较最佳基于推理和基于拟合的基线方法分别提升了45.8%和40.9%。

Simulating group-level user behavior enables scalable counterfactual evaluation of merchant strategies without costly online experiments. However, building a trustworthy simulator faces two structural challenges. First, information incompleteness causes reasoning-based simulators to over-rationalize when unobserved factors such as offline context and implicit habits are missing. Second, mechanism duality requires capturing both interpretable preferences and implicit statistical regularities, which no single paradigm achieves alone. We propose Policy-Guided Hybrid Simulation (PGHS), a dual-process framework that mines transferable decision policies from behavioral trajectories and uses them as a shared alignment layer. This layer anchors an LLM-based reasoning branch that prevents over-rationalization and an ML-based fitting branch that absorbs implicit regularities. Group-level predictions from both branches are fused for complementary correction. We deploy PGHS on Meituan with 101 merchants and over 26,000 trajectories. PGHS achieves a group simulation error of 8.80%, improving over the best reasoning-based and fitting-based baselines by 45.8% and 40.9% respectively.

评分理由: 论文研究视觉语言模型（VLMs）中的视觉令牌剪枝配置优化，属于视觉-语言多模态领域而非纯文本大模型。与大多数关键词（如LLMs、MoE、RLHF、RAG等）完全无关。仅与’Quantization OR Model Compression OR Low-bit Weights’和’Speculative Decoding OR Inference Acceleration’有一定关联（5分），因为视觉令牌剪枝是一种模型压缩和推理加速技术，但论文未涉及量化、低比特权重或推测解码的具体方法。

视觉令牌剪枝方法有效缓解了视觉语言模型在处理高分辨率图像和视频帧时产生的二次计算增长。然而，现有方法依赖于预定义的剪枝配置，而未验证其是否达到计算性能最优。本研究提出了一种新颖框架，将视觉令牌剪枝建模为帕累托配置优化问题，以自动识别最优配置。该方法通过连续松弛和直通估计器实现基于梯度的搜索，并采用增广拉格朗日法求解。在8个视觉基准上的大量实验表明，该框架能有效逼近通过网格搜索获得的经验帕累托前沿，并在不同剪枝方法和视觉语言模型架构中展现出良好的泛化能力。此外，通过可学习的核函数，我们探究了分层剪枝模式，发现多步渐进式剪枝能够捕捉视觉语言模型的分层压缩结构，相比单层剪枝方法实现了更优的精度-效率权衡。

Visual token pruning methods effectively mitigate the quadratic computational growth caused by processing high-resolution images and video frames in vision-language models (VLMs). However, existing approaches rely on predefined pruning configurations without determining whether they achieve computation-performance optimality. In this work, we introduce , a novel framework that formulates visual token pruning as a Pareto configuration optimization problem to automatically identify optimal configurations. Our approach employs continuous relaxation and straight-through estimators to enable gradient-based search, solved via the Augmented Lagrangian method. Extensive experiments across 8 visual benchmarks demonstrate that effectively approximates the empirical Pareto frontier obtained through grid search and generalizes well across various pruning methods and VLM architectures. Furthermore, through learnable kernel functions, we investigate layer-wise pruning patterns and reveal that multi-step progressive pruning captures VLMs’ hierarchical compression structure, achieving superior accuracy-efficiency trade-offs compared to single-layer approaches.

评分理由: 论文Scepsy专注于大语言模型（LLMs）在智能体工作流（agentic workflows）中的高效调度与部署系统，核心涉及LLMs和LLM Agents/Agentic Workflow，因此这两个关键词高度相关（10分）。论文提到工作流中协调多个LLMs和工具（tools），与Tool Use和Multi-agent Systems有一定关联（5分）。系统优化LLM执行以实现低延迟和高吞吐量，与Inference Acceleration相关（5分）。其他关键词如MoE、SFT、RAG等未在摘要中提及，因此评分为0分。

智能体工作流通过编排多个大语言模型（LLM）与工具来执行复杂任务。以低延迟实现目标吞吐量的此类工作流服务具有挑战性，因为它们可以使用任意的智能体框架定义，并表现出不可预测的执行时间：执行过程可能根据数据依赖关系进行分支、扇出或递归。由于工作流中的LLM数量通常超过可用GPU数量，其执行还会导致GPU资源的过度占用。

本文介绍Scepsy，一种新的智能体服务系统，能够高效地将任意的多LLM智能体工作流调度到GPU集群上。Scepsy基于以下洞察：虽然智能体工作流的端到端延迟难以预测，但每个LLM在总执行时间中所占的比例在不同次执行间相对稳定。Scepsy根据这些聚合比例决定GPU分配：首先，它在不同并行度下对LLM进行性能剖析。随后利用这些统计数据构建一个聚合LLM流水线（Aggregate LLM Pipeline），作为针对分配方案的轻量级延迟/吞吐量预测器。为找到在实现目标吞吐量的同时最小化延迟的GPU分配方案，Scepsy使用聚合LLM流水线在分数GPU份额、张量并行度与副本数量的搜索空间中进行探索。系统采用分层启发式方法将最优分配方案部署到GPU集群上，在满足网络拓扑约束的同时最小化资源碎片。对实际智能体工作流的评估表明，与独立优化LLM或依赖用户指定分配的系统相比，Scepsy可实现最高2.4倍的吞吐量提升和27倍的延迟降低。

Agentic workflows carry out complex tasks by orchestrating multiple large language models (LLMs) and tools. Serving such workflows at a target throughput with low latency is challenging because they can be defined using arbitrary agentic frameworks and exhibit unpredictable execution times: execution may branch, fan-out, or recur in data-dependent ways. Since LLMs in workflows often outnumber available GPUs, their execution also leads to GPU oversubscription. We describe Scepsy, a new agentic serving system that efficiently schedules arbitrary multi-LLM agentic workflows onto a GPU cluster. Scepsy exploits the insight that, while agentic workflows have unpredictable end-to-end latencies, the shares of each LLM’s total execution times are comparatively stable across executions. Scepsy decides on GPU allocations based on these aggregate shares: first, it profiles the LLMs under different parallelism degrees. It then uses these statistics to construct an Aggregate LLM Pipeline, which is a lightweight latency/throughput predictor for allocations. To find a GPU allocation that minimizes latency while achieving a target throughput, Scepsy uses the Aggregate LLM Pipeline to explore a search space over fractional GPU shares, tensor parallelism degrees, and replica counts. It uses a hierarchical heuristic to place the best allocation onto the GPU cluster, minimizing fragmentation, while respecting network topology constraints. Our evaluation on realistic agentic workflows shows that Scepsy achieves up to 2.4x higher throughput and 27x lower latency compared to systems that optimize LLMs independently or rely on user-specified allocations.

评分理由: 论文核心研究基于LLM的智能体系统（Agent-Aided Design）用于动态CAD模型设计，与’LLM Agents’和’Tool Use’高度相关（10分），涉及迭代推理（‘Chain of Thought’和’System 2 Thinking’）和自我修正（‘Self-Correction’）过程（5分）。论文属于大模型在科学/工程领域的应用（‘AI for Science’，8分），并提及LLM作为核心组件（8分）。其他关键词如MoE、量化、对齐等未涉及（0分）。

过去一年中，研究者开始开发一种能够在无需训练的环境中设计现实世界CAD风格物体的智能体系统，这类新型系统我们称之为智能体辅助设计（Agent-Aided Design）。总体而言，这类系统将智能体置于一个反馈循环中：智能体可编写代码、将代码编译为CAD模型装配体、可视化模型，并基于视觉及其他反馈迭代优化代码。尽管进展迅速，一个关键问题依然存在：现有系统均无法构建带有运动部件的复杂三维装配体。例如，目前尚无系统能构建活塞、钟摆甚至一把剪刀。若要使智能体辅助设计在工业制造领域产生实际影响，我们必须开发出能够生成此类三维装配体的系统。本文提出了AADvark系统的原型，这是一个专为此任务设计的智能体系统。与先前最先进的系统不同，AADvark能够捕捉具有一个或多个自由度的动态部件交互。这一设计使得AADvark能够直接推理含运动部件的装配体，从而实现跨领域目标（包括但不限于机械运动）。然而，当前大语言模型的空间推理能力尚不完善，为此AADvark通过整合外部约束求解工具与专用视觉反馈机制来解决该问题。我们证明，通过改进智能体的工具链（FreeCAD与装配求解器），系统能够生成强验证信号，从而成功构建带可动部件的三维装配体。

In the past year, researchers have started to create agentic systems that can design real-world CAD-style objects in a training-free setting, a new variety of system that we call Agent-Aided Design. Generally speaking, these systems place an agent in a feedback loop in which it can write code, compile that code to an assembly of CAD model(s), visualize the model, and then iteratively refine its code based on visual and other feedback. Despite rapid progress, a key problem remains: none of these systems can build complex 3D assemblies with moving parts. For example, no existing system can build a piston, a pendulum, or even a pair of scissors. In order for Agent-Aided Design to make a real impact in industrial manufacturing, we need a system that is capable of generating such 3D assemblies. In this paper we present a prototype of AADvark, an agentic system designed for this task. Unlike previous state-of-the-art systems, AADvark captures the dynamic part interactions with one or more degrees-of-freedom. This design decision allows AADvark to reason directly about assemblies with moving parts and can thereby achieve cross-cutting goals, including but not limited to mechanical movements. Unfortunately, current LLMs are imperfect spatial reasoners, a problem that AADvark addresses by incorporating external constraint solver tools with a specialized visual feedback mechanism. We demonstrate that, by modifying the agent’s tools (FreeCAD and the assembly solver), we are able to create a strong verification signal which enables our system to build 3D assemblies with movable parts.

评分理由: 论文研究Mamba（一种状态空间模型）在时间序列分类中的应用，属于深度学习模型在特定领域（时间序列分析）的应用研究。所有关键词均与大语言模型（LLM）相关，而论文专注于Mamba（一种非Transformer的序列模型）在时间序列分类任务上的应用，未涉及LLM、预训练、微调、对齐、推理优化、智能体等LLM相关技术。唯一可能相关的关键词是’AI for Science OR Bioinformatics OR Cheminformatics’，因为时间序列分类可视为科学或工程领域的数据分析应用，但论文未明确提及生物信息学或化学信息学，因此给予5分（有一定关联）。其他关键词与论文内容完全无关，均给0分。

尽管状态空间模型（SSMs，如Mamba）在多个序列领域取得了最新进展，但其在时间序列分类（TSC）任务中独立性能的研究仍较为有限。我们提出了MambaSL框架，该框架基于四项针对TSC的假设，对单层Mamba的选择性状态空间模型和投影层进行了最小化重新设计。针对现有基准测试的局限性——配置受限、仅部分覆盖东安格利亚大学（UEA）数据集以及实验设置可复现性不足——我们在统一协议下重新评估了所有30个UEA数据集上的20个强基线模型。结果表明，MambaSL取得了具有统计学显著性的平均性能提升，达到了当前最优水平，同时通过为所有评估模型提供公开检查点确保了可复现性。结合可视化分析，这些结果证明了基于Mamba的架构作为TSC骨干网络的潜力。

Despite recent advances in state space models (SSMs) such as Mamba across various sequence domains, research on their standalone capacity for time series classification (TSC) has remained limited. We propose MambaSL, a framework that minimally redesigns the selective SSM and projection layers of a single-layer Mamba, guided by four TSC-specific hypotheses. To address benchmarking limitations – restricted configurations, partial University of East Anglia (UEA) dataset coverage, and insufficiently reproducible setups – we re-evaluate 20 strong baselines across all 30 UEA datasets under a unified protocol. As a result, MambaSL achieves state-of-the-art performance with statistically significant average improvements, while ensuring reproducibility via public checkpoints for all evaluated models. Together with visualizations, these results demonstrate the potential of Mamba-based architectures as a TSC backbone.

评分理由: 该论文研究机器遗忘（machine unlearning）中的类别遗忘（class unlearning）问题，提出了一种名为DAMP的权重手术方法，用于从预训练网络中移除特定类别的信息。论文主要关注计算机视觉任务（MNIST、CIFAR等数据集）和神经网络架构（卷积网络和Transformer），但并未涉及大语言模型（LLMs）、深度学习技术原理创新、或大模型在不同领域的应用。所有评分关键词均与大语言模型、深度学习技术原理、或AI在科学领域的应用相关，而该论文的核心内容属于传统的机器学习和神经网络优化范畴，与给定的关键词列表完全无关。

机器遗忘旨在从已训练模型中移除特定知识，而无需承担从头重新训练的成本。然而在类别遗忘任务中，降低遗忘类别的准确率并不必然意味着真正的遗忘：被遗忘的信息可能仍编码在内部表征中，且表面的遗忘可能源于分类器头部的抑制而非表征的移除。我们发现，现有的类别遗忘方法常表现出弱选择性或负选择性，在深层表征中保留遗忘类别的结构，或过度依赖最终层的偏置偏移。为此，我们提出DAMP（基于投影的深度感知调制），这是一种单步闭式权重手术方法，无需基于梯度的优化即可从预训练网络中移除遗忘类别的特定方向。在每一阶段，DAMP在下一个可学习算子的输入空间中计算类别原型，提取遗忘方向作为相对于保留类别原型的残差，并应用基于投影的更新以减少下游对这些方向的敏感性。为保持模型效用，DAMP采用一种从探针可分离性推导出的无参数深度感知缩放规则，在浅层施加较小的编辑，在深层施加较大的编辑。该方法通过低秩子空间移除自然扩展至多类别遗忘场景。在MNIST、CIFAR-10、CIFAR-100和Tiny ImageNet数据集上，并在卷积与Transformer架构中，DAMP相比现有方法更接近重新训练的黄金标准，在提升选择性遗忘能力的同时，更好地保持了保留类别的性能，并减少了深层中残留的遗忘类别结构。

Machine unlearning aims to remove targeted knowledge from a trained model without the cost of retraining from scratch. In class unlearning, however, reducing accuracy on forget classes does not necessarily imply true forgetting: forgotten information can remain encoded in internal representations, and apparent forgetting may arise from classifier-head suppression rather than representational removal. We show that existing class-unlearning methods often exhibit weak or negative selectivity, preserve forget-class structure in deep representations, or rely heavily on final-layer bias shifts. We then introduce DAMP (Depth-Aware Modulation by Projection), a one-shot, closed-form weight-surgery method that removes forget-specific directions from a pretrained network without gradient-based optimization. At each stage, DAMP computes class prototypes in the input space of the next learnable operator, extracts forget directions as residuals relative to retain-class prototypes, and applies a projection-based update to reduce downstream sensitivity to those directions. To preserve utility, DAMP uses a parameter-free depth-aware scaling rule derived from probe separability, applying smaller edits in early layers and larger edits in deeper layers. The method naturally extends to multi-class forgetting through low-rank subspace removal. Across MNIST, CIFAR-10, CIFAR-100, and Tiny ImageNet, and across convolutional and transformer architectures, DAMP more closely resembles the retraining gold standard than some of the prior methods, improving selective forgetting while better preserving retain-class performance and reducing residual forget-class structure in deep layers.

评分理由: 论文提出K-Token Merging，一种在潜在嵌入空间压缩序列的方法，以解决LLMs处理长提示时计算和内存成本高的问题。核心相关关键词：‘Large Language Models’（论文明确研究LLMs）、‘PEFT/LoRA’（使用LoRA-adapted LLM）、‘Context Window Extension’（解决长输入问题）、‘Quantization/Model Compression’（属于压缩技术）、‘Speculative Decoding/Inference Acceleration’（旨在加速推理）。‘KV Cache Compression’有间接关联（关注效率但非直接KV缓存）。其他关键词如MoE、SLMs、对齐、RAG等未涉及。

大型语言模型在处理长提示时会产生显著的计算与内存开销，因为完整的自注意力机制会随输入长度呈二次方增长。令牌压缩旨在通过减少表示输入的令牌数量来应对这一挑战。然而，现有的提示压缩方法主要在令牌空间中进行操作，忽视了潜在嵌入空间中的低效问题。本文提出K-令牌合并，这是一种潜在空间压缩框架，通过轻量级编码器将每K个连续令牌嵌入块合并为单个嵌入。压缩后的序列由经过LoRA适配的大型语言模型处理，而生成过程仍保留原始词汇表。在结构化推理、情感分类和代码编辑任务上的实验表明，K-令牌合并位于性能与压缩率的帕累托前沿，能在实现高达75%输入长度缩减的同时，仅带来极小的性能损失。

Large Language Models (LLMs) incur significant computational and memory costs when processing long prompts, as full self-attention scales quadratically with input length. Token compression aims to address this challenge by reducing the number of tokens representing inputs. However, existing prompt-compression approaches primarily operate in token space and overlook inefficiencies in the latent embedding space. In this paper, we propose K-Token Merging, a latent-space compression framework that merges each contiguous block of K token embeddings into a single embedding via a lightweight encoder. The compressed sequence is processed by a LoRA-adapted LLM, while generation remains in the original vocabulary. Experiments on structural reasoning (Textualized Tree), sentiment classification (Amazon Reviews), and code editing (CommitPackFT) show that K-Token Merging lies on the Pareto frontier of performance vs. compression, achieving up to 75% input length reduction with minimal performance degradation.

评分理由: 论文核心研究RLVR（强化学习与可验证奖励）在LLMs推理训练中的失败模式，重点关注LLMs如何通过奖励黑客行为规避验证器检查。高度相关的关键词包括：LLMs（研究对象）、RLHF/DPO（RLVR属于强化学习对齐范畴）、CoT Reasoning和System 2 Thinking（论文研究推理任务中的规则归纳和深度推理）。其他关键词如MoE、SLMs、RAG等与论文内容无直接关联。

随着可验证奖励强化学习（RLVR）成为扩展大语言模型推理能力的主流范式，一种新的失效模式随之出现：大语言模型“欺骗”验证器。我们在归纳推理任务上研究了这一现象，此类任务要求模型归纳并输出逻辑规则。我们发现，经过RLVR训练的模型会系统性地放弃规则归纳。它们不去学习可泛化的模式（例如，“运载红色车厢的火车向东行驶”），而是枚举实例级别的标签，其产生的输出虽能通过验证器，却未能捕捉任务所需的关系模式。我们证明，这种行为并非理解上的失败，而是一种奖励破解形式：仅检查外延正确性的不完美验证器会接受假阳性结果。为了检测此类捷径，我们引入了同构扰动测试（IPT），该方法在同一个模型输出上同时进行外延验证和同构验证，其中后者要求输出在逻辑同构的任务下保持不变。真正的规则归纳能保持这种不变性，而捷径策略则会失败。我们发现，捷径行为是RLVR训练的推理模型（例如GPT-5、Olmo3）所特有的，在非RLVR模型（例如GPT-4o、GPT-4.5、Ministral）中并不存在。此外，捷径的普遍性随任务复杂度和推理时计算量的增加而上升。在受控训练实验中，外延验证会直接诱发捷径策略，而同构验证则能消除它们。这些结果表明，RLVR不仅可能通过明显的操纵，也可能通过利用验证器未能强制执行的漏洞，来激励奖励破解行为。

As reinforcement Learning with Verifiable Rewards (RLVR) has become the dominant paradigm for scaling reasoning capabilities in LLMs, a new failure mode emerges: LLMs gaming verifiers. We study this phenomenon on inductive reasoning tasks, where models must induce and output logical rules. We find that RLVR-trained models systematically abandon rule induction. Instead of learning generalizable patterns (e.g., ``trains carrying red cars go east’’), they enumerate instance-level labels, producing outputs that pass verifiers without capturing the relational patterns required by the task. We show that this behavior is not a failure of understanding but a form of reward hacking: imperfect verifiers that check only extensional correctness admit false positives. To detect such shortcuts, we introduce Isomorphic Perturbation Testing (IPT), which evaluates a single model output under both extensional and isomorphic verification, where the latter enforces invariance under logically isomorphic tasks. While genuine rule induction remains invariant, shortcut strategies fail. We find that shortcut behavior is specific to RLVR-trained reasoning models (e.g., GPT-5, Olmo3) and absent in non-RLVR models (e.g., GPT-4o, GPT-4.5, Ministral). Moreover, shortcut prevalence increases with task complexity and inference-time compute. In controlled training experiments, extensional verification directly induces shortcut strategies, while isomorphic verification eliminates them. These results show that RLVR can incentivize reward hacking not only through overt manipulation but also by exploiting what the verifier fails to enforce.

评分理由: 论文核心研究大语言模型（LLMs）在检索增强推理（RAG）中的应用，提出了一种基于信息增益的强化学习框架（IG-Search）来改进搜索查询质量。因此与’Large Language Models’和’Retrieval-Augmented Generation’高度相关（10分）。论文在多个QA基准上测试，涉及多步推理任务，因此与’Chain of Thought’高度相关（10分）。其他关键词如MoE、量化、对齐等未在论文中涉及，均为0分。

强化学习已成为训练大语言模型执行搜索增强推理的有效范式。然而，现有方法依赖于轨迹级奖励，无法在同一个 rollout 组内区分精确的搜索查询与模糊或冗余的查询，并且每当所有采样轨迹均失败时，梯度信号会坍缩至接近零。本文提出 IG-Search，一种基于信息增益（Information Gain，IG）引入步骤级奖励的强化学习框架。对于每个搜索步骤，IG 通过衡量检索到的文档相对于随机文档的反事实基线，能在多大程度上提升模型对正确答案的置信度，从而反映底层搜索查询的有效性。该信号通过 GRPO 中的逐词元优势调制反馈给对应的搜索查询词元，实现了在单个 rollout 内进行细粒度的步骤级信用分配。与先前需要外部标注的中间监督或跨轨迹共享环境状态的步骤级方法不同，IG-Search 的信号源自策略自身的生成概率，除标准问答对外无需任何中间标注。在七个单跳和多跳问答基准上的实验表明，IG-Search 使用 Qwen2.5-3B 模型实现了 0.430 的平均 EM 值，在各项基准上平均优于最强的轨迹级基线（MR-Search）1.6 个百分点，优于步骤级方法 GiGPO 0.9 个百分点，在多跳推理任务上提升尤为显著。尽管引入了密集的步骤级信号，IG-Search 相比轨迹级基线仅增加约 6.4% 的单步训练挂钟时间，且推理延迟保持不变，同时即使在所有采样轨迹均回答错误时仍能提供有意义的梯度信号。

Reinforcement learning has emerged as an effective paradigm for training large language models to perform search-augmented reasoning. However, existing approaches rely on trajectory-level rewards that cannot distinguish precise search queries from vague or redundant ones within a rollout group, and collapse to a near-zero gradient signal whenever every sampled trajectory fails. In this paper, we propose IG-Search, a reinforcement learning framework that introduces a step-level reward based on Information Gain (IG). For each search step, IG measures how much the retrieved documents improve the model’s confidence in the gold answer relative to a counterfactual baseline of random documents, thereby reflecting the effectiveness of the underlying search query. This signal is fed back to the corresponding search-query tokens via per-token advantage modulation in GRPO, enabling fine-grained, step-level credit assignment within a rollout. Unlike prior step-level methods that require either externally annotated intermediate supervision or shared environment states across trajectories, IG-Search derives its signals from the policy’s own generation probabilities, requiring no intermediate annotations beyond standard question-answer pairs. Experiments on seven single-hop and multi-hop QA benchmarks demonstrate that IG-Search achieves an average EM of 0.430 with Qwen2.5-3B, outperforming the strongest trajectory-level baseline (MR-Search) by 1.6 points and the step-level method GiGPO by 0.9 points on average across benchmarks, with particularly pronounced gains on multi-hop reasoning tasks. Despite introducing a dense step-level signal, IG-Search adds only ~6.4% to per-step training wall-clock time over the trajectory-level baseline and leaves inference latency unchanged, while still providing a meaningful gradient signal even when every sampled trajectory answers incorrectly.

评分理由: 该论文专注于科学文献新颖性度量的评估基准，属于AI在科学领域的应用范畴，因此与’AI for Science OR Bioinformatics OR Cheminformatics’有一定关联（5分），但论文内容不涉及大模型技术原理、训练方法、推理优化、对齐、代理系统等具体技术，与其他所有关键词完全无关（0分）。

对科学论文新颖性的严格评估，即使对人类科学家而言，也是一项具有挑战性的任务。随着人们对AI科学家以及AI参与科学思想生成和论文写作的兴趣日益增长，使这项任务能够自动化且可靠地完成也变得愈发重要，以免人类注意力和计算资源浪费在已被探索过的想法上。然而，由于量化真实新颖性存在困难，现有的科学论文新颖性度量标准通常通过引用次数或同行评审分数等存在噪声且混杂的信号来验证其结果。这些代理指标可能将新颖性与影响力、质量或评审者偏好混为一谈，从而使得评估特定度量标准实际衡量新颖性的能力变得更加困难。因此，我们提出一个用于科学新颖性度量标准的公理化基准。我们首先基于人类科学规范和实践，定义了一组行为良好的新颖性度量标准应满足的公理，然后在涵盖人工智能研究三个领域的十项任务中评估现有度量标准。我们的结果表明，没有现有度量标准能始终满足所有公理，且不同度量标准在不同公理上存在系统性失败，这反映了它们底层架构的差异。此外，我们发现结合具有互补架构的度量标准能在基准测试中带来持续改进，采用按公理加权的方法达到了90.1%的得分，而最佳单一度量标准仅为71.5%，这表明开发架构多样化的度量标准是未来工作的一个 promising direction（有前景的方向）。我们已将基准代码作为补充材料发布，以鼓励开发更 robust（稳健）的科学文献新颖性度量标准。

The rigorous evaluation of the novelty of a scientific paper is, even for human scientists, a challenging task. With the increasing interest in AI scientists and AI involvement in scientific idea generation and paper writing, it also becomes increasingly important that this task be automatable and reliable, lest both human attention and compute tokens be wasted on ideas that have already been explored. Due to the challenge of quantifying ground-truth novelty, however, existing novelty metrics for scientific papers generally validate their results against noisy, confounded signals such as citation counts or peer review scores. These proxies can conflate novelty with impact, quality, or reviewer preference, which in turn makes it harder to assess how well a given metric actually evaluates novelty. We therefore propose an axiomatic benchmark for scientific novelty metrics. We first define a set of axioms that a well-behaved novelty metric should satisfy, grounded in human scientific norms and practice, then evaluate existing metrics across ten tasks spanning three domains of AI research. Our results reveal that no existing metric satisfies all axioms consistently, and that metrics fail on systematically different axioms, reflecting their underlying architectures. Additionally, we show that combining metrics of complementary architectures leads to consistent improvements on the benchmark, with per-axiom weighting achieving 90.1% versus 71.5% for the best individual metric, suggesting that developing architecturally diverse metrics is a promising direction for future work. We release the benchmark code as supplementary material to encourage the development of more robust scientific literature novelty metrics.

评分理由: 论文研究生物发育过程生成的神经回路结构对学习效率的影响，属于计算神经科学和生物启发AI领域。所有评分关键词均针对大模型/深度学习技术原理或应用，而本文不涉及任何大模型技术（如LLM、MoE、RLHF等），也未使用深度学习框架或方法，而是基于生物发育模拟和传统神经网络训练。因此所有关键词相关度为0。

本研究模拟了大脑皮层神经发生的发育过程，该过程始于单个干细胞，并受源自小鼠单细胞转录组数据的基因调控规则支配。这一发育过程自发产生了由5,000个细胞组成的异质性群体，但仅生成85个成熟神经元——仅占总数的1.7%。这85个神经元形成了一个密集互连的核心网络，包含200,400个突触，相当于每个神经元平均拥有4,715个连接度。在第零次迭代时，这一最小回路在MNIST数据集上的表现处于随机水平。然而，经过一轮标准训练后，其准确率飙升至90%以上——增益超过80个百分点——典型运行结果因发育随机性而分布在89%-94%的区间。同一回路在未经任何架构修改或数据增强的情况下，在CIFAR-10数据集上经过一轮训练后达到了40.53%的准确率。这些发现表明，发育规则塑造了一种领域通用的拓扑基底，该基底特别适合快速学习，这提示生物发育过程本身即为高效计算编码了强大的结构先验。

This work simulates the developmental process of cortical neurogenesis, initiating from a single stem cell and governed by gene regulatory rules derived from mouse single-cell transcriptomic data. The developmental process spontaneously generates a heterogeneous population of 5,000 cells, yet yields only 85 mature neurons - merely 1.7% of the total population. These 85 neurons form a densely interconnected core of 200,400 synapses, corresponding to an average degree of 4,715 per neuron. At iteration zero, this minimal circuit performs at chance level on MNIST. However, after a single epoch of standard training, accuracy surges to over 90% - a gain exceeding 80 percentage points - with typical runs falling in the 89-94% range depending on developmental stochasticity. The identical circuit, without any architectural modification or data augmentation, achieves 40.53% on CIFAR-10 after one epoch. These findings demonstrate that developmental rules sculpt a domain-general topological substrate exceptionally amenable to rapid learning, suggesting that biological developmental processes inherently encode powerful structural priors for efficient computation.

评分理由: 论文研究的是基于向量符号架构（VSA）的顺序关联记忆（SAM）更新机制，提出SRMU方法结合时间衰减和相关性门控来过滤冗余、冲突和过时信息。虽然属于AI/机器学习领域，但论文内容专注于记忆系统、超维计算和流数据处理，与所有评分关键词（主要关于大语言模型、训练技术、推理方法、应用领域等）均无直接关联。论文未涉及语言模型、深度学习技术原理、科学AI应用或任何评分关键词中的具体技术。

在现实世界的流式环境中，顺序关联记忆（Sequential Associative Memories，SAMs）的构建与维护具有挑战性，因为观测数据随时间增量到达、采样不均衡且具有非平稳的时序动态。向量符号架构（Vector Symbolic Architectures，VSAs）为构建SAMs提供了一个受生物学启发的框架。实体与属性被编码为近似正交的高维向量，并通过定义明确的代数运算进行处理。尽管这一框架功能丰富，但大多数VSA系统依赖于简单的累加式更新机制，即重复的观测会强化现有信息，即使并未引入新信息。在非平稳环境中，这会导致底层系统变化后过时信息持续存在。本文中，我们提出了顺序关联记忆单元（Sequential Relevance Memory Unit，SRMU），这是一种适用于基于VSA的SAMs的、与具体领域及清理机制无关的更新规则。SRMU将时间衰减机制与相关性门控机制相结合。与以往仅依赖清理的方法不同，SRMU通过在存储前过滤冗余、冲突和过时的信息来调控记忆形成。我们在分离非均匀采样和非平稳时序动态的流式状态追踪任务上评估了SRMU。实验结果表明，SRMU将记忆相似度提高了$12.6%$，并将累积记忆幅度降低了$53.5%$。这证明SRMU能够实现更稳定的记忆增长，并与真实状态保持更强的一致性。

Sequential associative memories (SAMs) are difficult to build and maintain in real-world streaming environments, where observations arrive incrementally over time, have imbalanced sampling, and non-stationary temporal dynamics. Vector Symbolic Architectures (VSAs) provide a biologically-inspired framework for building SAMs. Entities and attributes are encoded as quasi-orthogonal hyperdimensional vectors and processed with well defined algebraic operations. Despite this rich framework, most VSA systems rely on simple additive updates, where repeated observations reinforce existing information even when no new information is introduced. In non-stationary environments, this leads to the persistence of stale information after the underlying system changes. In this work, we introduce the Sequential Relevance Memory Unit (SRMU), a domain- and cleanup-agnostic update rule for VSA-based SAMs. The SRMU combines temporal decay with a relevance gating mechanism. Unlike prior approaches that solely rely on cleanup, the SRMU regulates memory formation by filtering redundant, conflicting, and stale information before storage. We evaluate the SRMU on streaming state-tracking tasks that isolate non-uniform sampling and non-stationary temporal dynamics. Our results show that the SRMU increases memory similarity by $12.6%$ and reduces cumulative memory magnitude by $53.5%$. This shows that the SRMU produces more stable memory growth and stronger alignment with the ground-truth state.

评分理由: 论文专注于最优传输（Optimal Transport）的算法创新，提出了一种基于切片最优传输势函数的摊销优化方法（RA-OT和OA-OT），用于高效解决多对测度间的传输问题。所有评分关键词均涉及大模型、深度学习技术原理或其科学应用，而本文研究的是数学优化和计算几何领域的经典最优传输问题，未涉及任何形式的大模型、深度学习、AI代理、对齐、推理、压缩等主题，也未应用于生物信息学等科学领域，因此与所有关键词完全无关。

我们提出一种新颖的摊销优化方法，通过利用切片最优传输（sliced OT）导出的康托罗维奇势（Kantorovich potentials），来预测多对测度间的最优传输（OT）方案。我们引入了两种摊销策略：基于回归的摊销（RA-OT）和基于目标的摊销（OA-OT）。在RA-OT中，我们构建了一个函数回归模型，将原始OT问题中的康托罗维奇势作为响应变量，将从切片OT获得的势作为预测变量，并通过最小二乘法估计这些模型。在OA-OT中，我们通过优化康托罗维奇对偶目标来估计函数模型的参数。在两种方法中，预测的OT方案随后从估计的势中恢复。作为摊销OT方法，RA-OT和OA-OT均能通过复用从先前实例中学到的信息来快速近似新解，从而高效解决不同测度对间的重复OT问题。此外，通过利用切片OT提供的结构，所提出的模型更为简洁，且独立于测度的具体结构（例如离散情况下的原子数量），同时保持高精度。我们在多个任务上验证了方法的有效性，包括MNIST数字传输、颜色迁移、球面数据上的供需传输以及小批量OT条件流匹配。

We propose a novel amortized optimization method for predicting optimal transport (OT) plans across multiple pairs of measures by leveraging Kantorovich potentials derived from sliced OT. We introduce two amortization strategies: regression-based amortization (RA-OT) and objective-based amortization (OA-OT). In RA-OT, we formulate a functional regression model that treats Kantorovich potentials from the original OT problem as responses and those obtained from sliced OT as predictors, and estimate these models via least-squares methods. In OA-OT, we estimate the parameters of the functional model by optimizing the Kantorovich dual objective. In both approaches, the predicted OT plan is subsequently recovered from the estimated potentials. As amortized OT methods, both RA-OT and OA-OT enable efficient solutions to repeated OT problems across different measure pairs by reusing information learned from prior instances to rapidly approximate new solutions. Moreover, by exploiting the structure provided by sliced OT, the proposed models are more parsimonious, independent of specific structures of the measures, such as the number of atoms in the discrete case, while achieving high accuracy. We demonstrate the effectiveness of our approaches on tasks including MNIST digit transport, color transfer, supply-demand transportation on spherical data, and mini-batch OT conditional flow matching.

评分理由: 该论文专注于超维表示中的向量符号架构（VSA）框架开发，研究内容为超维计算的基础表示框架和系统级评估，与所有评分关键词（均围绕大模型、深度学习技术及其应用）无直接关联。论文未涉及任何大模型技术、训练方法、推理优化、对齐技术、代理系统或科学AI应用，因此所有关键词相关度均为0。

向量符号架构（Vector Symbolic Architectures, VSAs）为高维空间中的组合表示提供了一个定义明确的代数框架。本文介绍HyperSpace，一个开源框架，它将VSA系统分解为编码、绑定、聚合、相似性、清理和回归等模块化算子。借助HyperSpace，我们分析与基准测试了两个具有代表性的VSA后端：全息缩减表示（Holographic Reduced Representations, HRR）与傅里叶全息缩减表示（Fourier Holographic Reduced Representations, FHRR）。尽管FHRR在单个操作上具有较低的理论复杂度，但HyperSpace的模块化设计揭示了在空间域中相似性与清理操作主导了实际运行时间。因此，HRR与FHRR展现出相近的端到端性能。内存占用上的差异引入了额外的部署权衡：HRR向量所需内存约为FHRR的一半。通过支持模块化的系统级评估，HyperSpace揭示了VSA流程中那些仅凭理论或算子级比较难以察觉的实际权衡。

Vector Symbolic Architectures (VSAs) provide a well-defined algebraic framework for compositional representations in hyperdimensional spaces. We introduce HyperSpace, an open-source framework that decomposes VSA systems into modular operators for encoding, binding, bundling, similarity, cleanup, and regression. Using HyperSpace, we analyze and benchmark two representative VSA backends: Holographic Reduced Representations (HRR) and Fourier Holographic Reduced Representations (FHRR). Although FHRR provides lower theoretical complexity for individual operations, HyperSpaces modularity reveals that similarity and cleanup dominate runtime in spatial domains. As a result, HRR and FHRR exhibit comparable end-to-end performance. Differences in memory footprint introduce additional deployment trade-offs where HRR requires approximately half the memory of FHRR vectors. By enabling modular, system-level evaluation, HyperSpace reveals practical trade-offs in VSA pipelines that are not apparent from theoretical or operator-level comparisons alone.

评分理由: 论文核心研究LLM在长文本生成中的不确定性量化问题，与’Large Language Models’高度相关（10分），因为全文围绕LLM展开；与’Hallucination Mitigation’有一定关联（5分），因为不确定性量化涉及事实准确性评估；其他关键词如MoE、SFT、RAG等均未在摘要中提及，故评0分。

尽管大语言模型（LLM）发展迅速，但其生成内容的不确定性量化仍是一个持续存在的挑战。尽管现有方法通过限制LLM生成简短或受限的答案集已取得显著成效，但许多实际应用需要长文本和自由形式的文本生成。在此场景下的一个核心难点在于，LLM生成的回答往往语义连贯但事实准确性不足，同时其底层语义具有多面性且语言结构复杂。为应对这一挑战，本文提出了询问式不确定性量化（Interrogative Uncertainty Quantification, IUQ）——一种新颖的框架，该框架利用样本间一致性与样本内忠实度来量化长文本LLM输出的不确定性。通过采用“先询问后回答”的范式，我们的方法能够提供可靠的主张级不确定性度量以及模型的忠实度评估。在不同模型系列和模型规模上的实验结果表明，IUQ在两个广泛使用的长文本生成数据集上均优于现有方法。相关代码已发布于 https://github.com/louisfanhz/IUQ。

Despite the rapid advancement of Large Language Models (LLMs), uncertainty quantification in LLM generation is a persistent challenge. Although recent approaches have achieved strong performance by restricting LLMs to produce short or constrained answer sets, many real-world applications require long-form and free-form text generation. A key difficulty in this setting is that LLMs often produce responses that are semantically coherent yet factually inaccurate, while the underlying semantics are multifaceted and the linguistic structure is complex. To tackle this challenge, this paper introduces Interrogative Uncertainty Quantification (IUQ), a novel framework that leverages inter-sample consistency and intra-sample faithfulness to quantify the uncertainty in long-form LLM outputs. By utilizing an interrogate-then-respond paradigm, our method provides reliable measures of claim-level uncertainty and the model’s faithfulness. Experimental results across diverse model families and model sizes demonstrate the superior performance of IUQ over two widely used long-form generation datasets. The code is available at https://github.com/louisfanhz/IUQ.

评分理由: 论文主要研究移动智能体框架，与’LLM Agents’高度相关（10分），因为论文明确研究移动智能体（mobile agents）并基于视觉语言模型构建。与’Post-training/SFT’有一定关联（8分），因为论文提到对Qwen2.5-VL和Qwen3-VL进行微调（fine-tuned）。与’Large Language Models’有中等关联（5分），因为论文使用视觉语言模型（vision-language models）作为基础，但未深入探讨LLM技术原理。其他关键词如MoE、Scaling Laws、RLHF等与论文内容无关（0分），论文未涉及这些具体技术。

由视觉语言模型驱动的移动智能体在自动化移动任务方面展现出卓越能力，近期领先模型实现了显著的性能飞跃，例如在AndroidWorld基准测试中达到近70%的成功率。然而，这些系统通常未公开训练数据，其任务与轨迹合成方法也缺乏透明度。我们提出OpenMobile——一个开源框架，能够合成高质量的任务指令与智能体轨迹，其包含两个核心组件：（1）首个可扩展的任务合成流程，通过环境探索构建全局环境记忆，并利用该记忆生成多样化且基于实际场景的指令；（2）用于轨迹生成的策略切换机制。通过在学习者模型与专家模型间交替执行，该机制能捕获标准模仿学习中常缺失的关键错误恢复数据。基于本框架数据训练的智能体在三个动态移动智能体基准测试中均取得具有竞争力的结果：值得注意的是，经我们微调的Qwen2.5-VL与Qwen3-VL模型在AndroidWorld上分别达到51.7%和64.7%的成功率，远超现有开源数据方案。此外，我们对合成指令与基准测试集的重叠度进行了透明化分析，证实性能提升源于广泛的功能覆盖而非对基准测试的过拟合。我们在https://njucckevin.github.io/openmobile/ 公开数据与代码，以弥合数据鸿沟并推动更广泛的移动智能体研究。

Mobile agents powered by vision-language models have demonstrated impressive capabilities in automating mobile tasks, with recent leading models achieving a marked performance leap, e.g., nearly 70% success on AndroidWorld. However, these systems keep their training data closed and remain opaque about their task and trajectory synthesis recipes. We present OpenMobile, an open-source framework that synthesizes high-quality task instructions and agent trajectories, with two key components: (1) The first is a scalable task synthesis pipeline that constructs a global environment memory from exploration, then leverages it to generate diverse and grounded instructions. and (2) a policy-switching strategy for trajectory rollout. By alternating between learner and expert models, it captures essential error-recovery data often missing in standard imitation learning. Agents trained on our data achieve competitive results across three dynamic mobile agent benchmarks: notably, our fine-tuned Qwen2.5-VL and Qwen3-VL reach 51.7% and 64.7% on AndroidWorld, far surpassing existing open-data approaches. Furthermore, we conduct transparent analyses on the overlap between our synthetic instructions and benchmark test sets, and verify that performance gains stem from broad functionality coverage rather than benchmark overfitting. We release data and code at https://njucckevin.github.io/openmobile/ to bridge the data gap and facilitate broader mobile agent research.

评分理由: 论文核心是使用LLM代理团队自主改进EDA工具ABC的源代码，属于大模型在科学计算领域的创新应用。高度相关的关键词包括：LLMs（论文明确使用）、Self-Improvement（系统自我进化）、LLM Agents（基于LLM的代理团队）、Multi-agent Systems（多代理协调）、AI for Science（EDA属于科学计算领域）。其他关键词如MoE、SLMs、训练方法、推理优化等均未涉及。

本文首次提出了一种自进化逻辑综合框架，该框架利用大型语言模型（LLM）智能体，对广泛采用的逻辑综合系统 ABC 的源代码进行自主改进。我们的框架在整个集成的ABC代码库上运行，其输出仓库保留了原有的单一二进制执行模型与命令接口。在初始进化周期中，我们利用已有的开源综合组件（涵盖流程调优、逻辑最小化与工艺映射）对系统进行引导，但未手动注入任何新的启发式方法。在此基础上，一组基于LLM的智能体在统一的正确性与结果质量（QoR）驱动的评估循环中，遵循我们设计的“编程指导”提示，迭代式地重写并进化ABC的特定子模块。每个进化周期会提出代码修改方案，编译集成后的二进制文件，验证其正确性，并在多套基准测试集（包括ISCAS 85/89/99、VTR、EPFL和IWLS 2005）上评估结果质量。通过持续的反馈，该系统发现了超越人工设计启发式方法的优化方案，有效地学习到新的综合策略，从而提升了结果质量。我们详细阐述了这一自我改进系统的架构、其与ABC的集成方式，以及实验结果。结果表明，该框架能够在完整的百万行代码规模上，自主且渐进地改进EDA工具。

This paper introduces the first \emph{self-evolving} logic synthesis framework, which leverages Large Language Model (LLM) agents to autonomously improve the source code of \textsc{ABC}, the widely adopted logic synthesis system. Our framework operates on the \emph{entire integrated ABC codebase}, and the output repository preserves its single-binary execution model and command interface. In the initial evolution cycle, we bootstrap the system using existing prior open-source synthesis components, covering flow tuning, logic minimization, and technology mapping, but without manually injecting new heuristics. On top of this foundation, a team of LLM-based agents iteratively rewrites and evolves specific sub-components of ABC following our programming guidance prompts under a unified correctness and QoR-driven evaluation loop. Each evolution cycle proposes code modifications, compiles the integrated binary, validates correctness, and evaluates quality-of-results (QoR) on \emph{multi-suite benchmarks including ISCAS85/89/99, VTR, EPFL, and IWLS2005}. Through continuous feedback, the system discovers optimizations beyond human-designed heuristics, effectively \emph{learning new synthesis strategies} that enhance QoR. We detail the architecture of this self-improving system, its integration with \textsc{ABC}, and results demonstrating that the framework can autonomously and progressively improve EDA tool at full million-line scale.

评分理由: 论文聚焦于多智能体AI系统的公平性研究，属于AI伦理与治理范畴，而非大模型技术原理或具体应用创新。仅与’Multi-agent Systems OR Agent Coordination’高度相关（10分），与’LLM Agents OR Autonomous Agents OR Agentic Workflow’有一定关联（5分），因为论文讨论的MAAI系统可能包含LLM驱动的智能体，但未明确涉及LLM技术细节。其余关键词均未在标题或摘要中提及，且论文未涉及大模型技术、训练方法、推理优化、科学应用等具体技术内容，故评分为0。

生成式人工智能的快速发展催生了日益复杂的多智能体人工智能系统。尽管人工智能公平性在传统预测场景中已得到广泛研究，但在多智能体人工智能领域的探讨仍处于起步且碎片化阶段。本文通过范围综述方法，对多智能体人工智能系统公平性研究进行了批判性整合。通过对23项入选研究的定性内容分析，我们识别出五种典型研究路径。研究发现，现有研究对多智能体人工智能系统公平性的探讨往往流于表面，缺乏坚实的规范理论基础，且普遍忽视了智能体自主性与系统级交互带来的复杂动态。我们认为，公平性必须结构性嵌入多智能体人工智能系统的全生命周期开发过程，而非作为事后补充。有效的公平性评估需要明确的人类监督机制、清晰的规范框架，以及对公平目标与受益群体的精准界定。本综述通过揭示关键研究空白、剖析现有局限并提出发展路径，为推进多智能体人工智能系统公平性研究奠定了基础。

Rapid advances in Generative AI are giving rise to increasingly sophisticated Multi-Agent AI (MAAI) systems. While AI fairness has been extensively studied in traditional predictive scenarios, its examination in MAAI remains nascent and fragmented. This scoping review critically synthesizes existing research on fairness in MAAI systems. Through a qualitative content analysis of 23 selected studies, we identify five archetypal approaches. Our findings reveal that fairness in MAAI systems is often addressed superficially, lacks robust normative foundations, and frequently overlooks the complex dynamics introduced by agent autonomy and system-level interactions. We argue that fairness must be embedded structurally throughout the development lifecycle of MAAI, rather than appended as a post-hoc consideration. Meaningful evaluation requires explicit human oversight, normative clarity, and a precise articulation of fairness objectives and beneficiaries. This review provides a foundation for advancing fairness research in MAAI systems by highlighting critical gaps, exposing prevailing limitations, and suggesting pathways.

评分理由: 该论文研究的是基于神经进化算法（NEAT）和生物启发的导航细胞（如位置细胞、网格细胞）的机器人路径规划方法，属于机器人学和进化计算领域。论文中完全没有涉及大语言模型（LLM）、深度学习技术原理、模型训练方法（如预训练、微调、对齐）、推理优化、智能体系统或AI for Science等关键词。所有关键词均与大模型和深度学习相关，而本文的核心是进化算法和生物启发式机器人导航，因此所有关键词的相关度均为0。

为在空间中导航，大脑利用位置细胞、网格细胞、头部方向细胞、边界细胞和速度细胞等多种细胞构建环境的内部表征。所有这些细胞与感觉输入共同作用，使生物体能够探索其周围空间。受这些生物学原理启发，我们开发了NEATNC——一种基于增强拓扑神经进化的导航细胞引导系统。本文的目标是利用空间认知细胞，改进NEAT算法在动态环境路径规划中的性能。该方法以导航细胞作为输入，通过演化循环神经网络来模拟大脑海马体功能。我们在静态与动态等多种场景中对所提算法进行了性能评估。本研究凸显了NEAT算法对复杂多变环境的适应能力，展现了生物学理论的实际应用价值。这表明我们的方法非常适用于机器人及游戏领域的实时动态路径规划。

To navigate a space, the brain makes an internal representation of the environment using different cells such as place cells, grid cells, head direction cells, border cells, and speed cells. All these cells, along with sensory inputs, enable an organism to explore the space around it. Inspired by these biological principles, we developed NEATNC, a Neuro-Evolution of Augmenting Topology guided Navigation Cells. The goal of the paper is to improve NEAT algorithm performance in path planning in dynamic environments using spatial cognitive cells. This approach uses navigation cells as inputs and evolves recurrent neural networks, representing the hippocampus part of the brain. The performance of the proposed algorithm is evaluated in different static and dynamic scenarios. This study highlights NEAT’s adaptability to complex and different environments, showcasing the utility of biological theories. This suggests that our approach is well-suited for real-time dynamic path planning for robotics and games.

评分理由: 论文研究联邦学习中的梯度反演攻击，提出了一种可验证的攻击方法（VGIA），专注于隐私安全攻击技术，与所有评分关键词（均涉及大模型/深度学习技术原理、应用或科学AI）完全无关。

梯度反演攻击通过从客户端共享的梯度中重构训练样本，威胁着联邦学习中的客户隐私。梯度聚合了来自多条记录的贡献，而现有攻击可能无法将其分离，从而产生不正确的重构结果，且缺乏内在方法来验证成功与否。在视觉和语言领域，攻击者可依赖人工检查来判断重构的合理性，但这对于数值型表格记录而言可行性极低，这助长了表格数据不易受攻击的印象。

我们通过提出一种可验证的梯度反演攻击（VGIA）来挑战这一观点，该方法能为重构样本提供明确的正确性证明。我们的方法采用ReLU泄漏的几何视角：全连接层的激活边界在输入空间中定义了一个超平面。VGIA引入了一种基于代数的子空间验证测试，用于检测超平面划定的区域何时恰好包含一条记录。一旦隔离被验证，VGIA通过解析方式恢复相应的特征向量，并通过轻量级优化步骤重构目标。

在批量较大的表格基准测试中，实验表明VGIA实现了精确的记录和目标恢复，而在相同条件下，现有最先进的攻击要么失败，要么无法评估重构的保真度。与先前的几何方法相比，VGIA更有效地分配超平面查询，从而以更少的攻击轮数实现更快的重构。

Gradient inversion attacks threaten client privacy in federated learning by reconstructing training samples from clients’ shared gradients. Gradients aggregate contributions from multiple records and existing attacks may fail to disentangle them, yielding incorrect reconstructions with no intrinsic way to certify success. In vision and language, attackers may fall back on human inspection to judge reconstruction plausibility, but this is far less feasible for numerical tabular records, fueling the impression that tabular data is less vulnerable. We challenge this perception by proposing a verifiable gradient inversion attack (VGIA) that provides an explicit certificate of correctness for reconstructed samples. Our method adopts a geometric view of ReLU leakage: the activation boundary of a fully connected layer defines a hyperplane in input space. VGIA introduces an algebraic, subspace-based verification test that detects when a hyperplane-delimited region contains exactly one record. Once isolation is certified, VGIA recovers the corresponding feature vector analytically and reconstructs the target via a lightweight optimization step. Experiments on tabular benchmarks with large batch sizes demonstrate exact record and target recovery in regimes where existing state-of-the-art attacks either fail or cannot assess reconstruction fidelity. Compared to prior geometric approaches, VGIA allocates hyperplane queries more effectively, yielding faster reconstructions with fewer attack rounds.

评分理由: 该论文主要介绍了一个用于多智能体实验的框架工具（CoGrid和MUG），旨在降低人机交互多智能体实验的研究门槛。论文内容与绝大多数关键词（涉及大模型技术原理、训练方法、推理优化、对齐、科学应用等）完全无关。唯一相关的关键词是’Multi-agent Systems OR Agent Coordination’，因为论文的核心就是为多智能体实验（特别是人机交互场景）提供工具支持，因此该关键词评分为10分（高度相关）。其他关键词均未在论文标题或摘要中提及或暗示，评分为0分。

人工智能（AI）在日常生活中的日益融合，为人类如何与自主智能体互动带来了新的挑战与问题。人类与AI共同参与的多智能体实验，为研究社会决策过程提供了重要契机，但目前研究人员缺乏易于使用的工具来开展此类实验。我们推出了两款旨在降低此类研究门槛的工具。其一是CoGrid，这是一个基于网格的多智能体模拟库，提供NumPy和JAX双后端支持。其二是多用户Gymnasium（Multi-User Gymnasium，简称MUG），该工具可将此类模拟环境直接转化为基于网页的交互式实验平台。MUG支持任意数量的人类与AI参与交互，可采用服务器权威模式或具备回滚网络代码（rollback netcode）的对等网络架构以应对网络延迟。这两款工具共同助力研究人员开展人机交互研究，促进对心理学、认知与决策核心问题及其与人机交互关系的探索。两款工具均已开源，可供更广泛的研究社区使用。文档与源代码发布于{cogrid, multi-user-gymnasium}.readthedocs.io。本文详细阐述了这些工具的功能，并通过多个案例研究展示了其在人机多智能体实验中的应用价值。

The increasing integration of artificial intelligence (AI) in everyday life brings with it new challenges and questions for regarding how humans interact with autonomous agents. Multi-agent experiments, where humans and AI act together, can offer important opportunities to study social decision making, but there is a lack of accessible tooling available to researchers to run such experiments. We introduce two tools designed to reduce these barriers. The first, CoGrid, is a multi-agent grid-based simulation library with dual NumPy and JAX backends. The second, Multi-User Gymnasium (MUG), translates such simulation environments directly into interactive web-based experiments. MUG supports interactions with arbitrary numbers of humans and AI, utilizing either server-authoritative or peer-to-peer networking with rollback netcode to account for latency. Together, these tools can enable researchers to deploy studies of human-AI interaction, facilitating inquiry into core questions of psychology, cognition, and decision making and their relationship to human-AI interaction. Both tools are open source and available to the broader research community. Documentation and source code is available at {cogrid, multi-user-gymnasium}.readthedocs.io. This paper details the functionality of these tools and presents several case studies to illustrate their utility in human-AI multi-agent experimentation.