StableRCA 研究提出鲁棒图无关机制级根因分析方法

由厦门大学林晓宇等人提出的StableRCA方法，致力于解决系统故障排查中的根因分析问题。该方法创新性地引入图无关机制，无需依赖系统拓扑图即可实现鲁棒性根因定位。研究指出，现有根因分析方法过度依赖监控图结构，易因图数据扰动导致性能下降。StableRCA通过自注意力机制提取跨维度特征，并采用对比学习增强表征鲁棒性。在多个真实数据集上的实验表明，该方法在准确性和稳定性上均优于现有基线，尤其在动态变化环境中表现突出。该研究为复杂系统运维提供了新的自动化诊断方案。 #根因分析 #故障诊断 #机器学习 #系统运维 #StableRCA #鲁棒性 #研究前沿

跨时代自适应滚动优化

一篇题为《Cross-Epoch Adaptive Rollout Optimization for RL Post-Training》的学术论文于2026年6月4日提交至arXiv预印本平台。该论文由Yiming Zong等人撰写，提出了一种针对强化学习后训练阶段的跨时代自适应滚动优化方法。该方法旨在通过动态调整滚动策略，提升强化学习模型在训练后期阶段的效率与稳定性，为相关领域研究提供了新的技术思路。论文目前提供PDF、HTML及TeX源码等多种格式供研究者查阅。 #强化学习 #后训练 #自适应优化 #arXiv #机器学习 #AI #论文

从预测到自我

arXiv上近日发布了一篇题为“From Prediction to Self: Developmental Conditions for Agency in Minimal Neural Systems”的论文，作者为Evan Ye。该研究聚焦于极简神经系统中能动性的形成机制，提出个体从纯粹的预测性学习逐步发展出自我意识与自主行为的关键条件。论文通过构建最小神经模型，模拟了系统在最小化预测误差过程中如何动态调整内部表征，进而涌现出类似“自我”的代理能力。这一理论框架融合了预测编码与发育神经科学，为理解人工智能的自主性起源以及生物智能中自我意识的演化提供了新的思路。研究者认为，能动性并非预先给定，而是系统在与环境交互中自发构建的产物，相关结论对通用人工智能和认知科学领域具有重要参考价值。 #论文 #人工智能 #神经科学 #自我意识 #能动性 #预测编码 #认知科学 #arXiv

AsyncWebRL：面向视觉网页代理的高效多步强化学习新方法

来自arXiv的最新论文提出AsyncWebRL，一种针对视觉网页代理的高效多步强化学习框架。该方法通过异步训练与多步奖励设计，显著提升了智能体在复杂网页导航任务中的学习效率和成功率。实验表明，AsyncWebRL在多个基准测试中优于现有方法，为自动化网页操作和智能助手领域提供了新的技术路径。 #AI #强化学习 #网页代理 #异步训练 #arXiv #论文 #机器学习

CLaaS: Continual learning as a service for sample efficient online learning

Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Kion Fallah [ view email ] [v1] Thu, 4 Jun 2026 01:14:19 UTC (2,990 KB) Full-text links: Access Paper: View a PDF of the paper titled CLaaS: Continual learning as a service for sample efficient online learning, by Kion Fallah and 3 other authors View PDF HTML (experimental) TeX Source view license Current browse context: < prev | next > new | recent | 2026-06 Change to browse by: cs References & Citations NASA ADS Google Scholar Semantic Scholar export BibTeX citation Loading... BibTeX formatted citation × loading... Data provided by: Bookmark Bibliographic Tools Bibliographic and Citation Tools Bibliographic Explorer Toggle Bibliographic Explorer ( What is the Explorer? ) Connected Papers Toggle Co

nnected Papers ( What is Connected Papers? ) Litmaps Toggle Litmaps ( What is Litmaps? ) Toggle scite Smart Citations ( What are Smart Citations? ) Code, Data, Media Code, Data and Media Associated with this Article alphaXiv Toggle alphaXiv ( What is alphaXiv? ) Links to Code Toggle CatalyzeX Code Finder for Papers ( What is CatalyzeX? ) DagsHub Toggle DagsHub ( What is DagsHub? ) GotitPub Toggle ( What is GotitPub? ) Huggingface Toggle Hugging Face ( What is Huggingface? ) ScienceCast Toggle ScienceCast ( What is ScienceCast? ) Demos Demos Replicate Toggle Replicate ( What is Replicate? ) Spaces Toggle Hugging Face Spaces ( What is Spaces? ) Spaces Toggle ( What is ? ) Related Papers Recommenders and Search Tools Link to Influence Flower Influence Flower ( What are Influence Flowers? ) Core recommender toggle CORE Recommender ( What is CORE? ) IArxiv recommender toggle IArxiv Recommender ( What is IArxiv? ) Author Venue Institution Topic About arXivLabs arXivLabs: experimental projects with community collaborators arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs .

自回归扩散世界模型

一篇来自arXiv的论文提出了名为“自回归扩散世界模型”的新框架，用于对大型语言模型（LLM）智能体进行离线策略评估。该研究由Kaixuan Liu等人完成，于2026年6月4日提交。传统的离线评估方法在模拟环境和真实世界之间往往存在偏差，而该模型通过结合自回归生成与扩散过程，能够更准确地模拟智能体的行为轨迹和决策后果，从而提升评估的可靠性与泛化能力。这一方法有望加速LLM智能体在机器人控制、对话系统等领域的部署与优化，同时降低对在线试错的依赖。 #arXiv #论文 #自回归扩散 #世界模型 #LLM #离线策略评估 #AI安全 #智能体 #机器学习

Telegram必备的搜索引擎，极搜JISOU帮你精准找到，想要的群组、频道、视频、音乐

👉 t.me/jisou2?start=a_8247614025

多分辨率ConvLSTM框架用于挡土墙变形预测的现场验证

一项来自韩国学者的研究提出并验证了基于多分辨率卷积长短期记忆网络（ConvLSTM）的框架，用于预测挡土墙变形。该框架结合了多尺度空间特征与时间序列建模，能够更精准地捕捉土体位移的演化规律。研究团队利用实际工程现场监测数据进行了验证，结果表明该模型在预测精度和鲁棒性上优于传统方法，为岩土工程安全监测提供了新的智能化手段。相关论文已提交至arXiv预印本平台。 #挡土墙 #变形预测 #深度学习 #ConvLSTM #现场验证 #岩土工程 #AI #土木工程 #时间序列

Representation Learning Enables Scalable Multitask Deep Reinforcement Learning

Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Johan Obando-Ceron [ view email ] [v1] Thu, 4 Jun 2026 01:09:20 UTC (1,933 KB) Full-text links: Access Paper: View a PDF of the paper titled Representation Learning Enables Scalable Multitask Deep Reinforcement Learning, by Johan Obando-Ceron and 5 other authors View PDF HTML (experimental) TeX Source view license Current browse context: < prev | next > new | recent | 2026-06 Change to browse by: cs References & Citations NASA ADS Google Scholar Semantic Scholar export BibTeX citation Loading... BibTeX formatted citation × loading... Data provided by: Bookmark Bibliographic Tools Bibliographic and Citation Tools Bibliographic Explorer Toggle Bibliographic Explorer ( What is the Explorer? ) Connec

ted Papers Toggle Connected Papers ( What is Connected Papers? ) Litmaps Toggle Litmaps ( What is Litmaps? ) Toggle scite Smart Citations ( What are Smart Citations? ) Code, Data, Media Code, Data and Media Associated with this Article alphaXiv Toggle alphaXiv ( What is alphaXiv? ) Links to Code Toggle CatalyzeX Code Finder for Papers ( What is CatalyzeX? ) DagsHub Toggle DagsHub ( What is DagsHub? ) GotitPub Toggle ( What is GotitPub? ) Huggingface Toggle Hugging Face ( What is Huggingface? ) ScienceCast Toggle ScienceCast ( What is ScienceCast? ) Demos Demos Replicate Toggle Replicate ( What is Replicate? ) Spaces Toggle Hugging Face Spaces ( What is Spaces? ) Spaces Toggle ( What is ? ) Related Papers Recommenders and Search Tools Link to Influence Flower Influence Flower ( What are Influence Flowers? ) Core recommender toggle CORE Recommender ( What is CORE? ) IArxiv recommender toggle IArxiv Recommender ( What is IArxiv? ) Author Venue Institution Topic About arXivLabs arXivLabs: experimental projects with community collaborators arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs .

平衡图像压缩与生成

来自arXiv的一篇论文提出了一种名为“引导式分词”的新方法，旨在平衡图像压缩与生成任务。该方法通过自举（bootstrapping）机制优化分词器，使模型在压缩图像信息的同时保持高质量生成能力。研究团队由Haozhe Chi等六位作者组成，论文详细阐述了该技术如何在不牺牲生成质量的前提下提升压缩效率，为图像处理领域提供了新的思路。目前该论文已提交至arXiv，并可通过DOI获取。 #图像压缩 #图像生成 #引导式分词 #机器学习 #计算机视觉 #arXiv #论文

少即是多

来自 Haoze He 等学者的最新研究论文《Less is MoE: Trimming Experts in Domain-Specialist Language Models》提出，在混合专家模型（MoE）中，针对特定领域任务，许多专家模块可能存在冗余。该研究通过系统性地修剪不必要的专家，在不显著降低模型性能的前提下，大幅减少了模型参数量和推理成本。实验表明，经过修剪后的领域专家语言模型在保持甚至提升任务准确率的同时，计算效率得到显著提升。这一发现挑战了“更多专家更强”的传统认知，为构建轻量化、高效率的领域专用语言模型提供了新思路。 #机器学习 #MoE #专家修剪 #大模型 #自然语言处理 #效率优化 #AI #论文 #研究

What Objects Enable, Not What They Are: Functional Latent Spaces for Affordance Reasoning

Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Neel P. Bhatt [ view email ] [v1] Thu, 4 Jun 2026 00:26:04 UTC (1,147 KB) Full-text links: Access Paper: View a PDF of the paper titled What Objects Enable, Not What They Are: Functional Latent Spaces for Affordance Reasoning, by Rohan Siva and 8 other authors View PDF HTML (experimental) TeX Source view license Current browse context: < prev | next > new | recent | 2026-06 Change to browse by: cs References & Citations NASA ADS Google Scholar Semantic Scholar export BibTeX citation Loading... BibTeX formatted citation × loading... Data provided by: Bookmark Bibliographic Tools Bibliographic and Citation Tools Bibliographic Explorer Toggle Bibliographic Explorer ( What is the Explorer?

) Connected Papers Toggle Connected Papers ( What is Connected Papers? ) Litmaps Toggle Litmaps ( What is Litmaps? ) Toggle scite Smart Citations ( What are Smart Citations? ) Code, Data, Media Code, Data and Media Associated with this Article alphaXiv Toggle alphaXiv ( What is alphaXiv? ) Links to Code Toggle CatalyzeX Code Finder for Papers ( What is CatalyzeX? ) DagsHub Toggle DagsHub ( What is DagsHub? ) GotitPub Toggle ( What is GotitPub? ) Huggingface Toggle Hugging Face ( What is Huggingface? ) ScienceCast Toggle ScienceCast ( What is ScienceCast? ) Demos Demos Replicate Toggle Replicate ( What is Replicate? ) Spaces Toggle Hugging Face Spaces ( What is Spaces? ) Spaces Toggle ( What is ? ) Related Papers Recommenders and Search Tools Link to Influence Flower Influence Flower ( What are Influence Flowers? ) Core recommender toggle CORE Recommender ( What is CORE? ) IArxiv recommender toggle IArxiv Recommender ( What is IArxiv? ) Author Venue Institution Topic About arXivLabs arXivLabs: experimental projects with community collaborators arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs .

主导层ZO

一篇来自arXiv的论文揭示了在大语言模型零阶微调过程中，单个网络层对整体性能起主导作用。研究团队发现，通过仅优化这一“主导层”，即可达到与全模型微调相当的效果，大幅降低计算开销。该发现为资源受限场景下的模型适配提供了新思路，有望推动大模型在边缘设备上的高效部署。 #大语言模型 #零阶优化 #微调 #AI #机器学习 #arXiv论文 #高效部署

LEVANTE-bench：用认知任务多尺度比较视觉语言模型与儿童

研究人员提出了一项名为LEVANTE-bench的新基准，旨在通过认知任务在多个尺度上比较视觉语言模型（VLM）与儿童的表现。该基准的副标题“你的VLM比五年级学生聪明吗？”揭示了其核心问题。论文由Alvin Wei Ming Tan等五位作者撰写，目前已在arXiv上提交，等待DOI注册。LEVANTE-bench通过设计适合儿童认知水平的任务，系统评估VLM在感知、推理、语言理解等方面的能力，并与真实儿童数据进行对比。这一工作为理解当前视觉语言模型的认知局限性提供了新视角，也为未来模型改进和儿童认知研究搭建了桥梁。 #视觉语言模型 #认知科学 #儿童发展 #基准测试 #AI评估 #arXiv #机器学习

微软申请专利：Windows将内置AI数据收集控制开关

微软近日申请一项系统级隐私专利，计划在Windows操作系统中内置AI数据收集控制开关。该专利描述了一个类似摄像头指示灯的任务栏图标，实时显示AI数据收集状态，并提供管理界面让用户设置规则（如禁止在特定时段收集）、删除已收集数据或临时暂停。核心是一个运行在Windows内的“ML数据收集协调器”，作为用户与机器学习服务之间的中间人。此举直接回应了Windows Recall等AI功能引发的隐私争议，也顺应了欧盟AI法案等监管要求。若该功能落地，用户将能在一个统一位置控制所有Windows AI功能的数据收集，无需逐个应用查找设置。 #微软 #Windows #AI #隐私 #专利 #数据收集 #系统级控制 #Recall #欧盟AI法案

学习子空间压缩实现通讯高效的流水线并行

近日，一篇题为“Learned Subspace Compression for Communication-Efficient Pipeline Parallelism”的论文被提交至arXiv预印本平台。该论文由Paul Janson等人撰写，针对大规模深度学习模型训练中流水线并行存在的通信瓶颈问题，提出了一种基于学习的子空间压缩方法。该方法能够自动学习模型中间表示的潜在低维子空间，在节点间传输前进行高效压缩，从而显著减少通信数据量，提升并行训练的可扩展性。论文还探讨了该方法与现有梯度压缩技术的兼容性，为未来高效分布式训练系统设计提供了新思路。 #论文 #arXiv #分布式训练 #流水线并行 #通信压缩 #深度学习 #机器学习