论文精选 82
Beyond Scores: Understanding LLM-as-a-Judge Mechanisms in Summarization Evaluation· 超越评分：理解大语言模型裁判机制
LLM-based evaluators of natural language generation (NLG) quality are widely deployed as s…
领域：cs.CL作者：Himil Vasava、Ming Jiang
📰 AI 研究简报 · 9月3日
今日头条：OpenAI’s new reasoning technique alarms AI safety experts —— OpenAI’s new Astra model will use “recurrent depth,” a technique that allows the model to operate outside of the sequential thinking that characterizes most reasoning models.
共 68 条预计阅读 41 分钟Token 消耗 122,400
arXiv 最新论文
论文精选 81
CordisBench: Can Language Models Reason About Component Lifecycles in Dynamic Agent Harnesses?· CordisBench：语言模型能否推理动态代理生命周期？
Dynamic agent harnesses let language models change the software that shapes their own exec…
领域：cs.CL作者：Damien Sileo、Dimitri Kachler
论文精选 82
The Rise of Verbal Reinforcement Learning· 口头强化学习的兴起
Natural language is emerging as a primary feedback channel for improving language agents, …
领域：cs.CL作者：Kshitij Tayal、Arun Sharma、Genta Indra Winata
论文精选 84
StudentSim: Training LLM-based Student Simulators· StudentSim: 训练基于 LLM 的学生模拟器
AI tutors are most useful when they adapt to each student's strengths, weaknesses, and pre…
领域：cs.CL作者：Ke Yang、Chenglong Wang、Michel Galley
论文精选 85
Closing Cost-Quality Gap in Document VLMs: Difficulty-Aware Data Curation and Quality-Adjusted Deployment Economics· 关闭文档 VLM 成本-质量差距
Extracting structured fields from hundreds of millions of documents annually remains costl…
领域：cs.CL作者：Maksim Evdokimov、Matvey Ivanov、Dmitrii Tsiupin
论文精选 65已解读
Selective Agent Guidance via Entropy: Learning Autonomous Policies from Imperfect VLM Teachers· 通过熵选择性代理引导：从未完美的 VLM 教师学习自主策略
Vision-Language Models (VLMs) provide useful priors for interactive decision-making, but u…
领域：cs.AI作者：Matteo Merler、Giovanni Bonetta、Davide Zago
论文精选 65已解读
Can LLMs Discover Scientific Laws in Real and Parallel Worlds?· 大型模型能否发现科学定律？
Scientific equation discovery has long been central to scientific progress, proceeding thr…
领域：cs.AI作者：Yiming Huang、Ziche Liu、Zhuohang Wu
论文精选 65已解读
EvoSCM: Scientific Belief Revision Through Causal Model Evolution and Experimentation· EvoSCM：通过因果模型进化和实验的科学信念修订
Scientific agents must learn not only how to reason, but also what to believe. However, ex…
领域：cs.AI作者：Qing Zhao、Haowei Li、Weijian Deng
论文精选 65已解读
When Guardrails Look Effective: Construct Validity Failures in LLM Agent Commerce Evaluation· 当防护栏看起来有效时：LLM 代理商业评估中的构念效度失败
Interactive simulations increasingly evaluate policies in markets populated by language-mo…
领域：cs.AI作者：Peiying Zhu、Sidi Chang
论文精选 65已解读
Harness-of-Harness: Multi-Day Autonomous Software Development with Continual Improvement· 多日自主软件开发的持续改进框架
This paper studies autonomous software development, in which LLM-based coding agents trans…
领域：cs.AI作者：Haoyang Yan、Min-le Su、Hangfan Zhang
论文精选 65已解读
Parsing the Stream: A Live Trace Model for Long-Horizon Agents and Their Observers· 解析流：长视界代理及其观察者的实时追踪模型
A long-horizon agent's trace outgrows both of its consumers: the human observer monitoring…
领域：cs.AI作者：Egor Pakhomov、Erik Nijkamp
论文精选 65已解读
EdiTikZ: Scientific Figure Editing from Revision Trajectories
Vision-language models (VLMs) have shown strong performance in generating scientific figur…
领域：cs.AI作者：Christian Greisinger、Zhixue Zhao、Steffen Eger
论文精选 65已解读
Neuro-Symbolic Geometric Abstraction (NeuSOGA): From Observations to Symbolic Mathematical Representations· 神经符号几何抽象：从观察到数学符号表达
A fundamental challenge in artificial intelligence is the transformation of observations i…
领域：cs.AI作者：Qingde Li、Qingqi Hong、Jie Tian
论文精选 65已解读
EDGE: Error Dependency Graph-Guided Multi-Error Attribution in Multi-Agent LLM Systems· EDGE：基于错误依赖图的多错误归因
Large language model (LLM) agent failures often contain multiple related errors rather tha…
领域：cs.AI作者：Jun Hou、Priya Pitre、Yi Fang
论文精选 65已解读
SymFold: Synergizing Evolutionary and Structural Priors for Accurate Protein Inverse Folding· SymFold: 结合进化和结构先验的蛋白质逆折叠
Protein inverse folding aims to recover amino acid sequences for a given 3D protein struct…
领域：cs.AI作者：Handong Wang、Jiaxin Qi、Baisheng Lai
论文精选 65已解读
Cheap Verifiers, Large Blind Spots: Measuring the Reliability Cost of Cost-Saving Cascades· 廉价验证者，大盲点
Inference cascades cut cost by answering most queries with a cheap model and escalating a …
领域：cs.AI作者：Dushyant Rajput
论文精选 65已解读
LEAP: Likelihood Elicitation and Aggregation for LLM-based Probabilistic Forecasting· LEAP：基于LLM的概率预测的似然性提取与聚合
LLM-based forecasting systems have improved on real-world tasks such as financial markets …
领域：cs.AI作者：Yufei Chen、Yiran Zhao、Xiaogang Xu
论文精选 83
Automated Event Log Generation from Unstructured Text Using Finetuned LLMs· 从非结构化文本自动生成事件日志
Process mining (PM) provides a powerful framework for discovering and optimizing operation…
领域：cs.AI作者：Maximilian Seeth、Gabriel Marques Tavares、Daniel Schuster
论文精选 85
A Composable Evaluation System for Reproducible Omni-Modal Foundation Model Evaluation· 多模态基础模型的可复现评估系统
Building an omni-modal foundation model means evaluating it across text, image, video, and…
领域：cs.AI作者：Hodong Lee、Sanghee Park、Dohoon Ryu
论文精选 75
Analog-DB: An Agent-First Analog Integrated Circuit Database, From Blocks to Systems· Analog-DB：面向代理的模拟集成电路数据库
Sharing analog integrated circuit designs remains difficult: foundry non-disclosure agreem…
领域：cs.AI作者：Danial Noori Zadeh、Mohamed B. Elamien
论文精选 83
Making Prospective Memory SLM-Shaped: Typed Intention Stores for Small-Model Agents· 使小型模型具备前瞻性记忆能力
Prospective memory means carrying out a deferred intention at the right future cue while o…
领域：cs.AI作者：Jinqing Zhao、Chengcan Wu
论文精选 85
Dual Process Motion Planning· 双过程运动规划
Robotic systems are deeply embedded in both industry and everyday life, where they are exp…
领域：cs.AI作者：Jiayi Yan、Francesco Fabiano、Alessandro Abate
论文精选 82
Measuring the Behavioral Fidelity of Long-Horizon Human Activity Simulations· 测量长期人类活动模拟的行为保真度
As LLM-based human simulators are increasingly used for policy, evaluation, and training, …
领域：cs.AI作者：Yi Fei Cheng、Fan Yang、Iremsu Bas
论文精选 80
Prompt-Robust Language Models: Which Training Strategies Work?· 增强语言模型的提示鲁棒性：哪些训练策略有效？
Despite their strong performance, large language models remain highly sensitive to prompt …
领域：cs.AI作者：Frederic Sadrieh、Michal Štefánik
论文精选 85
H2Table: Hierarchical Hypergraph-Enhanced Large Language Models for Complex Table Reasoning· H2Table：层次超图增强的大语言模型
Tables are ubiquitous across diverse domains, yet reasoning over them remains a significan…
领域：cs.AI作者：Jia Ling、Yangfan Wang、Chen Tang
GitHub 热门开源项目
开源项目精选 82
reghilkentpilaspilas-stack/plumb-core· 嵌入式 Linux 的可插拔 AI 技能框架
Pluggable AI Skill Framework for Embedded Linux: 2026 Agentic Automation Guide
⭐ 56 stars语言 HTML
精选 AI 行业资讯
行业资讯精选 65
‘NBA 2K27’ With NVIDIA DLSS 5 Leads 28 New Games Coming to GeForce NOW
September is here with 28 more games streaming on GeForce NOW this month, led by a slam du…
行业资讯精选 65
Nvidia confirms it will buy Hugging Face for $12.9 billion
Nvidia said Hugging Face hosts over 3 million models and is used by over 18 million develo…
行业资讯精选 65
NVIDIA to Acquire Hugging Face
I’m excited to announce that NVIDIA has agreed to acquire Hugging Face for $12,930,300,000…
行业资讯精选 65
AI 群星闪耀之前，商汤先相信那群少年
2018年，王延森还是一名本科生，做生成对话方向。 <span leaf="" style="t
行业资讯精选 65
“没有Token的CS学生，应立即退学”
禁止古法编程、Token自费
行业资讯精选 65
腾讯WorkBuddy联名硬件来了！首批100多家伙伴入场
眼镜、耳机、麦克风都来了
行业资讯精选 65
全栈自研+生态协同：灵心巧手与产业链核心伙伴共建中国具身智能Tier1产业联盟
2026年9月2日，由灵心巧手与星源智共同发起，联合奥比中光、法奥机器人、福莱新材、禾赛科技、华威科、览沃科技、朗毅机器人、灵猴机器人、珞石机器人、迈尔微视、钕娲创造、泉智博、森云…
行业资讯精选 65
ZINOVA携手逐际动力，探索机器人在复杂建筑施工中的应用
近日，物理AI研发公司 ZINOVA 基于逐际动力TRON 2双臂具身机器人，联合建筑机器人公司RIC Robotics完成了混凝土倾覆板（Tilt-up Construction…
行业资讯精选 65
地平线征程智驾芯片量产突破1500万！
9月3日，地平线正式宣布征程®智驾芯片量产突破1500万，凭借量产规模第一与市场份额第一的双领先优势，持续领跑中国智驾芯片市场。 作为中国首款实现大规模前装量产、累计出货量最高的车…
行业资讯精选 65
世界生成模型来了3D头部玩家，可进入生产管线的场景级生成时代来临！
单图直出场景，顶会最佳论文落地
行业资讯精选 65
陈大年复出，入局大模型
首秀逼近DeepSeek万亿旗舰
行业资讯精选 65
机械臂做到第10步就容易出错？一个 2B 模型靠动态调整注意力解决了 | IJCAI 2026
<img data-src="https://static.leiphone.com/uploads/new/images/20260903/6a99477ae3a72.jpg" …
行业资讯精选 65
把 Gemini 和 GPT 放到《我的世界》当教练，谁能看懂机器人的微操？
<img data-src="https://static.leiphone.com/uploads/new/images/20260903/6a9946fa4e5e2.jpg" …
行业资讯精选 65
打破黑盒猜想：大模型通往真正「空间智能」的破局之路
<img class="rich_pages wxw-img" data-aistatus="1" data-croporisrc="https://mmbiz.qpic.cn/s…
行业资讯精选 65
26g 无边框带扬声器，加南打造面向日常的屏显 AI 眼镜
多数屏显 AI 眼镜，总要在重量、外观、音频体验之间做出取舍。 加南全新研发的带显示 AI 眼镜，把 26g 无感佩戴、无边框形态与内置扬声器融为一体。智能显示不再是显眼的科技装置…
行业资讯精选 65
IFA 2026 倒计时 ｜ 魔法原子以“本体+大脑+场景”全栈能力叩开欧洲工业大门
2026年9月4日至8日，全球领先的具身智能公司魔法原子（MagicLab）将亮相IFA2026，完整呈现其"本体硬件、物理AI大脑、场景交付"三层全链路闭环能力，同时面向欧洲企业…
行业资讯精选 65
拆解 Claude 5.1：38 小时不睡觉的背后，Anthropic 正在终结「模型论」
<img data-src="https://static.leiphone.com/uploads/new/images/20260903/6a99396d13d2f.jpg" …
行业资讯精选 65
实测飞猪AI：真正的突围才刚刚开始？
2026 年上半年，AI 行业的风向又变了。 在经历一轮疯狂堆算力、不计成本砸研发的扩张周期之后，全球科技巨头都不得不直面同一个现实问题：“AI砸进去的钱，究竟该如何收回来？” A…
行业资讯精选 65
视频生成算力新选择：SmarCo GC3如何绕过“存储墙”这道硬门槛
大模型生成视频，正在成为这个算力饥渴时代最典型的应用场景之一。一段由AI生成的1080P视频，背后往往是数十亿参数神经网络的反复迭代，是海量张量运算在时空维度上的密集交织，也是无数…
行业资讯精选 65
「小」鸭子火了之后，端侧 AI 还有一个更「大」的问题
最近，一只只有 25 厘米高的机器鸭，意外把端侧 AI 的想象力拉到了聚光灯下。 Hugging Face 旗下 Pollen Robotics 发布的 Microduck，是一款…
行业资讯精选 65
全场景智享硬核SUV 传祺越7开启预售 权益价17.18万-21.78万
9月2日，“都市自在 山海从容”传祺越7预售发布会于广州白云国际会议中心举办。广州这座活力之城，既是大湾区产业创新高地，亦是中国汽车产业的重要策源地。广汽传祺是广州首个乘用车自主品…
行业资讯精选 82
一个模型场景通吃！它石智航AWE3.7的泛化能力有点狠· 一个模型场景通吃！
从工厂一路干到家庭
行业资讯精选 65
连尚集团入选上海市首批算力生态合作伙伴名单
近日，据雷峰网了解，上海市数据局、上海市通信管理局正式公布2026年第一批算力生态合作伙伴名单。上海连尚网络科技集团有限公司（以下简称“连尚集团”）入选“算力应用方”名单，成为上海…
行业资讯精选 83
WorkBuddy金融版正式发布，让Agent真正进入金融业务现场
9月 3 日，腾讯WorkBuddy金融版正式上线 。 WorkBuddy金融版 是 专门 面向金融机构打造的AI智能工作台， 可为券商、银行、保险等 行业 提供开箱即用的专家与技…
行业资讯精选 83
神秘具身团队又放出一连串很炸的Demo视频…自进化模型，技术路线曝光
神秘团队展示自进化模型Demo视频
行业资讯精选 82
千问办公落地长安汽车，“研产供销服”都在用
9月2日，记者获悉，长安汽车已在研发、生产制造、供应链、销售、服务等业务场景接入千问办公。 作为中国汽车行业三大央企之一，长安汽车积极拥抱AI技术浪潮，提出“无AI，不长安”。此前…
行业资讯精选 65
宇树机器狗被曝灰产链：421Wh电池贴假标签改成97Wh，「偷渡」上飞机；罕见！小米华为发布时间撞档，网友：神仙打架；月之暗面启动港股IPO
<
行业资讯精选 65
Palo Alto Networks paid $500M for Thrive-backed Console, sources say· Palo Alto Networks 以 5 亿美元收购 Console
The acquisition also leaves Sequoia-backed Serval as the de facto startup leader in AI IT …
行业资讯精选 65
TechCrunch Disrupt 2026’s new Real World AI Stage features Nvidia, robots, and extinct animals· TechCrunch Disrupt 2026：现实世界AI舞台
On our new Real World AI stage, we’ll be focusing on the intersection between the digital …
行业资讯精选 85
OpenAI’s new reasoning technique alarms AI safety experts· OpenAI新推理技术引发AI安全专家担忧
OpenAI’s new Astra model will use “recurrent depth,” a technique that allows the model to …
行业资讯精选 75
Pangram’s Max Spero on why AI detection is harder than ‘Real or Fake’· 为什么 AI 检测比'真假'更难
The internet has a trust problem, and it’s not just because social media feeds are filling…
行业资讯精选 75
We’re ‘dangerously close’ to dead internet theory, says Pangram’s CEO· 我们正‘危险地接近’互联网死亡理论
The internet has a trust problem, and it’s not just because social media feeds are filling…
行业资讯精选 80
US government sides with OpenAI on issue of training LLMs on copyrighted material· 美国政府支持 OpenAI 训练大模型
"The United States has a strong interest in continuing to develop a robust and competitive…
行业资讯精选 65
India’s richest man now wants to turn aging computers into AI-ready PCs· 印度首富欲将旧电脑升级为AI电脑
Jio is betting it can turn an aging computer into an AI-ready PC for as little as about $1…
行业资讯精选 85
HiddenLayer nabs $100M as enterprises rush to secure their AI deployments· HiddenLayer 获 1 亿美元融资保护企业 AI 部署
HiddenLayer has raised a $100M Series B from Delta-v Capital, Ten Eleven Ventures, Morgan …
行业资讯精选 82
PSA: Amazon’s shopping AI can now tell you if that message is a scam
Amazon is adding a scam-detection feature to Alexa for Shopping that can verify whether su…
行业资讯精选 80
北京协和医学院与SKG签署合作协议，开展睡眠呼吸暂停检测算法临床研究
9月1日，SKG与北京协和医学院于签署合作协议，双方将围绕睡眠呼吸暂停检测算法开展联合研究。该项目纳入三家三甲医院进行多中心临床验证，计划以医学专家标注的多导睡眠监测（PSG）数据…
行业资讯精选 75
燧原科技9月2日申购：腾讯、小米、兆易创新等参与战略配售
9月2日，上海燧原科技股份有限公司（以下简称“燧原科技”）正式开启科创板新股申购，公司本次公开发行股票数量为4303.52万股，其中网上发行688.55万股，网上申购代码78780…
行业资讯精选 82
ATV Big Air Tour turned 3 days of work into 3 hours with ChatGPT· ATV Big Air Tour 用 ChatGPT 极速提升工作效率
ATV Big Air Tour uses ChatGPT Work to speed up marketing, merchandising, and more. It even…
YouTube 热门 AI 视频
AI 视频精选 69
How Watermarks Track AI Generated Content - Computerphile
YouTube · Computerphile
📺 Computerphile▶ 1.5k 次观看
AI 视频精选 69
Claude Fable AI Is Much Stranger Than The Headlines Suggest
YouTube · Two Minute Papers
📺 Two Minute Papers▶ 2.4k 次观看
AI 视频精选 82
The Industrialization of Intelligence· 智能的工业化
YouTube · NVIDIA
📺 NVIDIA▶ 1.2k 次观看