论文精选 60已解读
Can Language Models Play the Wikipedia Game?· 语言模型能否玩维基百科游戏？
Language models increasingly support or automate information retrieval from document colle…
领域：cs.AI作者：Alexander Hickey、Joanna Knight、James Bishop
📰 AI 研究简报 · 10月13日
今日头条：梁文锋投了Swin Transformer作者！DeepSeek开始布局具身大脑 —— Winverse AI
共 65 条预计阅读 39 分钟Token 消耗 117,000
arXiv 最新论文
论文精选 83
RSI-Anything: Unlocking Recursive Self-Improvement with World Models· RSI-Anything：解锁递归自我改进的世界模型
Recursive self-improvement (RSI) aims to transform interaction experience into persistent …
领域：cs.CL作者：Taoyong Cui、Pinxian Zeng、Wanghan Xu
论文精选 82
When VLMs Destroy What They Know: Spatial Representations Fall Below Lexical Priors· 当大模型破坏已知：空间表示逊于词汇先验
Vision-Language Models (VLMs) increasingly analyze and reason about images, yet it remains…
领域：cs.CL作者：Shubh Garg、Chen Shani
论文精选 83
Structural Automata for Neural Semantic Parsing· 神经语义解析的结构自动机
Neural semantic parsers generate meaning representations token by token, yet well-formedne…
领域：cs.CL作者：Xiao Zhang
论文精选 82
Task Context Shapes Component Roles in Language Models· 任务上下文如何塑造语言模型组件角色
Large language models reuse the same attention heads across tasks, but identifying these s…
领域：cs.CL作者：Sovesh Mohapatra、Dani S Bassett
论文精选 60已解读
RATED: Reliability-Aware AI Agent Tool Retrieval via Execution Dynamics· RATED：基于执行动态的可靠性感知AI代理工具检索
As AI agents increasingly rely on external tools, accurate retrieval is essential for reli…
领域：cs.AI作者：Weiliang Chen、Song Liao
论文精选 60已解读
X-AgentRed: Red-Teaming Personal Agents in Evolving Multi-Service Ecosystems· X-AgentRed：在多服务生态系统中对个人代理进行红队测试
Personal agents increasingly connect to services across the web and applications by using …
领域：cs.AI作者：Yeonsung Jung、Zhiyuan Wang、Zifan Wang
论文精选 60已解读
Studying Comics at Scale with Artificial Intelligence: The PAGE Framework and Three Studies of American Comics and Japanese Manga· 用 AI 研究大规模漫画：PAGE 框架及案例
Comics are one of the great visual records of the twentieth century: hundreds of thousands…
领域：cs.AI作者：Emilio Ferrara
论文精选 82
Can Language Models Learn When to Abstain? Policy Learning and Calibrated Answer Selection· 语言模型的选择性回答
A useful language model should know when to answer and when to say "I don't know." Can we …
领域：cs.CL作者：Juan Terven、Diana-Margarita Córdova-Esparza、Ruben Gomez
论文精选 60已解读
UkrMeme: Multimodal Explainable Persuasion Detection in Ukrainian Memes· 乌克兰讽刺图的多模态可解释劝导技术检测
Memes play a central role in modern online communication, especially in contexts of politi…
领域：cs.AI作者：Nazar Andrushko、Vera Schmitt、Veronika Solopova
论文精选 60已解读
A Privileged Teacher That Only Scores Cannot Search: Critique-Conditioned On-Policy Self-Distillation for Reasoning· 无法搜索的特权教师：基于批评的在线策略自蒸馏
On-policy self-distillation promises to replace the sparse outcome signal of verifiable-re…
领域：cs.AI作者：Mayug Maniparambil、Ayush Rai、Ekansh Verma
论文精选 60已解读
Adaptive DHiSparse: Speculative Decoding over a Load-Adaptive KV Hierarchy· 自适应DHiSparse：基于负载适应的稀疏KV层次结构的投机解码
Long-context serving runs out of KV memory before the hardware reaches its useful throughp…
领域：cs.AI作者：Naji Alazhar、Wassel Alazhar
论文精选 60已解读
A Baseline Assessment for the Agentic Web Technology Stack· LLM基代理技术栈的基线评估
The fast pace in which protocols, frameworks, and technologies are being proposed for LLM-…
领域：cs.AI作者：Vagner Figueredo de Santana、Diogo Cortiz
论文精选 60已解读
A Step Harder, Not a Step Too Far: The Benefits and Limits of Curriculum Difficulty Progression in GRPO· 课程难度递进在 GRPO 中的益处和限制
Curriculum learning typically organizes training samples from easy to hard. In GRPO, mediu…
领域：cs.AI作者：Ronghan Li、Fei Chen、Mingyu Lu
论文精选 60已解读
A Multi-Agent LLM System for Automated Tuning of Specialized Neural Network Models· 多智能体大模型系统自动调优神经网络
Many machine learning approaches are two-level: an expert selects a class of methods, with…
领域：cs.AI作者：Ivan Moloshnikov、Dmitry Kunitsyn、Alexander Sboev
论文精选 60已解读
Learning Size Independent Constraint Models from Input Data and Positive and Negative Example Solutions
In this paper we describe a Constraint Acquisition Tool that generates size independent co…
领域：cs.AI作者：Helmut Simonis
论文精选 60已解读
Humanize Olympiad Agents: A Multi-Session Workflow for Olympiad Problem Solving with Solve, Review, Wiki, Verify· 人类化奥赛代理
Large language model (LLM) agents are rapidly improving on long-horizon scientific and oly…
领域：cs.AI作者：Zhengyang Zhang、Jing Xiong、Hongzhou Lin
论文精选 85
MENTOR: Learning from Failure with On-Policy Pedagogical Intervention· MENTOR：从失败中学习的同策教育干预
Reinforcement Learning from Verifiable Rewards (RLVR) is a key driver for Large Reasoning …
领域：cs.AI作者：Guanlan Yue、Hang Yu、Xiaoyong Huang
论文精选 83
Constructing Pareto Frontiers of Optimal Reasoning Termination Methods for Overthinking in Reasoning Models· 构建推理模型最优终止方法的帕累托前沿
Reasoning models can benefit from test-time scaling, but continued reasoning may become in…
领域：cs.AI作者：Weixin Guan、Liang Li、Shiyu Xie
论文精选 82
MAGMA-Bench: Evaluating Long-Horizon Embodied Agents under Dynamic Tasks and Execution Failures· MAGMA-Bench: 动态任务和执行失败下的长时实体代理评估
Long-horizon embodied agents must maintain task-relevant information, coordinate heterogen…
领域：cs.AI作者：Loan Bernat、Matthieu Grard、Ariane Herbulot
论文精选 82
Beyond Domain-Routed Supervision: Cross-Teacher Guarding for Multi-Teacher On-Policy Distillation· 超越领域路由监督：多教师在线策略蒸馏的跨教师保护
Multi-teacher on-policy distillation (MOPD) extends on-policy distillation (OPD) by routin…
领域：cs.AI作者：Shi-Yu Tian、Yong Liu、Ximan Liu
论文精选 85
Causal Effect Identifiability in Causal Abstractions: A Strict Hierarchy of Criteria· 因果抽象中的因果效应可识别性
Identifying causal effects from observational data typically requires a fully specified ca…
领域：cs.AI作者：Clément Yvernes、Marianne Clausel、Emilie Devijver
论文精选 85
Retrieval-Augmented Multimodal Language Model with Heterogeneous Evidence Integration for Clinical Prognosis
Clinical prognosis requires integrating structured measurements, physiological waveforms, …
领域：cs.AI作者：Hamza Shafiq、Jialu Tang、Ting Dang
论文精选 82
Humanize MLE Agents: Multi-Agent Session Alternation Beyond Single-Agent Limits· 超越单智能体限制的多智能体会话交替
Existing LLM-based agent approaches for long-horizon machine learning engineering (MLE) of…
领域：cs.AI作者：Changye Li、Zijian Zhang、Muheng Li
论文精选 82
Branching On-Policy Distillation· 分支策略蒸馏
On-policy distillation (OPD) provides teacher supervision on student-generated trajectorie…
领域：cs.AI作者：Dongyang Chen、Shudong Liu、Shubo Li
GitHub 热门开源项目
开源项目精选 75
deeplethe/solid-glass· solid-glass 设计系统
A design system for AI applications. Material, physical, tactile, technical.
⭐ 229 stars语言 TypeScript
开源项目精选 82
zhuyansen/new-term-radar· 新兴 AI 术语雷达
Daily radar for new AI terms: names of new models, products and repos, caught on X, Huggin…
⭐ 544 stars语言 Python
精选 AI 行业资讯
行业资讯精选 65
SpaceXSI bundles X Premium, Grok, and Cursor into one subscription· SpaceX 推出整合 AI 工具的 X 订阅服务
The new X Subscription combines X's paid features with AI tools like Grok and Cursor, offe…
行业资讯精选 81
Amex introduces a framework to prevent AI agents from making unauthorized purchases· Amex 推出防止 AI 未经授权购物的框架
American Express is introducing a new framework designed to ensure AI agents only make pur…
行业资讯精选 65
Just 5 years in, interactive chatbot platform Character.AI files for bankruptcy· 五岁的互动聊天平台Character.AI申请破产
The startup will sell some of its assets to Disney for $20 million.
行业资讯精选 80
All Systems Go: Global PC App Ecosystem Ready for NVIDIA RTX Spark· NVIDIA RTX Spark：全球PC应用生态的新起点
NVIDIA RTX Spark is a new beginning for the PC — a reinvention combining the capabilities …
行业资讯精选 82
魔法原子具身Agent框架拿下MolmoSpaces全球第一
让VLA从“演示可用”逼近“产线可用”
行业资讯精选 85
梁文锋投了Swin Transformer作者！DeepSeek开始布局具身大脑
Winverse AI
行业资讯精选 75
独家丨字节最早天使投资人，投了一家 AI 音乐硬件公司
<img data-src="https://static.leiphone.com/6a6c4981e5bcb.jpg" alt="图片" class="rich_pages w…
行业资讯精选 80
华为哈勃投了个98年博士！让机器人皮肤变色，团队平均23岁
让材料感知压力，造出超薄高密度电子皮肤
行业资讯精选 65
Claude团队又搞事：实测AI全自主复制进智能家电
拉马克进化的幽灵
行业资讯精选 70
“模态进化”：第八届亿邦产业互联网年会暨首届产业智能体大会双会合体定档· 模态进化大会定档
由亿邦动力主办的“2026第八届亿邦产业互联网年会暨首届产业智能体大会”正式定档
行业资讯精选 75
其域创新以 62 亿个高斯、150 亿个点云，建成全球最大的空地一体 3D 高斯模型“闽宁镇”
它覆盖 23.5 平方公里，相当于 3291 个标准足球场，由 18.6 万张航拍与地面影像、150 亿个雷达点云训练而成，规模达 62 亿个高斯椭球。镇口的牌坊、红顶白墙的街区、…
行业资讯精选 85
字节跳动 Seed 单独设立大模型安全与对齐部门
字节跳动 Seed 近日发出内部邮件，宣布将成立“Seed Model-Alignment&Preparedness”部门，由严林负责，汇报给 Seed 负责人吴永辉。原 Resp…
行业资讯精选 75
From Earth to Orbit, France’s Ecosystem Delivers Agentic Computing at Scale With NVIDIA· 法国生态系统大规模交付代理计算
Companies in France are deploying agentic computing across industries spanning space, fina…
行业资讯精选 82
千问确认将推出「千问Go!」，已开放体验申请
10月13日，千问官方公众号发布消息，Personal Agent「千问Go!」的体验申请页面已在千问APP上线。用户可通过APP内指定入口提交申请，千问将逐步为更多用户开放体验资…
行业资讯精选 82
9大赛道，23份企业真实考卷！上海具身智能大考，邀千队共解题
赢奖金、拿融资、送出海
行业资讯精选 80
退差价、约保养、订球场、跟客服磨赔偿，千问Go什么都想管
丢一句话进去，事儿就办完了
行业资讯精选 85
单轮23亿，累计43亿，无问芯穹刷新国产AI Infra赛道融资双重纪录
算力、模型、行业，三层生意一起做
行业资讯精选 75
1688全面接入AI电商工作台Accio
10月13日，阿里旗下数字供应链平台1688宣布，即日起，平台逾1亿年度活跃买家和超100万源头厂商，均可直接开通全球电商AI工作台Accio。 接入Accio后，买家不仅能用其完…
行业资讯精选 82
「千问Go！」即将发布，刚帮用户在淘宝申请52元价保退款
10月13日消息，千问Personal Agent「千问Go！」即将发布。一位用户使用后分享，通过千问Go为自己申请航班延误理赔，已获取70元理赔款；也有用户表示，千问Go已帮忙申…
行业资讯精选 82
Meta releases its first frontier model, Muse Wonder· Meta 发布首个前沿模型 Muse Wonder
Meta's Superintelligence Lab released a pair of new state-of-the-art models on Monday, Mus…
行业资讯精选 75
WindBorne wins a hefty US Air Force contract for its balloon data· 风生赢得美国空军大量合同
A Silicon Valley startup that uses weather balloons and AI to predict the weather won a US…
行业资讯精选 75
Honeyjar offers AI tools to improve PR and comms at every company· Honeyjar 提供 AI 工具改善公司公关
Journalists, entrepreneurs, and even some public relations professionals will tell you tha…
行业资讯精选 81
Microsoft president says UK is best placed to drive AI regulation· 微软总裁称英国最适合推动 AI 监管
Microsoft President Brad Smith calls for urgent AI regulation, saying milk and washing mac…
行业资讯精选 75
This AI startup says it can generate permit-ready blueprints in minutes· Blueprints AI 快速生成建筑蓝图
Blueprints AI, a bootstrapped startup based in Santa Monica, developed software that produ…
行业资讯精选 85
OCP Global Summit: Amid Ecosystem Adoption, NVIDIA Opens Vera Rubin Architecture to Help Scale AI Factories Worldwide· NVIDIA 在 OCP 峰会上开放 Vera Rubin 架构
As AI factories grow from megawatts to gigawatts, building them takes more than advanced c…
行业资讯精选 80
产业出题，科技答题：2026 IEIC国际具身智能技能大赛，用真实岗位任务考核机器人落地能力
<img data-src="https://static.leiphone.com/uploads/new/images/20261012/6accea8c03d8f.jpg" …
行业资讯精选 85
安踏×火山引擎：下单缩至10分钟、月均建单1000+，Agent跑进全球电商业务
Agent，正在成为安踏全球化业务中的“数字同事”。 作为全球前三的体育用品集团，安踏集团多品牌、跨市场经营的背后，是大量商品、内容、订单和售后流程。如今，Agent 正在进入内容…
行业资讯精选 80
2026没过完，已有480亿美元投进了物理AI
日均10亿元流入
行业资讯精选 75
具身智能回到产业现场， 思灵机器人Agile Robots携12款展品亮相2026工博会
思灵机器人Agile Robots携12款展品参展第26届中国国际工业博览会。
行业资讯精选 82
百度世界2026将于11月12日在北京举办，主题为「下一幕，智能组织」
10月12日，百度官方宣布，百度世界2026将于11月12日在北京·国家会议中心二期举办。目前，大会官网已开启售票通道（https://baiduworld2026.events.…
行业资讯精选 75
吉利银河战舰700上市：限时焕新价16.98万元起，押注电混与智能越野
10月10日，吉利首款新能源硬派SUV银河战舰700在杭州正式上市，推出三种动力、六大版型，上市限时焕新价16.98万元起。 作为吉利银河拓展硬派SUV品类的首款产品，战舰700将…
行业资讯精选 82
阶跃 Step 5 Preview 接入腾讯 WorkBuddy，独家限免 2 周！
10 月 12 日消息，腾讯旗下全场景 AI 办公工作台 WorkBuddy 已接入阶跃最新旗舰模型 Step 5 Preview。即日起，该模型在 WorkBuddy 独家限时免…
行业资讯精选 82
微软用阿里Qwen打造决策模型，或将推广至所有领域
10月12日，微软基于阿里Qwen，推出全新决策模型Microsoft-Decision-1。微软官方透露称，该模型基于Qwen3.5-9B后训练而成，在结构化决策任务中，该模型在…
行业资讯精选 82
费莫一刘念邱：智驾无图的经验，也能迁移到机器人 | 物理AI 50人
刘念邱很难被归进典型的具身创业者模板里。 他不是从机器人实验室出来的学者，也不是过去两年才进入 Physical AI 的连续创业者。他曾在英特尔做过新工艺和设备研发，创业做过物联…
YouTube 热门 AI 视频
AI 视频精选 65
A little something from Jensen Huang ✍️· 黄仁勋的 AI 新视频
YouTube · NVIDIA
📺 NVIDIA▶ 1.1k 次观看
AI 视频精选 82
OpenAI Just Changed What It Means To Do Math· OpenAI 改变数学计算方式
YouTube · Two Minute Papers
📺 Two Minute Papers▶ 2.7k 次观看
AI 视频精选 65
The NVIDIA GTC pregame show is back.· NVIDIA GTC 预热秀回归
YouTube · NVIDIA
📺 NVIDIA▶ 118 次观看
AI 视频精选 80
Review Code with Codex Before You Merge· 合并前用Codex审查代码
YouTube · OpenAI
📺 OpenAI▶ 2k 次观看