论文精选 82
EngramEdit: Decoupled Knowledge Updates in LLMs through Conditional Memory· EngramEdit：通过条件记忆实现 LLM 知识更新
Conditional memory architectures such as DeepSeek Engram use input n-grams to look up lear…
领域：cs.CL作者：Hongru Cai、Ran Wei、Wenjie Wang
📰 AI 研究简报 · 10月9日
今日头条：How Sophos cuts threat response time by 96% with OpenAI Daybreak —— YouTube · OpenAI
共 68 条预计阅读 41 分钟Token 消耗 122,400
arXiv 最新论文
论文精选 65已解读
RoboJEPA: Scaling Robotic Latent World Models· RoboJEPA：机器人潜在世界模型的扩展
Latent world models have shown a remarkable ability to predict future states and to plan i…
领域：cs.AI作者：Artem Zholus、Nicolas Beltran-Velez、Jianhao Yuan
论文精选 65已解读
SciExam for ENSO: Can AI Agents Build Climate Models?· SciExam for ENSO: AI代理能否构建气候模型？
Language-model agents are increasingly asked to carry out open-ended scientific research, …
领域：cs.AI作者：Yinling Zhang、Langchen Liu、Dongbin Xiu
论文精选 83
Your Prompt Should Do More: Effects of Retrieval Instructions in Embedding Models
Prompted embedding models have recently received increasing attention, particularly for re…
领域：cs.CL作者：Amanda Myntti、Jenna Kanerva、Veronika Laippala
论文精选 65已解读
RECAST: Learning to Compute the Right Context through Adaptive Evidence Routing· RECAST：通过适应性证据路由学习构建上下文
Large language models are increasingly applied to tasks grounded in long, heterogeneous in…
领域：cs.AI作者：Yilun Hao、Krishna Sayana、Isabella Ye
论文精选 65已解读
Validity Without Ground Truth: What Stated-Preference Economics Offers the Evaluation of Language Models· 无真实标签的有效性
Many of the questions now put to large language models have no correct answer to score aga…
领域：cs.AI作者：Daniel Robert Kling Alexander、Catherine Louise Kling
论文精选 65已解读
EmbodiedRSI: Active Continual Robot Learning Through Hypothesis-Guided Co-Evolution· EmbodiedRSI: 假设引导的主动持续机器人学习
Robot foundation models provide strong visuomotor control, yet their performance can degra…
领域：cs.AI作者：Python Song、Zhixuan Liang、Kelsey Fu
论文精选 65已解读
Before They Can Solve: Predicting Post-Training Coding-Agent Performance from Base Models· 预测基础模型的代理后训练表现
How can we predict which base checkpoint is worth an expensive round of agentic post-train…
领域：cs.AI作者：Tan Yu、Alexander Bukharin、Khushi Bhardwaj
论文精选 82
PHRBench: A Behavioral Evaluation of Post-Hallucination Reasoning in LLMs· 幻觉后推理的基准测试
Hallucinated information can propagate through multi-stage LLM systems and become part of …
领域：cs.CL作者：Linghao Meng、Feng He、Xuan Yang
论文精选 65已解读
RunningTab: Direct Workspace Interaction with Environment-Side Tabs
Much knowledge work produces new deliverables from files a workspace already holds, and LL…
领域：cs.AI作者：Jinheon Baek、Soyeong Jeong、Yumin Choi
论文精选 83
CoTrace: Data Recipes for Training Terminal Agents with Harness-Model Co-Evolution· 终端智能体用共演化训练的数据配方
Terminal-agent capability depends jointly on model weights and the runtime harness that fo…
领域：cs.CL作者：Jixuan Chen、Jiaxin Zhang、Qinyuan Ye
论文精选 65已解读
SOTA: Stock Options Trading Agents Guided by Option-Implied Return Distributions· SOTA: 基于期权隐含收益分布的交易代理
As option markets grow and AI advances, agentic systems for option trading are gaining inc…
领域：cs.AI作者：Yizhen Xie、Mengyang Liu
论文精选 65已解读
Reasoning-Token Spikes Under Prompted Untruthful Responding in Large Language Models· 提示下大语言模型的推理令牌激增现象
Monitoring the chain-of-thought of reasoning artificial intelligence (AI) models remains a…
领域：cs.AI作者：Maverick Morales、Tomáš Dominik、Vermut Gao
论文精选 83
Document-Level Text Simplification in Estonian Using Large Language Models· 爱沙尼亚语文档级文本简化
Document-level text simplification involves transformations that go beyond sentence-intern…
领域：cs.CL作者：Meeri-Ly Muru、Eduard Barbu
论文精选 65已解读
Open-MMUnlearning: Unifying Methods and Evaluation for MLLM Unlearning· Open-MMUnlearning: 多模态大模型遗忘统一方法与评估
As multimodal large language models (MLLMs) become more capable and widely deployed, conce…
领域：cs.AI作者：Junkai Chen、Yuhao He、Qianshan Wei
论文精选 65已解读
The Handover Problem: Governing Autonomy Transitions in Human-AI Collaboration· 人机协作中的自主权转移问题
Human-machine systems rarely operate at a fixed level of AI autonomy. As operators and AI …
领域：cs.AI作者：Vicente Pelechano、Antoni Mestre、Manoli Albert
论文精选 65已解读
AI Safety Considerations for Agents With Limited Time to Act· 有限时间内行动的AI代理的安全考量
In the wake of the increasingly public discussion about AI alignment, recent work has trie…
领域：cs.AI作者：Leo Zeitler、Jack Richings、Victoria Nockles
论文精选 82
Stale, Misattributed, or Late: Where Personal Memory Fails Before Generation
Personal memory for language agents is usually judged by whether the final an- swer is cor…
领域：cs.AI作者：Haonan Deng、Park Sinchaisri
论文精选 82
OOM-RL II: Reality Is an Oracle, Not a Debugger Provenance-Constrained Diagnosis in Continually Evolving Agent-Engineered Systems· OOM-RL II：现实是预言者，而非调试器
Reality may establish that an outcome occurred without identifying which evolving procedur…
领域：cs.AI作者：Kun Liu、Liqun Chen
论文精选 83
GAGR-Lab: Evaluating Joint Spatial-Geometric and Analytic Function Reasoning· GAGR-Lab: 评估联合推理能力
Joint spatial-geometric and analytic function reasoning requires translating a perceived s…
领域：cs.AI作者：Jingyao Zhang、Yun Li、Lu Han
论文精选 82
Agentic AI-Assisted Modeling for Production Scheduling: Assessment in Constraint Programming· 代理式 AI 辅助生产调度建模
Developing optimization models for production scheduling requires substantial expert effor…
领域：cs.AI作者：Ángel Sánchez-Fernández、Javier Pernas-Álvarez、Diego Crespo-Pereira
论文精选 83
UniSkill: Learning Actor-Aligned Skill Proposals for an Evolving Policy· UniSkill: 学习共享策略的技能提案
Large language model agents can improve across tasks by retaining reusable skills distille…
领域：cs.AI作者：Yifei Lu、Cheng Liu、Dianzhi Yu
论文精选 82
RewardWeaver: Long-Horizon Interactive Learning for Language Agents via Self-Evolving Reward Adaptation· 长期交互学习的自我进化奖励适应框架
Reinforcement learning with verifiable rewards (RLVR) has driven substantial progress in d…
领域：cs.AI作者：Hengbo Xiao、Boyao Zhang、Purui Liu
论文精选 82
SkillSandbox: Skill Verification via Dynamic Scenario Synthesis· 技能沙盒：通过动态情景合成进行技能验证
Self-evolving agents distill task-solving experience into skills for future reuse, but the…
领域：cs.AI作者：Serin Kim、Kwangwook Seo、Dokyung Song
论文精选 83
HGP:An on-device personalized agent memory via hybrid graph storage· HGP：结合图存储的设备端个性化代理记忆
LLM-based agents face challenges in personalized interactive tasks due to heterogeneous, m…
领域：cs.AI作者：Ran Zhou、Xueming Han、Jiaheng Liu
GitHub 热门开源项目
开源项目精选 82
nachisama/ai-data-extractor· AI 数据提取器
Free open-source extractor for AI coding assistant chat histories. Supports Claude Code, C…
⭐ 120 stars语言 Python
开源项目精选 70
nalinbhardwaj/botscent· 检测网站AI访问
Botscent tells your site if an AI agent, not a person, is browsing it, and which one. 5 KB…
⭐ 75 stars语言 TypeScript
开源项目精选 85
ShadowSafin/Synapass
AI inference gateway and control plane for multi-provider routing, policies, caching, tool…
⭐ 75 stars语言 Go
精选 AI 行业资讯
行业资讯精选 82
联想天禧自研代码智能体TianxiCode斩获SWE-bench-Live全球第一
联想天禧AI自主研发的专业代码智能体框架TianxiCode 以71%的问题解决率登顶全球第一名
行业资讯精选 83
0.2秒急停、秒级重规划！因果智能走进真实世界
这是一台机器人正在关闭微波炉门时，因人手突然插进来而紧急悬停的时间
行业资讯精选 75
字节找到了DeepSeek时强时弱的原因
答不答得对，得看Token站位
行业资讯精选 82
Sophos cuts threat investigation time by 96% with OpenAI Daybreak· Sophos 使用 OpenAI Daybreak 减少威胁调查时间
Discover how Sophos uses OpenAI’s Daybreak to cut cyber-threat investigation time by 96% a…
行业资讯精选 80
《柳叶刀》研究表明：AI 有望改善医患关系· AI 有望改善医患关系
Google 研究成果首次登上《柳叶刀》主刊
行业资讯精选 82
再裁员5000人！保时捷明确回归燃油车，高端车型计划涨价20%；腾讯拟发行50亿美元离岸债券，或筹钱加码AI；OpenAI全面上线GPT-6· 腾讯拟发行50亿美元债券，加码AI
<
行业资讯精选 82
跨域顶层监管层｜监管者自身的跨域审计规则
本文为《碳硅道统·跨域迁移治理法典》七层体系官方连载第五篇，是整套理论从企业合规升维到公共治理、从行业自律升维到国家制度、从被动治理升维到元监管的顶层范式层。 解决全球AI治理终极…
行业资讯精选 85
灵巧操作头号玩家：Sharpa把指尖「触觉」进化到全面「体感」
机器人摆脱展会、舞台Demo，产生真实商业价值，究竟需要哪些东西？
行业资讯精选 83
清华具身模型登顶全球第一！突围GPT-6、英伟达，不靠外挂和额外数据
星动纪元选择将视频预测与动作学习分阶段训练，重点不是「视频、动作一锅炖」，而是把两者「解耦」，重新「排序」。
行业资讯精选 82
TRAE全面升级：双端融合，打造全链路开发平台
10 月 9 日，TRAE宣布全面升级：TraeWork与TraeCode双端融合，统一连接Agent与IDE两种模式，覆盖从任务推进到深度编码的开发流程。 过去，TraeWork…
行业资讯精选 82
豆包工作新增画布功能，并接入豆包2.1 Lite模型
近期，豆包工作持续更新，推出创作画布功能，并在模型上新增了轻量级的豆包2.1 Lite模型。 据介绍，画布功能主要针对复杂创作任务，速度更快，交互体验更好。该功能将素材、设计方案与…
行业资讯精选 80
openJiuwen发布并开源企业级AgentOS，加速智能体规模落地企业· 开放智能发布AgentOS
多Agent协同还能自我进化
行业资讯精选 80
代码造世界，扩散绘现实：AgentGarten让智能体在实时试炼场中边玩边进化
让AI反复试错的“练兵场”来了
行业资讯精选 80
陶哲轩带头宣战！人类数学家联合抵制OpenAI· 数学家抵制OpenAI
彻底撕破脸了
行业资讯精选 80
不等Gemini 4了！谷歌发布办公Agent，支持调用Claude· 谷歌发布办公Agent，支持调用Claude
新的“缝合怪”已经出现，怎么能够停滞不前
行业资讯精选 82
Into the Omniverse: How Developers Turn Ideas Into Simulations With Frontier AI Agents· 构建模拟应用：开发者如何利用前沿AI代理
Turning a simulation idea into a working application means assembling assets, connecting p…
行业资讯精选 82
Fired OpenAI safety researchers dispute misconduct claims, warn of chilling effect· 被解雇的 OpenAI 安全研究人员反驳不当行为指控
Three fired OpenAI safety researchers dispute allegations of mishandling sensitive informa…
行业资讯精选 65
Ben Affleck is an AI nerd, and the internet is impressed· 本·阿弗莱克是 AI 爱好者，网友纷纷赞叹
Ben Affleck is going viral for his deep knowledge of AI, from neural networks and transfor…
行业资讯精选 82
Popular AI leaderboard Arena nearly doubles valuation to $3.1B valuation in 10 months· 受欢迎的 AI 排行榜 Arena 估值10个月内接近翻倍至31亿美元
The company behind the popular LMArena leaderboard has raised $200 million led by Lightspe…
行业资讯精选 75
OpenAI’s revenue is reportedly $20 billion less than previously projected· OpenAI 收入大缩水
It had previously been reported that the AI lab's annualized revenue was some $70 billion,…
行业资讯精选 82
Google brings agentic AI to Gemini, starting with businesses· Google为Gemini引入代理AI
Google is turning Gemini into an AI agent that can plan, execute tasks, and work across bu…
行业资讯精选 75
Anthropic changes usage policy to ban model abuse and election interference· Anthropic更新使用政策
Anthropic's updated usage policy explicitly prohibits users from repeatedly abusing Claude…
行业资讯精选 75
OpenAI’s math solutions aren’t meeting the field’s standards yet· OpenAI的数学解决方案尚未达标
OpenAI's flood of proofs deviated from the guidelines set by a group of mathematical resea…
行业资讯精选 75
Natura’s $99 smart ring puts AI agents on your finger· Natura 的 99 美元智能戒指将 AI 代理戴在指尖
Natura’s $99 Interface smart ring lets you summon AI agents with the press of a finger to …
行业资讯精选 82
How Oracle turns days of work into minutes with ChatGPT and Codex· 甲骨文如何用 ChatGPT 和 Codex 加速工作
Across recruiting, engineering, and operations, Oracle turns specialist knowledge into fas…
行业资讯精选 82
Goodfire says its new ‘inside-out’ monitors catch rogue AI agents at a fraction of the cost· Goodfire 推出新 AI 监控方法
Goodfire just launched what it says is a cheaper way to keep AI agents in check: Instead o…
行业资讯精选 82
Claude Haiku 5.5 降本背后：操作能力暴涨，复杂编程为何仍差一截？
<img data-src="https://static.leiphone.com/uploads/new/images/20261008/6ac7b88ded3c4.jpg" …
行业资讯精选 82
Hear from Ambrosia Energy and Bloom Energy execs on where the AI infrastructure boom is creating opportunity at TechCrunch Disrupt 2026· 能源公司高管在 TechCrunch Disrupt 2026 讨论 AI 基础设施
Ambrosia Energy CEO Ben Longmier and Bloom Energy SVP Bill Thayer join the Smart Systems S…
行业资讯精选 85
Cal AI’s 19-year-old founder just raised $10M for his new AI startup· 19岁Cal AI创始人获1000万美元投资
Zach Yadegari, the teen co-founder of popular Cal AI calorie tracking app, has launched a …
行业资讯精选 82
Pollo AI turns creative ideas into campaigns with OpenAI· Pollo AI：将创意转变为广告
With GPT-5.6, GPT-6 Astra, and GPT‑Image‑2.5, Pollo AI helps creators turn bold ideas into…
行业资讯精选 65
LegalOn halves Codex costs while maintaining development speed· LegalOn 降低 Codex 成本
LegalOn cut estimated daily Codex costs by 65% while maintaining development speed. It mat…
行业资讯精选 82
Disrupting AI-enabled “false front” operations· 叫停 AI 伪造的虚假影响行动
OpenAI disrupted two AI-enabled influence operations that used false-front journalists and…
YouTube 热门 AI 视频
AI 视频精选 85
How Sophos cuts threat response time by 96% with OpenAI Daybreak· Sophos 如何用 OpenAI Daybreak 将威胁响应时间缩短 96%
YouTube · OpenAI
📺 OpenAI▶ 857 次观看
AI 视频精选 80
Building with Gemma 4· 构建 Gemma 4
YouTube · Google for Developers
📺 Google for Developers▶ 399 次观看
AI 视频精选 85
More visual answers from GPT-6 with Intelligent UI· GPT-6推出智能UI
YouTube · OpenAI
📺 OpenAI▶ 740 次观看
AI 视频精选 75
How Oracle Uses ChatGPT Work to Transform Recruitment· 甲骨文如何用ChatGPT改变招聘
YouTube · OpenAI
📺 OpenAI▶ 275 次观看
AI 视频精选 82
Get a more visual answer from GPT-6 with Intelligent UI· GPT-6通过智能UI提供更直观回答
YouTube · OpenAI
📺 OpenAI▶ 261 次观看
AI 视频精选 82
Play a game against GPT-6 with Intelligent UI· 用智能界面与 GPT-6 游戏
YouTube · OpenAI
📺 OpenAI▶ 545 次观看
AI 视频精选 85
Make a game using GPT-6 with Intelligent UI· 用 GPT-6 制作智能界面游戏
YouTube · OpenAI
📺 OpenAI▶ 2.2k 次观看
AI 视频精选 82
📷 Searching for images from text with EmbeddingGemma 2· 用 EmbeddingGemma 2 进行文本搜索图片
YouTube · Google for Developers
📺 Google for Developers▶ 113 次观看