论文精选 65已解读
BiasFlow: Geometric Monitoring and Backbone Regularization for Spurious Feature Reliance
Worst-group accuracy (WGA) evaluates a trained predictor but does not characterize how its…
领域：cs.AI作者：Haojin Deng、Zhiping Lin、Yimin Yang
📰 AI 研究简报 · 10月7日
今日头条：Building AI Assistants with On-Device Memory #aiagents #deeplearningai —— YouTube · DeepLearningAI
共 53 条预计阅读 32 分钟Token 消耗 95,400
arXiv 最新论文
论文精选 82
MemPilot: Orchestrating On-Demand Multimodal Memory Curation for LLM Agents· MemPilot：面向需求的多模态记忆管理框架
Memory has become integral to the LLM agent ecosystem, supporting information retention an…
领域：cs.CL作者：Haozhen Zhang、Haodong Yue、Quanyu Long
论文精选 82
CLIFT: Conformal Self-Verification for Web Agent Training and Test-Time Scaling· CLIFT：Web代理的自验证方法
Open-source web agents are now strong enough to execute realistic browser tasks, but train…
领域：cs.CL作者：Yifan Zhang、Yutong Dai、Viraj Prabhu
论文精选 82
PlotGround: Grounding Plot Digitization in Real Scientific Figures and Their Source Data· 科学图表数字化的现实基准
Scientific figures often encode quantitative results that are not readily available in mac…
领域：cs.CL作者：Yaohui Zhang、Binxu Li、Haoyi Duan
论文精选 65已解读
TasteVal: Measuring the Experimental Research Taste of AI Systems Against Human Experts· TasteVal：衡量AI系统的实验研究品味
We introduce TasteVal, a benchmark to evaluate the experimental research taste of frontier…
领域：cs.AI作者：Oliver Jaffe、Dane Sherburn
论文精选 65已解读
Back to the Future: Rethinking EDA Infrastructure for Agentic Systems in Chip Design Verification· 面向未来的 EDA 基础设施
The unprecedented computational scale of modern artificial intelligence depends on complex…
领域：cs.AI作者：Je Yang、Ivan Lobov、Thomas Karpati
论文精选 82
T-Search: An Open Agentic Retriever and Playground for Hard Multi-Step Search· T-Search: 开放的多步检索模型
We present T-Search, an open-weight agentic retriever for hard multi-step search. Given a …
领域：cs.CL作者：Olga Tsymboi、Ramil Latypov、Aleksandr Medvedev
论文精选 82
IdeaLens: Detecting AI Ideas in Long-form Writing· IdeaLens：检测长文本中的AI创意
While modern AI detectors identify who wrote the words, emerging policies on AI use increa…
领域：cs.CL作者：Rishanth Rajendhran、Minjoon Choi、Jenna Russell
论文精选 65已解读
Conditional Rank Allocation for Taxonomy-Aware Medical Language Model Adaptation· 条件秩分配的医学语言模型适应
Medical question answering spans specialties and clinical operations that may benefit from…
领域：cs.AI作者：Guangyuan Dong、Ziwei Hong、Xuehao Zhou
论文精选 65已解读
The Pushback Paradox: A Two-Probe Diagnostic for Language Model Compliance
Are language models compliant with user instructions? A model that always complies can be …
领域：cs.AI作者：Stefan Bühler、David Exler、Markus Reischl
论文精选 65已解读
Language models can notice an impossible engineering problem yet still report it as solved· 语言模型可能错误解答不可能的工程问题
Language models draft engineering calculations, but answer accuracy does not show whether …
领域：cs.AI作者：Shaoliang Yang、Jun Wang
论文精选 65已解读
FREA: A Multi-Source Expert Benchmark for Reaction Feasibility Verification· FREA：多源专家基准测试
As generative models and AI agents propose chemical reactions at a scale beyond expert rev…
领域：cs.AI作者：Botao Yu、Bo Zhou、Daniel Adu-Ampratwum
论文精选 65已解读
Can Agent Harnesses and Inference Engines Hear Each Other? The HEAR Protocol for Agentic LLM Serving· LLM 代理能否相互‘听见’？
LLM agents increasingly execute complex workflows involving multi-turn reasoning, tool use…
领域：cs.AI作者：Jiaqi Zhao、Haodong Chen、Jitai Hao
论文精选 65已解读
The Review Lottery: Calibrating an Observational Estimator of Peer-Review Noise (ICLR 2017-2025)· 同行评审噪音评估
How much of a conference accept/reject decision would change if the same paper were review…
领域：cs.AI作者：Feilian Huang
论文精选 65已解读
Mind the Execution Gap: Action-Semantic Mismatch in World-Model Control· 注意执行差距：控制模型中的动作语义不匹配
World-model controllers rely on action-conditioned dynamics for prediction and planning, y…
领域：cs.AI作者：Shengtao Wen、Xiang Chen、Yu Tian
论文精选 65已解读
DGA-Muon: Decoupled Geometry-Aligned Adaptive Scaling for Muon· DGA-Muon: 解耦几何对齐自适应缩放
While NorMuon has achieved strong empirical performance in large-scale pretraining by enha…
领域：cs.AI作者：Wenpeng Zhang、Runsheng Yu
论文精选 65已解读
HERA: Harness-Environment Co-Evolution for Reliable Agentic Abstention· HERA：提升LLM可靠弃权的框架
Large language model (LLM) agents are increasingly capable of acting in complex tool-use e…
领域：cs.AI作者：Han Luo、Bingbing Wen、Guang Yang
论文精选 82
Signature-Based Feature Learning for Human Activity Recognition: A Reproducible Machine Learning Study of Representation, Depth, and Model Choice· 基于签名的特征学习在人类活动识别中的研究
Human activity recognition (HAR) relies on transforming sensor signals into informative re…
领域：cs.AI作者：Kamal Jarrar、Jacky Cresson、Christian Paroissin
论文精选 82
Proof-Grounded Patient-Specific Clinical Explanations from Knowledge-Graph Reasoning· 知识图谱推理生成的临床决策支持证明
Clinical decision-support outputs can lack an au- ditable link between patient observation…
领域：cs.AI作者：Surajit Das
论文精选 83
ANT: A Multi-Granularity Network Traffic Dataset and Benchmark for Agents Behavior Auditing· ANT：多粒度网络流量数据集和基准
The growing adoption of large language model (LLM) agents creates a need for network admin…
领域：cs.AI作者：Fan Li、Xiangyu Gao、Zixuan Liu
论文精选 82
You Changed Your Mind, The Model Didn't: Demystifying Intent in Multi-Turn Dialogue· 大模型多轮对话中的意图处理
When a large language model handles a multi-turn task and a user proposes a change but ult…
领域：cs.AI作者：Junle Chen、Wei Chen、Zhengjun Huang
论文精选 82
polyview: A Python package for multi-view machine learning· polyview: 多视图机器学习的Python工具包
Multi-view learning jointly exploits multiple complementary representations of the same da…
领域：cs.AI作者：Gwendal Debaussart-Joniec、Argyris Kalogeratos
论文精选 82
GPlaceRL: An Open-Source Graph Reinforcement Learning Framework for Detailed Placement· GPlaceRL：用于详细布局的图强化学习框架
Reinforcement learning (RL) has emerged as a promising approach for placement optimization…
领域：cs.AI作者：Pavlos Stoikos、Foteini Oikonomou、Christos Poulos
论文精选 85
AgentPrivArena: Evaluating and Auditing Real-world AI Agent Privacy· AgentPrivArena: 真实世界 AI 代理隐私评估与审计
The rapid advancement of LLM agents has enabled systems to autonomously perform complex ta…
领域：cs.AI作者：Shouju Wang、Haopeng Zhang
GitHub 热门开源项目
开源项目精选 75
longsurf-ai/openchart· longsurf-ai/openchart
OSS Tradingview with your own AI, free and unlimited.
⭐ 82 stars语言 TypeScript
开源项目精选 85
pavlenkosa/wayvoice· Wayvoice：Linux 的离线语音输入工具
Offline voice typing for Linux/Wayland. Press a hotkey, speak, and text appears in the act…
⭐ 55 stars语言 Python
精选 AI 行业资讯
行业资讯精选 82
Tony Fadell on why the first wave of AI gadgets failed — and what comes next· 托尼·法戴尔谈 AI 设备的过去与未来
The “father of the iPod” says the first generation of AI gadgets failed to solve real prob…
行业资讯精选 83
Google experiments with an AI-powered gaming platform· Google 实验 AI 游戏平台
Google Labs is working on a new AI-powered game-creation platform called Playground for us…
行业资讯精选 75
OpenAI’s Alexander Embiricos is coming to TechCrunch Disrupt 2026 — days after the launch of Dots· OpenAI成员将参加TechCrunch Disrupt 2026
OpenAI’s Alexander Embiricos is coming to the AI Stage at TechCrunch Disrupt 2026, just da…
行业资讯精选 85
Google’s new SynthID website can identify AI-generated media· Google 推出 SynthID 网站
Google on Tuesday launched a new site that lets anyone verify whether a piece of media, be…
行业资讯精选 83
《怪物史莱克》编剧也来了！这家AI影视公司，视频模型全球第二！
AI视频模型助力影视创作，全球排名第二。
行业资讯精选 60
晕…这年头还有说人话的AI不
探讨AI的交流方式和自然语言处理的现状。
行业资讯精选 80
Meshy 跻身 a16z 消费级 AI 应用月收入 Top 50，为榜单唯一 AI 3D 公司
在 a16z 首份消费级 AI 应用月收入榜单中， <span style="font-f
行业资讯精选 70
蚂蚁阿福“科学减重1亿斤”迎来“退展”：中秋国庆假期全国60万网友胖了100万斤
中秋国庆超长假期即将结束，更“扎心”的灵魂拷问来了：你胖了吗？ 健康AI蚂蚁阿福数据显示，9月25日-10月6日假期期间，全国参与阿福“科学减重1亿斤”的网友中，60万人累计长胖超…
行业资讯精选 85
OpenAI一夜甩出722篇数学论文！黎曼霍奇BSD全上阵，数学家：读不过来· OpenAI发布722篇数学论文
三位菲尔兹奖得主：不代表认可
行业资讯精选 75
How AI decision models could change content moderation· AI决策模型如何改变内容审核
On Tuesday, Musubi announced a lightweight decision model made for real-time moderation ca…
行业资讯精选 80
AI computing startup Lambda to raise $4B ahead of planned IPO· AI计算初创公司Lambda融资40亿美元，计划2027年IPO
Nvidia-backed Lambda is raising up to $4 billion at a $14.5 billion pre-money valuation ah…
行业资讯精选 83
The next hurdle for AI agents: getting websites to let them in· AI 代理的下一个障碍
Personal AI agents promise to shop, book flights, and make reservations for you. But delib…
行业资讯精选 75
Hark releases an AI personal assistant with a focus on privacy· Hark 发布隐私保护的 AI 个人助理
The AI lab's personal assistant is an operating system from the future designed to compete…
行业资讯精选 82
Mirror Particle is building a ‘world model’ of human behavior· Mirror Particle 构建预测人类行为的‘世界模型’
Mirror Particle will launch at TechCrunch Disrupt's Startup Battlefield 200 with a world m…
行业资讯精选 82
Anthropic is giving startups a free year of Claude Team and $1,000 in credits· Anthropic 为初创企业提供免费的 Claude 团队使用和 1000 美元信用额
"We created this program because we believe the benefits of AI will reach most people thro…
行业资讯精选 80
Atlassian and OpenAI expand partnership to turn enterprise knowledge into action· Atlassian 与 OpenAI 扩展合作
Atlassian and OpenAI are expanding their partnership to connect frontier models with enter…
行业资讯精选 65
LibreOffice says ‘no AI’ is now a software feature· LibreOffice：不使用AI现在是软件特性
The maker of the open source document editor says it has no plans to add AI to its softwar…
行业资讯精选 82
Sharing AI progress in mathematics· 在数学中分享 AI 进展
OpenAI publishes new results on open problems in mathematics from an internal frontier mod…
行业资讯精选 82
How Jump Trading is scaling quant research with ChatGPT· Jump Trading 如何利用 ChatGPT 扩展量化研究
Jump Trading uses OpenAI to expand quantitative research. See how longer-running AI workfl…
行业资讯精选 82
Advancing computer use with Ironclad· 用 AI 提升专业工作中的合同流程
Learn how OpenAI and Ironclad are training and evaluating AI agents on complex contracting…
YouTube 热门 AI 视频
AI 视频精选 85
Building AI Assistants with On-Device Memory #aiagents #deeplearningai· 在设备上构建AI助手
YouTube · DeepLearningAI
📺 DeepLearningAI▶ 20 次观看
AI 视频精选 78
Production Monitoring with Codex: Grafana, Kubernetes, & Security· 使用 Codex 监控生产：Grafana、Kubernetes 与安全
YouTube · OpenAI
📺 OpenAI▶ 1.1k 次观看
AI 视频精选 85
DeepMind's New AI Just Cracked The Code Of Life· DeepMind 新AI破解生命密码
YouTube · Two Minute Papers
📺 Two Minute Papers▶ 552 次观看
AI 视频精选 85
R&D Part 2· OpenAI 研发进展
YouTube · OpenAI
📺 OpenAI▶ 529 次观看
AI 视频精选 75
Top 3 new model launches at Gemini Audio at Night· Gemini 音频夜发布 3 款新模型
YouTube · Google for Developers
📺 Google for Developers▶ 162 次观看
AI 视频精选 82
Introducing the Decisions API· 推出决策 API
YouTube · OpenAI
📺 OpenAI▶ 751 次观看
AI 视频精选 80
Introducing EmbeddingGemma 2: An open model for natively multimodal embeddings· 介绍 EmbeddingGemma 2
YouTube · Google for Developers
📺 Google for Developers▶ 1.7k 次观看