论文精选 82
Elucidating the Design Space of LM Data Attribution· 阐明大模型数据归属的设计空间
Data attribution methods for Language Models (LMs) are typically evaluated using downstrea…
领域：cs.CL作者：Steven Yu、Sheng-Yu Wang、Kimia Hamidieh
📰 AI 研究简报 · 10月31日
今日头条：世界模型首次指导设计芯片！超大核NPU700TOPS，99%算力可用 —— AI算力利用“九九成”
共 56 条预计阅读 34 分钟Token 消耗 100,800
arXiv 最新论文
论文精选 83
ShallowDefend: Harnessing Shallow Safety Alignment for Multi-Turn Defender
Safety alignment has enabled large language models to reliably refuse harmful requests dir…
领域：cs.CL作者：Qiyao Ma、Xiaodong Liu、Weiwei Yang
论文精选 85
Testimony from Timelines: Evidence-Grounded Significance Testing of Temporal Clinical Hypotheses· 时间线证据在临床假设检验中的应用
Ask a language model whether sales tend to dip after a product recall, or whether one side…
领域：cs.CL作者：Chengyang He、Wenlong Zhang、Koduvayur Subbalakshmi
论文精选 83
When to Assist and What to Suggest: Joint Gating and Memory-Grounded Hint Generation in One Compact Decoder· 何时辅助与建议：紧凑解码器中的联合门控和基于记忆的提示生成
Proactive assistants offer short, memory-grounded hints during a conversation, deciding wh…
领域：cs.CL作者：A. J. A. Jubayer Talukder、Mohammad Ariful Haque
论文精选 65已解读
Collaborative AI agents for autonomous experimental research with solid-state quantum spins· 基于固态量子自旋的协作AI代理自主实验研究
An autonomous research laboratory could conduct an entire study, from formulating question…
领域：cs.AI作者：El Mustapha Mansouri、Eikichi Kimura、Masato Koga
论文精选 65已解读
Pre-Crime for Agents, at the Last Second: Probing When a Recruited Agent Gives In· 预判智能体在最后时刻泄露信息
When AI agents share channels and credentials, one agent can press another to break its op…
领域：cs.AI作者：Leonidas Raghav
论文精选 65已解读
VeriCUA: Verification as Agentic Exploration for Computer-Using Agents· 面向代理探索的 CUA 验证框架
Building and scaling computer-using agents (CUAs) across desktop, mobile, and web environm…
领域：cs.AI作者：Qiushi Sun、Da Yu、Alexander Knop
论文精选 83
Multi-View MoE: Letting Every Router Shape Expert Computation· 多视图 MoE：让每个路由器塑造专家计算
Sparse mixture-of-experts (MoE) models expand parameter capacity by activating a small sub…
领域：cs.CL作者：Zelin Zhao、Tianchi Ma、Junyuan Shang
论文精选 65已解读
Comparing Human Role-Players and LLM-Simulated Clients in Online Counselling Training: An Analysis of Counselling Patterns· 人类角色扮演者与LLM模拟客户在在线辅导培训中的比较
This study investigates the capabilities of Large Language Models to simulate counselling …
领域：cs.AI作者：Eric Rudolph、Philipp Steigerwald、Jens Albrecht
论文精选 65已解读
FARMS: A Verifiable Foundation for Automated Computational Science· FARMS：自动计算科学的可验证基础
Automated computational science could accelerate discovery by selecting, adapting and deve…
领域：cs.AI作者：Di Wu、Ling Liang、Haizhao Yang
论文精选 65已解读
Advancing Clinical Genomic Analysis with a Skill-Centred Agent· 用技能中心代理推进临床基因组分析
Clinical genomic analysis requires experts to integrate evolving guidelines, databases, an…
领域：cs.AI作者：Qiaoyu Zheng、Weike Zhao、Yanjie Fan
论文精选 65已解读
Ramanujan Challenge Results: AI-Assisted Evaluation Pipeline for Multi-Format Proofs· 拉马努金挑战赛结果：多格式证明的人工智能辅助评估管道
As AI's mathematical capabilities rapidly improve, meaningful benchmarks are shifting from…
领域：cs.AI作者：Alan Sorani、Ashvni Narayanan、Michael Shalyt
论文精选 65已解读
Retrieval as Traversal: Value-Indexed Memory for Stateful Long-Context Agents
Long-context models repeatedly solve closely related retrieval problems. Attention or glob…
领域：cs.AI作者：Daniel Strauß、Zhenshan Bing、Alois Knoll
论文精选 65已解读
On the Fundamental Limits of Prefill-Based Decoding Length Prediction· 预填充解码长度预测的基本限制
Predicting how many tokens a large language model (LLM) will decode, using only prefill-ph…
领域：cs.AI作者：Haowen Shi、Xuzhe Yan、Fahao Chen
论文精选 65已解读
Efficient Scalable Oversight via Rewinding· 通过重播提高可扩展监督效率
Scalable oversight asks how a weak verifier can check the outputs of computations performe…
领域：cs.AI作者：Ziyi Guan、Yael Tauman Kalai、Ron Rothblum
论文精选 65已解读
Algebraic World Model: From Exploration to Proofs· 代数世界模型：从探索到证明
AI has made remarkable progress in mathematics, solving challenging problems with machine-…
领域：cs.AI作者：Guancheng Pan、Qixin Tan、Yang Yuan
论文精选 65已解读
Agentic University: Collaborative Agents for Personalized Learning· Agentic University: 个性化学习的协作智能体
Large language models support personalized tutoring, content generation, and adaptive lear…
领域：cs.AI作者：Jingru Fan、Tianle Zhou、Yunjie Zhou
论文精选 82
FIRME: Registry-Anchored Memory Graphs for Evolving, Heterogeneous Investor Knowledge· 金融投资者知识的注册记忆图
Financial AI workflows that reuse investor knowledge need persistent memory that integrate…
领域：cs.AI作者：Yunbai Zhang、Yunyi Gao、Han-Tai Shiao
论文精选 81
Training language models in simulation for psychiatric intake· 强化学习训练语言模型进行精神科接诊
In psychiatric practice, the clinical intake interview remains a primary diagnostic instru…
领域：cs.AI作者：Samuel Schmidgall、Tim Strother、Alex Bijamov
论文精选 85
Prefix Executability: Evaluating Tool-Using Agents Beyond Final Success· 工具使用代理的前缀可执行性评估
Tool-use benchmarks often score agents by final task success, but terminal outcomes can ob…
领域：cs.AI作者：Amir H. Rezaeian、Weiyi Sun、Yassine Benajiba
论文精选 85
Retrieving from Your Own Mistakes: On-Policy Training of Embedding Models for Multi-Hop Retrieval-Augmented Generation
Complex questions in retrieval-augmented generation require evidence spread over several d…
领域：cs.AI作者：Chenjie Qiu、Zixuan Zeng、Shaobin Shi
论文精选 85
Beyond Selected Breakthroughs: Evaluating Language Models on Randomly Sampled Mathematical Conjectures· 超越特定突破：在随机数学猜想上评估语言模型
The last year has seen a shift in AI-assisted mathematics, with frontier language models p…
领域：cs.AI作者：Helen Jenne、Davis Brown、Jesse He
论文精选 82
Multistep and Multivariate Time Series Explainability with GradientSHAP: A Wind Power Case Study· 多步骤多变量时间序列解释性研究：风力发电案例
As the effects of climate change become increasingly apparent, the demand for renewable en…
领域：cs.AI作者：Robbie Southam、Zhong Fan、Ke Li
论文精选 82
Neighborhood Elites: Quality-Diversity Search for Uncertainty Quantification in Question Answering· 邻域精英：问答中不确定性量化的质量多样性搜索
Large language models (LLMs) are increasingly capable, yet they produce wrong answers indi…
领域：cs.AI作者：Mesut Ceylan、Arda Baris Basaran、Elgar Fleisch
论文精选 83
Distributional Reward Calibration and Budget-Neutral Credit Assignment for Multimodal RLVR· 多模态强化学习中的分布奖励校准和预算中立信用分配
We revisit advantage estimation in Group Relative Policy Optimization (GRPO) and observe t…
领域：cs.AI作者：Lang Chen、Zongxin Xu、Borui Wang
GitHub 热门开源项目
开源项目精选 82
boardshogunfluctuate/Capri-AI· Capri-AI：桌面AI代理
⚡ Tell your PC what to do. Capri is a cyberpunk AI agent that works with your files, proje…
⭐ 196 stars
开源项目精选 82
EntranceStarfish/VoiceStudioV2· 本地 AI 语音工作室
Local AI voice studio: real-time voice changer, text-to-speech, voice cloning & dubbing wi…
⭐ 128 stars
开源项目精选 75
hhtower/price-comparison
AI 订阅与卡网比价，聚合 ChatGPT / Codex、Claude、Gemini、Grok、Cursor 等代充、成品号与兑换码报价，提供价格走势、官方地区价参考和 AI /…
⭐ 84 stars语言 TypeScript
开源项目精选 65
CCharlie-xiu/wuch-ai· Wuch-ai 界面设计
说一句，界面在画布上长出来 · https://wuch-ai.net
⭐ 67 stars语言 JavaScript
精选 AI 行业资讯
行业资讯精选 82
从“懂回答”到“能执行”：Zoho如何让AI走进企业业务
10月20日，在以“助力中国企业全球化成长”为主题的Zoho北京媒体沟通会上， Zoho不仅向外界展示了其最新的企业级AI全景战略，更试图回答一个当前行业的普遍痛点：当AI大模型喧…
行业资讯精选 85
世界模型首次指导设计芯片！超大核NPU700TOPS，99%算力可用
AI算力利用“九九成”
行业资讯精选 75
刚刚，Claude宣布取消5小时限制！首批Ultra用户已「解封」
200美元Ultra用户先用
行业资讯精选 82
智能驾驶进入物理AI时代？7位行业嘉宾给出最新判断
10月26日，第八届智能驾驶大会暨2026环秀湖物理AI产业发展推介会在苏州相城举行。 上一届，车路云一体化、智能驾驶出海和无人驾驶商业化是讨论重点，而到了今年，大会的讨论边界进一…
行业资讯精选 75
Nvidia weighs playing the cosigner for Nscale’s AI data center in West Virginia· 英伟达考虑为 Nscale 的 AI 数据中心提供贷款担保
When investors get nervous about AI spending, it helps to have the chipmaker that sells th…
行业资讯精选 65
Ollie, a privacy-focused AI assistant, is shutting down· 注重隐私的AI助手Ollie将关闭
As an increasing number of AI assistants flood the market, the founder of Ollie, an assist…
行业资讯精选 85
阿里巴巴与 Apple 达成合作，将千问集成至 Apple 智能· 阿里巴巴与 Apple 达成合作，将千问集成至 Apple 智能
10月31日，阿里巴巴与Apple达成合作，将千问集成至 Apple 智能，中国用户可以在Siri、写作工具及视觉智能中，直接调用千问的文本与图像理解、内容生成等能力。这些功能将于…
行业资讯精选 65
Intel’s stock crashes, knocking $12B from government’s stake· 英特尔股价暴跌，政府持有的股份价值缩水120亿美元
Intel is being hurt by PC sales as the AI industry gobbles memory and other components cau…
行业资讯精选 65
Anthropic expected to receive high ‘junk’ rating from credit agencies following IPO, report says· Anthropic IPO 后可能获高垃圾级信用评级
The new report says the AI lab will likely qualify for a high junk rating.
行业资讯精选 75
Cracks are emerging in the AI spending boom· AI 投资热潮出现裂痕
Big Tech is pouring billions into AI infrastructure, but mounting debt, negative cash flow…
行业资讯精选 75
Zest’s new AI agent finds restaurants and books reservations over text· Zest 的 AI 餐厅预订助手
Can't agree on where to eat? Zest's AI concierge can combine everyone's tastes, put restau…
行业资讯精选 85
Google’s first data center satellite is online and writing poetry· 谷歌首颗数据中心卫星上线写诗
The first TPU in space is a running a Google DeepMind Gemma 4 model and writing cheesy poe…
行业资讯精选 65
Theranos whistleblower Tyler Shultz warns that the AI boom is ‘a great opportunity to go deceive people’· 警惕 AI 领域的‘巨大机会’
The cautionary Theranos tale takes on a new significance in the AI era, when companies tha…
行业资讯精选 75
Things That Go Beep in the Night: Agentic AI Joins the Search for Extraterrestrial Signals· 夜半的哔哔声：代理 AI 加入搜索外星信号
While science fiction lore tells us that humans will meet extraterrestrial life in the for…
行业资讯精选 80
榜一，拿来吧你！智元这波双模型杀穿多榜榜单
xs，还要给机器人打绩效
行业资讯精选 80
李飞飞挖的三个「坑」：从 ImageNet 到空间智能，这一次她会怎么填？
<img class="rich_pages wxw-img" data-aistatus="1" data-croporisrc="https://mmbiz.qpic.cn/s…
行业资讯精选 83
全行业两年攒下百万小时数据，影溯想 1 个月生产完：全国首个具身仿真数采中心落地杭州
<img data-src="https://static.leiphone.com/uploads/new/images/20261030/6ae4767186efc.jpg" …
行业资讯精选 82
AI零帧起手写算子，反超DeepSeek人类手写巅峰：快了9.8%！
之前还连登昇腾三榜第一
行业资讯精选 82
“超能模式”重磅登场！联想天禧AI 4.5全端升级，三大场景打造“职场超能搭档”
10月30日，联想天禧AI 4.5迎来AIPC端正式发布。
行业资讯精选 75
Making it easier for websites to express AI training preferences· 让网站更容易表达 AI 训练偏好
OpenAI will honor a shared robots.txt signal to help website operators communicate their A…
行业资讯精选 80
Community safety report and new protections for teens· 社区安全报告和对青少年的新保护
OpenAI is strengthening violence prevention in ChatGPT with expanded safeguards, earlier i…
行业资讯精选 80
好莱坞最缺的答案，被这家 10 亿美元估值的 AI 影视独角兽写进了片单
创立 4 年，估值站上 10 亿美元——资本市场定价的不是一部电影，而是一套内容生产基础设施。
行业资讯精选 80
Fair use benefits everyone· 版权公平使用支持AI创新
OpenAI shares excerpts from court briefs by creators, academics, accessibility advocates, …
行业资讯精选 85
Credit Saison achieves 954% ROI in a ChatGPT Enterprise pilot· Credit Saison 通过 ChatGPT 试点实现 954% ROI
Credit Saison’s ChatGPT Enterprise pilot delivered 954% ROI. See how leaders and employees…
YouTube 热门 AI 视频
AI 视频精选 82
NVIDIA GTC Berlin 2026 Event Highlights: AI in Action Across Europe· NVIDIA GTC柏林2026：欧洲AI应用亮点
YouTube · NVIDIA
📺 NVIDIA▶ 68 次观看
AI 视频精选 85
OpenAI Claims It Just Did The Impossible· OpenAI 声称实现不可能任务
YouTube · Two Minute Papers
📺 Two Minute Papers▶ 5.3k 次观看
AI 视频精选 83
How Gen uses ChatGPT Work to turn data Into decisions | OpenAI· Gen 如何使用 ChatGPT 将数据转化为决策
YouTube · OpenAI
📺 OpenAI▶ 235 次观看