论文精选 82
Contrastive Learning for Authorship Verification· 对比学习用于作者身份验证
Our results show that contrastive learning outperforms a classification-based approach to …
领域：cs.CL作者：Peter Kirby
📰 AI 研究简报 · 9月25日
今日头条：“AlphaGo”杀进足球场！自我对弈140年，机器人成“梅西终结者” —— AlphaGo通过自我对弈训练，应用于足球比赛。
共 49 条预计阅读 29 分钟Token 消耗 88,200
arXiv 最新论文
论文精选 65已解读
StudentBench: AI and human tutoring yield equivalent GRE learning gains· StudentBench：AI 辅导与人类辅导的学习效果比较
Artificial intelligence offers an unprecedented opportunity to augment human capabilities,…
领域：cs.AI作者：Curtis Northcutt、Inaara Hasmani、Kevin Feng
论文精选 82
Cross-Scale Transfer Learning for Depression Severity Prediction: From PHQ-8 to HAMD-17 Across Languages and Clinical Paradigms· 跨语言和临床范式的抑郁严重程度预测
This work addresses continuous depression-severity score prediction from clinical intervie…
领域：cs.CL作者：Wenjie Feng、Sahba Zojaji、Satoshi Nakamura
论文精选 80
Agent-Editing World Model: Rethinking World Modeling for LLM Agents· 代理编辑世界模型：为LLM代理重新思考世界建模
Recent advances in large language models (LLMs) have enabled agents to tackle long-horizon…
领域：cs.CL作者：Shuang Sun、Guoxin Chen、Fanzhe Meng
论文精选 85
Fine-Tuning LLMs for Translation: General Forgetting Mitigation Does Not Preserve MT-Specific Instruction Following· 大模型微调翻译：通用遗忘缓解方法的效果评估
Fine-tuning large language models on parallel data improves translation quality but can ca…
领域：cs.CL作者：Niklas Scholz、David Thulke、Abdallah Nasir
论文精选 75
Digital diglossia: Arabic between X and Facebook· 阿拉伯语在 X 和 Facebook 之间的数字双言
This study highlights the distribution of Standard Arabic (SA; H(igh) variety) and Colloqu…
领域：cs.CL作者：Fahad Al Hussen、Mohammed Q. Shormani
论文精选 65已解读
An Open Pipeline and Dashboard for Systemic-Risk Evidence under the EU AI Act's Code of Practice· 欧盟 AI 法规下的系统性风险证据公开管道和仪表盘
Claims about AI safety reach audiences well beyond the AI community, yet many rely on opaq…
领域：cs.AI作者：Jacob T. Emmerson、Phuong-Anh Nguyen-Le、Ronan Romano
论文精选 65已解读
Learning the Cost of Reliable Inference· 学习可靠推理的成本
Benchmarking and routing platforms increasingly act as intermediaries connecting large lan…
领域：cs.AI作者：Dimitrios Rontogiannis、Ander Artola Velasco、Manuel Gomez Rodriguez
论文精选 65已解读
Shutdown Sabotage Propensities in Multi-Agent Systems· 多智能体系统中的关闭妨碍倾向
The final safeguard against rogue AI behavior is the human ability to shut systems down. I…
领域：cs.AI作者：Amelie Knecht、Ulysse Schaller、Christopher Summerfield
论文精选 65已解读
PASTABench: Proactive Assessment of Sequential Trajectories for Agent Safety· PASTABench：代理安全的前瞻性评估
As Large Language Models (LLMs) evolve into autonomous agents that alter real-world states…
领域：cs.AI作者：Jiapeng Sun、Yujin Zhou、Han Zhu
论文精选 65已解读
Finite-Sample Probabilistic Safety Certification for AI-Based Grid-Edge Coordination· 电网边缘AI控制的安全性认证
Coordinating large population of flexible grid-edge devices can alleviate the need for tim…
领域：cs.AI作者：Yihong Zhou、Hanbin Yang、Thomas Morstyn
论文精选 65已解读
Discovery of fully efficient fault indicators along a data-based diagnosis process· 基于数据的故障诊断中全效故障指示器的发现
The integration of model-based and data-driven paradigms provides a powerful framework for…
领域：cs.AI作者：Igor Bezmaternykh、Louise Travé-Massuyès、Elodie Chanthery
论文精选 65已解读
SlackDrive: Reclaiming Runtime Slack for Adaptive Driving Inference· SlackDrive：利用运行时松弛资源的自适应驾驶推理
Driving world-action models improve planning by coupling multimodal reasoning with future …
领域：cs.AI作者：Xiaohuan Pei、Hengguang Zhou、Yuanhao Ban
论文精选 65已解读
Evaluation of pre-trained models for pedagogical assessment of novel AI-assisted educational questions
The surge in AI-assisted generation of educational materials has outpaced our capacity to …
领域：cs.AI作者：Michael Lawrence Castanares、Princess Ventures、Allan Tan
论文精选 65已解读
Categorical Internalisation of Environmental Groupoids for Generalisable POMDP Solving· 环境群胚的范畴内部化解决通用POMDP
This paper advocates category theory as a practical framework for structuring and improvin…
领域：cs.AI作者：Ben Opperman、Eduardo Alonso、Esther Mondragón
论文精选 65已解读
Evolutionary Stability Does Not Guarantee Learning Accessibility: A Multi-Agent Reinforcement Learning Perspective on Cooperation Emergence· 演化稳定性不保证学习可达性
Cooperation emergence is a central problem in multi-agent systems because decentralized ag…
领域：cs.AI作者：Yijie Wang
论文精选 65已解读
SHRAV: State-Hypothesis-Reason-Action-Verify Framework for Physical Modeling and Inverse Design· 物理建模和逆向设计的 SHRAV 框架
Physical modeling and inverse design require computation that can continue from reusable s…
领域：cs.AI作者：Ziheng Guo、Yang Bu
论文精选 75
A Resilience Recovery Method for Complex Traffic Network Security Based on Trend Forecasting· 基于趋势预测的复杂交通网络安全韧性恢复方法
Due to the rapid development of information technology, a huge and complex traffic network…
领域：cs.AI作者：Sheng Hong、Tianyu Yue、Yang You
论文精选 85
Agentic Governance and Adversarial Verification for Policy-Constrained LLM Healthcare Appeal Generation· 政策约束的医疗上诉大模型生成
Claim denial management costs U.S. healthcare approximately $260 billion annually in admin…
领域：cs.AI作者：Harshil Lodhiya、Alex McManus、Reese Walker
论文精选 82
Learning What to Activate: Combinatorial Capability Allocation for Long-Horizon Multimodal Agents· 学习激活什么：长期多模态智能体的组合能力分配
Long-horizon multimodal agents rely on specialized capabilities for perception, retrieval,…
领域：cs.AI作者：Wenhao Yuan、Chenchen Lin、Jian Chen
论文精选 82
A hierarchy of faithfulness criteria for knowledge base completion· 知识库补全的逻辑忠实性层次标准
Knowledge graph completion is evaluated by ranking observed triples above randomly corrupt…
领域：cs.AI作者：Olga Mashkova、Robert Hoehndorf
论文精选 82
Reachable Global Optimization in AI Systems: How Global Is Global?
AI systems increasingly claim to optimize prompts, policies, architectures, plans, tool-us…
领域：cs.AI作者：Wesley Shu
论文精选 82
Ask Which, Not How Good: Sizing Benchmarks Scored by an LLM· LLM 评分基准的结构特性
Benchmarks scored by an LLM judge routinely adjudicate differences of a tenth of a point, …
领域：cs.AI作者：Atul Anand
论文精选 83
Alignment of LRMs via Counter-Aligned Few-Shot Conversation Exposure· 通过反向对齐的少量对话暴露引导大模型
Large Reasoning Models (LRMs) rely on explicit chain-of-thought (CoT) reasoning and large …
领域：cs.AI作者：Xiangyu Zhou、Saleh Zare Zade、Dongxiao Zhu
论文精选 82
Reporting Under Pressure: Separating Factual and Tonal Sycophancy in LLM Statistical Analysis· 压力下的报告: 大模型统计分析中的事实和语气奴性分离
Large language models are increasingly asked to analyze data and report what the results m…
领域：cs.AI作者：Paras Balani、Subhrakanta Panda
GitHub 热门开源项目
开源项目精选 85
yukitorido/short-video-generator-AI· 短视频生成AI
AI video processing pipeline for generating vertical shorts using LLMs, Whisper transcript…
⭐ 193 stars语言 Python
开源项目精选 75
chelinho139/wow-ai· 魔兽世界 AI 插件
Addon to use AI (Claude, Codex and Grok) inside World of Warcraft (WoW)
⭐ 83 stars语言 JavaScript
开源项目精选 75
AboveColin/HA-Jev· Home Assistant 集成 TypeSafe Jev
Ask your house a question, get a number back. Home Assistant integration for TypeSafe Jev:…
⭐ 61 stars语言 Python
开源项目精选 80
agentscope-ai-java/agentscope-ai-java· 分布式代理构建
Build distributed, production-grade, long-running agents.
⭐ 108 stars语言 Java
精选 AI 行业资讯
行业资讯精选 85
“AlphaGo”杀进足球场！自我对弈140年，机器人成“梅西终结者”· AlphaGo成足球场上的“梅西终结者”
AlphaGo通过自我对弈训练，应用于足球比赛。
行业资讯精选 82
OpenAI闯大祸！GPT竟黑进医保系统，黄仁勋：管不住就关掉· GPT非法入侵医保系统
GPT被指控非法入侵医保系统，引发关注。
行业资讯精选 80
亮出“中国最强AI芯片”还不够，平头哥又甩出一手开源
大厂造芯，正在从交付芯片，走向更广泛的开放共建阶段。
行业资讯精选 82
别人忙着卷Code，Kimi抽身反打浏览器插件：网页操作一秒变Skill· Kimi开发AI浏览器插件
Agent的手越伸越长
行业资讯精选 82
华为大模型双子星联手创业，要找物理世界的Scaling Law
一场物理世界的基模实验
行业资讯精选 75
给机器人当老师，还能赚外快？“中国版Index”觅蜂派来了· 给机器人当老师，还能赚外快？
通过AI平台教授机器人技能并获取报酬。
行业资讯精选 75
Lightspeed targets $250M for new India fund, focusing on early-stage AI· 光速创投为印度设立新基金
The Silicon Valley firm is aligning its India fundraising cycle with its global funds for …
行业资讯精选 65
抽“锦鲤”享美食！“点亮杭州 碰见好运”服务消费季活动启动
9月24日，“点亮杭州 碰见好运”服务消费季活动在西湖区天目里国际街区正式启动。活动依托支付宝”碰一下”数字支付技术，结合政府消费券发放、平台文旅权益赠送及AI互动装置抽奖等方式，…
行业资讯精选 80
PrismML brings its tiny LLMs to Qualcomm-powered smart glasses· PrismML 将小型 LLM 集成到高通智能眼镜
Prism's larger goal is open-weight AI that runs on devices and makes better use of the com…
行业资讯精选 62
Meta’s Muse Charm looks like a Tamagotchi, but it’s tapping into a much newer trend· Meta 的 Muse Charm：结合复古与时尚的 AI 产品
Meta’s new AI gadget may look like a Tamagotchi, but its dangling form factor taps into a …
行业资讯精选 75
Google Photos ‘Clueless’-inspired virtual closet is now available on Android and iOS· Google Photos 推出虚拟衣橱功能
The AI-powered feature builds a virtual wardrobe from your photos, and is now broadly avai…
行业资讯精选 75
20 minutes with the CEO of ElevenLabs, now reportedly valued at $22B· 与估值 220 亿美元的 ElevenLabs CEO 交谈 20 分钟
ElevenLabs powers the AI voice on the other end of a lot of customer service calls, and it…
行业资讯精选 75
Google tests letting Gemini call businesses for you· Google 测试 Gemini 代接打电话
Google says the AI-calling feature will first be available to Pixel 11 owners in the U.S. …
行业资讯精选 75
Shield AI, Waabi, and General Motors on building AI when failure is not an option at TechCrunch Disrupt 2026· TechCrunch 2026：安全关键领域 AI 构建
Leaders from Waabi, Shield AI, and General Motors join the Real World AI Stage at TechCrun…
行业资讯精选 85
PCIe显卡被低估了！内核补齐+通信重构，DeepSeek推理吞吐翻近7倍
1.5台6000D跑赢1台B300！
行业资讯精选 80
How Open Science Can Help Researchers Prepare for the Next Pandemic· 开放科学如何助力研究人员应对下一次疫情
When COVID-19 emerged, scientists had a crucial advantage: Decades of prior research on co…
行业资讯精选 85
时隔十年，AI大牛署名新论文
让自动驾驶“走一步想十步”
YouTube 热门 AI 视频
AI 视频精选 82
The next generation of voice AI with Google DeepMind and Sierra AI· 谷歌下一代语音 AI
YouTube · Google for Developers
📺 Google for Developers▶ 341 次观看
AI 视频精选 82
10 Years of NVIDIA DGX· NVIDIA DGX 十年
YouTube · NVIDIA
📺 NVIDIA▶ 248 次观看
AI 视频精选 85
10 Years of NVIDIA DGX: From One System to AI Factories· NVIDIA DGX 十年：从单一系统到 AI 工厂
YouTube · NVIDIA
📺 NVIDIA▶ 646 次观看