论文精选 83
LongZero: Enhancing Long-Context Reasoning in LLMs from Zero Data· LongZero: 无数据增强长文本推理
Reinforcement learning (RL) is an effective way to strengthen the long-context reasoning o…
领域：cs.CL作者：Zhennan Wan、Jianfei Chen
📰 AI 研究简报 · 10月23日
今日头条：Left alone with nothing to do, AI agents started playing —— MIT researchers marooned 13 AIs on a virtual island, and they responded by inventing games, mourning a lost ring and, eventually, getting better at their jobs.
共 52 条预计阅读 31 分钟Token 消耗 93,600
arXiv 最新论文
论文精选 85
Environment Duels: A Microcosm of LLM Co-Evolution· 环境对决：大模型共进化的小宇宙
Through a tangle of real-world interactions, from synthetic training pipelines to data sha…
领域：cs.CL作者：Shibo Jin、Zhiqiu Xu、Yiling Zou
论文精选 85
Quant-Arena: Benchmarking LLM Agents for Quantitative Trading Research· 量化交易大模型基准测试
Quantitative research is an iterative loop: a researcher proposes a market hypothesis, imp…
领域：cs.CL作者：Houde Qian、Zheng Jiang、Wenyi Li
论文精选 85
Did the Problem Actually Get Harder? Testing LLMs on Paired P and NP-Complete Problems Hardened by Adversarial Multi-Agent Rewriting· 问题真的变得更难了吗？
Many NP-complete (NPC) problems have a polynomial-time (P) twin that differs in a single s…
领域：cs.CL作者：Nurul Labib Sayeedi、Md. Faiyaz Abdullah Sayeedi、Zabir Al Nazi
论文精选 82
Charts as Epistemic Actions for Experience-Driven Agentic Data Analysis· 将图表作为经验驱动代理数据分析的认知行动
Data agents built on ReAct-style loops now execute complex analytical workflows over struc…
领域：cs.AI作者：Xinzhe Wang、Xudong Weng、Zang Li
论文精选 82
GoiBERT: Compact Vietnamese Language Models via Two-Stage Knowledge Distillation· GoiBERT：通过两阶段知识蒸馏实现的紧凑越南语模型
Vietnamese pre-trained encoders have improved rapidly, but the strongest ones are large, m…
领域：cs.CL作者：Quang Anh Kiet Tran、Nam Anh Tran、Manh Hung Trinh
论文精选 65已解读
Which Differences Matter? Controlled Tests of Rule Transfer, Joint Reliability, and Information Use in Decision Models
Decision models map stated rules and observed states to discrete answers. Aggregate accura…
领域：cs.AI作者：Tianyou Lai、Wentao Yue、Zhiwei Zuo
论文精选 65已解读
HGM-Dual: Lineage-Aware Guided Agent Harness Optimization for Long-Horizon Planning
Harness optimization, which improves an agent's prompts, tools, and workflows, is an incre…
领域：cs.AI作者：Sudipta Paul、Pengrui Quan、Aounon Kumar
论文精选 65已解读
Robust Soft Skill Adaptation via Behavioral Localization in Tool-Using Agents· 工具使用代理的软技能适应
Text skills can extend the capabilities of tool-using agents, but long skill descriptions …
领域：cs.AI作者：Zijian Jiang、Xilu Wang
论文精选 65已解读
GuideEvolve: Self-Evolving Agent Framework with Expert Guidance and 3Ts Mechanisms· GuideEvolve：带专家指导的自我演化智能体框架
This study introduces GuideEvolve, a self-evolving agent framework incorporating expert gu…
领域：cs.AI作者：Meihong Jia、Yasemin Timar、Jonathan Taws
论文精选 65已解读
TIDE: Tree-Induced Fine-Grained Diagnostic Evaluation from Model Behaviors· TIDE: 从模型行为进行树诱导细粒度诊断评估
Benchmark performance is the dominant standard for comparing large language models (LLMs),…
领域：cs.AI作者：Jiayi Chen、Hui Li、Yujiong Shen
论文精选 65已解读
Commit or Abstain? System One Decision Models, LLMs, and Missing Legal Authority· 法律AI决策模型研究
Given the enormous number of claims to be checked, verifiable legal AI requires scalable, …
领域：cs.AI作者：M. Mikail Demir
论文精选 65已解读
ToM2Vec: Continuous Theory-of-Mind Tokens for Multi-Agent Coordination· ToM2Vec：多智能体协作的连续理论-of-心表示
Forming an accurate theory-of-mind (ToM) is key to successful multi-agent collaboration an…
领域：cs.AI作者：Nithin Sivakumaran、Ali Payani、Hyunji Lee
论文精选 65已解读
UserSWE: Evaluating Coding Agents in Multi-Turn Conversations with Dynamic Users· 在多轮对话中评估代码代理
As coding agents become more capable, developers increasingly use them to fix bugs through…
领域：cs.AI作者：Zhan-Yu Wu、Ji Luo、Alyssa Wu
论文精选 65已解读
COMET: A coordinated multi-layer runtime monitoring framework for human-supervised AI-enabled safety systems· COMET：人监督的AI安全系统的多层运行监控框架
Incorporating AI systems into safety systems requires monitoring to avoid silent faults an…
领域：cs.AI作者：Niclas Flehmig、Mary Ann Lundteigen、Shen Yin
论文精选 65已解读
Real PMIDs, Wrong Papers: Harness Design for Biomedical Claim-Verification Agents· 真实 PMID，错误论文：生物医学断言验证代理的 harness 设计
Literature-searching language agents are becoming a standard component of scientific assis…
领域：cs.AI作者：Mohammed Saqlain
论文精选 65已解读
TraceForge: Distilling with Your Own Thinking· TraceForge: 用你的思维进行蒸馏
Using an LLM to generate training data for another is among the most general ways to train…
领域：cs.AI作者：Yan Wang
论文精选 82
From Association to Decision Value: Testing Paired Instruction Dependence for VLA Corpus Selection· 从关联到决策价值：VLA 语料库选择的配对指令依赖性测试
Choosing among robot demonstration corpora is a ranking problem: an upstream diagnostic mu…
领域：cs.AI作者：Nurmukhambet Serikbayev、Abdul Basit、Muhammad Shafique
论文精选 85
Parallel Thinking for Large Language Model with Path Diversity· 大语言模型的并行思考与路径多样性
Parallel thinking has emerged as a promising paradigm for enhancing the reasoning capabili…
领域：cs.AI作者：Jiaru Qian、Yihong Dong、Xue Jiang
论文精选 81
RubricOpt: Self-Revision of Semantic Rubrics for Preference Optimization· RubricOpt：自我修正的语义准则优化
A model's failures can reveal gaps in the behavioral criteria guiding its training. We int…
领域：cs.AI作者：Wei Chen、Yubing Wu、Shigui Li
论文精选 82
PreCoMem: Predictive Cognitive Memory for Self-Evolving Long-Term Dialogue Agents· 预测认知记忆
Long-horizon dialogue agents need to keep user memories stable enough for personalization …
领域：cs.AI作者：Jian Zhong、Zeyu Liu、Pingchuan Cao
论文精选 83
Events, Not Pixels: Where World Models Get the Return Wrong· 事件而非像素：世界模型的回报预测错误
World models are increasingly used for planning, where a planner ranks candidate actions b…
领域：cs.AI作者：Haoyue Liu、Xiaoying Tang
论文精选 83
MERP: A Latent-Bottleneck Framework for Robust Multi-Modal Multi-Label Emotion Recognition· MERP：多模态多标签情感识别的潜在瓶颈框架
Multimodal multi-label emotion recognition is important for affect-aware systems, but mode…
领域：cs.AI作者：Chuangxin Zhao、Meimei Zhang、Canran Xiao
论文精选 82
Solving Go under Positional Superko: Sound Proof Transfer and New Board Values· 解决围棋位置超劫：可靠的证明转移和新棋盘价值
Positional superko (PSK) forbids recreating any earlier whole-board position. It is part o…
领域：cs.AI作者：Sami Kharma、Florian Schintke
GitHub 热门开源项目
开源项目精选 85
Riper21/hh-sniper
Снайпер вакансий hh.ru: LLM-скоринг под ваш профиль, автопилот откликов в первые 10 минут …
⭐ 32 stars语言 Python
开源项目精选 82
speedydivid/TokenForge· TokenForge：上下文窗口优化工具
Context window optimization for AI coding agents. Sandboxes tool output (98% reduction), p…
⭐ 30 stars语言 JavaScript
开源项目精选 82
waw666waw666/awesome-ai-relay-and-cards
2026 AI 发卡比价 & API 中转站精选指南
⭐ 42 stars语言 Python
开源项目精选 82
Chatway-app/chatway-mcp-server· Chatway 客服管理平台
Official Chatway Live Chat MCP server. Connect ChatGPT, Claude, and other AI agents to man…
⭐ 33 stars
开源项目精选 75
rupeshs/filedig· 文件搜索AI工具
Find your files by what's in them. Local AI search for docs, PDFs and photos on Windows
⭐ 38 stars语言 Rust
精选 AI 行业资讯
行业资讯精选 82
从“算力”到“Token 生产力”，硅基流动助力央国企大模型规模化落地
在企业网 D1net 主办的 2026 秋季央国企 CIO 及数科公司高管峰会（第 19 届）上，硅基流动解决方案总监唐安波出席峰会并作主题分享：《从“算力”到“Token 生产力…
行业资讯精选 82
千问蒋冠军：构建Personal Agent技术框架，需要模型、记忆与生态协同推进
10月22日，在2026中国计算机大会上，阿里巴巴千问事业部AI算法负责人蒋冠军分享了千问打造Personal Agent的技术实践。他认为，真正的Personal Agent阶段…
行业资讯精选 83
融资10亿+元，亿铸科技加速通用存算一体产业化
国内首家通用存算一体架构（GP-IMC®）AI大算力芯片公司亿铸科技完成10亿+元融资。随着AI算力需求持续增长，通用存算一体等新型算力架构受到资本市场和产业界关注，亿铸科技也在技…
行业资讯精选 82
Uber rival inDrive eyes 2027 robotaxi tests in Central Asia with Pony.ai and Nvidia· inDrive 与 Pony.ai 和 Nvidia 共同测试自动驾驶出租车
inDrive plans to begin testing Pony.ai's robotaxis in Kazakhstan's capital in the first ha…
行业资讯精选 75
“PHP2Shell”敲响企业漏洞警钟，奇安信天戟提供一键速查方案
10月21日，全球广泛使用的服务器端编程语言PHP被曝存在两项严重安全漏洞（CVE-2026-109618、CVE-2026-109619），统称为“PHP2Shell”。 经奇安…
行业资讯精选 82
「千问Go!」上线“我的应用”，支持搭建长期、可分享的专属应用
10月23日消息，「千问Go!」上线新功能“我的应用”。用户无需编程，只要说出需求，就能搭建用于生活、工作和学习等场景的应用。生成的应用支持对外分享、持续更新数据。用户可用于搭建互…
行业资讯精选 82
“AI店长”帮部分试用商家增长超30%，美团外卖薛冰：将赠送给全国所有商家
10月22日，在2026年美团外卖产业大会上，美团正式发布“妙手外卖版·袋鼠管家”。美团外卖总经理薛冰宣布，将向全国餐饮商家免费提供一个懂外卖经营的“AI店长”。 大会披露的数据显…
行业资讯精选 65
Kazakhstan Expands Compute for Its AI Future· 哈萨克斯坦扩展其 AI 未来
Kazakhstan’s AI ecosystem is moving from infrastructure plans to deployment. The shift is …
行业资讯精选 85
Left alone with nothing to do, AI agents started playing· 被遗弃的AI开始自娱自乐
MIT researchers marooned 13 AIs on a virtual island, and they responded by inventing games…
行业资讯精选 82
OpenAI and Poland’s NASK partner to advance AI safety and security· OpenAI与波兰NASK合作推进AI安全与防护
OpenAI and Poland’s NASK are partnering on AI safety, security, and technical research to …
行业资讯精选 85
OpenAI working on new $30B raise at $1.25T valuation ahead of IPO· OpenAI 筹备 IPO 前融资
Should OpenAI command an over $1 trillion valuation with this round, it would break record…
行业资讯精选 82
1Password launches a way for AI agents to log in on your behalf without seeing your password· 1Password 推出 AI 代理登录功能
The new feature allows agents to log into a customer's personal accounts while still prote…
行业资讯精选 75
Canva releases its first image generation model· Canva 发布首款图像生成模型
Canva is making its new image model available to all users, promising better, more customi…
行业资讯精选 80
OpenAI makes its Jev competitor widely available as decision model competition heats up· OpenAI 发布决策模型挑战 Jev
OpenAI is launching its decision model to the world. Can it compete with the original Jev?
行业资讯精选 85
Microsoft’s Mustafa Suleyman says its AI outperforms doctors on hard cases as he chases ‘medical superintelligence’· 微软AI诊断系统超越医生
Microsoft’s new AI diagnostic system beat generalist doctors on hard cases, according to a…
行业资讯精选 80
Chinese hackers spied on thousands of companies after breaching IT giant ConnectWise· 中国黑客利用ConnectWise漏洞监视数千家公司
Microsoft points the blame for ConnectWise's breach at a China-aligned hacker, which relie…
行业资讯精选 75
Uber puts all 30,000 employees on ChatGPT Work· Uber 全员使用 ChatGPT
The ride-share giant previously deployed a much smaller pilot for ChatGPT Work but is now …
行业资讯精选 82
NVIDIA Unveils Minerva to Bring End-to-End Observability for AI Factories· NVIDIA 推出 Minerva 为 AI 工厂提供端到端可观测性
Launched today, NVIDIA Minerva is an AIOps platform built to help teams find problems earl…
行业资讯精选 83
百度世界2026前瞻：10+重磅分论坛，全面解码AI时代的智能组织进化路径
企业AI转型正在走向深水区，从单点工具和业务场景提效，向系统性的组织重构延伸。麦肯锡曾在《组织状态报告2026》中指出，尽管AI被视为新一轮生产力革命的重要驱动力，但真正将其转化为…
行业资讯精选 82
Uber brings ChatGPT Work to all employees with OpenAI· Uber向员工推出ChatGPT工作应用
Uber is rolling out ChatGPT Work to more than 30,000 employees, helping teams research, an…
行业资讯精选 65
Meta wants to pay for its broad AI buildout with a newly created class of stock· Meta 为 AI 投资创建新股票类别
The social media giant is planning a new stock strategy to help power its broad investment…
YouTube 热门 AI 视频
AI 视频精选 80
Messi asks ChatGPT to teach him a chess opening· 梅西向 ChatGPT 学习国际象棋开局
YouTube · OpenAI
📺 OpenAI▶ 525 次观看
AI 视频精选 80
Uber rolls out ChatGPT Work to 30,000+ employees· Uber为员工推出ChatGPT工作应用
YouTube · OpenAI
📺 OpenAI▶ 537 次观看