AAI Search
AI 资讯日报 · 2026-09-25

2026-09-25 · AI 资讯日报

今日新收录 165 条公开资讯，按模型 / 产品 / 行业 / 论文 / 观点 自动归类汇编（非 AI 生成，点击可溯源原文）。

今日精选6 条

  1. 1.
    论文研究Contrastive Learning for Authorship Verification— arXiv

    Our results show that contrastive learning outperforms a classification-based approach to authorship verification under the tested settings. We identify loss function, batch size, training duration, p…

  2. 2.
    论文研究Where Should I Join? Robot Group Joining via Language-Guided Goal Prediction— arXiv

    Social navigation typically assumes a specified goal and focuses on reaching it while respecting social conventions, whereas robot group joining requires predicting where to join based on the group's…

  3. 3.
    论文研究Even Sharper Bounds for Transductive Learning and Its Applications— arXiv

    We introduce Sharper Transductive Local Complexity (STLC), a localized complexity method for transductive learning under uniform sampling without replacement. The construction starts from a Bernstein-…

  4. 4.
    论文研究Can LLMs Reason About Runtime Behavior? A Repository-Level Dynamic Benchmark— arXiv

    Large language models (LLMs) are increasingly used in coding tasks, but their ability to reason about code execution remains unclear. Existing repository-level QA benchmarks mainly evaluate static cod…

  5. 5.
    论文研究Nonequilibrium Phases of Repulsive Self-Attention: Chaos, Attention Condensation, and Emergent Locality— arXiv

    We study the nonequilibrium dynamics of a minimal recurrent transformer with $N$ normalized tokens, $Q=K=I$, and a negative value map $V=-I$. Similarity-based attention selects nearby representations,…

  6. 6.
    论文研究Order-Invariant Answers, Order-Sensitive Representations in Mathematical Reasoning— arXiv

    Reordering a set of mathematical rules without changing its meaning should preserve the correct answer, but must a model's internal representations stay invariant too? We investigate this question usi…

模型发布/更新6 条

  1. 1.
    Opus 5.5 for Work— Anthropic

    AI 导读 · Anthropic把Opus推向办公场景，意味着大模型竞争正从聊天转向企业生产力落地。

  2. 2.
    Project Swap: What happens when agents trade for us?— Anthropic

    AI 导读 · AI代理自主交易若普及，信任、责任与市场公平将面临全新挑战。

  3. 3.
    Introducing Gemini 3.8 Live with Live Avatar— Google DeepMind
  4. 4.
    ChatGPT Ads expands to Southeast Asia and Taiwan— OpenAI

    ChatGPT Ads is expanding to Southeast Asia and Taiwan, giving eligible businesses new ways to reach people across more than 60 countries.

  5. 5.
    Google's Suncatcher project aims to put AI data centers in orbit powered by solar energy— The Decoder

    AI 导读 · 把耗电巨大的算力送上太空，用太阳能就地供电，若成真将重塑数据中心选址逻辑。

  6. 6.
    Black Forest Labs launches FLUX 3 Action, an open robotics AI model— The Decoder

    AI 导读 · 视觉语言模型正从生成内容走向驱动机器人行动，开源七亿参数让具身智能门槛大幅降低。

产品发布/更新6 条

  1. 7.
    Google is sending an AI satellite into space next week— The Verge

    Google is getting ready to launch a satellite with its AI processors to test how well they perform in space, as reported earlier by The New York Times. The move is part of Google's…

  2. 8.
    Muse sure looks a lot like OpenClaw— The Verge

    AI 导读 · Meta消费级智能体登顶榜单，AI代理赛道正从技术演示转向用户争夺战。

  3. 9.
    Loopera-ai/loopera— GitHub

    面向基本面因子研究的智能体-A hypothesis-driven AI agent for fundamental factor research, with evidence-gated validation and research memory.

  4. 10.
    消息称 OpenAI 将在数日内预览网络安全专用模型 GPT-6 Cyber— IT之家

    AI 导读 · 垂直安全模型竞争打响，AI攻防能力或成下一个战略高地。

  5. 11.
    Meta 的 AI 智能体 Muse 被发现可导出虚拟机大量文件— IT之家

    AI 导读 · AI智能体隔离机制形同虚设，数据安全风险敲响警钟。

  6. 12.
    雷蛇推出清姬 Kiyo V2 Pro 网络摄像头，支持 4K 60FPS 录制— IT之家

    AI 导读 · 索尼传感器加AI取景，直播画质向单反看齐，创作者值得关注。

行业动态6 条

  1. 13.
    Meta takes down a critical video about meta AI Glasses after filming at Meta— Hacker News
  2. 14.
    Estimating suicide risk from text— MIT News

    AI 导读 · 语言模型能从日常文字中识别自杀高风险者，让干预更及时，但隐私与误判边界值得警惕。

  3. 15.
    The promise and peril of using visual AI to study cities— MIT News

    In their new book, “How AI Sees the City,” the leaders of MIT’s Senseable City Lab examine the technology’s implications for researching urban life.

  4. 16.
    Feds Target AI Critics as "Foreign Agents"— Hacker News
  5. 17.
    Show HN: Whiteboard (YC W26) – An open-source IDE for thoughtful software design— Hacker News

    AI 导读 · 开源IDE让人类与AI代理同台设计架构，或重塑协作编程的入口。

  6. 18.
    Google’s Project Suncatcher to put ML infrastructure in space— Hacker News

    https://www.nytimes.com/2026/09/24/technology/google-suncatc... ( https://archive.ph/BGFnP )

论文研究6 条

  1. 19.
    Contrastive Learning for Authorship Verification— arXiv

    Our results show that contrastive learning outperforms a classification-based approach to authorship verification under the tested settings. We identify loss function, batch size, training duration, p…

  2. 20.
    Where Should I Join? Robot Group Joining via Language-Guided Goal Prediction— arXiv

    Social navigation typically assumes a specified goal and focuses on reaching it while respecting social conventions, whereas robot group joining requires predicting where to join based on the group's…

  3. 21.
    Even Sharper Bounds for Transductive Learning and Its Applications— arXiv

    We introduce Sharper Transductive Local Complexity (STLC), a localized complexity method for transductive learning under uniform sampling without replacement. The construction starts from a Bernstein-…

  4. 22.
    Can LLMs Reason About Runtime Behavior? A Repository-Level Dynamic Benchmark— arXiv

    Large language models (LLMs) are increasingly used in coding tasks, but their ability to reason about code execution remains unclear. Existing repository-level QA benchmarks mainly evaluate static cod…

  5. 23.
    Nonequilibrium Phases of Repulsive Self-Attention: Chaos, Attention Condensation, and Emergent Locality— arXiv

    We study the nonequilibrium dynamics of a minimal recurrent transformer with $N$ normalized tokens, $Q=K=I$, and a negative value map $V=-I$. Similarity-based attention selects nearby representations,…

  6. 24.
    Order-Invariant Answers, Order-Sensitive Representations in Mathematical Reasoning— arXiv

    Reordering a set of mathematical rules without changing its meaning should preserve the correct answer, but must a model's internal representations stay invariant too? We investigate this question usi…

技巧与观点6 条

  1. 25.
    Accelerating vision-language models with LFM2.5-VL-DSpark— HuggingFace Blog
  2. 26.
    TechCrunch Disrupt 2026: Cal AI’s Zach Yadegari on how to create viral growth and capitalize on it— TechCrunch

    Zach Yadegari joins the Builders Stage at TechCrunch Disrupt 2026 to share how he capitalized on viral growth. Save up to $200 before September 25. Save 50% on a second pass.

  3. 27.
    Speaker-labeled transcription with WhisperX on SageMaker AI— AWS ML

    The AWS WhisperX Deep Learning Container packages Whisper, wav2vec2 forced alignment, and speaker diarization into a GPU-ready image. Learn how to deploy it to Amazon SageMaker AI…

  4. 28.
    Build a multi-account AI agent with AgentCore Gateway and MCP— AWS ML

    Build a multi-account architecture that keeps each team's data in its own AWS account while giving AI agents a unified way to query across them. A central platform account runs the…

  5. 29.
    Aderant builds intelligent ticket triage with Amazon Nova— AWS ML

    Learn how Aderant built an intelligent ticket triage system on Amazon Nova Lite through Amazon Bedrock, automating context gathering, classification, routing, and knowledge enrichm…

  6. 30.
    Foundries vs Navigators: Lowering the Cost of Science— Latent Space

    Guest Post: In science, thinking has gotten cheap but doing has not. This asymmetry is reshaping how research companies operate, largely inconspicuously.

快讯

日报生成时间：2026-09-28
← 2026-09-24最新日报2026-09-26 →