AAI Search
AI 资讯日报 · 2026-09-03

2026-09-03 · AI 资讯日报

今日新收录 158 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Beyond Scores: Understanding LLM-as-a-Judge Mechanisms in Summarization Evaluation— arXiv

    LLM-based evaluators of natural language generation (NLG) quality are widely deployed as scoring tools and as automated training signals, yet the internal procedure by which they assign a rating remai…

  2. 2.
    论文研究Efficient SWE Agent Benchmarking via Trajectory-Aware Evaluation— arXiv

    Evaluating software engineering agents on realistic benchmarks is costly, since each task may require multi-step code exploration, modification, and test execution. Existing efficient evaluation metho…

  3. 3.
    论文研究Adaptive Critical Token-Aware Retrieval for Repository-Level Code Generation— arXiv

    The repository-level code generation task requires synthesizing code that satisfies task requirements while remaining consistent with the target repository context. Since real-world repositories often…

  4. 4.
    论文研究CordisBench: Can Language Models Reason About Component Lifecycles in Dynamic Agent Harnesses?— arXiv

    Dynamic agent harnesses let language models change the software that shapes their own execution. This flexibility brings a new reasoning burden: a local plugin change can propagate through dependencie…

  5. 5.
    论文研究The Rise of Verbal Reinforcement Learning— arXiv

    Natural language is emerging as a primary feedback channel for improving language agents, capable of conveying intent, preferences, and causal structure in forms interpretable by both humans and moder…

  6. 6.
    论文研究Facet-0: A Robotic Foundation Model for Contact-Rich Precise Manipulation— arXiv

    Real-world robotic assembly at sub-millimeter tolerances demands spatial precision, compliant interaction, and robustness to contact failures. We present Facet-0, a robotic foundation model that predi…

模型发布/更新6 条

  1. 1.
    Proactive cyber defense for governments and enterprises— Google AI

    The Fairwind Program is a limited access program for governments and trusted partners to use our cyber defense tools.

  2. 2.
    ATV Big Air Tour turned 3 days of work into 3 hours with ChatGPT— OpenAI

    ATV Big Air Tour uses ChatGPT Work to speed up marketing, merchandising, and more. It even turned merchandise photos into an inventory website in 15 minutes.

  3. 3.
    Anthropic Releases Claude Fable 5.1 and Claude Mythos 5.1: 52.6% on Terminal-Bench-Science and 75% Cheaper Cache Reads— MarkTechPost

    Anthropic has released Claude Fable 5.1 and Claude Mythos 5.1, the same model behind two different safeguard layers. Fable 5.1 is generally available on the Claude API, AWS, Google…

  4. 4.
    Google DeepMind Releases Gemini 3.8 Flash and Gemini 3.8 Flash Cyber: One Core Model, Two Access Envelopes— MarkTechPost

    Google released Gemini 3.8 Flash and Gemini 3.8 Flash Cyber on September 2, 2026. Both variants run on the same foundational intelligence, split by safety mitigations rather than m…

  5. 5.
    Download Meta AI for Mac: Free Desktop App | AI at Meta— Meta AI

    Download Meta AI for Mac: Free Desktop App | AI at Meta AI at Meta

  6. 6.
    Google releases Gemini 3.8 Flash, its third Flash model in six weeks— Ars Technica

    AI 导读 · 谷歌密集发布Flash模型，暗示其轻量级AI战略提速，市场关注度陡增。

产品发布/更新6 条

  1. 7.
    Perplexity Releases Hybrid Compute on Mac: Cloud Agents Orchestrate Down to a Local Model, Gated On Device— MarkTechPost

    Agentic assistants have a structural problem: the context that makes them useful — deal documents, privileged files, client records — is exactly the context users cannot send to a…

  2. 8.
    Gradium AI Releases New Default TTS Model: 81.0% Hard-Case Pass Rate at 216 ms Time-to-First-Audio— MarkTechPost

    Speed and accuracy usually pull against each other in text-to-speech. Gradium AI's new default model reports both: an 81.0% human-rated pass rate on 500 hard sentences across five…

  3. 9.
    HanyuanWang/LiveStream-Agent-Studio— GitHub

    面向抖音直播电商的 Windows 本地 AI Agent Studio，贯通主播发现、直播洞察、直播复盘与短视频内容编导的统一智能工作流。

  4. 10.
    Tiger3807861189/GLM-5.3-Flash-J-Space-Capability-Realization-Report— GitHub

    GLM-5.3-Flash × J-Space capability realization — benchmark presentation of the J-Space Cognition Suite

  5. 11.
    ZekunCheng/novoweave— GitHub

    NovoWeave — a conceptual generative protein-design framework (non-functional pseudocode).

  6. 12.
    SkyProduction天工工作台：从剧本到成片，一套工作台把精品短剧创作真正跑起来— 量子位

    SkyProduction（天工工作台）全新版本于8月31日正式上线

行业动态6 条

  1. 13.
    From MIT to IBM, expediting AI and quantum deployment— MIT News

    AI 导读 · 产学研融合加速AI与量子计算落地，MIT与IBM合作模式成行业标杆。

  2. 14.
    Muse Spark 1.3— Hacker News

    AI 导读 · 多模态推理新突破，Meta开源模型或改写行业格局。

  3. 15.
    System helps humans predict when self-driving cars will make mistakes— MIT News

    A new method, called CW-Net, translates the reasoning process of an autonomous vehicle’s AI system into understandable concepts that explain its behavior.

  4. 16.
    Gemini 3.8 Flash and 3.8 Flash Cyber— Hacker News

    https://deepmind.google/models/model-cards/gemini-3-8-flash/

  5. 17.
    Fable 5.1 World Modeling— Hacker News

    AI 导读 · 多模态世界模型新突破，预示AI从感知迈向认知推理的关键一步。

  6. 18.
    Three sites made 215,128 “best software” pages for AI. Perplexity cites them— Hacker News

论文研究6 条

  1. 19.
    Beyond Scores: Understanding LLM-as-a-Judge Mechanisms in Summarization Evaluation— arXiv

    LLM-based evaluators of natural language generation (NLG) quality are widely deployed as scoring tools and as automated training signals, yet the internal procedure by which they assign a rating remai…

  2. 20.
    Efficient SWE Agent Benchmarking via Trajectory-Aware Evaluation— arXiv

    Evaluating software engineering agents on realistic benchmarks is costly, since each task may require multi-step code exploration, modification, and test execution. Existing efficient evaluation metho…

  3. 21.
    Adaptive Critical Token-Aware Retrieval for Repository-Level Code Generation— arXiv

    The repository-level code generation task requires synthesizing code that satisfies task requirements while remaining consistent with the target repository context. Since real-world repositories often…

  4. 22.
    CordisBench: Can Language Models Reason About Component Lifecycles in Dynamic Agent Harnesses?— arXiv

    Dynamic agent harnesses let language models change the software that shapes their own execution. This flexibility brings a new reasoning burden: a local plugin change can propagate through dependencie…

  5. 23.
    The Rise of Verbal Reinforcement Learning— arXiv

    Natural language is emerging as a primary feedback channel for improving language agents, capable of conveying intent, preferences, and causal structure in forms interpretable by both humans and moder…

  6. 24.
    Facet-0: A Robotic Foundation Model for Contact-Rich Precise Manipulation— arXiv

    Real-world robotic assembly at sub-millimeter tolerances demands spatial precision, compliant interaction, and robustness to contact failures. We present Facet-0, a robotic foundation model that predi…

技巧与观点6 条

  1. 25.
    Real-Time Intelligence with IBM Time Series Models on Confluent— HuggingFace Blog
  2. 26.
    These Russian Mathematicians Taught AI Models How to Talk to Each Other Without Using Words— Wired

    AI 导读 · 让不同AI模型直接互通，绕开语言瓶颈，或开启多模型协作新范式。

  3. 27.
    Accessing OpenAI models on Amazon Bedrock from Australia with global cross-Region inference— AWS ML

    AI 导读 · 云厂商联手AI巨头，跨区域推理落地澳洲，标志多模型生态竞争加剧。

  4. 28.
    Modernizing and scaling support operations with generative AI on AWS— AWS ML

    AI 导读 · 用生成式AI将培训视频转为标准作业流程，并借RAG优化工单处理，展现云上客服自动化的落地范式。

  5. 29.
    How an AWS team detects dashboard content failures at scale using Amazon Bedrock— AWS ML

    AI 导读 · 用AI检测仪表盘静默故障，填补运维盲区，是数据可靠性领域的新解法。

  6. 30.
    From code to diagrams: Agentic architecture documentation with Amazon Bedrock AgentCore— AWS ML

    AI 导读 · 自动生成架构图打通代码与文档，企业级效率革新值得关注。

快讯

日报生成时间：2026-09-06
← 2026-09-02最新日报2026-09-04 →