AAI Search
AI 资讯日报 · 2026-08-28

2026-08-28 · AI 资讯日报

今日新收录 174 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CritICL: Inference-Time Weak-to-Strong Generalization from Small Language Model Failure Modes— arXiv

    Recent advances in inference-time scaling have significantly improved the reasoning performance of large language models (LLMs). However, these methods typically rely on repeated generation or externa…

  2. 2.
    论文研究SWE-Prime: Fewer Trajectories, Better Performance— arXiv

    To improve large language models' ability to resolve real-world software issues, prior work has focused on constructing large-scale agent trajectory datasets and performing supervised fine-tuning (SFT…

  3. 3.
    论文研究From Static to Dynamic: Benchmarking Real-World Code Review with MCR-Bench— arXiv

    In real-world software development, code review typically involves iterative interactions between developers and reviewers to improve software quality, making the process costly and time-consuming. Al…

  4. 4.
    论文研究RedEvoAgent: Automatic Red-Teaming Agent with Experience-Driven Skill Evolution— arXiv

    LLM-based agents are increasingly deployed in product-level execution harnesses, where jailbreaks can trigger harmful tool use and persistent state changes, creating greater risks than unsafe text gen…

  5. 5.
    论文研究Mechanistic Reaction Prediction via Discrete Flow Matching on Graph-Structured Electron Occupation— arXiv

    Chemical reactions are fundamentally transformations in electron space, yet most machine learning approaches model them either through \textit{de novo} generation of product molecules or through heuri…

  6. 6.
    论文研究Stochastic Estimation of Transduced Language Models— arXiv

    Transduced language models (TLMs) compose a pretrained \emph{source} language model with a functional finite-state transducer to induce a language model over \emph{target} strings. Computing the proba…

模型发布/更新6 条

  1. 1.
    Beneficial Deployments— Anthropic

    AI 导读 · Anthropic新部署方案，AI安全落地的关键一步。

  2. 2.
    Expanding our support for scientists— Anthropic

    AI 导读 · AI实验室加大科研扶持，或催生更多前沿突破。

  3. 3.
    Gemini Omni 1.1 Flash lets you build with more control— Google DeepMind
  4. 4.
    3 new ways to plan and book travel in Search— Google AI

    Book hotels and track airfares, plus view miles and rewards with AI Mode in Google Search.

  5. 5.
    Delivering Vera: NVIDIA’s First CPU Built for Agents Is Shipping Now— NVIDIA

    NVIDIA Vice President of Hyperscale and HPC Ian Buck hand-delivers Vera CPU systems across the AI ecosystem as Vera begins shipping at scale.

  6. 6.
    Piloting the world's first double-blind AI evaluations— Google DeepMind

    Piloting the world's first double-blind AI evaluations

产品发布/更新6 条

  1. 7.
    deeplethe/utopia— GitHub

    World's first open-source enterprise world model.

  2. 8.
    kulkarnirohit123/cra-agent— GitHub

    Autonomous agentic AI for CRA (Cyber Resilience Act) compliance: scans repos, triages findings, opens Jira tickets, and auto-fixes vulnerabilities via PR.

  3. 9.
    calmrocks/ai-engineer-notebooks— GitHub

    Hands-on, framework-free Colab notebooks for the AI Engineer / Forward Deployed Engineer (FDE) skill set — model APIs, structured output, tool calling, RAG, eva…

  4. 10.
    高德发布首个无长程依赖的万帧级流式3D重建模型ABot-Recon，以12帧重建万帧3D场景— 量子位

    AI 导读 · 突破实时重建瓶颈，开启万帧级3D场景流式处理新纪元。

  5. 11.
    34克、全天候AI，1996元起，雷鸟iO热卖加速智能眼镜全民普及— 量子位

    AI 导读 · 轻量低价直击痛点，智能眼镜或迎来真正大众化拐点。

  6. 12.
    全球首款专为iPhone用户打造的Agent耳机正式发布，有道AI耳机让声音变成生产力— 量子位

    AI 导读 · 把AI助手塞进耳朵，重新定义耳机价值，值得关注。

行业动态6 条

  1. 13.
    Looking beyond natural sequences— MIT News

    A new machine-learning framework aims to improve the success rate of computational protein design while moving away from results that reproduce sequences found in nature.

  2. 14.
    Please stop flooding our projects with AI slop to furnish your CV— Hacker News

    AI 导读 · 警惕AI沦为简历注水工具，损害真实项目价值。

  3. 15.
    Anthropic locks in 45-billion-dollar compute deal with Nscale ahead of IPO— The Decoder

    Anthropic has struck a deal worth around 45 billion dollars with British cloud startup Nscale, Bloomberg reports. The article Anthropic locks in 45-billion-dollar compute deal with…

  4. 16.
    Google's Gemini 3.5 Transcribe turns speech to text in 85 languages while auto-correcting your verbal stumbles— The Decoder

    Google's new Gemini 3.5 Transcribe recognizes over 85 languages, strips filler words, and corrects slips of the tongue in real time. It hits a 4.0 percent word error rate in stream…

  5. 17.
    Claude Cowork now runs its own browser inside the desktop app— The Decoder

    Anthropic is building its own browser into Claude Cowork, running inside the desktop app. The article Claude Cowork now runs its own browser inside the desktop app appeared first o…

  6. 18.
    AI shopping agents aren't ready to buy on your behalf, study finds— The Decoder

    Letting an AI do your shopping might not get you the best deal. Researchers at the Wharton School show how erratic AI shopping agents really are: a single external source like Wire…

论文研究6 条

  1. 19.
    CritICL: Inference-Time Weak-to-Strong Generalization from Small Language Model Failure Modes— arXiv

    Recent advances in inference-time scaling have significantly improved the reasoning performance of large language models (LLMs). However, these methods typically rely on repeated generation or externa…

  2. 20.
    SWE-Prime: Fewer Trajectories, Better Performance— arXiv

    To improve large language models' ability to resolve real-world software issues, prior work has focused on constructing large-scale agent trajectory datasets and performing supervised fine-tuning (SFT…

  3. 21.
    From Static to Dynamic: Benchmarking Real-World Code Review with MCR-Bench— arXiv

    In real-world software development, code review typically involves iterative interactions between developers and reviewers to improve software quality, making the process costly and time-consuming. Al…

  4. 22.
    RedEvoAgent: Automatic Red-Teaming Agent with Experience-Driven Skill Evolution— arXiv

    LLM-based agents are increasingly deployed in product-level execution harnesses, where jailbreaks can trigger harmful tool use and persistent state changes, creating greater risks than unsafe text gen…

  5. 23.
    Mechanistic Reaction Prediction via Discrete Flow Matching on Graph-Structured Electron Occupation— arXiv

    Chemical reactions are fundamentally transformations in electron space, yet most machine learning approaches model them either through \textit{de novo} generation of product molecules or through heuri…

  6. 24.
    Stochastic Estimation of Transduced Language Models— arXiv

    Transduced language models (TLMs) compose a pretrained \emph{source} language model with a functional finite-state transducer to induce a language model over \emph{target} strings. Computing the proba…

技巧与观点6 条

  1. 25.
    Claude Science AMA: How to accelerate scientific discovery— Anthropic

    Claude Science AMA: How to accelerate scientific discovery Anthropic

  2. 26.
    Build agentic creative workflows with Amazon Quick and fal— AWS ML

    AI 导读 · 用Agent串联生成式AI工具，打通创作流程，效率提升看得见。

  3. 27.
    Deepgram deepens Amazon SageMaker AI observability with Enhanced Metrics— AWS ML

    Self-hosted speech AI carries an observability trade-off: the numbers that drive capacity planning and cost management stay locked inside the vendor container. Deepgram closes that…

  4. 28.
    Reduce ASR inference costs by 75% with NVIDIA MPS on Amazon EC2— AWS ML

    Serving automatic speech recognition (ASR) models at scale is costly when each request uses only a fraction of a GPU. Learn how NVIDIA CUDA Multi-Process Service (MPS) with NVIDIA…

  5. 29.
    Breaking Claude Code Opus 5 Auto Mode— Simon Willison

    AI 导读 · 自动模式成防注入攻击新防线，Anthropic押注AI编码安全，行业风向标值得紧盯。

  6. 30.
    Who let the agents in— Ben's Bites

    It’s you, and it’s getting easier

快讯

日报生成时间：2026-09-06
← 2026-08-27最新日报2026-08-29 →