NLNEXT LAB RADAR
AI 资讯日报 · 2026-07-29

2026-07-29 · AI 资讯日报

今日新收录 167 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$π\mathbf{R}^2$: Reactive Real-time Flow Policies— arXiv

    Generalist manipulation policies increasingly take the form of action-chunking flow policies built on large pretrained backbones. Such chunks run open-loop, so the policy cannot react to sensory input…

  2. 2.
    论文研究Spend Experts Where You Are Unsure: Confidence-Adaptive Routing for Mixture-of-Experts LoRA— arXiv

    Mixture-of-Experts (MoE) variants of Low-Rank Adaptation (LoRA) route every token to a fixed number of experts $k$. Tokens differ in how uncertain the model is about them, so a single k over-spends on…

  3. 3.
    论文研究Re-thinking Mammography Transfer Learning: The Dataset-Informed Transfer Learning (DITL) Framework for Breast Cancer Screening and Lesion Diagnosis— arXiv

    Enhancing classification performance in mammography remains a persistent challenge across both small curated datasets and large-scale clinical cohorts. Conventional transfer learning approaches often…

  4. 4.
    论文研究VetClaw: An Edge-Cloud Multimodal Agentic System for Veterinary Disease Screening— arXiv

    We present VetClaw, an edge-cloud multimodal agentic system for early veterinary disease screening. VetClaw uses a camera module as an edge sensing device and sends captured images, together with opti…

  5. 5.
    论文研究Desktop-Delta Bench: Do Computer-Use Models Understand Desktop GUI Transitions?— arXiv

    Computer-use agents (CUAs) increasingly act through desktop GUIs to complete long-horizon tasks. Current benchmarks primarily measure end-task success or single-frame grounding. Neither isolates wheth…

  6. 6.
    论文研究Reinformed Dreamer: An Asymmetric World Model Efficiently Trained through Latent Guidance— arXiv

    Much like humans benefit from guidance while learning, reinforcement learning algorithms may benefit from additional supervision beyond rewards. Leveraging additional information during training to le…

模型发布/更新6 条

  1. 1.
    Cryptanalysis of 7-Round AES via the Algebraic Structure of its S-box— Anthropic

    Cryptanalysis of 7-Round AES via the Algebraic Structure of its S-box www-cdn.anthropic.com

  2. 2.
    Mythos Preview’s Chain of Thought in Discovering the AES Möbius Bridge— Anthropic

    Mythos Preview’s Chain of Thought in Discovering the AES Möbius Bridge www-cdn.anthropic.com

  3. 3.
    Claude Opus 5 System Card— Anthropic

    Claude Opus 5 System Card www-cdn.anthropic.com

  4. 4.
    GPT Transcribe improves on its predecessor but can't catch ElevenLabs, Google, or Mistral on error rates— The Decoder

    OpenAI has released GPT Transcribe and GPT Live Transcribe, two new speech recognition models available through its API. The article GPT Transcribe improves on its predecessor but…

  5. 5.
    How enabling two settings tripled our scores on the ARC-AGI-3 benchmark— OpenAI

    How two API settings improved GPT-5.6 performance on ARC-AGI-3, boosting scores and efficiency by retaining reasoning and enabling compaction.

  6. 6.
    Accelerating scientific discovery with ChatGPT for Academic Researchers— OpenAI

    OpenAI is giving 100,000 academic researchers free access to ChatGPT's most advanced AI models to accelerate scientific research, collaboration, and discovery.

产品发布/更新6 条

  1. 7.
    OpenAI open-sources Codex Security CLI to help developers find and fix vulnerabilities from the command line— The Decoder

    OpenAI has released Codex Security CLI, an open-source tool that automatically detects and fixes vulnerabilities in code repositories. Previously known internally as "Aardvark," th…

  2. 8.
    drumih/turbo-fieldfare— GitHub

    Gemma 4 26B-A4B inference in ~2 GB of RAM on any M-series MacBook

  3. 9.
    mikehasa/agentacct— GitHub

    See what your coding agents did and what it cost. Breaks each task down into work steps — tools used, files changed, tests run, time and tokens spent. Local-fir…

  4. 10.
    xuzhougeng/wisp-science— GitHub

    Open-source, local-first desktop AI research workbench for scientific computing with Python/R, MCP bioinformatics tools, SSH/WSL/GPU runtimes, and OpenAI/Anthro…

  5. 11.
    微软确认 Copilot“超级应用”年内问世，整合智能体、AI 编程、聊天— IT之家

    AI 导读 · 微软将AI聊天、编程与智能体整合为单一应用，或重塑用户与AI的交互方式。

  6. 12.
    让AI先吃，这个国民级健康难题有解了— 36氪

    在科学减重这件事上，继战略投资薄荷健康之后，蚂蚁阿福又放了个大招。 7月29日，蚂蚁阿福宣布升级上线“AI拍饮食”功能，用户拍食物照片发给阿福，就能自动估算营养和热量信息，获得健康饮食搭配建议和运动消耗计划等，还支持一健存入个人健康档案，建立“热量账户”。吃得对不对，多不多，一目了然。 阿福此次推出“AI拍��食”功能，并非是一项孤立动作。 一个月前，36氪…

行业动态6 条

  1. 13.
    AI's top startups are barely publishing their research— Hacker News

    AI 导读 · 顶尖AI初创公司闭门造车，研究透明度骤降，创新生态面临风险。

  2. 14.
    Document-borne AI worms can self-propagate through Copilot for Word— Hacker News
  3. 15.
    LearnVector – Andrew Ng's AI company building one‑to‑one learning experiences— Hacker News
  4. 16.
    Self-hosting Kimi K3: 20% more hardware cost, 20% better task resolution— Hacker News
  5. 17.
    Hint, a new AI startup co-founded by Martha Stewart, offers an AI assistant for homeowners— TechCrunch

    AI home management startup Hint, co-founded by Martha Stewart, wants to become an “AI for your home,” combining property records, maintenance schedules, home documents, and an AI a…

  6. 18.
    Encore AI raises $30M to build AI agents that learn from customer calls— TechCrunch

    The startup analyzes calls, messages, and CRM data to identify effective sales techniques and turn them into playbooks for AI agents.

论文研究6 条

  1. 19.
    $π\mathbf{R}^2$: Reactive Real-time Flow Policies— arXiv

    Generalist manipulation policies increasingly take the form of action-chunking flow policies built on large pretrained backbones. Such chunks run open-loop, so the policy cannot react to sensory input…

  2. 20.
    Spend Experts Where You Are Unsure: Confidence-Adaptive Routing for Mixture-of-Experts LoRA— arXiv

    Mixture-of-Experts (MoE) variants of Low-Rank Adaptation (LoRA) route every token to a fixed number of experts $k$. Tokens differ in how uncertain the model is about them, so a single k over-spends on…

  3. 21.
    Re-thinking Mammography Transfer Learning: The Dataset-Informed Transfer Learning (DITL) Framework for Breast Cancer Screening and Lesion Diagnosis— arXiv

    Enhancing classification performance in mammography remains a persistent challenge across both small curated datasets and large-scale clinical cohorts. Conventional transfer learning approaches often…

  4. 22.
    VetClaw: An Edge-Cloud Multimodal Agentic System for Veterinary Disease Screening— arXiv

    We present VetClaw, an edge-cloud multimodal agentic system for early veterinary disease screening. VetClaw uses a camera module as an edge sensing device and sends captured images, together with opti…

  5. 23.
    Desktop-Delta Bench: Do Computer-Use Models Understand Desktop GUI Transitions?— arXiv

    Computer-use agents (CUAs) increasingly act through desktop GUIs to complete long-horizon tasks. Current benchmarks primarily measure end-task success or single-frame grounding. Neither isolates wheth…

  6. 24.
    Reinformed Dreamer: An Asymmetric World Model Efficiently Trained through Latent Guidance— arXiv

    Much like humans benefit from guidance while learning, reinforcement learning algorithms may benefit from additional supervision beyond rewards. Leveraging additional information during training to le…

技巧与观点6 条

  1. 25.
    not much happened today— AI News

    **OpenAI's agent security incident expanded beyond Hugging Face, affecting four additional accounts and highlighting the need for stronger enterprise hardening measures like sandbo…

  2. 26.
    AI Worming through Word— Simon Willison

    AI 导读 · 提示注入攻击升级，AI蠕虫可在Word中自我复制，威胁文档安全。

  3. 27.
    Adding a custom MCP server to Claude and ChatGPT— Simon Willison

    TIL: Adding a custom MCP server to Claude and ChatGPT Connecting a custom MCP server to Claude and ChatGPT's standard chat interfaces is possible, but can take quite a few steps. T…

  4. 28.
    Authenticate with Private Key JWT using Amazon Bedrock AgentCore Identity— AWS ML

    This post explains how Private Key JWT client authentication works in AgentCore Identity and reviews the supported grant flows. We then walk through creating an AWS KMS signing key…

  5. 29.
    Generate Autonomous Business Insights with AI Agent and MCP Servers— AWS ML

    Learn how Amazon Bedrock AgentCore delivers autonomous, cross-system business intelligence through configuration rather than custom code. Using pre-built MCP server connectors, fin…

  6. 30.
    Automating customer retention workflows in Amazon Quick— AWS ML

    Learn how to build a no-code customer retention pipeline in Amazon Quick that detects at-risk customers from call transcripts and CSAT data, scores them by retention priority with…

快讯

日报生成时间：2026-08-16
← 2026-07-28最新日报2026-07-30 →