2026-07-22 科技日报

扫描 571 篇候选内容 · 覆盖 650 个信息源 · 纳入 558 篇

📌 今日要点

• The State of Simulation for Physical AI: An Overview
• Grabette: an open system to record robot-manipulation data
• WAIC 重磅成果｜上海仪电智算牵头成立“智算系统架构联盟”并发布《超节点系统架构规范》
• WAIC 重磅成果｜上海市教育算力专区正式开启试运行
• WAIC 重磅成果｜智爱赛思全面升级并发布科研专属 Token Plan
• WAIC 重磅成果｜仪电智算云在国家人工智能应用中试基地建设中展现全栈服务能力

📝 科技简讯

• The State of Simulation for Physical AI: An Overview — Hugging Face Blog
• Grabette: an open system to record robot-manipulation data — Hugging Face Blog
• WAIC 重磅成果｜上海仪电智算牵头成立“智算系统架构联盟”并发布《超节点系统架构规范》 — 量子位
• WAIC 重磅成果｜上海市教育算力专区正式开启试运行 — 量子位
• WAIC 重磅成果｜智爱赛思全面升级并发布科研专属 Token Plan — 量子位
• WAIC 重磅成果｜仪电智算云在国家人工智能应用中试基地建设中展现全栈服务能力 — 量子位
• 国内首个！阿里健康氢离子达成 NEJM、JAMA、BMJ 三大医学顶刊内容合作 — 量子位
• 阿里 Qoder 上线全新安全能力，为每位用户配备一位专属安全工程师 — 量子位
• Introducing Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber — Google DeepMind
• Google’s Gemini 3.6 Flash targets enterprise agent token costs — AI News (Jack Clark)
• The AI Slot Machine Effect: Why Generative Feeds Disrupt Deep Work And How to Reclaim Focus — AI News (Jack Clark)
• Bristol Myers Squibb buys Nvidia AI system for drug discovery — AI News (Jack Clark)
• Chinese open-weight models are cheap. Washington is deciding what that costs. — AI News (Jack Clark)
• Run the Mythos Enhanced Coding Model Locally with llama.cpp and Pi — KDnuggets
• 5 Free Courses to Go From AI Beginner to Practitioner — KDnuggets
• OpenAI Models Behind HuggingFace Cybersecurity Incident — LessWrong
• WeirdChat: A catalog of unexpected AI behaviors, discovered automatically — LessWrong
• OpenAI Shares Some Alignment Problems — LessWrong
• Blogging Technology Interlude — LessWrong
• Ending Soon: Fundamental Uncertainty $2,000 Essay Contest — LessWrong
• Exploring self-distilled reasoning for supervised fine-tuning with Amazon Nova — AWS Machine Learning
• Reinforcement Learning-Guided NSGA-II Enhanced with Gray Relational Coefficient for Multi-Objective Optimization: Application to NASDAQ Portfolio Optimization — ArXiv ML (cs.LG)
• DocOCR-Eval: A Correction-Based Framework for OCR Tool Selection Without Ground Truth — ArXiv ML (cs.LG)
• Fully-sensorized smart-eyewear platform for on-device Machine Learning — ArXiv ML (cs.LG)
• LLM Unlearning for Cyber Defense: A Survey on Methods, Challenges, and Emerging Threats — ArXiv ML (cs.LG)
• Operator-Aware Mixed-Precision Tolerance Calibration for Tensor Kernels — ArXiv ML (cs.LG)
• RouteCost: A Production-Inspired Multi-Stage Framework for Pre-Order Shipping Cost Estimation in E-Commerce — ArXiv ML (cs.LG)
• Orthogonal Gradient Constraints Shape Noisy-Label Memorization Dynamics — ArXiv ML (cs.LG)
• 刚刚，中国 AI 交卷了！不加一张卡，暴省数百万 — 新智元
• 300 万粉女神全是 AI 合成！伪造孤儿院，跨国「假慈善」一夜塌房 — 新智元
• 5.83 倍加速！PolicyTrim：让 VLA 机器人少走弯路、更快完成任务 — 新智元
• 一颗「脑」装进 27000 台真机！遍布 50 国 — 新智元
• GPT-5.6-Sol 在网络攻防能力上，超越了 Mythos！开源模型也排好了 — 新智元
• Meta is testing an AI bedtime story app for people with no imagination — TechCrunch AI
• OpenAI says Hugging Face was breached by its own pre-release models — TechCrunch AI
• Jack Dorsey is taking on Slack with Buzz, a group chat platform for teams and their AI agents — TechCrunch AI
• AI and the rise of the universal entertainment app — TechCrunch AI
• Data centers expected to use 4x more electricity by 2035 — TechCrunch AI
• Google releases three new Gemini models — but no 3.5 Pro — TechCrunch AI
• US threatens sanctions against Chinese AI models over IP theft — TechCrunch AI

📊 数据概览