2026-09-24 科技日报

扫描 587 篇候选内容 · 覆盖 650 个信息源 · 纳入 573 篇

📌 今日要点

• How to Use NVIDIA Warp and MjWarp to Accelerate Robotics Simulation and Learning Workflows
• Advancing Private AI Compute with secure, server-side memory
• Gemini 3.8 text-to-speech says hello
• “As a Language Model…”: Chat Template Switches LLM Self-Referential Voice and Activation Steering Reproduces It
• Federating Quantum and Classical Computing: A Privacy-Preserving Hybrid Approach
• Entropy Can Flow, or It Can Guide. Be Entropy. LEDFlow: Introducing Entropy-guided Generation Order into Uniform Discrete Flow

📝 科技简讯

• How to Use NVIDIA Warp and MjWarp to Accelerate Robotics Simulation and Learning Workflows — Hugging Face Blog
• Advancing Private AI Compute with secure, server-side memory — Google DeepMind
• Gemini 3.8 text-to-speech says hello — Google DeepMind
• “As a Language Model…”: Chat Template Switches LLM Self-Referential Voice and Activation Steering Reproduces It — ArXiv ML (cs.LG)
• Federating Quantum and Classical Computing: A Privacy-Preserving Hybrid Approach — ArXiv ML (cs.LG)
• Entropy Can Flow, or It Can Guide. Be Entropy. LEDFlow: Introducing Entropy-guided Generation Order into Uniform Discrete Flow — ArXiv ML (cs.LG)
• The Probabilistic Structure of Large Language Models — ArXiv ML (cs.LG)
• Stable Unsupervised Continual Chunking with Sheaf SyncMap — ArXiv ML (cs.LG)
• Brain-Inspired Hierarchical Modularity for General Continual Learning — ArXiv ML (cs.LG)
• Dual-GNN Multilevel Coarsening for Maximum Independent Set — ArXiv ML (cs.LG)
• Everything Claude Opus 5.5 Actually Ships With — KDnuggets
• Why Most Data Science Notebooks Die After Day One: How to Build Ones That Survive — KDnuggets
• High-Performance Data Processing with Polars: A KDnuggets Cheat Sheet — KDnuggets
• 「一人剧组」杀疯了！LibTV 1.5，一下午拍完一集爆款 AI 漫剧 — 新智元
• Claude Opus 5.5 深夜发布，减价 40%！大语言模型正式开始吞噬多模态 — 新智元
• 罗福莉带小米登顶全球开源榜首！大规模 RL 立功，斩杀 Grok 4.7 — 新智元
• 00 后华人天才少女、Bengio 高徒，从 DeepMind 转会 OpenAI！ — 新智元
• Meta 刚登顶，千问就交卷！3 亿人的个人 AI 助理要来了 — 新智元
• From portal-hopping to instant answers: HEMA’s journey with MCP and Amazon Bedrock — AWS Machine Learning
• Agentic conversational video intelligence built on AWS — AWS Machine Learning
• Use open weight models as your AI coding agent with Amazon Bedrock — AWS Machine Learning
• Google Beam expands with new regions, partners, and customers — Google AI Blog
• U.S. TRANSCOM deploys randomised AI to secure military logistics — AI News (Jack Clark)
• AI Agents Are Becoming a New Malware Distribution Channel — AI News (Jack Clark)
• Meta made a Tamagotchi-like wearable for its Muse AI agent — TechCrunch AI
• Google’s new Flash TTS models let you design AI voices from scratch using text descriptions — The Decoder
• YouTube adds AI tools to Creator Studio with script coaching, smart thumbnails, and Gemini editing — The Decoder
• Anthropic engineer explains why Claude’s writing got worse although the model got smarter — The Decoder
• Nvidia-backed Nscale keeps its biggest customer, Bytedance, out of its IPO filing — The Decoder
• Meta’s AI agent Muse draws 500,000 users in a week along with claims it copied OpenClaw — The Decoder
• Inside Basecamp Research, the AI startup turning evolution into training data — The Decoder
• Why I’m scared of RL — AI Alignment Forum
• WorkspaceBench: Evaluating Interpretability Methods for the Global Workspace — AI Alignment Forum
• 联想亮相阿里云栖大会：联想天禧 AI 把超级组织落地到端侧 — 量子位
• 让 Token 生产更高效：异构混推的关键技术演进与创新实践 — 量子位
• 汇智智能发布 Hellome：国内首个 FDE 直连智能体服务平台，把 AI 交付周期压进“周” — 量子位
• 达卯科技完成新一轮融资，算电协同核心软件层“稀缺标的” — 量子位
• 它石智航顶尖团队引领中国具身智能率先迈入规模化落地阶段 — 量子位
• 斑马智能发布端模型 AutoOmni2.0，让元神 AI 更懂“我的世界” — 量子位
• DeepSeek 新论文公开 Agent 训练！梁文锋署名 — 量子位

📊 数据概览