2026-09-30 科技日报

扫描 469 篇候选内容 · 覆盖 650 个信息源 · 纳入 456 篇

📌 今日要点

• 反超 Seedance 2.0！RunningHub 发布增强版 H3，1 秒只要 0.1 元
• OpenAI 紧急叫停 GPT-6.1！
• AI 找出数学反例推翻论文，作者确认！453 篇手稿，AI 开始自己出题了
• Salesforce Achieves UK Cyber Essentials and Cyber Essentials Plus Certifications
• Agentic AI and the Future of Work: What It Means for Salesforce Practitioners
• Dreamforce 2026: 5 Ways Marketing Cloud Is Building a New Model for Growth

📝 科技简讯

• 反超 Seedance 2.0！RunningHub 发布增强版 H3，1 秒只要 0.1 元 — 新智元
• OpenAI 紧急叫停 GPT-6.1！ — 新智元
• AI 找出数学反例推翻论文，作者确认！453 篇手稿，AI 开始自己出题了 — 新智元
• Salesforce Achieves UK Cyber Essentials and Cyber Essentials Plus Certifications — Salesforce AI Blog
• Agentic AI and the Future of Work: What It Means for Salesforce Practitioners — Salesforce AI Blog
• Dreamforce 2026: 5 Ways Marketing Cloud Is Building a New Model for Growth — Salesforce AI Blog
• Cash Flow Management Tips: What’s Your Strategy, Business Owners? — Salesforce AI Blog
• Bringing AI to Autonomous Systems — From Cognition to Collective Intelligence — arXiv AI (cs.AI)
• ScopeBench: Do Agents Preserve Engagement Boundaries Under Goal Pressure? — arXiv AI (cs.AI)
• When Is a Multi-Agent Code Judge Actually Grounded? Two Label-Free Measurements, and a Judge That Declines to Guess — arXiv AI (cs.AI)
• Bridging LLM Agents and Data Spaces: An Architectural Mediation Approach using the Model Context Protocol — arXiv AI (cs.AI)
• Stealth Apart, Harm Together: Skill Cascading Attacks on Skill-Based Agent Systems — arXiv AI (cs.AI)
• How AI is creating a UK super-museum and revealing our planet’s history — BBC Science
• Oura pulls $15bn stock market listing days after announcement — BBC Science
• Hurricane Polo makes landfall on Mexico’s Pacific coast — BBC Science
• Warning more homes will be uninsurable due to flood risk — BBC Science
• Watch: Bangkok locals create floating market during floods — BBC Science
• Lives ‘will be lost’ unless UK does more to combat wildfires, committee chairwoman says — BBC Science
• OpenAI connects the Dots — Platformer (Casey Newton)
• Two New Birds Louder Than Jackhammers Just Dropped — 404 Media
• Surveillance Finds a Way — 404 Media
• Surveillance Company Tells Cops It Wants to Add Facial Recognition to Flock Cameras — 404 Media
• These Tech Workers Made ChatGPT Drive a Toyota Corolla — 404 Media
• One More Note on Agents, Meta Connect, Meta Enterprise Platform — Stratechery (Ben Thompson)
• OpenAI unveils AI assistant ‘dots’ while safety worries delay new model — BBC Tech
• NVIDIA Kumo Tabular Sets a New Accuracy-Efficiency Frontier for Tabular Prediction — Hugging Face Blog
• Getting the Source Right, Not Just the Fact: Source-Aware Verification for MCP Agents — Hugging Face Blog
• Amazon Bedrock expands Claude model availability to in-country inferencing in India — AWS Machine Learning
• Introducing Anthropic models on Amazon Bedrock for in-region inference in Seoul and Singapore — AWS Machine Learning
• Bring near-Astra intelligence to everyday work with GPT-6.1 Sol on Amazon Bedrock — AWS Machine Learning
• Prompt engineering fundamentals for Amazon Quick — AWS Machine Learning
• Prompt engineering by Quick component: Patterns and pitfalls — AWS Machine Learning
• Building an AI-powered contract intelligence platform with Amazon Quick and Amazon Bedrock AgentCore — AWS Machine Learning
• How Condé Nast built multimodal video discovery with Amazon Bedrock — AWS Machine Learning
• HybridInfer: Thermal-Aware Reinforcement-Learning Tier Routing for On-Device, Edge, and Cloud LLM Inference — ArXiv ML (cs.LG)
• When the Preconditioning Exponent Turns Negative: Learning-Rate Coupling and Cross-Environment Generalization — ArXiv ML (cs.LG)
• ENAS: An Efficient Hardware-Aware Neural Architecture Search Framework for TinyML on Resource-Constrained Microcontrollers — ArXiv ML (cs.LG)
• Offline Policy Evaluation as a decision support tool for designing Adaptive Experiments — ArXiv ML (cs.LG)
• Cosine Similarity Is Not Evidence: Measuring the Noise Floor of Interpretability Transfer Under Quantization — ArXiv ML (cs.LG)
• Why Clipping Matters in AdaGrad? Toward a High-Probability Theory under Generalized Smoothness — ArXiv ML (cs.LG)

📊 数据概览