2026-09-29 科技日报

扫描 547 篇候选内容 · 覆盖 650 个信息源 · 纳入 532 篇

📌 今日要点

• Holo4: powering generalist computer-use agents
• One year in: How Microsoft Research Asia – Singapore is advancing research, partnership and talent for real-world impact
• Gemini 3.5 Transcribe vs OpenAI’s GPT-Transcribe
• 3 Numba Tricks for Python Runtime Optimization
• Watch the winning trailer from the Future Vision XPRIZE, The Gifted.
• Fixed-weight models are adversarially vulnerable: hence misaligned

📝 科技简讯

• Holo4: powering generalist computer-use agents — Hugging Face Blog
• One year in: How Microsoft Research Asia – Singapore is advancing research, partnership and talent for real-world impact — Microsoft Research
• Gemini 3.5 Transcribe vs OpenAI’s GPT-Transcribe — KDnuggets
• 3 Numba Tricks for Python Runtime Optimization — KDnuggets
• Watch the winning trailer from the Future Vision XPRIZE, The Gifted. — Google AI Blog
• Fixed-weight models are adversarially vulnerable: hence misaligned — AI Alignment Forum
• Clipboard Normalizer in Mac App Store — LessWrong
• “China needs more AI safety.” More of what, exactly? — LessWrong
• Thank you to the AI safety people — LessWrong
• gpuc: A single-user GPU queue that doesn’t require sudo — LessWrong
• AI Companies Are Not (Necessarily) Liable for Unintended AI Cyberattacks — LessWrong
• Grok 4.7 is now available on Amazon Bedrock — AWS Machine Learning
• Introducing Claude Sonnet 5.5 on AWS — AWS Machine Learning
• Build real-time voice applications with vLLM-Omni on SageMaker AI – Part 1 — AWS Machine Learning
• Generate images and video with vLLM-Omni on SageMaker AI – Part 2 — AWS Machine Learning
• Implementing synthetic monitoring using Amazon Nova Act — AWS Machine Learning
• Automating Amazon Textract adapter lifecycle management across accounts — AWS Machine Learning
• Peak XV ups Surge seed investment ceiling to $5M, unveils 18-startup cohort — TechCrunch AI
• OpenAI reportedly ditches model over safety concerns — TechCrunch AI
• Source: Inference provider Modal Labs closing in on $750M round at$15.75B valuation — TechCrunch AI
• AMD will acquire Fei-Fei Li’s World Labs for $8.2 billion — TechCrunch AI
• Shopify opens checkout to browser-based AI agents — TechCrunch AI
• The AI boom took over Climate Week and not everyone is happy about it — TechCrunch AI
• Nvidia launches new platform for reining in rogue AI agents — TechCrunch AI
• TinyCVIO: A Constellation-Aided Visual-Inertial Odometry System for Nanodrones — ArXiv RO (cs.RO)
• A Mechanistic Study of AI-Text Detection Neurons in Frozen BERT: Sparse Probing and Activation Patching on RAID — ArXiv CL (cs.CL)
• Manifold Projection and Iterative Autoencoder Refinement for Masked Language Modeling — ArXiv CL (cs.CL)
• Not All Memories Are Equal: Hierarchical Collaborative Memory for Validity-Aware Retrieval in LLM Agents — ArXiv CL (cs.CL)
• Auditing and Repairing LLM-as-Judge Failures in a Production Text-to-SQL Pipeline — ArXiv CL (cs.CL)
• A Survey on Fake Review Detection: From Pre-trained Language Models to Large Language Models — ArXiv CL (cs.CL)
• AlphaEarth distinguishes cities but compresses urban variation — ArXiv CV (cs.CV)
• LiTe-GS: Oracle-Efficient Next Best View Selection for 3D Gaussian Splatting — ArXiv CV (cs.CV)
• CSCWD: Cross-Scale Channel-wise Knowledge Distillation for Lightweight Tiny Object Detection on Edge Devices — ArXiv CV (cs.CV)
• What Improves Multimodal Misinformation Detection? Answers from a Large-Scale Empirical Study — ArXiv CV (cs.CV)
• ProCAP: Probabilistic Cross-Attentive Prompt Learning for Vision-Language Models — ArXiv CV (cs.CV)
• What an Unforgettable Dreamforce 2026: Unleashing IT & HR Service Innovation — Salesforce AI Blog
• Want to Maximize Revenue and Cut Costs? Align Your Sales and Service Teams — Salesforce AI Blog
• Extending Admin Control to the Runtime Layer with NVIDIA OpenShell and Slack — Salesforce AI Blog
• 21 Things We Announced at Dreamforce 2026 — Salesforce AI Blog
• Sneak Peek Into the Small Business Service Playbook (3 Key Insights) — Salesforce AI Blog

📊 数据概览