2026-08-07 科技日报

扫描 552 篇候选内容 · 覆盖 650 个信息源 · 纳入 536 篇

📌 今日要点

• Mind the Cap: Output-Budget Regimes Change the Measured Multilingual Reasoning Gap
• OpenAI improves GPT-5.6 Sol in ChatGPT and restricts free users to its weakest model
• Deepmind’s talent drain likely comes down to chip shortages, a conflict of interest, and Google’s bureaucracy
• Microsoft’s AI revenue reportedly depends on OpenAI for 70 percent
• Claude Code is the fastest agent framework but costs nearly three times more than the cheapest rival
• Qwen3.8 Max catches Claude Opus 4.8 but Kimi K3 still scores higher for 25 percent less

📝 科技简讯

• Mind the Cap: Output-Budget Regimes Change the Measured Multilingual Reasoning Gap — ArXiv CL (cs.CL)
• OpenAI improves GPT-5.6 Sol in ChatGPT and restricts free users to its weakest model — The Decoder
• Deepmind’s talent drain likely comes down to chip shortages, a conflict of interest, and Google’s bureaucracy — The Decoder
• Microsoft’s AI revenue reportedly depends on OpenAI for 70 percent — The Decoder
• Claude Code is the fastest agent framework but costs nearly three times more than the cheapest rival — The Decoder
• Qwen3.8 Max catches Claude Opus 4.8 but Kimi K3 still scores higher for 25 percent less — The Decoder
• The company that made open weights mainstream now competes on discounts — The Decoder
• OpenAI reportedly slows research after its own models secretly coordinated hacks for weeks undetected — The Decoder
• Advancing Utility Pole and Sign Detection Through Deep Learning — ArXiv CV (cs.CV)
• LoRetta: A Foundation Model and Extensive Dataset for Global-Scale Remote Sensing Dense Image Matching — ArXiv CV (cs.CV)
• GEB-Bench: Abstract Structures Told in Many Voices — ArXiv CV (cs.CV)
• Perception Before Reasoning: Dynamic Latent Reasoning for Video Understanding and Question Answering — ArXiv CV (cs.CV)
• Teaching Foundation Models to Read mmWave: Pose-Guided Kinematic Representation for Human Behavior Understanding — ArXiv CV (cs.CV)
• Why do models task game? — AI Alignment Forum
• Some SHA-256 hashes — LessWrong
• Have models report provable security bugs in their environment — LessWrong
• Why do models task game? — LessWrong
• Contra Oster on Alcohol in Pregnancy. Part 1. The pharmacokinetics of alcohol metabolism — LessWrong
• Side-Effects of Length Penalty in RL — LessWrong
• Agentic AI in HR: Why Employee Trust Starts with the Data Foundation — Salesforce AI Blog
• 2026 Customer Success Awards — Salesforce AI Blog
• 4 Steps to Eliminate Identity Debt and Build Reliable Agentic AI with Data 360 — Salesforce AI Blog
• What is Emergent Leadership And Why Does Your Growing Business Need it? — Salesforce AI Blog
• A Long-Run Persistence Theory for AI Systems under the Redundancy-Adjusted Artificial Age Score (AAS) — arXiv AI (cs.AI)
• The LLM Proposes, the Executive Disposes: A Self-Verifying Agent Instrument that Dissociates Commitment Drift from Binding Drift in Long-Horizon Agents — arXiv AI (cs.AI)
• Artificial Analysis 榜单：阿里 Qwen3.8Agentic 能力得分全球第一 — 量子位
• 这人谁啊？哈萨比斯都让位了 — 量子位
• MiniMax H3 登顶开源社区第一，定义视频模型领域“斩杀线” — 量子位
• 超级算力枢纽远景乌兰察布星河基地投产，全球最大 AI 算力超级单体落地 — 量子位
• 谷歌传奇 Jeff Dean 今天离职，中国 AI 下一秒交卷！ — 新智元
• 谷歌 AI 换帅！13 年老将接棒 DeepMind，新掌门却不叫 CEO — 新智元
• 清华校友出手！视频界首个千亿 MoE 开源，6B 激活挤进全球第 6 — 新智元
• 美国 AI 安全新规出炉！最强闭源模型自愿送测，开放权重直接放行 — 新智元
• AI 能接管实验室了？中国科大最新研究给出真实物理世界的压力测试 — 新智元
• OpenAI’s new AI smart speaker will reportedly sell for between $300 and$400 — TechCrunch AI
• ChatGPT brings unlimited text chats to free users — TechCrunch AI
• Naïve raises $28.5M to automate the grunt work of setting up and running a company — TechCrunch AI
• Gen Z dating apps like Ditto ditch swiping in favor of AI matchmaking — TechCrunch AI
• OpenAI says Apple’s own security practices undermine its trade secrets case — TechCrunch AI
• Amid legal battles, Suno says it will start watermarking songs — TechCrunch AI

📊 数据概览