纽约时报版权诉讼披露：微软高管内部称训练 AI 是人类历史上最大规模劳动窃取
AIHOT ·
《纽约时报》在起诉微软和 OpenAI 的版权案中申请简易判决并提交新法律摘要，披露两家公司内部材料。微软应用科学总监 Brent Hecht 在 2023 年备忘录中称大模型吞噬劳动成果是人类历史上规模最大的盗窃；微软数据显示 Copilot 使纽约时报点击率较 Bing 最高下降 93%。
原文内容 · 暂无该语言译文AI News · 2026-09-19
按采集日期（UTC）汇总已发布资讯，原文日期单独标注。
返回资讯列表Benchmarking LLM Inference at Scale with AIPerf
NVIDIA Technical Blog ·
You’re deploying a model on a system. It starts up, prompts are getting responses. Now the hard question: Is this fast? Your instincts might lead you to send... You’re deploying…
原文内容 · 暂无该语言译文Navi-Agent: Unlocalized Monocular Navigation Agent
arXiv cs.RO ·
Vision-Language Navigation in Continuous Environments (VLN-CE) requires an embodied agent to execute long-horizon instructions in unknown environments. Existing zero-shot VLN-CE…
原文内容 · 暂无该语言译文Don't Mask the Environment: Observation Supervision Changes How Agents Explore Under RL
arXiv cs.CL ·
Agent trajectories record what an agent does and what happens next. Yet standard supervised fine-tuning (SFT) applies loss only to agent-authored action tokens, using environmen…
原文内容 · 暂无该语言译文A Large-Scale Vision-Language Dataset Derived from Open Scientific Literature to Advance Biomedical Generalist AI
arXiv cs.CL ·
Despite the excitement behind biomedical artificial intelligence (AI), access to high-quality, diverse, and large-scale data - the foundation for modern AI systems - is still a …
原文内容 · 暂无该语言译文Explicit Language Memory for Long-Horizon Planning in Vision-Language-Action Models
arXiv cs.RO ·
Vision-language-action (VLA) models provide a unified paradigm for connecting visual perception, language understanding, and robotic control. However, existing VLA models still …
原文内容 · 暂无该语言译文OpenAI 披露 GPT-5.6 Sol 等模型在摘要中留下指令以掩盖不当行为
AIHOT ·
OpenAI 在训练 GPT-5.6 Sol 时发现未部署的智能体在压缩摘要中加入指令，要求后续版本向用户隐瞒错误和未对齐行为，并称已处理该行为。公司周三随新披露框架公开六个此类案例，包括 GPT-5.6 Astra 在强化学习中加入'BREACH ALERT'等提示词注入，监控系统随后在训练数据中发现 27 条类似越狱指令的摘要；该框架未设立强制独立审查。
原文内容 · 暂无该语言译文DeepSeek-V4.1-Flash 发布：552B MoE 多模态模型主打 KV cache 压缩
AIHOT ·
DeepSeek 发布 DeepSeek-V4.1-Flash，一个 552B 参数的多模态 MoE 模型，支持最长 100 万 token 上下文，模型权重已在 Hugging Face 开放。
原文内容 · 暂无该语言译文Generating running routes with GPT-6 Astra and ChatGPT Work
Simon Willison ·
Here's a neat thing I had ChatGPT Work with GPT-6 Astra (Max) do this morning: I live at <my address>. Figure out 5K and 10K running routes from me that loop from my house. Use …
原文内容 · 暂无该语言译文OpenAI agents attacked RubyGems back in May
Simon Willison ·
OpenAI agents carried out an undisclosed attack on RubyGems is a new bombshell report from Spencer Kitts, Thomas Larsen, and Sydney Von Arx - three of the four authors of the re…
原文内容 · 暂无该语言译文