日报精选日报精选

不靠堆砌数据，让机器人看懂真实物理世界 | 招商局狮子山实验室发布「驭物™ Lion-ER-35B-A3B」具身推理模型

仅激活3B参数，十项基准全线提升！

Palette: teaching a small model to build decks in your design language

Palette: teaching a small model to build decks in your design language Hugging Face开源/仓库数据/训练模型发布

让Agent首次拥有“视听双流记忆”：openJiuwen Agent-Memory看懂长视频，关键事件一个不落

长视频记忆精度飙升，问答时延最高降97%

NVIDIA Shows 3 Ways to Speed Up MoE Inference on DGX Spark, Lifting Ling-3.0-flash Throughput 42.5%

NVIDIA shows how 3 stacked optimizations speed up MoE inference on a single DGX Spark. On Ant Group inclusionAI's 124B Ling-3.0-flash, GSM8K decode throughput rose 42.5% with no de

Nvidia RTX Spark encrypted firmware slows inference speed to 35% on Linux

HN ⭐ 259 · 138 评论

Nvidia RTX Spark encrypted firmware slows inference speed to 35% on Linux NVIDIA 英伟达端侧行业动态部署/工程

Postman silently auto-installs AI plugins and attempts to share all API data

HN ⭐ 1393 · 461 评论

Postman silently auto-installs AI plugins and attempts to share all API data MCP/工具调用安全/对齐行业动态

Rapid LLM development is outrunning the research meant to explain it, study finds

Stanford researchers tested 50 widely cited claims about LLMs against 20 current models. 44 percent of the findings are already "expired," and another 34 percent have weakened sign

Rapid LLM development is outrunning the research meant to explain it, study finds 现象/趋势论文/研究评测/基准

StateBench: Testing how language models track state changes over long documents

StateBench: Testing how language models track state changes over long documents 推理论文/研究评测/基准