Topic

#linear-attention

3 篇笔记

10-27

vLLM Semantic Router released d3-ultra, a 26.09B Apache-2.0 multimodal decision model scoring 87.2 on its own SR-Text benchmark.

marktechpost.com· 3 分钟 ·#clippings·#AI·#LLM

07-27

Twenty-two thousand five hundred and eighty. That’s how many GPT-2 (2019) models fit inside KimiK3 (2026). We scaled up by a factor of 22,58...

x.com· 15 分钟 ·#clippings·#AI·#LLM

05-15

五一本打算更新一年前写的开源引擎 KVCache 实现文档，但梳理下来发现，光是 KVCache 演进的理论部分就已足够撑起一篇独立文章。在 Agent 改变搜索习惯、重塑软件行业的今天，多轮交互带来的长上下文让 KVCache 持…

zhuanlan.zhihu.com· 16 分钟 ·#clippings·#LLM·#KVCache