KV Cache on Yiwen Cai

KV Cache on Yiwen Caihttps://yiwen-cai.github.io/tags/kv-cache/Recent content in KV Cache on Yiwen CaiHugo -- gohugo.iozh-cncaiyiwen.cs@foxmail.com (Yiwen Cai)caiyiwen.cs@foxmail.com (Yiwen Cai)© 2026 Yiwen CaiFri, 10 Jul 2026 22:20:09 +0800长上下文 LLM 推理中的 KV Cache 优化综述：系统管理、缓存压缩与架构协同https://yiwen-cai.github.io/notes/llm-inference/kv-cache-optimization-survey/Fri, 10 Jul 2026 00:00:00 +0000caiyiwen.cs@foxmail.com (Yiwen Cai)https://yiwen-cai.github.io/notes/llm-inference/kv-cache-optimization-survey/以 KV Cache Size = 2×L×T×H_kv×D_h×bytes 为锚点，系统梳理 2023–2026 年 KV Cache 优化全景：PagedAttention 与 KV-aware serving 的系统管理、DapQ/LaProx/TurboQuant/LongFlow 等压缩新方法、MHA→MLA→DeepSeek V4 CSA/HCA→GDN→MLRA 的架构演进，并讲清四条路线（压维度/压序列/少访问/固定状态）的边界。SparseSpec：加速推理模型的稀疏自推测解码https://yiwen-cai.github.io/notes/papers/sparsec-speculative-decoding/Sat, 20 Jun 2026 00:00:00 +0000caiyiwen.cs@foxmail.com (Yiwen Cai)https://yiwen-cai.github.io/notes/papers/sparsec-speculative-decoding/针对推理语言模型长输出的 memory-bound 瓶颈，用同一模型做 self-speculative decoding——verification 阶段顺手 dump 出 attention scores 做 Top-K，作为后续 draft 的动态稀疏模式，零训练、无损、最高 2.13× 加速。