英伟达发布 BlueField-4 DPU：集成 64 核 Arm CPU，支持 800G 网络
"KV缓存" 相关文章
Not All Heads Matter: A Head-Level KV Cache Compression Method with Integrated Retrieval and Reasoning
Attention Is All You Need for KV Cache in Diffusion LLMs
LouisKV: Efficient KV Cache Retrieval for Long Input-Output Sequences
PatternKV: Flattening KV Representation Expands Quantization Headroom
AdaptCache: KV Cache Native Storage Hierarchy for Low-Delay and High-Quality Language Model Serving
SemShareKV: Efficient KVCache Sharing for Semantically Similar Prompts via Token-Level LSH Matching
EpiCache: Episodic KV Cache Management for Long Conversational Question Answering
KV缓存不再爆！清华姚期智团队重写注意力维度，长上下文更省更强 | NeurIPS 2025 Spotlight
AIBrix v0.4.0 发布：P/D 解耦与专家并行支持、KVCache v1 连接器、KV 事件同步与多引擎支持
LLM 推理经济学
Understanding and Coding the KV Cache in LLMs from Scratch
Where do LLMs spend their FLOPS?
Transformer inference tricks
Neural Attention Search
AI代理的上下文工程：构建Manus的经验教训
AIBrix v0.4.0 发布：P/D 解耦与专家并行支持、KVCache v1 连接器、KV 事件同步与多引擎支持
KVComp: A High-Performance, LLM-Aware, Lossy Compression Framework for KV Cache
Huawei CloudMatrix: A Peer-to-Peer AI Datacenter Architecture for Scalable and Efficient LLM Serving
AIBrix v0.4.0 发布
