Unverified50% confidenceFactExact time
MQA令所有查询头共享同一组Key-Value以压缩KV Cache
1
Sources
50%
Confidence
Long-term
Relevance
7/15/2026
First Seen
Sources
Related Claims
VerifiedGrouped Query Attention (GQA) and Multi-Query Attention (MQA) reduce KV cache memory usage during inference by having multiple query heads share the same group of key/value heads, and are standard optimizations in models such as Llama 3 and Mistral.65% similarUnverifiedQKV分别代表Query（查询）、Key（键）和Value（值），命名借鉴自信息检索系统概念59% similarUnverifiedGQA让多个查询头共享同一组键值头，Llama 2 70B 和 Llama 3 系列均采用了 GQA58% similarUnverifiedThe Slot+Scope unique key design draws on the concept of Composite Primary Keys in relational databases.58% similarUnverifiedsqlite-utils 4.1 的 upsert 命令能够自动推断已有表的主键，对已定义主键的表可省略 --pk 参数57% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/511288API
curl https://kongchang.com/api/v1/knowledge/claims/511288MCP
get_claim(id=511288)