Unverified60% confidenceFactExact time
GQA was first proposed by Google during the training of PaLM 2 and subsequently adopted by open-source models such as Llama 2/3 and Mistral.
2
Sources
60%
Confidence
Long-term
Relevance
8/20/2026
First Seen
Sources
Related Claims
UnverifiedGQA由Google在2023年提出，将多个查询头共享同一组Key-Value头，已被LLaMA 3、Mistral等主流模型采用69% similarVerifiedGQA (Grouped Query Attention) is an efficient variant of standard multi-head attention, proposed by Google in 2023, and has been adopted by mainstream models such as LLaMA 3 and Mistral.69% similarVerifiedMainstream open-source models such as Meta's LLaMA and Google's Gemma make extensive use of quantization and knowledge distillation techniques63% similarUnverifiedOpen-source systems such as SpiceDB and OpenFGA are designed with Google Zanzibar as their prototype63% similarUnverifiedGLM 4.7和Minimax M2.1均采用开源策略，模型权重公开可用63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/778809API
curl https://kongchang.com/api/v1/knowledge/claims/778809MCP
get_claim(id=778809)