RAG(Retrieval-Augmented Generation,检索增强生成)把外部资料先检索成相关片段,再交给模型生成。经典论文把它拆成两种记忆:模型参数里的参数化记忆,以及可从外部索引取回的非参数化记忆。1
References
- 1
- 2Retrieval | OpenAI API
developers.openai.com
- 3Retrieval augmented generation (RAG) for projects
support.anthropic.com
- 4Local on-device LLM with RAG through the browser
huggingface.co


Comments (2)
Sign in to comment.