Unverified50% confidenceFactExact time
RAG由Meta AI的Lewis等人于2020年正式提出，最初架构用密集向量检索器DPR从Wikipedia召回文档段落再喂给seq2seq生成器
1
Sources
50%
Confidence
Long-term
Relevance
7/15/2026
First Seen
Sources
Related Claims
UnverifiedRAG paper authors Lewis et al. proposed an end-to-end trainable architecture that combines a Dense Passage Retriever (DPR) with a seq2seq generative model.74% similarVerifiedRAG (Retrieval-Augmented Generation) was proposed by Meta in 202071% similarUnverifiedThe core contribution of RAG lies in combining dense vector retrieval (DPR) with a sequence-to-sequence generation model (seq2seq).67% similarUnverifiedRAG架构核心思路是将检索与生成两个模块解耦，检索模块从外部知识库实时拉取相关内容，生成模块以检索结果作为锚点组织输出66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/518300API
curl https://kongchang.com/api/v1/knowledge/claims/518300MCP
get_claim(id=518300)