Unverified90% confidenceFactTime unknown
RAG论文的核心洞察是参数化知识（存储在模型权重中）与非参数化知识（存储在外部检索库中）可以有机结合
1
Sources
90%
Confidence
Long-term
Relevance
6/1/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedRAG (Retrieval-Augmented Generation) was proposed by Lewis et al. in 2020, with the core idea of decoupling parametric knowledge from non-parametric knowledge and dynamically retrieving relevant documents to concatenate into the prompt at inference time.76% similarUnverifiedIt is recommended to use Tongyi Qianwen's Embedding model in Dify for RAG functions such as knowledge base retrieval.69% similarUnverified传统大模型依赖参数化知识（将文本编码进模型权重参数），而RAG引入非参数化知识（将外部文档库作为可实时检索的动态知识源）69% similarUnverifiedRAG检索增强生成由Lewis等人于2020年在Facebook AI Research提出，核心在于将参数化知识与非参数化知识解耦，使知识更新无需重新训练模型69% similarVerifiedRAG effectively mitigates the two major problems of large language models—knowledge cutoff and hallucination—by forcing the model to extract answers from trusted documents.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/18819API
curl https://kongchang.com/api/v1/knowledge/claims/18819MCP
get_claim(id=18819)