Unverified50% confidenceFactExact time
Advanced RAG introduces Query Rewriting and HyDE (Hypothetical Document Embeddings) before retrieval to optimize query quality, and improves relevance after retrieval through Cross-Encoder re-ranking and context compression.
1
Sources
50%
Confidence
Long-term
Relevance
9/10/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedQuery Rewriting and Reranking are identified as the two most critical optimization techniques in Advanced RAG.75% similarUnverifiedDify supports hybrid retrieval (vector retrieval + BM25 keyword retrieval, fused via the RRF algorithm) and Reranker re-ranking in its RAG pipeline.72% similarUnverified重排模型会对查询和每个候选文档进行更精细的相关性打分，将最相关的文档排在前列，从而提升检索质量70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/886988API
curl https://kongchang.com/api/v1/knowledge/claims/886988MCP
get_claim(id=886988)