Unverified50% confidenceFactExact time
高级RAG变体包括HyDE（假设性文档嵌入）、Multi-Query RAG和Contextual Compression
1
Sources
50%
Confidence
Long-term
Relevance
7/19/2026
First Seen
Sources
Related Claims
Unverified进阶的RAG变体包括HyDE（假设性文档嵌入）、重排序（Reranker）、多跳检索等技术75% similarUnverified进阶的RAG变体如HyDE、RAG-Fusion和GraphRAG正在将RAG从简单语义搜索升级为具备深度推理能力的知识问答系统72% similarUnverifiedModular RAG引入了查询重写、HyDE（假设性文档嵌入）等检索前处理策略以提升检索召回率71% similarUnverifiedDocument Context采用RAG（检索增强生成）范式，先用向量检索找到最相关片段，再连同用户问题送给LLM69% similarUnverifiedDify的RAG模块在索引阶段将私有文档切分为文本块，通过嵌入模型转化为高维向量并存入向量数据库（如Milvus、Weaviate、pgvector等）69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/561312API
curl https://kongchang.com/api/v1/knowledge/claims/561312MCP
get_claim(id=561312)