Unverified85% confidenceFactTime unknown
实际生产中通常会结合向量相似度和关键词匹配两种方式进行混合检索
1
Sources
85%
Confidence
Long-term
Relevance
6/3/2026
First Seen
Sources
RAG系统全流程解析：从向量索引到生产优化
bilibili大模型本地部署
Related Entities
Related Claims
Verified关键词搜索基于精确字符匹配无法理解同义词，语义搜索依赖向量嵌入可捕捉语义关联，两者结合的混合检索效果优于单一方法73% similarVerified混合检索（Hybrid Search）将向量检索与BM25关键词检索的结果通过RRF（倒数排名融合）等算法合并，兼顾语义理解与精确匹配70% similarUnverified混合检索（Hybrid Search）结合稠密向量语义相似性与BM25稀疏关键词匹配，两路结果经Re-Ranker模型重排后注入LLM上下文68% similarUnverifiedVector retrieval excels at capturing semantic similarity (e.g., 'automobile' and 'car'), while BM25 excels at exact keyword matching (e.g., product model numbers, proper nouns)67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/38317API
curl https://kongchang.com/api/v1/knowledge/claims/38317MCP
get_claim(id=38317)