Partially Verified65% confidenceFactTime unknown
RAG系统的召回主要有两种方式：向量相似度计算和关键词匹配
3
Sources
65%
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
UnverifiedRAG工程实现包含离线索引与在线检索两个阶段，在线阶段通过余弦相似度或点积相似度检索Top-K相关片段拼接进Prompt79% similarUnverifiedRAG技术演进已从朴素的向量相似度检索发展到混合检索（Hybrid Search）、重排序（Reranking）和图谱增强检索等更高阶方案78% similarVerifiedRAG (Retrieval-Augmented Generation) relies on vector databases such as Pinecone and Weaviate, and embedding models to achieve semantic-level similarity matching78% similarUnverified主流 RAG 系统通常采用混合检索策略，将关键词检索与向量检索结合并通过重排序模型二次过滤77% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/38316API
curl https://kongchang.com/api/v1/knowledge/claims/38316MCP
get_claim(id=38316)