Unverified50% confidenceFactExact time
Vanilla RAG handles simple queries efficiently but performs poorly on relational reasoning or multi-hop reasoning.
1
Sources
50%
Confidence
Long-term
Relevance
9/11/2026
First Seen
Sources
Related Entities
Related Claims
Unverified大多数RAG（检索增强生成）实现只解决了'检索到什么'的问题，却忽视了检索到的内容是否可信、是否互相矛盾的更深层问题70% similarUnverifiedRAG选型决策规则：数据简单更新少只需相似检索选向量RAG，强调实体关系多跳推理选Graph RAG，需要可导航可溯源自动更新选编译式RAG67% similarUnverifiedThe answer quality of RAG depends largely on the retrieval step rather than on the model's capabilities themselves.67% similarUnverifiedMost RAG (Retrieval-Augmented Generation) tools can provide answers, but struggle to let users quickly confirm which specific sentence in the document the answer came from.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/897334API
curl https://kongchang.com/api/v1/knowledge/claims/897334MCP
get_claim(id=897334)