Unverified50% confidenceFactExact time
Vector retrieval (coarse ranking) is fast but lacks deep semantic understanding and may recall irrelevant content, so re-ranking is needed for fine ranking.
1
Sources
50%
Confidence
Long-term
Relevance
10/28/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedTraditional parallel retrieval is fast but often sacrifices accuracy, while sequential search yields more precise results but introduces noticeable latency.75% similarUnverifiedLexical overlap metrics are fast to compute, reproducible, and deterministic, but may underestimate semantically correct answers that involve synonym substitution, word-order changes, or paraphrasing.75% similarUnverified当向量库中存在大量语义相近文档时，相似度分数趋于集中，导致Top-K召回结果区分度下降，即语义粘连问题74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1106818API
curl https://kongchang.com/api/v1/knowledge/claims/1106818MCP
get_claim(id=1106818)