Unverified50% confidenceCautionExact time
更换或升级嵌入模型时，旧向量与新向量处于不同语义空间无法直接比较，往往需要全量重建索引
1
Sources
50%
Confidence
Long-term
Relevance
8/27/2026
First Seen
Sources
Related Entities
Related Claims
Unverifiedembedding模型的版本更新可能导致语义向量空间发生偏移，使原本高排名的相关文档被不太相关的文档取代74% similarUnverified向量索引方案存在索引随代码变更过时、以及语义相似度可能错误召回功能相似但实现无关代码的缺陷67% similarUnverifiedFailure of vector indexes to reflect data changes in a timely manner leads to two types of problems: stale retrieval (generating answers based on outdated documents) and ghost data (source documents have been deleted but vectors remain in the index and are still retrieved).66% similarUnverifiedGenerating document vectors and query vectors must use the same embedding model, as the vector spaces of different models are mutually incompatible, and mixing them will cause retrieval to fail.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/809974API
curl https://kongchang.com/api/v1/knowledge/claims/809974MCP
get_claim(id=809974)