hard20 pts · first answer countsembeddingsvector-search

Upgrading the embedding model

You switch to a better embedding model for new documents only and keep old vectors in the same index. Search quality becomes very poor. Why?

A.New models always produce worse embeddingsB.Vectors from different embedding models live in different spaces and are not comparable; you must re-embed the whole corpus (and queries) with one modelC.The vector database needs a restart after any changeD.The index needs more replicas

Discussion 0

Join free to post in the discussion.

Upgrading the embedding model

Discussion 0

Related questions