Gemini-embedding-001 vs text-multilingual-embedding-002

Google Cloud Build with AI AI Solutions
1

Hi, in my solution I have replaced the model with the new one recommended by google gemini-embedding-001. The problem is that the vectors generated have different values compared to the previous model text-multilingual-embedding-002. This causes problems in comparing them. I have database with many embedding and for which I am looking for similarities and now for new documents I find nothing.

Example: one document, two models - different vectors - different distance - not similar.

Is that true that both models give different not comparable vectors or I’m doing something wrong.

2

Different models give different outputs. Each model has its own vector space. If you’re using a new embedding model, you have to re-embed everything.

3

ive been wanting to try that for a minute