Hello. I’m currently using your product for a RAG system. I’m using the model paraphrase-multilingual-MiniLM-L12-v2 to generate embeddings. However, when I query the vector database with a relevant question — one that definitely has an answer in the database — I get a very low score. Could you please advise how I can fix this?
Hi @7wafer7 ,
Can you give an example of what you’re seeing? What are the responses you’re getting back and what is the expected response based on ground truth?
Also, are you using any metadata filters? It’s possible you’re filtering out the most relevant results.
Cory