Find answers from the community

Updated 2 months ago

Reramker

Hybrid Search and Re-ranking:

Hello team, I'm planning to implement Qdrant Hybrid Search: https://docs.llamaindex.ai/en/stable/examples/vector_stores/qdrant_hybrid.html

At the moment, I'm using Cohere for re-ranking at the final stage. However, the document mentions, "A fusion algorithm is applied to rank and order the nodes from different vector spaces (relative score fusion in this case)." Does this mean that the search already includes a built-in re-ranker, and therefore, I wouldn't need to use Cohere if I opt for this?
L
S
6 comments
No, it's just using an algorithm to normalize and order the nodes retrieved from dense and sparse embeddings.

Not quite reranking, and it's purely simple math
Got it, thanks @Logan M

Typically with Cohere I'd first do a top-10, sent to Cohere and get top-2 from it. If I were to combine Hybrid with Cohere, how would you recommend I proceed?

I see that Hybrid has similarity_top_k and sparse_top_k. I'm wondering how to put these together with Cohere re-ranking.
Yea it's maybe a tad confusing.

If you only set similarity_top_k, it will retrieve k samples from the dense and k samples from the sparse, rank, and then return the set top k. Then from there you would rerank

The sparse top k let's you control how much is retrieved from the sparse embeddings collection
Is 50% a reasonable split?
Add a reply
Sign up and join the conversation on Discord