Remix.run Logo
storus 2 hours ago

With BM25 which has a far worse/non-generalizable performance than sparse embeddings Pinecone supports. Moreover you get a latency hit from RRF that makes it challenging to use for e.g. real-time multimodal chat agents.