The community member is working on RAG (Retrieval Augmented Generation) but is not satisfied with the results as the documents (or nodes) they are retrieving are very short, only containing one or two sentences. They are asking the community for suggestions on the choice of embedding model and retriever to improve the results.
In the comments, another community member suggests checking the RAG 101 blog post to improve the documents, and referring to the MTEB leaderboard for an embedding model.
Hello Everyone, I am currently working on RAG. But the documents (or nodes) I want to retrieve are all very small (like only contains one or two sentences in each node), the result is not very satisfactory. So do you guys have any good suggestions on the choice of embed model and retriever?
