advanced rag techniques

[andrecajuscarvalho]

No description provided.

[JdFSilva]

What's the difference between the hybrid approach VS RAG fusion? is it just the way that the retrievals (sparse and dense) are merged?

[JdFSilva]

I'm not sure I understood the difference between Hyde and RRR, both generate a new prompt from the user's prompt. Is it just because one generates a possible answer to the prompt and the other rewrites the prompt itself?

[andrecajuscarvalho]

What's the difference between the hybrid approach VS RAG fusion? is it just the way that the retrievals (sparse and dense) are merged?

Yes, the hybrid approach simply combines dense and sparse retrieval results using fixed weights, while RAG Fusion refines this by generating sub queries and applying Reciprocal Rank Fusion to dynamically rank and merge the results.

[JdFSilva]

Right now we're using the Naive approach right?
To change it to a Hybrid one, do we only have to add the Sparse Retrieval and the Fusion? From what I gathered, we already everything besides those parts

[andrecajuscarvalho]

I'm not sure I understood the difference between Hyde and RRR, both generate a new prompt from the user's prompt. Is it just because one generates a possible answer to the prompt and the other rewrites the prompt itself?

Hyde works by creating a draft answer based on the question, and then uses that draft to find documents that match its ideas. In contrast, RAG Fusion makes extra related questions from the original query and uses those to sort and rank the documents it retrieves.

[andrecajuscarvalho]

Right now we're using the Naive approach right?
To change it to a Hybrid one, do we only have to add the Sparse Retrieval and the Fusion? From what I gathered, we already everything besides those parts

Yes 👍! To switch to a Hybrid approach, we would add a sparse retrieval component (like BM25) and then fuse its results with the dense results! This extra step provides a richer context for the LM...