RAG-Fusion: Ask the Question Many Ways, Then Fuse the Ranks (RRF)
RAG-Fusion is a technique designed to improve the accuracy of retrieval-augmented generation (RAG) systems by addressing the limitations of single-query phrasing. It involves having a large language model generate multiple variations of a user's question, performing a vector search for each variation, and then fusing the results using reciprocal rank fusion (RRF). This method prioritizes documents that appear with high ranks across multiple queries, leading to more robust retrieval than relying on a single, potentially suboptimal, phrasing. AI
IMPACT Improves retrieval robustness in RAG systems by using multiple query phrasings and rank fusion, reducing reliance on single-query accuracy.