DMQR-RAG: Diverse Multi-Query Rewriting for RAG

TL;DR

This paper introduces DMQR-RAG, a framework that rewrites user queries diversely to improve document retrieval and response quality in retrieval-augmented generation systems, addressing noise and intent deviations.

Contribution

It proposes four query rewriting strategies and an adaptive selection method to enhance RAG performance, validated through extensive experiments.

Findings

Improved retrieval relevance and response accuracy in RAG systems.

Effective reduction of query noise and intent deviation impacts.

Demonstrated robustness across academic and industry datasets.

Abstract

Large language models often encounter challenges with static knowledge and hallucinations, which undermine their reliability. Retrieval-augmented generation (RAG) mitigates these issues by incorporating external information. However, user queries frequently contain noise and intent deviations, necessitating query rewriting to improve the relevance of retrieved documents. In this paper, we introduce DMQR-RAG, a Diverse Multi-Query Rewriting framework designed to improve the performance of both document retrieval and final responses in RAG. Specifically, we investigate how queries with varying information quantities can retrieve a diverse array of documents, presenting four rewriting strategies that operate at different levels of information to enhance the performance of baseline approaches. Additionally, we propose an adaptive strategy selection method that minimizes the number of rewrites while optimizing overall performance. Our methods have been rigorously validated through extensive experiments conducted in both academic and industry settings.