Multi-Meta-RAG: Improving RAG for Multi-Hop Queries using Database Filtering with LLM-Extracted Metadata
This addresses the issue of answering multi-hop questions in RAG systems, which is incremental as it builds on existing RAG methods with domain-specific enhancements.
The paper tackles the problem of poor performance of traditional retrieval-augmented generation (RAG) on multi-hop queries by introducing Multi-Meta-RAG, which uses database filtering with LLM-extracted metadata to improve document selection, resulting in greatly improved results on the MultiHop-RAG benchmark.
The retrieval-augmented generation (RAG) enables retrieval of relevant information from an external knowledge source and allows large language models (LLMs) to answer queries over previously unseen document collections. However, it was demonstrated that traditional RAG applications perform poorly in answering multi-hop questions, which require retrieving and reasoning over multiple elements of supporting evidence. We introduce a new method called Multi-Meta-RAG, which uses database filtering with LLM-extracted metadata to improve the RAG selection of the relevant documents from various sources, relevant to the question. While database filtering is specific to a set of questions from a particular domain and format, we found out that Multi-Meta-RAG greatly improves the results on the MultiHop-RAG benchmark. The code is available at https://github.com/mxpoliakov/Multi-Meta-RAG.