Multi-Head RAG: Solving Multi-Aspect Problems with LLMs

TL;DR

Multi-Head RAG (MRAG) enhances retrieval-augmented generation by using multi-head attention activations to better capture diverse semantic aspects, significantly improving retrieval success for complex, multi-aspect queries.

Contribution

Introducing MRAG, a novel approach that uses Transformer multi-head attention as retrieval keys to improve multi-aspect document retrieval in RAG systems.

Findings

Up to 20% higher retrieval success ratios on real-world tasks.

MRAG outperforms 18 RAG baselines in complex query retrieval.

Improved downstream language model generation quality.

Abstract

Retrieval-Augmented Generation (RAG) improves Large Language Models (LLMs) by retrieving supporting documents into the prompt, but existing methods do not explicitly target queries that require fetching multiple documents with substantially different content. Such multi-aspect queries are challenging because relevant documents can be far apart in embedding space, making joint retrieval difficult. We introduce Multi-Head RAG (MRAG), which addresses this gap with a simple yet powerful idea: using Transformer multi-head attention activations rather than the standard decoder-layer embedding, as retrieval keys. It leverages the observation that different heads capture different semantic aspects. This yields multi-aspect embeddings for both documents and queries, improving retrieval accuracy on complex queries. We show MRAG's design advantages over 18 RAG baselines, up to 20% higher retrieval success ratios for real-world use cases, and improved downstream LLM generation. MRAG integrates seamlessly with existing RAG frameworks and benchmarks.