Reasoning in Trees: Improving Retrieval-Augmented Generation for Multi-Hop Question Answering

TL;DR

This paper introduces RT-RAG, a hierarchical reasoning tree framework that improves multi-hop question answering by reducing errors in query decomposition and evidence collection, significantly outperforming existing methods.

Contribution

The paper proposes RT-RAG, a novel structured approach that decomposes questions into reasoning trees and employs bottom-up evidence collection to enhance multi-hop QA accuracy.

Findings

RT-RAG outperforms state-of-the-art methods by 7.0% F1 and 6.0% EM.

Structured reasoning trees improve decomposition accuracy.

Iterative query refinement reduces error propagation.

Abstract

Retrieval-Augmented Generation (RAG) has demonstrated significant effectiveness in enhancing large language models (LLMs) for complex multi-hop question answering (QA). For multi-hop QA tasks, current iterative approaches predominantly rely on LLMs to self-guide and plan multi-step exploration paths during retrieval, leading to substantial challenges in maintaining reasoning coherence across steps from inaccurate query decomposition and error propagation. To address these issues, we introduce Reasoning Tree Guided RAG (RT-RAG), a novel hierarchical framework for complex multi-hop QA. RT-RAG systematically decomposes multi-hop questions into explicit reasoning trees, minimizing inaccurate decomposition through structured entity analysis and consensus-based tree selection that clearly separates core queries, known entities, and unknown entities. Subsequently, a bottom-up traversal strategy employs iterative query rewriting and refinement to collect high-quality evidence, thereby mitigating error propagation. Comprehensive experiments show that RT-RAG substantially outperforms state-of-the-art methods by 7.0% F1 and 6.0% EM, demonstrating the effectiveness of RT-RAG in complex multi-hop QA.