Self-Critique Guided Iterative Reasoning for Multi-hop Question Answering

TL;DR

This paper introduces SiGIR, a novel iterative reasoning method guided by self-critique feedback, significantly improving multi-hop question answering accuracy by enabling models to self-evaluate and refine their reasoning steps.

Contribution

We propose SiGIR, a self-critique guided iterative reasoning framework that enhances multi-hop reasoning by enabling models to self-evaluate and iteratively improve their answers.

Findings

Surpasses previous SOTA by 8.6% on three datasets

Effective self-evaluation improves reasoning accuracy

Enables question decomposition and branching exploration

Abstract

Although large language models (LLMs) have demonstrated remarkable reasoning capabilities, they still face challenges in knowledge-intensive multi-hop reasoning. Recent work explores iterative retrieval to address complex problems. However, the lack of intermediate guidance often results in inaccurate retrieval and flawed intermediate reasoning, leading to incorrect reasoning. To address these, we propose Self-Critique Guided Iterative Reasoning (SiGIR), which uses self-critique feedback to guide the iterative reasoning process. Specifically, through end-to-end training, we enable the model to iteratively address complex problems via question decomposition. Additionally, the model is able to self-evaluate its intermediate reasoning steps. During iterative reasoning, the model engages in branching exploration and employs self-evaluation to guide the selection of promising reasoning trajectories. Extensive experiments on three multi-hop reasoning datasets demonstrate the effectiveness of our proposed method, surpassing the previous SOTA by $8.6\%$. Furthermore, our thorough analysis offers insights for future research. Our code, data, and models are available at Github: https://github.com/zchuz/SiGIR-MHQA.