BeamAggR: Beam Aggregation Reasoning over Multi-source Knowledge for Multi-hop Question Answering

TL;DR

BeamAggR is a novel reasoning framework that enhances multi-hop question answering by aggregating and prioritizing multi-source knowledge, significantly improving accuracy over state-of-the-art methods.

Contribution

We introduce BeamAggR, a new multi-hop QA method that parses questions into trees, explores multiple reasoning paths, and aggregates knowledge to improve retrieval and reasoning accuracy.

Findings

Outperforms SOTA methods by 8.5% on four datasets

Effectively aggregates multi-source knowledge for better reasoning

Enhances answer accuracy through probabilistic path exploration

Abstract

Large language models (LLMs) have demonstrated strong reasoning capabilities. Nevertheless, they still suffer from factual errors when tackling knowledge-intensive tasks. Retrieval-augmented reasoning represents a promising approach. However, significant challenges still persist, including inaccurate and insufficient retrieval for complex questions, as well as difficulty in integrating multi-source knowledge. To address this, we propose Beam Aggregation Reasoning, BeamAggR, a reasoning framework for knowledge-intensive multi-hop QA. BeamAggR explores and prioritizes promising answers at each hop of question. Concretely, we parse the complex questions into trees, which include atom and composite questions, followed by bottom-up reasoning. For atomic questions, the LLM conducts reasoning on multi-source knowledge to get answer candidates. For composite questions, the LLM combines beam candidates, explores multiple reasoning paths through probabilistic aggregation, and prioritizes the most promising trajectory. Extensive experiments on four open-domain multi-hop reasoning datasets show that our method significantly outperforms SOTA methods by 8.5%. Furthermore, our analysis reveals that BeamAggR elicits better knowledge collaboration and answer aggregation.