文献类型：期刊文献

英文题名：KG-o1: Enhancing Multi-hop Question Answering in Large Language Models via Knowledge Graph Integration

作者：Wang, Nan[1]; Fan, Yongqi[1]; Zhu, Yansha[1]; Wang, ZongYu[2]; Cao, Xuezhi[2]; He, Xinyan[2]; Jiang, Haiyun[3]; Ruan, Tong[1]; Liu, Jingping[1]

机构：[1] School of Information Science and Engineering, East China University of Science and Technology, Shanghai, China; [2] Meituan, Shanghai, China; [3] School of Automation and Perception, Shanghai Jiao Tong University, China

年份：2025

外文期刊名：arXiv

收录：EI(收录号：20250392773)

语种：英文

外文关键词：Graph theory - Knowledge graph - Question answering

摘要：Large Language Models (LLMs) face challenges in knowledge-intensive reasoning tasks like classic multi-hop question answering, which involves reasoning across multiple facts. This difficulty arises because the chain-of-thoughts (CoTs) generated by LLMs in such tasks often deviate from real or a priori reasoning paths. In contrast, knowledge graphs (KGs) explicitly represent the logical connections between facts through entities and relationships. This reflects a significant gap. Meanwhile, large reasoning models (LRMs), such as o1, have demonstrated that long-step reasoning significantly enhances the performance of LLMs. Building on these insights, we propose KG-o1, a four-stage approach that integrates KGs to enhance the multi-hop reasoning abilities of LLMs. We first filter out initial entities and generate complex subgraphs. Secondly, we construct logical paths for subgraphs and then use knowledge graphs to build a dataset with a complex and extended brainstorming process, which trains LLMs to imitate long-term reasoning. Finally, we employ rejection sampling to generate a self-improving corpus for direct preference optimization (DPO), further refining the LLMs’ reasoning abilities. We conducted experiments on two simple and two complex datasets. The results show that KG-o1 models exhibit superior performance across all tasks compared to existing LRMs. The datasets, models, and code for KG-o1 are publicly available1 Copyright ? 2025, The Authors. All rights reserved.