Knowledge Graph-Guided Retrieval Augmented Generation

• DOI: 10.48550/arxiv.2502.06864
• Publication Date: 2025-02-07
• Authors: Xiangrong Zhu; Yuexiang Xie; Yi Liu; Yuanman Li; Wei Hu
• Keywords: Knowledge graph

Retrieval-augmented generation (RAG) has emerged as a promising technology for addressing hallucination issues in the responses generated by large language models (LLMs). Existing studies on RAG primarily focus applying semantic-based approaches to retrieve isolated relevant chunks, which ignore their intrinsic relationships. In this paper, we propose novel Knowledge Graph-Guided Retrieval Augmented Generation (KG$^2$RAG) framework that utilizes knowledge graphs (KGs) provide fact-level relationships between improving diversity and coherence of retrieved results. Specifically, after performing retrieval seed KG$^2$RAG employs KG-guided chunk expansion process KG-based organization deliver important well-organized paragraphs. Extensive experiments conducted HotpotQA dataset its variants demonstrate advantages compared existing RAG-based approaches, terms both response quality quality.