Retrieval-Augmented Generation (RAG) has recently emerged as a powerful framework to enhance Large Language Models (LLMs) by leveraging external knowledge sources. However, traditional RAG systems relying on flat data structures often struggle with complex relationships and semantic reasoning. This survey explores the integration of knowledge graphs (KGs) into RAG systems, which offers structured information storage, enhanced semantic understanding, and dynamic update capabilities. We systematically review current methodologies across the stages of graph construction, graph retrieval, and augmented generation, highlighting key innovations and challenges. Additionally, we examine datasets, evaluation metrics, and experimental frameworks, and propose future research directions to advance RAG systems. This paper serves as a comprehensive resource for understanding the state-of-the-art and identifying open problems in this emerging field.
Keywords: Retrieval-Augmented Generation, Large Language Models, Knowledge Graphs
Abstract:
Submission Number: 2
Loading