A Survey of Graph Retrieval-Augmented Generation for Customized Large Language Models

TL;DR

<3-5 sentence high-level summary>GraphRAG addresses the domain adaptation gap of large language models by introducing graph-structured representations for knowledge organization, retrieval, and integration. It systematically analyzes three GraphRAG paradigms—knowledge-based, index-based, and hybrid—and provides detailed pipelines for constructing graphs, retrieving subgraphs, and integrating retrieved knowledge into LLM prompts or training. The survey covers theoretical foundations, implementation patterns across domains, benchmark datasets, and a rich set of open-source resources, highlighting challenges in knowledge quality, privacy, and efficiency. The work offers practical guidance for deploying graph-enhanced LLMs in production and identifies promising directions for scalable, interpretable, and faithful domain-specific AI systems.

Abstract

Large language models (LLMs) have demonstrated remarkable capabilities in a wide range of tasks, yet their application to specialized domains remains challenging due to the need for deep expertise. Retrieval-Augmented generation (RAG) has emerged as a promising solution to customize LLMs for professional fields by seamlessly integrating external knowledge bases, enabling real-time access to domain-specific expertise during inference. Despite its potential, traditional RAG systems, based on flat text retrieval, face three critical challenges: (i) complex query understanding in professional contexts, (ii) difficulties in knowledge integration across distributed sources, and (iii) system efficiency bottlenecks at scale. This survey presents a systematic analysis of Graph-based Retrieval-Augmented Generation (GraphRAG), a new paradigm that revolutionizes domain-specific LLM applications. GraphRAG addresses traditional RAG limitations through three key innovations: (i) graph-structured knowledge representation that explicitly captures entity relationships and domain hierarchies, (ii) efficient graph-based retrieval techniques that enable context-preserving knowledge retrieval with multihop reasoning ability, and (iii) structure-aware knowledge integration algorithms that leverage retrieved knowledge for accurate and logical coherent generation of LLMs. In this survey, we systematically analyze the technical foundations of GraphRAG and examine current implementations across various professional domains, identifying key technical challenges and promising research directions. All the related resources of GraphRAG, including research papers, open-source data, and projects, are collected for the community in https://github.com/DEEP-PolyU/Awesome-GraphRAG.

Figures (4)

• Figure 1: The development trends in the field of GraphRAG with representative works.
• Figure 2: A comprehensive overview of traditional RAG and two typical GraphRAG workflows. Non-graph RAG organizes the corpus into chunks, ranks them by similarity, and retrieves the most relevant text for generating responses. Knowledge-based GraphRAG extracts detailed knowledge graphs from the corpus using entity recognition and relation extraction, offering fine-grained, domain-specific information. Index-based GraphRAG summarizes the corpus into high-level topic nodes, which are linked to form an index graph while fact-linking maps topics to text. This two-layer structure combines efficient topic retrieval with detailed text knowledge, offering scalability and performance compared to Knowledge-based GraphRAG.
• Figure 3: The taxonomy for existing GraphRAG methods in the survey.
• Figure 4: The illustration of the comparison between traditional RAG and GraphRAG.