Graph RAG
Graph RAG (Graph Retrieval-Augmented Generation) is an information retrieval and generation framework that combines knowledge graphs with retrieval-augmented generation (RAG) techniques. Its core mechanism involves constructing text corpora as graph structures—where nodes represent entities and edges represent relationships—and leveraging graph topology for multi-hop reasoning and relational information aggregation during retrieval. This approach addresses limitations of conventional RAG in handling complex relational queries and global semantic understanding, with retrieved results passed to a large language model to generate more accurate and contextually rich responses.
Core Facts
The technical evolution of RAG has gone through three stages: Naive RAG, Advanced RAG, and Modular RAG.
Graph RAG was popularized by Microsoft Research in 2024.
Graph RAG, which combines knowledge graphs with RAG, is considered one of the best-performing RAG solutions currently available.
Timeline (last 90 days)
The cost of Graph RAG is the relatively high expense of building and maintaining the graph, as well as its strong dependence on the accuracy of entity extraction.
The 'knowledge graph + large model' solution adopted by GitDecode can be understood as a Graph RAG practice.
All Facts (7)
The technical evolution of RAG has gone through three stages: Naive RAG, Advanced RAG, and Modular RAG.
80%VerifiedGraph RAG was popularized by Microsoft Research in 2024.
70%UnverifiedGraph RAG通常以Neo4j等图数据库作为底层支撑
80%UnverifiedGraph RAG和Hybrid RAG等变体进一步提升了复杂知识场景下的检索精度
75%Partially VerifiedGraph RAG, which combines knowledge graphs with RAG, is considered one of the best-performing RAG solutions currently available.
65%UnverifiedThe cost of Graph RAG is the relatively high expense of building and maintaining the graph, as well as its strong dependence on the accuracy of entity extraction.
50%UnverifiedThe 'knowledge graph + large model' solution adopted by GitDecode can be understood as a Graph RAG practice.
50%