GraphRAG

A form of retrieval-augmented generation that retrieves connected entities and relationships from a knowledge graph to ground an LLM's answers.

1 articleLast mentioned

GraphRAG is a form of retrieval-augmented generation (RAG) that retrieves from a knowledge graph—a structure of entities and the relationships between them—and supplies the connected entities and relations to an LLM as grounding, instead of relying only on vector similarity search over documents. Because it can gather information scattered across documents by following relationships, it is used for questions that span an entire corpus or require multiple reasoning steps.

The term became widely known when Microsoft Research released an open-source project of the same name in 2024. That project uses an LLM to extract entities and relationships from documents into a graph, partitions the graph into communities, and prepares summaries of each community for use in answering questions. The term has since also been used generically for any RAG approach that uses graph-based retrieval.

This entry is based on AIPOST articles and widely known facts. If something is wrong, please send us a correction request.

Articles covering this entry

  • NVIDIA Agent Toolkit Bets on Harness Engineering and Knowledge GraphsAI News ·

    Ontologies versus GraphRAG

    Read article

Found a mistake or missing an entry?

Wiki entries are based on AIPOST articles and widely known facts. If an entry is wrong, send us a correction request and we will check and fix it. Names from new articles are added to the wiki after the article is published.