PathRAG
PathRAG is an advanced rag technique that extends GraphRAG and LightRAG by retrieving along paths in knowledge graphs to handle multi-hop queries and complex relationships.
Evolution of RAG Systems
From the Discover AI channel - Graph RAG evolved:
- Traditional RAG: Limited by document chunking and lack of semantic connections despite optimization techniques (e.g., chunking strategies, pre/post-retrieval optimization).
- GraphRAG: Uses knowledge graphs to model entity relationships for contextual retrieval.
- LightRAG: Efficient, lightweight variant of GraphRAG optimized for scalability.
- PathRAG: Retrieves along graph paths to capture multi-hop relationships, enabling deeper context for complex queries.
For implementation details, see the video.
2026 04 14 Discover AI channel Graph RAG evolved