I recently surfaced above the waterline from a deep dive on Microsoft’s GraphRAG paper, hoping to learn lots on the effect of knowledge graphs in RAG. Instead I found questionable metrics and an oddly unsubstantiated AI tool.So I ran my own study using Neo4j to measure a knowledge graph’s overall effect within a RAG system.Knowledge graphs may not significantly impact context retrieval — all knowledge graph RAG methods I examined showed similar context relevancy scores to those of FAISS (~0.74)....