NODES AI 2026 - Adaptive GraphRAG: A Framework for Knowledge Graph Quality, Consistency, Evolution

In this session, Joshua Yu presents "Adaptive GraphRAG: A Framework for Knowledge Graph Quality, Consistency, and Evolution" at NODES AI. The talk addresses the critical challenge of maintaining the accuracy, consistency, and trustworthiness of knowledge graphs (KGs) as new data is introduced. He introduces Adaptive GraphRAG, a practical framework based on Neo4j, which focuses on continuous enrichment and evolution of KGs. The speaker discusses techniques for detecting and repairing coreferences, entity drift, and semantic contradictions, as well as enforcing entity consistency across ingestion cycles. This talk provides valuable insights for those deploying GraphRAG or enterprise knowledge systems, offering mechanisms to ensure high-quality, self-improving knowledge graphs over time.