A Unified Framework for Context-Aware and Relation-Aware Graph Retrieval-Augmented Generation

The paper introduces HyGRAG, a hierarchical graph framework for Retrieval-Augmented Generation (RAG) that integrates contextual and relational information from external knowledge sources. It employs hybrid graphs with chunk and entity nodes, allowing for dynamic updates and efficient retrieval across multiple abstraction levels. Experimental results indicate a 9.7% improvement in multi-hop reasoning accuracy, highlighting its potential for enhancing the performance of large language models in knowledge-intensive applications.