In recent years, Large Language Models (LLMs) have significantly advanced the field of natural language processing by enabling powerful text generation and understanding capabilities. However, these models often suffer from limitations such as hallucination, lack of real-time knowledge, and poor handling of complex relational queries. Retrieval-Augmented Generation (RAG) has emerged as a solution to enhance LLMs by integrating external knowledge sources. Despite its effectiveness, traditional RAG systems rely primarily on vector similarity search, which fails to capture deep relationships between entities. This paper introduces Graph Retrieval-Augmented Generation (GraphRAG), a novel approach that integrates knowledge graphs with LLMs to enable context-aware question answering. By representing data as nodes and relationships, GraphRAG enhances semantic understanding, supports multi-hop reasoning, and improves answer accuracy. The proposed system combines graph traversal techniques with embeddingbased retrieval to provide enriched contextual information to LLMs. Experimental observations demonstrate that GraphRAG significantly reduces hallucinations, improves response relevance, and provides explainable outputs compared to traditional RAG systems. The approach is particularly effective in domains such as healthcare, academic research, and enterprise knowledge management, where understanding relationships between entities is critical.
Vijay Kumar Verma (Sun,) studied this question.