Retrieval-Augmented Generation with Hierarchical Knowledge

2025-03-13

AI Summary

Existing RAG methods rely on flat knowledge retrieval, overlooking the inherent hierarchical structure of human cognition—thereby limiting semantic understanding and generalization in domain-specific tasks. Method: We propose HiRAG, the first framework to systematically model and leverage knowledge hierarchy for RAG enhancement. It constructs a graph-based hierarchical knowledge index during indexing and introduces a multi-granularity routing mechanism during retrieval to enable semantics-aware, layered matching, tightly integrated with LLM-based generation. Contribution/Results: HiRAG breaks away from the conventional flat paradigm, substantially improving complex reasoning and cross-granularity generalization. It achieves state-of-the-art performance across multiple domain-specific benchmarks, outperforming all existing RAG methods. The implementation is publicly available.

Technology Category

• Knowledge Representation and Reasoning: Knowledge Acquisition
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
• Search and Optimization: Metareasoning and Metaheuristics

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Querying, indexing, and retrieval in Web-related graphs

Abstract

Graph-based Retrieval-Augmented Generation (RAG) methods have significantly enhanced the performance of large language models (LLMs) in domain-specific tasks. However, existing RAG methods do not adequately utilize the naturally inherent hierarchical knowledge in human cognition, which limits the capabilities of RAG systems. In this paper, we introduce a new RAG approach, called HiRAG, which utilizes hierarchical knowledge to enhance the semantic understanding and structure capturing capabilities of RAG systems in the indexing and retrieval processes. Our extensive experiments demonstrate that HiRAG achieves significant performance improvements over the state-of-the-art baseline methods. The code of our proposed method is available at href{https://github.com/hhy-huang/HiRAG}{https://github.com/hhy-huang/HiRAG}.

Problem

• Enhances RAG systems using hierarchical knowledge
• Improves semantic understanding and structure capturing
• Outperforms state-of-the-art baseline methods

Innovation

• Hierarchical knowledge enhances RAG systems.
• HiRAG improves semantic understanding and structure.
• Significant performance gains over baseline methods.