ArchRAG: Attributed Community-based Hierarchical Retrieval-Augmented Generation
Researchers have introduced ArchRAG, a novel graph-based Retrieval-Augmented Generation (RAG) framework designed to enhance the performance of Large Language Models (LLMs) in question-answering tasks. While current state-of-the-art RAG methods utilize graph data to capture semantic relationships, they often struggle with accurately identifying relevant information and incur high token costs during online retrieval. ArchRAG addresses these limitations by augmenting questions using attributed communities and implementing a new LLM-based hierarchical clustering method. The system constructs a specialized hierarchical index structure for these communities, enabling more effective online retrieval of pertinent graph data. Experimental results indicate that ArchRAG significantly outperforms existing approaches in both accuracy and token efficiency. This development represents a significant advancement in optimizing external knowledge integration for LLMs, offering a more cost-effective and precise solution for complex query resolution. The paper, authored by Shu Wang and colleagues, was published on arXiv, contributing to the fields of Information Retrieval and Artificial Intelligence.
Wire timeline
ArchRAG: Attributed Community-based Hierarchical Retrieval-Augmented Generation
Researchers have introduced ArchRAG, a novel graph-based Retrieval-Augmented Generation (RAG) framework designed to enhance the performance of Large Language Models (LLMs) in question-answering tasks. While current state-of-the-art RAG methods utilize graph data to capture semantic relationships, they often struggle with accurately identifying relevant information and incur high token costs during online retrieval. ArchRAG addresses these limitations by augmenting questions using attributed communities and implementing a new LLM-based hierarchical clustering method. The system constructs a specialized hierarchical index structure for these communities, enabling more effective online retrieval of pertinent graph data. Experimental results indicate that ArchRAG significantly outperforms existing approaches in both accuracy and token efficiency. This development represents a significant advancement in optimizing external knowledge integration for LLMs, offering a more cost-effective and precise solution for complex query resolution. The paper, authored by Shu Wang and colleagues, was published on arXiv, contributing to the fields of Information Retrieval and Artificial Intelligence.
cs.AI updates on arXiv.org