T-Retriever: Tree-based Hierarchical Retrieval Augmented Generation for Textual Graphs
This addresses limitations in graph-based RAG for applications requiring hierarchical data, such as complex query answering, by improving retrieval coherence and relevance, though it appears incremental as it builds on existing RAG approaches.
The paper tackled the problem of hierarchical information management in graph-based Retrieval-Augmented Generation (RAG) by introducing T-Retriever, a framework that reformulates attributed graph retrieval as tree-based retrieval, resulting in significant outperformance over state-of-the-art RAG methods on diverse graph reasoning benchmarks.
Retrieval-Augmented Generation (RAG) has significantly enhanced Large Language Models' ability to access external knowledge, yet current graph-based RAG approaches face two critical limitations in managing hierarchical information: they impose rigid layer-specific compression quotas that damage local graph structures, and they prioritize topological structure while neglecting semantic content. We introduce T-Retriever, a novel framework that reformulates attributed graph retrieval as tree-based retrieval using a semantic and structure-guided encoding tree. Our approach features two key innovations: (1) Adaptive Compression Encoding, which replaces artificial compression quotas with a global optimization strategy that preserves the graph's natural hierarchical organization, and (2) Semantic-Structural Entropy ($S^2$-Entropy), which jointly optimizes for both structural cohesion and semantic consistency when creating hierarchical partitions. Experiments across diverse graph reasoning benchmarks demonstrate that T-Retriever significantly outperforms state-of-the-art RAG methods, providing more coherent and contextually relevant responses to complex queries.