Beyond Chunks and Graphs: Retrieval-Augmented Generation through Triplet-Driven Thinking

2025-08-04

AI Summary

Existing RAG approaches face a trade-off between performance and efficiency: multi-hop RAG achieves strong reasoning capability but incurs high LLM invocation overhead, while graph-based RAG suffers from expensive computation, error-prone graph construction, and redundant retrieval. To address these limitations, we propose T²RAG—a novel framework that abandons conventional text chunking and explicit graph structures. Instead, it introduces a lightweight, atomic triplet-based knowledge base. Given a query, an LLM automatically decomposes it into retrievable placeholder-augmented triplets; T²RAG then iteratively performs triplet matching and evidence aggregation to enable end-to-end retrieval-augmented reasoning. This triplet-driven mechanism eliminates complex graph construction and substantially reduces retrieval redundancy. Evaluated on six benchmark datasets, T²RAG achieves an average 11% accuracy improvement and a 45% reduction in retrieval token cost, outperforming state-of-the-art multi-hop and graph-based RAG methods in overall effectiveness and efficiency.

Technology Category

• Knowledge Representation and Reasoning: Computational Complexity of Reasoning
• Data Mining & Knowledge Management: Representing, Reasoning, and Using Provenance, Trust
• Natural Language Processing: Question Answering

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: Foundation models and LLMs for Web-related graphs

Abstract

Retrieval-augmented generation (RAG) is critical for reducing hallucinations and incorporating external knowledge into Large Language Models (LLMs). However, advanced RAG systems face a trade-off between performance and efficiency. Multi-round RAG approaches achieve strong reasoning but incur excessive LLM calls and token costs, while Graph RAG methods suffer from computationally expensive, error-prone graph construction and retrieval redundancy. To address these challenges, we propose T$^2$RAG, a novel framework that operates on a simple, graph-free knowledge base of atomic triplets. T$^2$RAG leverages an LLM to decompose questions into searchable triplets with placeholders, which it then iteratively resolves by retrieving evidence from the triplet database. Empirical results show that T$^2$RAG significantly outperforms state-of-the-art multi-round and Graph RAG methods, achieving an average performance gain of up to 11% across six datasets while reducing retrieval costs by up to 45%. Our code is available at https://github.com/rockcor/T2RAG

Problem

• Balancing performance and efficiency in RAG systems
• Reducing LLM calls and token costs in multi-round RAG
• Avoiding complex graph construction in Graph RAG methods

Innovation

• Uses triplet-driven knowledge base
• Decomposes questions into searchable triplets
• Iteratively resolves triplets with evidence