TableRAG: A Retrieval Augmented Generation Framework for Heterogeneous Document Reasoning
This addresses the challenge of multi-hop reasoning over mixed text and table data for question-answering systems, representing a domain-specific advancement.
The paper tackles the problem of applying Retrieval-Augmented Generation (RAG) to heterogeneous documents with text and tables, where existing methods disrupt tabular structure and hinder reasoning, and proposes TableRAG, an SQL-based framework that outperforms baselines on public datasets and a new benchmark, establishing a new state-of-the-art.
Retrieval-Augmented Generation (RAG) has demonstrated considerable effectiveness in open-domain question answering. However, when applied to heterogeneous documents, comprising both textual and tabular components, existing RAG approaches exhibit critical limitations. The prevailing practice of flattening tables and chunking strategies disrupts the intrinsic tabular structure, leads to information loss, and undermines the reasoning capabilities of LLMs in multi-hop, global queries. To address these challenges, we propose TableRAG, an SQL-based framework that unifies textual understanding and complex manipulations over tabular data. TableRAG iteratively operates in four steps: context-sensitive query decomposition, text retrieval, SQL programming and execution, and compositional intermediate answer generation. We also develop HeteQA, a novel benchmark designed to evaluate the multi-hop heterogeneous reasoning capabilities. Experimental results demonstrate that TableRAG consistently outperforms existing baselines on both public datasets and our HeteQA, establishing a new state-of-the-art for heterogeneous document question answering. We release TableRAG at https://github.com/yxh-y/TableRAG/tree/main.