Problem
Research questions and friction points this paper is trying to address.
Multi-Table Question Answering
Retrieval-Augmented Generation
Complex Database Queries
Large Language Models
Structured Data Reasoning
DBRAG: Multi-Table Retrieval-Augmented Generation for Complex Database Queries
🤖 AI Summary
This study addresses the challenge of information fragmentation across multiple tables and the absence of automated retrieval and relational analysis in complex database queries. To this end, it proposes a Retrieval-Augmented Generation (RAG) framework tailored for multi-table question answering. The framework constructs compact contexts through offline table indexing and achieves precise multi-table retrieval by integrating large language model-based reranking with row-level summary augmentation. Furthermore, it incorporates program-aided reasoning to support full-table operations, thereby overcoming the limitations of conventional single-table analysis. Evaluations on benchmark datasets such as Spider demonstrate that the proposed approach significantly improves table retrieval accuracy and multi-table question answering performance, offering a novel paradigm for the efficient exploration of complex structured data.
📝 Abstract
Recent advancements in large language models have introduced new capabilities for reasoning over structured data, particularly through program-aided tools that can analyze tables. However, many existing methods address single-table scenarios or assume that the relevant tables are already provided. In practice, users often issue complex data exploration queries over entire databases, where relevant information may be distributed across multiple relations. In this work, we introduce DBRAG, a retrieval-augmented generation framework tailored for multi-table question answering. DBRAG first retrieves candidate tables using an offline table index, enriches their summaries with query-relevant rows, and uses an LLM to rerank the candidates. A program-aided reasoner then selects the required tables and executes operations over their full contents, keeping the initial prompt context compact. Experiments on the Spider, GeoQuery, and ATIS datasets used in this study demonstrate improvements in table retrieval and multi-table question answering.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Retrieval-Augmented Generation
Multi-Table Question Answering
Program-Aided Reasoning
LLM Reranking
Database Queries
🔎 Similar Papers
No similar papers found.
💼 Related Jobs
