FGTR: Fine-Grained Multi-Table Retrieval via Hierarchical LLM Reasoning

2026-03-13

AI Summary

Existing large language model–driven table retrieval approaches are largely confined to single-table queries, relying on coarse-grained encoding that struggles to support multi-table joins and suffers from limited accuracy and efficiency. This work proposes a human-like hierarchical reasoning mechanism: it first identifies schema elements relevant to the query, then performs fine-grained retrieval of corresponding cell contents, and finally reconstructs a concise subtable precisely aligned with the query intent. By departing from the conventional paradigm of holistic table encoding, this method achieves, for the first time, fine-grained retrieval across multiple tables. Evaluated on the Spider and BIRD benchmarks, it improves the F₂ score by 18% and 21%, respectively, substantially outperforming current state-of-the-art methods.

Technology Category

• Knowledge Representation and Reasoning: Common-Sense Reasoning
• Natural Language Processing: Information Extraction
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval

Application Category

• Search and Retrieval-Augmented AI: Large language models for search
• Graph Algorithms and Modeling for the Web: Querying, indexing, and retrieval in Web-related graphs
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

With the rapid advancement of large language models (LLMs), growing efforts have been made on LLM-based table retrieval. However, existing studies typically focus on single-table query, and implement it by similarity matching after encoding the entire table. These methods usually result in low accuracy due to their coarse-grained encoding which incorporates much query-irrelated data, and are also inefficient when dealing with large tables, failing to fully utilize the reasoning capabilities of LLM. Further, multi-table query is under-explored in retrieval tasks. To this end, we propose a hierarchical multi-table query method based on LLM: Fine-Grained Multi-Table Retrieval FGTR, a new retrieval paradigm that employs a human-like reasoning strategy. Through hierarchical reasoning, FGTR first identifies relevant schema elements and then retrieves the corresponding cell contents, ultimately constructing a concise and accurate sub-table that aligns with the given query. To comprehensively evaluate the performance of FGTR, we construct two new benchmark datasets based on Spider and BIRD . Experimental results show that FGTR outperforms previous state-of-the-art methods, improving the F_2 metric by 18% on Spider and 21% on BIRD, demonstrating its effectiveness in enhancing fine-grained retrieval and its potential to improve end-to-end performance on table-based downstream tasks.

Problem

• table retrieval
• multi-table query
• fine-grained retrieval
• large language models
• coarse-grained encoding

Innovation

• fine-grained retrieval
• multi-table query
• hierarchical reasoning
• LLM-based table retrieval
• schema-aware retrieval