SHARE: An SLM-based Hierarchical Action CorREction Assistant for Text-to-SQL

2025-05-31

AI Summary

Existing text-to-SQL self-correction methods face two key bottlenecks: excessive computational overhead from recursive LLM invocations and difficulty in performing fine-grained, interpretable error localization and correction on declarative SQL. This paper proposes a hierarchical action-based error correction framework powered by small language models (SLMs), which decomposes SQL generation into explicit, stepwise action trajectories and introduces a two-stage progressive correction mechanism. Innovatively, it employs a cascaded three-tier SLM pipeline coupled with a hierarchical self-evolutionary training strategy, enabling explicit modeling of reasoning paths without reliance on large-scale annotated data. Experiments demonstrate substantial improvements in SQL correction accuracy and robustness across diverse LLMs, particularly excelling in low-resource and privacy-sensitive settings where strong generalization capability is retained.

Technology Category

• Natural Language Processing: Code Generation / Program Synthesis from Natural Language
• Search and Optimization: Learning to Search
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Search and Retrieval-Augmented AI: Large language models for search
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Current self-correction approaches in text-to-SQL face two critical limitations: 1) Conventional self-correction methods rely on recursive self-calls of LLMs, resulting in multiplicative computational overhead, and 2) LLMs struggle to implement effective error detection and correction for declarative SQL queries, as they fail to demonstrate the underlying reasoning path. In this work, we propose SHARE, an SLM-based Hierarchical Action corREction assistant that enables LLMs to perform more precise error localization and efficient correction. SHARE orchestrates three specialized Small Language Models (SLMs) in a sequential pipeline, where it first transforms declarative SQL queries into stepwise action trajectories that reveal underlying reasoning, followed by a two-phase granular refinement. We further propose a novel hierarchical self-evolution strategy for data-efficient training. Experimental results demonstrate that SHARE effectively enhances self-correction capabilities while proving robust across various LLMs. Furthermore, our comprehensive analysis shows that SHARE maintains strong performance even in low-resource training settings, which is particularly valuable for text-to-SQL applications with data privacy constraints.

Problem

• Reduces computational overhead in text-to-SQL self-correction
• Improves error detection and correction for declarative SQL queries
• Enhances performance in low-resource and data privacy settings

Innovation

• Uses SLM-based hierarchical action correction
• Transforms SQL queries into stepwise actions
• Implements data-efficient hierarchical self-evolution training