Beyond Static Rules: Automated Discovery of Latent Vulnerabilities in Text-to-SQL

2026-07-04

AI Summary

This work addresses the latent reliability flaws of existing Text-to-SQL systems in real-world scenarios, which are difficult to uncover through current diagnostic methods reliant on static rules. To this end, we propose SAGE, a novel framework that enables the first automated and systematic exploration of model vulnerabilities in Text-to-SQL translation. SAGE leverages large language models to generate vulnerability hypotheses and designs targeted perturbations guided by a dynamically evolving Vulnerability Codex, iteratively validating and cataloging failure modes. Extensive experiments across multiple state-of-the-art open-source large language models reveal numerous previously unknown failure cases, exposing their structural fragility. Furthermore, lightweight fine-tuning using the generated adversarial samples substantially enhances model robustness and demonstrates the feasibility of cross-model transferable repairs.

Abstract

While Large Language Models (LLMs) have achieved remarkable success in Text-to-SQL tasks, their deployment in real-world environments is hindered by latent reliability issues. Identifying these latent weaknesses is critical for building trustworthy database interfaces, yet current diagnostic approaches rely heavily on static, expert-defined rules, which lack the capability for systematic and automated exploration. To bridge this gap, we propose SAGE (Systematic Automated Guided Exploration), a novel framework designed to autonomously uncover latent failure patterns in LLM-based Text-to-SQL generation. Specifically, SAGE generates vulnerability hypotheses for given samples and references a continuously evolving Vulnerability Codex to design targeted perturbations, thereby iteratively verifying and documenting potential defects. Extensive experiments on state-of-the-art open-source LLMs demonstrate that SAGE uncovers a substantial number of failure cases, highlighting the significant fragility of current models. Furthermore, our analysis reveals that the Vulnerability Codex exhibits strong cross-model transferability, indicating that the discovered patterns represent generalized structural weaknesses. Finally, we explore SAGE's potential for remediation. Although preliminary, lightweight fine-tuning on the generated samples yields promising improvements, suggesting a scalable pathway for closing the reliability loop in future work.

Problem

• Text-to-SQL
• latent vulnerabilities
• reliability
• automated discovery
• LLMs

Innovation

• Automated Vulnerability Discovery
• Text-to-SQL
• Large Language Models
• Systematic Perturbation
• Vulnerability Codex