Rethinking Text-to-SQL: Dynamic Multi-turn SQL Interaction for Real-world Database Exploration

TL;DR

This paper introduces DySQL-Bench, a new benchmark for evaluating multi-turn, dynamic Text-to-SQL systems in realistic scenarios, highlighting current models' limitations in adapting to evolving user intents.

Contribution

The paper presents DySQL-Bench, an automated, multi-domain benchmark for real-world interactive Text-to-SQL tasks, and a multi-turn evaluation framework to assess model adaptability.

Findings

GPT-4o achieves only 58.34% accuracy on the benchmark.

The benchmark covers 13 domains with 1,072 tasks.

Current models struggle with dynamic, multi-turn interactions.

Abstract

Recent advances in Text-to-SQL have achieved strong results in static, single-turn tasks, where models generate SQL queries from natural language questions. However, these systems fall short in real-world interactive scenarios, where user intents evolve and queries must be refined over multiple turns. In applications such as finance and business analytics, users iteratively adjust query constraints or dimensions based on intermediate results. To evaluate such dynamic capabilities, we introduce DySQL-Bench, a benchmark assessing model performance under evolving user interactions. Unlike previous manually curated datasets, DySQL-Bench is built through an automated two-stage pipeline of task synthesis and verification. Structured tree representations derived from raw database tables guide LLM-based task generation, followed by interaction-oriented filtering and expert validation. Human evaluation confirms 100% correctness of the synthesized data. We further propose a multi-turn evaluation framework simulating realistic interactions among an LLM-simulated user, the model under test, and an executable database. The model must adapt its reasoning and SQL generation as user intents change. DySQL-Bench covers 13 domains across BIRD and Spider 2 databases, totaling 1,072 tasks. Even GPT-4o attains only 58.34% overall accuracy and 23.81% on the Pass@5 metric, underscoring the benchmark's difficulty. All code and data are released at https://github.com/Aurora-slz/Real-World-SQL-Bench .