arXiv cs.AI· Rui Zhang, Yujia Su, Ruixuan Li, Yuhua Li, Yixin Su·· 11 小时前AI 评分38
VCoT-SQL：用双链式推理验证训练提升 Text-to-SQL 生成
VCoT-SQL: Enhancing Text-to-SQL Generation through Dual Chain-of-Thought Validation Training
AI 导读
VCoT-SQL 框架通过中间步骤验证的 CoT 监督和专用 SQL 验证器，让小型开源模型在 Text-to-SQL 任务上超越前沿闭源模型。该框架在 BIRD dev 上达到 66.82% 执行准确率、Spider test 上达 86.8%，使基础 7B 模型在 BIRD 上提升 7%。其 CoT 合成仅依赖 7B 模型，却优于使用 72B 级甚至闭源教师模型生成数据的方法。
正文 · 原文
Abstract:Open-source LLMs provide a flexible alternative to closed-source models for Text-to-SQL but still struggle with complex queries. Finetuning with chain-of-thought (CoT) data improves reasoning, yet standard datasets lack CoT annotations, requiring synthetic generation. Existing synthesis methods only verify the final answer, allowing flawed reasoning steps to corrupt the training data. We propose VCoT-SQL, a framework that improves Text-to-SQL generation through intermediate-step validated CoT supervision and a dedicated SQL validator, which enables small open-source models to outperform frontier closed-source models. VCoT-SQL trains two complementary models on curated reasoning traces: an SQL generator using clause-level CoT templates where reasoning steps are verified via execution and schema alignment checks, and an SQL validator using diagnostic CoT templates to detect syntactic, semantic, and type level errors. We further design a hierarchical schema filtering pipeline that integrates and refines complementary retrieval strategies to distill relevant database elements. VCoT-SQL achieves 66.82% execution accuracy on challenging dataset BIRD dev and 86.8% on Spider test, improving the base 7B model by 7% on BIRD. Notably, our CoT synthesis relies on 7B models yet outperforms methods that use 72B-scale or even proprietary closed source teachers for data generation, demonstrating that rigorous step-level validation can be more effective than scaling up teacher model size. Our dataset and code are available at this https URL.
| Comments: | Accepted to EMNLP 2026 (Main Conference) |
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2611.00815 [cs.AI] |
| (or arXiv:2611.00815v2 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2611.00815 arXiv-issued DOI via DataCite |
|
| Related DOI: | https://doi.org/10.18653/2026.emnlp-main.2686
DOI(s) linking to related resources |
Submission history
From: Rui Zhang [view email]
[v1]
Sat, 31 Oct 2026 10:55:52 UTC (1,657 KB)
[v2]
Wed, 4 Nov 2026 10:37:32 UTC (1,657 KB)
来源：arXiv cs.AI · arxiv.org