Research

Paper

AI LLM March 24, 2026

Optimizing Small Language Models for NL2SQL via Chain-of-Thought Fine-Tuning

Authors

Anshul Solanki, Sanchit Latawa, Koushik Chakraborty, Navneet Kamboj

Abstract

Translating Natural Language to SQL (NL2SQL) remains a critical bottleneck for democratization of data in enterprises. Although Large Language Models (LLMs) like Gemini 2.5 and other LLMs have demonstrated impressive zero-shot capabilities, their high inference costs limit deployment at scale. This paper explores the efficacy of fine-tuning both large and small language models on NL2SQL tasks. Our research reveals a counter-intuitive scaling phenomenon. Fine-tuning large models (Gemini 2.5 Flash/Lite) on standard datasets yields negligible returns, often leading to overfitting on complex queries. Conversely, small models (Qwen) show significant gains. Fine-tuning improved the small model baseline from 36% to 45%, and further enriching the dataset with explicit Chain-of-Thought (CoT) reasoning surged accuracy to 54.5%(Fig 2). While this is still lower than the accuracy of large models like Gemini 2.5 , it does serve the business goal of significant cost reduction, latency in inference time and also meeting the business critical performance accuracy threshold.This paper demonstrates that transferring reasoning patterns enables compute-efficient smaller models to approach production-grade performance.

Metadata

arXiv ID: 2603.22942
Provider: ARXIV
Primary Category: cs.AI
Published: 2026-03-24
Fetched: 2026-03-25 06:02

Related papers

Raw Data (Debug)
{
  "raw_xml": "<entry>\n    <id>http://arxiv.org/abs/2603.22942v1</id>\n    <title>Optimizing Small Language Models for NL2SQL via Chain-of-Thought Fine-Tuning</title>\n    <updated>2026-03-24T08:39:06Z</updated>\n    <link href='https://arxiv.org/abs/2603.22942v1' rel='alternate' type='text/html'/>\n    <link href='https://arxiv.org/pdf/2603.22942v1' rel='related' title='pdf' type='application/pdf'/>\n    <summary>Translating Natural Language to SQL (NL2SQL) remains a critical bottleneck for democratization of data in enterprises. Although Large Language Models (LLMs) like Gemini 2.5 and other LLMs have demonstrated impressive zero-shot capabilities, their high inference costs limit deployment at scale. This paper explores the efficacy of fine-tuning both large and small language models on NL2SQL tasks.\n  Our research reveals a counter-intuitive scaling phenomenon. Fine-tuning large models (Gemini 2.5 Flash/Lite) on standard datasets yields negligible returns, often leading to overfitting on complex queries. Conversely, small models (Qwen) show significant gains. Fine-tuning improved the small model baseline from 36% to 45%, and further enriching the dataset with explicit Chain-of-Thought (CoT) reasoning surged accuracy to 54.5%(Fig 2). While this is still lower than the accuracy of large models like Gemini 2.5 , it does serve the business goal of significant cost reduction, latency in inference time and also meeting the business critical performance accuracy threshold.This paper demonstrates that transferring reasoning patterns enables compute-efficient smaller models to approach production-grade performance.</summary>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.AI'/>\n    <published>2026-03-24T08:39:06Z</published>\n    <arxiv:comment>9 pages , 3 fifures</arxiv:comment>\n    <arxiv:primary_category term='cs.AI'/>\n    <author>\n      <name>Anshul Solanki</name>\n    </author>\n    <author>\n      <name>Sanchit Latawa</name>\n    </author>\n    <author>\n      <name>Koushik Chakraborty</name>\n    </author>\n    <author>\n      <name>Navneet Kamboj</name>\n    </author>\n  </entry>"
}