PDC & DM-SFT: A Road for LLM SQL Bug-Fix Enhancing

TL;DR

This paper presents PDC and DM-SFT, two novel methods to significantly improve large language models' ability to fix SQL bugs, outperforming larger existing models.

Contribution

Introduction of PDC and DM-SFT, innovative data construction and fine-tuning techniques that enhance SQL bug-fixing capabilities of code LLMs.

Findings

Models with proposed methods outperform larger existing models.

PDC effectively expands training data for bug fixing.

DM-SFT reduces training steps and improves bug-fixing accuracy.

Abstract

Code Large Language Models (Code LLMs), such as Code llama and DeepSeek-Coder, have demonstrated exceptional performance in the code generation tasks. However, most existing models focus on the abilities of generating correct code, but often struggle with bug repair. We introduce a suit of methods to enhance LLM's SQL bug-fixing abilities. The methods are mainly consisted of two parts: A Progressive Dataset Construction (PDC) from scratch and Dynamic Mask Supervised Fine-tuning (DM-SFT). PDC proposes two data expansion methods from the perspectives of breadth first and depth first respectively. DM-SFT introduces an efficient bug-fixing supervised learning approach, which effectively reduce the total training steps and mitigate the "disorientation" in SQL code bug-fixing training. In our evaluation, the code LLM models trained with two methods have exceeds all current best performing model which size is much larger.