BugPilot: Complex Bug Generation for Efficient Learning of SWE Skills

TL;DR

This paper presents BugPilot, a novel synthetic bug generation method that creates realistic, diverse bugs to improve training of SWE agents, leading to better performance with less data.

Contribution

Introducing BugPilot, a new bug generation approach that produces more realistic bugs, enhancing supervised learning efficiency for SWE models.

Findings

Generated bugs outperform existing datasets in training efficiency.

Models trained on BugPilot bugs achieve state-of-the-art results.

Synthetic bugs closely mimic human-authored code edits.

Abstract

High quality bugs are key to training the next generation of language model based software engineering (SWE) agents. We introduce a novel method for synthetic generation of difficult and diverse bugs. Our method instructs SWE Agents to introduce a feature into the codebase whereby they may unintentionally break tests, resulting in bugs. Prior approaches often induce an out-of-distribution effect by generating bugs intentionally (e.g. by introducing local perturbation to existing code), which does not reflect realistic development processes. We perform qualitative analysis to demonstrate that our approach for generating bugs more closely reflects the patterns found in human-authored edits. Through extensive experiments, we demonstrate that our bugs provide more efficient training data for supervised fine-tuning, outperforming other bug datasets by 2% with half the training data (1.2k vs. 3k bugs). We train on our newly generated bugs in addition to existing bug datasets to get FrogBoss a state-of-the-art 32B parameter model on SWE-bench Verified with a pass@1 of 54.6% and FrogMini a state-of-the-art 14B model on SWE-bench Verified with a pass@1 of 45.3% on SWE-bench Verified averaged over three seeds.