LLMs as Continuous Learners: Improving the Reproduction of Defective Code in Software Issues

TL;DR

EvoCoder is a multi-agent continuous learning framework that enhances the reproduction of defective code in software issues by enabling LLMs to adapt to evolving errors through reflection and hierarchical experience management.

Contribution

It introduces EvoCoder, a novel framework with reflection and hierarchical experience pools, improving issue code reproduction and adapting to unique, evolving errors.

Findings

20% improvement in issue reproduction rates

Significant boost in overall issue-resolving pipeline accuracy

Effective adaptation to repository-specific and emerging errors

Abstract

Reproducing buggy code is the first and crucially important step in issue resolving, as it aids in identifying the underlying problems and validating that generated patches resolve the problem. While numerous approaches have been proposed for this task, they primarily address common, widespread errors and struggle to adapt to unique, evolving errors specific to individual code repositories. To fill this gap, we propose EvoCoder, a multi-agent continuous learning framework for issue code reproduction. EvoCoder adopts a reflection mechanism that allows the LLM to continuously learn from previously resolved problems and dynamically refine its strategies to new emerging challenges. To prevent experience bloating, EvoCoder introduces a novel hierarchical experience pool that enables the model to adaptively update common and repo-specific experiences. Our experimental results show a 20\% improvement in issue reproduction rates over existing SOTA methods. Furthermore, integrating our reproduction mechanism significantly boosts the overall accuracy of the existing issue-resolving pipeline.