codefuse-ai/Awesome-Code-LLM

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

Closed this issue · 1 comments

yingweima2022 commented

Thank you for your great work on this survey! It provides a clear and comprehensive roadmap in the code LLM domain.

Our team has recently made some progress on code agent and issue reproduction. We propose a method called EvoCoder, a multi-agent continuous learning framework that focuses on improving the reproduction of defective code in software issues. This aligns with the broader goals discussed in your survey.

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

Preprint: arXiv:2411.13941

Geralt-Targaryen commented

Thanks for sharing your work! We have added it in section 3.3.