CodeR: Issue Resolving with Multi-Agent and Task Graphs
This addresses the challenge of automating code repository maintenance for developers and researchers, representing an incremental improvement over existing methods.
The paper tackles the problem of resolving GitHub issues, such as bug fixes and feature additions, using a multi-agent framework with pre-defined task graphs, achieving a 28.33% success rate on the SWE-bench lite benchmark with single submissions.
GitHub issue resolving recently has attracted significant attention from academia and industry. SWE-bench is proposed to measure the performance in resolving issues. In this paper, we propose CodeR, which adopts a multi-agent framework and pre-defined task graphs to Repair & Resolve reported bugs and add new features within code Repository. On SWE-bench lite, CodeR is able to solve 28.33% of issues, when submitting only once for each issue. We examine the performance impact of each design of CodeR and offer insights to advance this research direction.