An Evaluation of Role-Based Multi-Agent Code Generation on Repository-Scale Problems
arXiv:2607.04212 · doi:10.1109/MS.2026.3701516
Abstract
Role-based multiagent code generation aims to make LLMs more effective on repository-scale problems, moving beyond small programming tasks. We evaluate this approach on 12 Java repositories, finding greater similarity to developer code than single LLMs, but a persistent gap from human implementations.
8 pages, 1 figure, 2 tables. Accepted for publication in the IEEE Software Special Issue on Engineering Agentic Systems