Automatic program repair (APR) aims at reducing the manual efforts required to identify and fix errors in source code. Before the rise of Large Language Model (LLM)-based agents, a common strategy was simply to increase the number of generated patches, sometimes to the thousands, which usually yielded better repair results on benchmarks. More recently, self-iterative capabilities enabled LLMs to refine patches over multiple rounds guided by feedback. However, literature often focuses on many iterations and disregards different numbers of outputs.

With software maintenance accounting for 50% of the cost of developing software, enhancing code quality and reliability has become more critical than ever. In response to this challenge, this doctoral research proposal aims to explore innovative solutions by focusing on the deployment of agents powered by Large Language Models (LLMs) to perform software maintenance tasks. The iterative nature of agents, which allows for continuous learning and adaptation, can help surpass common challenges in code generation. One distinct challenge is the last-mile problems, errors at the final stage of producing functionally and contextually relevant code. Furthermore, this project aims to surpass the inherent limitations of current LLMs in source code through a collaborative framework where agents can correct and learn from each other’s errors. We aim to use the iterative feedback in these systems to further fine-tune the LLMs underlying the agents, becoming better aligned to the task of automated software improvement. Our main goal is to achieve a leap forward in the field of automatic software improvement by developing new tools and frameworks that can enhance the efficiency and reliability of software development.