Autonomous Program Improvement

Autonomous program improvement aims to automate software maintenance and evolution tasks, such as bug fixing and feature addition, without direct human intervention. Current research focuses on leveraging large language models (LLMs) combined with program analysis techniques, often within agent-based frameworks, to infer program intent and generate effective code modifications. This approach shows promise for reducing software development costs and improving efficiency, particularly in addressing real-world issues like those found on platforms like GitHub. The development of robust quality assessment methods, especially for automatically generated code and virtualized data, is also a critical area of ongoing investigation.