In recent years, advancements in large language models (LLMs) have shown great potential in automating various software development tasks, including code completion, test generation, and bug fixing. However, due to the complexity and diversity of real-world software systems, the application of LLMs in automated bug fixing still faces many challenges. To address these issues, we propose MarsCode Agent, a new framework that leverages LLMs to automatically identify and fix errors in software code. MarsCode Agent combines the powerful capabilities of LLMs with advanced code analysis techniques to accurately locate faults and generate patches. Our approach follows a systematic process, including planning, bug reproduction, fault localization, candidate patch generation, and validation, to ensure high-quality bug fixes. Evaluations on SWE-bench and deployments in real-world enterprise code repositories demonstrate that MarsCode Agent achieves a high success rate in bug fixing compared to most existing automated methods.

I am a Senior Researcher at ByteDance (字节跳动) and a part-time postgraduate student mentor of School of Computer Science, Fudan University. I received my PhD degree from Laboratory for Foundations of Computer Science (LFCS), The University of Edinburgh under supervision of Dr. Ajitha Rajan.

At ByteDance, I am on the Software Engineering Lab. My research interest lies in the area of software testing, program repair and optimisations, and the synergy with machine learning and compiler techniques. I am also responsible for academic development and university collaboration.

I am passionate about building practical software testing, analysis, and debugging systems to predict, detect, diagnose, and fix bugs for all kinds of software systems.

Outside of work, I enjoy going to the gym.