In this talk, we spotlight Trae Agent’s remarkable achievement of securing the top position on the SWE-bench Verified leaderboard with a 75.2% success rate. Trae Agent, an intelligent LLM-based assistant, has demonstrated exceptional capabilities in autonomously debugging complex issues, implementing robust fixes, and navigating intricate codebases. This session will explore the methodologies behind Trae Agent’s performance, including its innovative patch generation and selection strategies, and the integration of multiple LLMs. Additionally, we will discuss the significance of making Trae Agent open-source, fostering community collaboration, and accelerating the evolution of AI in software development.

I am a Principal Research Scientist at ByteDance (字节跳动). 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 lead the Trae Research team (ByteDance Software Engineering Lab), where we conduct research on AI agents for software engineering including the application and evaluation of AI agents, and training LLMs for agents. 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.