I am currently an Assistant Professor in the Siebel School of Computing and Data Science at UIUC. I received my Ph.D degree in Computer Science from Peking University (advised by Prof. Lu Zhang and Prof. Dan Hao). I was a Postdoctoral Fellow at Purdue University working with Prof. Lin Tan.

My research interests lie primarily in Software Engineering and its intersection with Artificial Intelligence and Programming Languages, with a focus on LLM4Code, Agent&SE, Vulnerability Detection, and Software Testing&Debugging. More recently, my research has centered on two complementary directions: (1) Agent4SE: developing agentic solutions for tackling long-horizon software engineering tasks for complex software systems, and (2) SE4Agent: developing reliability assurance techniques for agentic systems.

News

• I will be attending ICSE 🇧🇷 from April 14–18. I’ll be at the LLM4Code workshop on April 14, the SE.next panel on April 17, and present our work on reducing false positives in bug detection on April 15. Feel free to stop by and chat if you’re interested in our work or any of these events!
• Check out our work TransAgent on code translation. TransAgent is a multi-agent system achieving stronger self-fixing capabilities via program execution alignment.
• I am co-chairing the third International Workshop on Large Language Models for Code (LLM4Code 2026) co-located with ICSE 2026 in Rio de Janeiro, Brazil. Looking forward to your submissions by 31 October!
• I am co-chairing AIware 2025 (co-located with ASE 2025 in Seoul, South Korea). Looking forward to your submissions!
• Check our our work BAMAS on constructing budget-aware multi-agent system.
• Check out our survey on agents for software engineering. [Preprint] [Github]
• I will join the Siebel School of Computing and Data Science at UIUC as an Assistant Professor in Spring 2026!
• Agents are emerging as a new software paradigm and automatically maintaining agent systems is challenging. Check out AgentIssue-Bench, the first reproducible benchmark of agent issue resolution tasks. We find that existing Software Engineering agents perform poorly in resolving agent issues. Check out the preprint "Can Agents Fix Agent Issues?"
• Our work (INFERROI) on enhancing traditional static analysis with LLMs in resource leak detection is released. INFERROI extends the knowledge boundary of traditional static analysis with the API specifications inferred by LLMs, which has detected previously-unknown resource leaks in open-source projects.
• We are organizing the second International Workshop on Large Language Models for Code (LLM4Code 2025) co-located with ICSE 2025. Looking forward to your high-quality submissions by Nov 18!
• I attended the Dagstuhl Seminar on Automated Programming and Program Repair and gave a talk on LLM-based agents for software engineering.
• I attended No. 176 Shonan Meeting on "Foundation Models and Software Engineering: Challenges and Opportunities".
• Check out our manually-crafted benchmark ClassEval for evaluating LLMs on class-level code generation. [ClassEval Leaderboard] [Benchmark Github] [Hugging Face]
• We are launching the first International Workshop on Large Language Models for Code (LLM4Code 2024) co-located with ICSE 2024. Looking forward to your submissions!
• Our work won ACM SIGSOFT Distinguished Paper Award at ESEC/FSE 2023.
• Check out our preprint for evaluating instruction-tuned LLMs on code comprehension tasks.
• Check out our preprint on LLM-based unit test generation.