I lead the AISE Lab at the School of Software, Tsinghua University. My research interests primarily lie in AI Agent, Software Engineering for/with AI, and Trustworthy AI.My recent research focuses on improving the capabilities, efficiency, and trustworthiness of LLM-powered software systems, particularly AI agents.

📍Openings: I am recruiting Post-docs, PhD, Master's, and undergraduate students. I welcome self-motivated students with strong backgrounds in AI and/or Software Engineering. My group is well-funded and offers excellent research opportunities. Both on-site and remote research internships are available. If interested, please send me your CV via email.

Education & Experience

• 2025.12 - Present, Tsinghua University, Assistant Professor
• 2024.02 - 2025.11, Nanyang Technological University, Research Fellow (working with Prof. Yang Liu)
• 2021.09 - 2024.02, University College London, Research Fellow (working with Prof. Mark Harman and Federica Sarro)
• 2016.09 - 2021.07, Peking University, Ph.D. in Computer Software and Theory (advised by Prof. Hong Mei and Xuanzhe Liu, Outstanding Doctoral Dissertation Award of China Computer Federation)
• 2012.09 - 2016.07, Peking University, B.S. in Computer Science and B.A. in Economics

News

• October 9, 2026. Delighted to be included in the World's Top 2% Scientists list by Stanford University and Elsevier!
• October 8, 2026. How to make AI agents more reliable? Our recent work includes HarnessTester, which has uncovered over 100 bugs in real-world agentic systems, and Live-Harness-Bench, a live benchmark for evaluating agents' ability to repair real-world harness bugs.
• September 29, 2026. Can software engineering agents learn repair skills that transfer across repositories? Check out our latest work, "XRepoSkill: Learning Transferable Skills for Software Engineering Agents," in collaboration with Tencent!
• August 22, 2026. Honored to receive the KDD 2026 Best Reviewer Award!
• July 31, 2026. Curious why LLMs become less safe when deployed as agents? Check out our work, "Tool Specifications Matter: Uncovering and Mitigating Safety Risks in AI Agents."
• July 2, 2026. Check out our work SWE-Doctor on software issue resolution. SWE is a software issue resolution agent that guides patch generation with runtime diagnoses derived from multi-faceted bug reproduction test executions.
• June 30, 2026. Two papers about security of code LLMs are accepted by TSE.
• April 6, 2026. Our paper on experience-driven early termination for cost-efficient software engineering agents has been accepted by ACL 2026 as a Findings paper.
• March 25, 2026. Our paper on fairness testing of LLMs in role-playing scenarios has been accepted by FSE 2026.
• January 26, 2026. Our paper "Promptware Engineering: Software Engineering for Prompt-Enabled Systems" has been accepted by TOSEM.
• January 8, 2026. Our study on bugs in LLM inference engines has been accepted by TOSEM.
• December 22, 2025. Our paper "TransAgent: Enhancing LLM-Based Code Translation via Fine-Grained Execution Alignment" has been accepted by FSE 2026.
• December 3, 2025. I have joined Tsinghua University as a tenure-track Assistant Professor.
• November 8, 2025. Our paper "BAMAS: Structuring Budget-Aware Multi-Agent Systems" has been accepted to AAAI 2026 as an oral presentation.
• November 1, 2025. Our survey about agents for software engineering has been accepted by TOSEM.
• September 18, 2025. Our paper "Can Agent Fix Agent Issues?" has been accepted by NeurIPS 2025.