Jia Li (李佳)
Portrait

I am the tenure-track Assistant Professor of College of AI at Tsinghua University. I obtained my PhD degree from Peking University in 2025, supervised by Prof. Zhi Jin and Prof. Ge Li.

Now, I lead the Tsinghua AI Agent Group and work closely with Prof. Shi-Min Hu. My research interests include trustworthy agents, safe agents, and software engineering.

I am looking for highly-motivated research interns, master students, PhD students and Postdoctoral researchers to join my group. Feel free to email me with your resume attached.

Google Scholar GitHub Twitter
News
2026
Our paper 'To See is Not to Master: Teaching LLMs to Use Private Libraries for Code Generation' has been accepted by TOSEM.
Oct 29
Our three papers are directly accepted by ICSE 2027 and one paper is under major revision.
Oct 21
Our paper Beyond Autoregression: An Empirical Study of Diffusion Large Language Models for Code Generation has been accepted by TOSEM.
Sep 10
Our paper DiffuTester has been accepted by EMNLP 2026 as a main conference paper.
Aug 21
Our paper Structured Chain-of-Thought Prompting for Code Generation is selected as the featured article in TOSEM. Only four papers are selected as the featured article in TOSEM.
Jul 31
Our two papers (LAVE and PSC) have been accepted by ISSTA 2026.
Jun 25
Our two papers (PackMonitor and Chameleon) have been directly accepted by ISSTA 2026.
Apr 16
Our two papers (VulAgent and EvoCoT) have been accepted by ACL 2026.
Apr 06
Our two papers (VulInstruct and EffiSkel) have been accepted by FSE 2026.
Mar 25
2025
Our paper (DAVSP) has been accepted by AAAI 2026.
Nov 08
Selected Publications (view all )
An Empirical Study of Proactive Coding Assistants in Real-World Software Development

Lehui Li*, Ruixuan Jia*, Guoye Yang, Jia Li# (* equal contribution, # corresponding author)

The 49th International Conference on Software Engineering (ICSE 2027), October 2026 Direct Acceptance (Top 5%)

[Paper]

An Empirical Study of Proactive Coding Assistants in Real-World Software Development

Lehui Li*, Ruixuan Jia*, Guoye Yang, Jia Li# (* equal contribution, # corresponding author)

The 49th International Conference on Software Engineering (ICSE 2027), October 2026 Direct Acceptance (Top 5%)

[Paper]

Grammar-Constrained Decoding Can Jailbreak LLMs into Generating Malicious Code

Yitong Zhang*, Shiteng Lu*, Jia Li# (* equal contribution, # corresponding author)

The 49th International Conference on Software Engineering (ICSE 2027), October 2026 Direct Acceptance (Top 5%)

[Paper] [Data]

Grammar-Constrained Decoding Can Jailbreak LLMs into Generating Malicious Code

Yitong Zhang*, Shiteng Lu*, Jia Li# (* equal contribution, # corresponding author)

The 49th International Conference on Software Engineering (ICSE 2027), October 2026 Direct Acceptance (Top 5%)

[Paper] [Data]

CoACT: Action-Preserving Observation Compression for Coding Agents

Haorui Chen, Yuancheng Zhu, Yitong Zhang, Jia Li# (# corresponding author)

The 49th International Conference on Software Engineering (ICSE 2027), October 2026 Direct Acceptance (Top 5%)

[Paper] [Code]

CoACT: Action-Preserving Observation Compression for Coding Agents

Haorui Chen, Yuancheng Zhu, Yitong Zhang, Jia Li# (# corresponding author)

The 49th International Conference on Software Engineering (ICSE 2027), October 2026 Direct Acceptance (Top 5%)

[Paper] [Code]

Environmental Injection Attacks against GUI Agents in Realistic Dynamic Environments

Yitong Zhang, Ximo Li, Liyi Cai, Jia Li# (# corresponding author)

The 35th SIGSOFT International Symposium on Software Testing and Analysis (ISSTA 2026), Just Accepted, April 2026 Direct Acceptance (10%)

[Paper] [Code]

Environmental Injection Attacks against GUI Agents in Realistic Dynamic Environments

Yitong Zhang, Ximo Li, Liyi Cai, Jia Li# (# corresponding author)

The 35th SIGSOFT International Symposium on Software Testing and Analysis (ISSTA 2026), Just Accepted, April 2026 Direct Acceptance (10%)

[Paper] [Code]

DAVSP: Safety Alignment for Large Vision-Language Models via Deep Aligned Visual Safety Prompt

Yitong Zhang, Jia Li#, Liyi Cai, Ge Li (# corresponding author)

The 40th Annual AAAI Conference on Artificial Intelligence (AAAI 2026), 40(44), 38111-38119, 2026 Oral (Top 3%)

[Paper] [Code]

DAVSP: Safety Alignment for Large Vision-Language Models via Deep Aligned Visual Safety Prompt

Yitong Zhang, Jia Li#, Liyi Cai, Ge Li (# corresponding author)

The 40th Annual AAAI Conference on Artificial Intelligence (AAAI 2026), 40(44), 38111-38119, 2026 Oral (Top 3%)

[Paper] [Code]

Structured Chain-of-Thought Prompting for Code Generation

Jia Li, Ge Li, Yongmin Li, Zhi Jin

ACM Transactions on Software Engineering and Methodology (TOSEM), Volume 34, Issue 2, Pages 1-23, 21 January 2025 Featured Article

[Paper]

Structured Chain-of-Thought Prompting for Code Generation

Jia Li, Ge Li, Yongmin Li, Zhi Jin

ACM Transactions on Software Engineering and Methodology (TOSEM), Volume 34, Issue 2, Pages 1-23, 21 January 2025 Featured Article

[Paper]

All publications