Hi, my name is Chunlin Gong (巩春林). I am currently a PhD student at the University of Minnesota, in the Department of Computer Science and Engineering, advised by Prof. Mattia Fazzini and Prof. Caiwen Ding.
My primary research interests lie in trustworthy agentic systems—building AI agents that are safe, reliable, and aligned with human values. I study how to ensure the security and robustness of LLM-based agents in open-ended environments, and explore alignment methods to make these systems more trustworthy in real-world deployment.
I also had the privilege of interning at the Institute of Automation, Chinese Academy of Sciences (CASIA) and Shanghai AI Lab. I am deeply grateful for the guidance and support from Prof. Shu Wu, Prof. Xingcheng Xu, and Prof. Zhao Tong.
🔥 News
- 2025.12: 🎉I joined the Shanghai AI Lab to conduct research on safety alignment strategies in collaboration with CASIA.
- 2025.08: 🎉I joined the software engineering research group at the University of Minnesota, to study safetyissues related to logs.
- 2025.05: 🎉 I joined the Institute of Automation, Chinese Academy of Sciences (CASIA), to research content Safety in social media.
- 2024.9: 🏠Thanks to Prof. Wang, School of Control Science and Engineering, Shandong University. Our project has been approved by the Shandong Provincial Natural Science Foundation! This will be the starting point for my research.
📝 Publications (* Equal Contribution)
Zhao Tong*,Chunlin Gong*, Yiping Zhang, Haichao Shi, Qiang Liu, Xingcheng Xu, Shu Wu, Xiao-Yu Zhang
- We connect chain-of-thought safety, mechanistic interpretability, and targeted attention-head fine-tuning for AI safety in reasoning LLMs. Across five models, updating 0.64%-1.95% of parameters improves CoT safety by 67.1 percentage points on News and 55.0 on HarmBench on average.
- Paper · Preprint · Code · Project
- Research perspective: AI Safety and Mechanistic Interpretability of Chain-of-Thought Reasoning
Zhao Tong*, Chunlin Gong*, Yimeng Gu, Haichao Shi, Qiang Liu, Shu Wu, Xingcheng Xu, Xiao-Yu Zhang
- CogniDir strengthens comment-aware fake news detection with mechanism-specific malicious comments and adaptive adversarial training mixtures. InfoDirichlet Resampling updates exposure across attack groups using vulnerability feedback.
- Paper · Preprint · Code · Project
- Research perspective: Robust Fake News Detection under Adversarial Social Context
🎖 Honors and Awards
- 2025 University of Minnesota UROP Scholarship💴
- 2024 Shandong Provincial Natural Science Foundation💴, Funded (Sole student member).
-
2023 National Bronze Medal🥉, The International Collegiate Programming Contest (ACM-ICPC).
- 2022 Second Prize🥈, Shandong Division, National Olympiad in Informatics(NOIP).
📖 Educations
- 2026–Present, Ph.D. in Computer Science, University of Minnesota, Twin Cities.
- 2023–2026, B.A. in Computer Science, University of Minnesota, Twin Cities.
- 2020–2023, Shandong Experimental High School.
💻 Research Experience
- Collaborate with CASIA to analyze and develop defense strategies for multimodal large-scale safety issues.
- Research on software engineering security focuses on anomaly detection in log files.
- Content Safety in social media and LLM safety