I’m a 2nd year PhD student in Computer Science at Simon Fraser University, fortunate to be supervised by Dr. Linyi Li in the Trustworthy AI (TAI) Lab.

I am broadly interested in large language models. In particular, I am interested in model architecture (model depth, looped models). At the same time, I care about LLM security, especially the potential threats that frontier LLMs may pose to society.

🔥 News

• 2026.01: Completed PhD coursework and breadth exam requirements.
• 2025.06:  🎉🎉 Participated in a hackathon with developers from my team at SAP about AI model governance.

📝 Publications

AutoRed: Automated Attack Scenario Generation Framework for Red Teaming of LLMs
Zhe Wang, M. A. Tayebi
IEEE BigData 2024
We present AutoRed, a red-teaming framework that proactively generates adversarial prompts to challenge LLMs’ refusal strategies and safety alignment.

📖 Educations

• 2025.09 - Present, Ph.D., Computer Science at Simon Fraser University
• 2022.09 - 2024.05, M.Sc., Computer Science at Simon Fraser University
• 2018.09 - 2022.07, B.Sc., Mathematics and Applied Mathematics at The Chinese University of Hong Kong, Shenzhen

💬 Invited Talks

• 2025.06, The Art of LLM Evaluation, AI Lunch and Learn at SAP Canada, Vancouver office.

💻 Internships

• 2026.01 - 2026.09, Mitacs Research Intern, RBC Borealis, Canada.
  • Identified a state-supervision mismatch in looped models such as tiny recursive models (TRMs): with deep supervision and gradient detachment, early steps are trained on answers they cannot yet derive.
  • Designed adaptive supervision that trains each step only on what the model can already derive, improving Sudoku-Extreme accuracy by 7.9 points with up to 1.79× faster training.
  • Resulted in a manuscript submitted to ICLR 2027 for peer review.
• 2024.09 - 2025.08, Data Science Intern, SAP, Canada.
• 2023.07 - 2023.09, Machine Learning Research Intern, Hanglok-Tech, China.

🖋️Services

• Conference Reviewer: WWW, EMNLP