Zeming Wei

Zeming Wei

Peking University

I’m Zeming Wei (魏泽明), an undergraduate at School of Mathematical Sciences, Peking University. I was also a visiting student at UC Berkeley in Fall, 2023. My research focuses on the trustworthiness of AI, specifically focusing on mechanism interpretability, adversarial robustness, and generative safety.

If you are interested in collaborating (or just chatting) with me, feel free to email me.

🔥 News

  • 2025.05:  🎉 1 Paper (as first author) accepted by ICML 2025.
  • 2025.04:  🌟 Our new preprint on discussing Risks in LLM-based Agents is available online.
  • 2025.03:  🎖 I’m selected for the Somersault Cloud Talent Program (筋斗云人才计划) by ByteDance (internship track).
  • 2024.12:  🎉 2 Papers (as corresponding author; 1 as Oral) accepted by ICASSP 2025.
  • 2024.12:  🏅 I received the Academic Rising Star Award (Top 5 undergraduates university-wide, blog) by School of Computer Science, Peking University.
  • 2024.11:  🏅 I received the May 4th Scholarship (五四奖学金), which is the highest honor scholarship of Peking University (only 1 undergraduate awardee in School of Mathematical Sciences, Top 0.1%).
  • 2024.10:  ✨ My research grant (as Principal Investigator) is approved by Beijing Natural Science Foundation.
  • 2024.09:  🎉 1 Paper (as first author) accepted by SETTA 2024.
  • 2024.09:  🎉 3 Papers accepted by NeurIPS 2024.
  • 2024.07:  🎡 I attended ICML 2024 at Vienna and illustrated our poster.
  • 2024.05:  🎉 1 Paper (as corresponding author) accepted by ICML 2024.
  • 2023.12:  💯 I achieved a full GPA (4.0/4.0) during my study at UC Berkeley (with 1 A and 2 A+ grades).
  • 2023.10:  🔗 I serve as a fellow of Berkeley AI Safety Initiative for Students (BASIS).
  • 2023.09:  🏅 I received the Exceptional Award for Academic Innovation of Peking University (only 1 undergraduate awardee in School of Mathematical Sciences, Top 0.1%).
  • 2023.08:  🎉 1 Paper (as first author) accepted by Journal of Logical and Algebraic Methods in Programming.
  • 2023.06:  🍁 I attended CVPR 2023 at Vancouver and illustrated our poster.
  • 2023.05:  🥈 Won Second prize in Chinese Mathematics Competitions for Undergraduates (National final, Top 0.2%).
  • 2023.02:  🎉 1 Paper (as first author) accepted by CVPR 2023.
  • 2022.12:  🥇 Won First prize in Chinese Mathematics Competitions for Undergraduates (Beijing Division), and qualified for the finals.

📝 Selected Preprints

  • Jailbreak and Guard Aligned Language Models with Only Few In-Context Demonstrations
    Zeming Wei, Yifei Wang, Ang Li, Yichuan Mo, Yisen Wang
    Preprint (Cited 200+ times)
    [pdf]

  • Position: Agent-Specific Trustworthiness Risk as a Research Priority
    Zeming Wei, Tianlin Li, Xiaojun Jia, Yang Liu, Meng Sun
    Preprint
    [pdf]

📝 Selected Publications

(${}^{\boldsymbol\dagger}$: Corresponding Author; *: Equal Contribution)

  • Identifying and Understanding Cross-Class Features in Adversarial Training
    Zeming Wei, Yiwen Guo, Yisen Wang
    ICML 2025
    [pdf] [arxiv] [code]

  • Boosting Jailbreak Attack with Momentum
    Yihao Zhang*, Zeming Wei*${}^{\boldsymbol\dagger}$
    ICASSP 2025 (Oral)
    [pdf] [arxiv] [code]

  • On the Duality Between Sharpness-Aware Minimization and Adversarial Training
    Yihao Zhang*, Hangzhou He*, Jingyu Zhu*, Huanran Chen, Yifei Wang, Zeming Wei${}^{\boldsymbol\dagger}$
    ICML 2024
    [pdf] [arxiv] [code]

  • Weighted Automata Extraction and Explanation of Recurrent Neural Networks for Natural Language Tasks
    Zeming Wei, Xiyue Zhang, Yihao Zhang, Meng Sun
    Journal of Logical and Algebraic Methods in Programming
    [pdf] [arxiv] [code]

  • CFA: Class-wise Calibrated Fair Adversarial Training
    Zeming Wei, Yifei Wang, Yiwen Guo, Yisen Wang
    CVPR 2023
    [pdf] [arxiv] [code]

💻 Projects

  • Adversarial Safety Testing and Defense of AI Foundation Models
    Principal Investigator, Beijing Natural Science Foundation (Grant No. QY24035)
    2024.10 - 2026.09

🎖️ Talent Programs

  • Somersault Cloud Talent Program (筋斗云人才计划), internship track, ByteDance
  • Elite Ph.D. Program in Mathematics, Peking University
  • Elite Ph.D. Program in Applied Mathematics, Center for Machine Learning Research, Peking University

🏅 Honors and Awards

  • Excellent Graduate of Beijing Municipal (Top 5%), Beijing Municipal Education Commission, 2025
  • Excellent Graduate of Peking University, Peking University, 2025
  • May 4th Scholarship (五四奖学金, Top 0.1%), the highest honor scholarship of Peking University, 2024
  • Academic Rising Star Award (Top 5 undergraduates university-wide) [blog], School of Computer Science, Peking University, 2024
  • Merit Student (Top 10%), Peking University, 2024
  • Spotlight Award (Best Paper), 1st ICML Workshop on In-Context Learning, 2024
  • ICML Travel Award, 2024
  • Exceptional Award for Academic Innovation (Top 0.1%), Peking University, 2023
  • Merit Student (Top 10%), Peking University, 2023
  • Second prize, Chinese Mathematics Competitions for Undergraduates (National Final, Top 0.2%), 2023
  • First prize, Chinese Mathematics Competitions for Undergraduates (Beijing Division), 2022
  • Merit Student (Top 10%), Peking University, 2022
  • Award for Contribution in Student Organizations, Peking University, 2021

📖 Educations

  • 2023.08 - 2023.12, Visiting Student, University of California Berkeley
  • 2021.06 - 2025.06 (expected), Undergraduate Student, School of Mathematical Sciences, Peking University
  • 2020.09 - 2021.06, Undergraduate Student, College of Engineering, Peking University
  • 2017.09 - 2020.06, Senior High School Student, Beijing No.4 High School

💼 Academic Service

  • Reviewer (Conference): NeurIPS, ICLR, ICML, AISTATS, ECCV, CVPR, AAAI
  • Reviewer (Journal): TIFS, TMLR
  • Reviewer (Workshop): XAIA (@NeurIPS 2023), ICL (@ICML 2024), IAI (@NeurIPS 2024)
  • Fellow, Berkeley AI Safety Initiative for Students (BASIS), UC Berkeley

🔗 Links

(Alphabetical Order)