Yik Siu Chan

profile_pic.jpeg

Hi! My name is Yik Siu, and I am a first-year PhD student in Computer Science at the University of Maryland, College Park. I am co-advised by Professors Sarah Wiegreffe and Hal Daume III. I am interested in the interpretability and training dynamics of AI systems, specifically language models.

I have been fortunate to collaborate with Yanai Elazar, Atticus Geiger, Marzyeh Ghassemi, and Professors Stephen Bach and Ellie Pavlick at Brown University, where I did my master’s. Before that, I did my bachelor’s in Computer Science and Economics at Wellesley College. I am grateful to the UWC Davis Scholarship and the Brown CS Department for supporting my studies.

papers

  1. NeurIPS 2026
    Pretraining Data Statistics Shape the Phases of Learning Entity Comparison in Language Models
    Yik Siu Chan, Jing Huang, Yanai Elazar, and Atticus Geiger
    NeurIPS, ICML Mechanistic Interpretability Workshop, 2026
  2. NeurIPS FoRLM
    Can We Predict Alignment Before Models Finish Thinking? Towards Monitoring Misaligned Reasoning Models
    Yik Siu Chan*, Zheng-Xin Yong*, and Stephen H. Bach
    NeurIPS Foundations of Reasoning in Language Models, 2025
    arXiv Code
  3. ICML 2025
    Speak Easy: Eliciting Harmful Jailbreaks from LLMs with Simple Interactions
    Yik Siu Chan*, Narutatsu Ri*, Yuxin Xiao*, and Marzyeh Ghassemi
    The Forty-Second International Conference on Machine Learning, 2025
    arXiv Code
  4. ICML 2025
    MIB: A Mechanistic Interpretability Benchmark
    Aaron Mueller*, Atticus Geiger*, Sarah Wiegreffe, Dana Arad, Iván Arcuschin, Adam Belfki, Yik Siu Chan, Jaden Fiotto-Kaufman, and 15 more authors
    The Forty-Second International Conference on Machine Learning, 2025
    arXiv Website