[00]
I'm a Computer Science undergraduate student at UC Santa Cruz researching LLM interpretability, explainability, and safety with Prof. Leilani Gilpin.
Update: I'll be joining the University of Cambridge as a PhD in Computer Science this fall!
[01]
My primary research focus is in mechanistic interpretability and safety. I'm particularly interested in applying interpretability techniques for real-world applications.
I began my research journey in 2023 in Prof. Gilpin's lab (AIEA), where I co-first authored a paper on LLM explainability that now has 100+ citations. In this past year, I collaborated with Yilun Zhou and Prof. Ziyu Yao on mechanistic interpretability research, resulting in a paper accepted to EMNLP 2025. I've also explored broader LLM applications through work with Prof. Ben Nye and Joel Walsh at the USC Institute for Creative Technologies in 2024, writing a paper focusing on fine-tuning vs. few-shot approaches for automated grading. I'm now working on monitoring LLM outputs and reasoning traces using mechanistic tools, and would love to talk more about my research--please don't hesitate to get in touch.
[02]
[02.1]
A mechanistic study of mental-math behavior in language models, arguing that the final token solution depends on information routed forward from earlier token positions.
[02.2]
An evaluation of model-written explanations that asks when self-explanations track the underlying computation and when they behave more like polished post-hoc justifications.
[02.3]
A comparison of fine-tuning and few-shot methods for AI-based short-answer grading.
[03]
coming soon :)
[04]
If you're interested in my work, would like to collaborate, or just want to chat, please feel free to reach out.