I’m currently a MATS fellow working with SaferAI and a PhD student in applied math at Cornell advised by Lionel Levine and Ziv Goldfeld. Before that, I studied applied math at UCLA.
I’m interested in understanding and measuring AI values. I developed EigenBench to measure a model’s values through social consensus, and I’m building ValueArena to make these scores more legible to the general public. I’ve also studied how the methods that instill these values can sometimes induce side effects.
Lately, I’ve been thinking about how agent swarms become misaligned while collaborating on difficult tasks, and what makes oversight effective in these settings. Related to this, I’m curious how well agents can deliberate on a shared charter dictating their future societies.
I’m also interested in whether high-level, human-understandable causal models can explain the internal workings of LLMs—a question I explore in PLOT.
Outside research, I love playing piano and chamber music. I also enjoy playing racket sports.
If you’d like to chat or collaborate, please feel free to reach out!