About

I’m a master’s student at Princeton Department of Computer Science, advised by Prof. Sanjeev Arora. I’m part of the Princeton Language and Intelligence Group. I obtained my Bachelor’s degree at Computer Science from Princeton University and had the fortune to work with Prof. Danqi Chen and Prof. Kai Li.

I work on understanding and developing recipes for language model behavior at scale. Model behavior includes instruction following, skill association, self-correction, to name a fiew. Unlike factual knowledge, these cognitive behaviors are somewhat implicit in the pretraining data and should be studied and improved in isolation. I’m particularly excited to seeing when and where does behavior learning occur. These days, I’m working on synthetic data as a potential approach to answer some of these questions.

News

• [Oct. 2024] Our paper about skill-mix data generation is accepted to NeurIPS MATH-AI 24!
• [Sept. 2024] Started teaching COS226: Algorithms and Data Structures.
• [Sept. 2024] Graduated with Distinguished Service award for AI@Princeton Club.
• [Oct. 2023] Our paper on meta-evaluation of instruction-following is accepted to ICLR2024!
• [March 2023] We made the very first AI TigerTrek happen! Flying to SF to meet amazing researchers :)

Publications