Published paper 'Updating CLIP to Prefer Descriptions Over Captions' at EMNLP 2024 conference; co-authored 'Causal Abstraction: A Theoretical Foundation for Mechanistic Interpretability'.
Research Experience
Worked as a Data Scientist at Microsoft; conducting research at Stanford University on language models.
Education
PhD student at Stanford University, advised by Dr. Christopher Potts.
Background
First-year PhD student at Stanford NLP, excited about understanding how AI models learn and process language. Research focuses on causally-motivated explanation and evaluation of language models, currently exploring how language models represent and resolve uncertainty.