About me

I am Yilun Kuang, a first-year PhD student in Data Science at NYU CDS & NYU CILVR Lab advised by Andrew Gordon Wilson. My research interests includes Large Language Models, Diffusion Models, Self-Supervised Learning, Multimodal Vision-Language Learning, Probabilistic Generative Models, NeuroAI & AI for Science, Generalization Theory, and Bayesian Deep Learning.

Prior to starting PhD, I graduated magna cum laude with high honors from NYU with a BA in Mathematics. I was fortunate to work with SueYeon Chung and Eero Simoncelli on manifold geometry/efficient coding inspired self-supervised learning at the Center for Computational Neuroscience of Flatiron Institute, Simons Foundation.

Outside of research, I enjoy playing ping pong, ultimate frisbee, basketball, and reading about philosophy, politics, and economics.

Curriculum Vitae

[CV]

Publications

Conference Papers

Token-Level Generalization Bounds for Large Language Models
Sanae Lotfi*, Yilun Kuang* (equal contribution), Marc Finzi*, Brandon Amos, Micah Goldblum, Andrew Gordon Wilson.
Under Review.

Non-Vacuous Generalization Bounds for Large Language Models
Sanae Lotfi*, Marc Finzi*, Yilun Kuang* (equal contribution), Tim G. J. Rudner, Micah Goldblum, Andrew Gordon Wilson.
International Conference on Machine Learning (ICML), 2024.

Learning Efficient Coding of Natural Images with Maximum Manifold Capacity Representations
Thomas Edward Yerxa, Yilun Kuang, Eero P Simoncelli, SueYeon Chung.
Neural Information Processing Systems (NeurIPS), 2023. [Poster]

Workshop Papers

Unsupervised Learning on Spontaneous Retinal Activity Leads to Efficient Neural Representation Geometry
Andrew Ligeralde, Yilun Kuang (equal contribution), Thomas Edward Yerxa, Miah N Pitcher, Marla Feller, SueYeon Chung.
NeurIPS 2023 Workshop: UniReps: Unifying Representations in Neural Models.

Non-Vacuous Generalization Bounds for Large Language Models
Sanae Lotfi*, Marc Finzi*, Yilun Kuang* (equal contribution), Tim G. J. Rudner, Micah Goldblum, Andrew Gordon Wilson.
NeurIPS 2023 Workshop: Self-Supervised Learning - Theory and Practice.

Learning a Visual Representation by Maximizing Manifold Capacity
Thomas Yerxa, Yilun Kuang, Eero Simoncelli, SueYeon Chung.
Computational and Systems Neuroscience (COSYNE), 2023. [Poster]