Physics-AI Fellow
Publications
Scaling laws and representation learning in simple hierarchical languages: Transformers versus convolutional architectures
– Physical Review E
(2025)
112,
065312
(doi: 10.1103/qtd6-nl8p)
MEMOIR: Lifelong Model Editing with Minimal Overwrite and Informed Retention for LLMs
– Advances in Neural Information Processing Systems
(2025)
The Physics of Data and Tasks: Theories of Locality and Compositionality in Deep Learning
(2025)
Probing the latent hierarchical structure of data via diffusion models
– Journal of Statistical Mechanics: Theory and Experiment
(2025)
2025,
084005
(doi: 10.1088/1742-5468/aded6c)
How compositional generalization and creativity improve as diffusion models are trained
– Proceedings of the 42nd International Conference on Machine Learning, PMLR 267
(2025)
Bigger Isn't Always Memorizing: Early Stopping Overparameterized Diffusion Models
(2025)
Lines: Post-training layer scaling prevents forgetting and enhances model merging
– International Conference on Learning Representations
(2025)
How Compositional Generalization and Creativity Improve as Diffusion Models are Trained
(2025)
A phase transition in diffusion models reveals the hierarchical nature of data
– Proceedings of the National Academy of Sciences of the United States of America
(2025)
122,
e2408799121
(doi: 10.1073/pnas.2408799121)
- 1 of 3