Search

Yash Sharma

Pretraining Frequency Predicts Compositional Generalization of CLIP on Real-World Tasks
Provably Learning Object-Centric Representations
Jacobian-based Causal Discovery with Nonlinear ICA
Self-Supervised Learning with Data Augmentations Provably Isolates Content from Style
Contrastive Learning Inverts the Data Generating Process
Towards Nonlinear Disentanglement in Natural Data with Temporal Sparse Coding
Unmasking the inductive biases of unsupervised object representations for video sequences

Published with Wowchemy — the free, open source website builder that empowers creators.