Alexei Efros

3 papers UC Berkeley

• Fast Data Attribution for Text-to-Image Models
• Vision Transformers Don't Need Trained Registers
• Visual Jenga: Discovering Object Dependencies via Counterfactual Inpainting