Song Wen

Implicit In-context Learning

Discrete Inversion: A Controllable Latent Space for Multinomial Diffusion and Masked Generative Models

Improving Compositional Text-to-image Generation with Large Vision-Language Models