1 paper
Feihong Yan, Shaoyu Liu, Haixuan Wang +4
Visual Autoregressive (VAR) models have emerged as a strong alternative to diffusion for image synthesis, yet their fixed training resolution prevents direct generation at higher r…