1 paper
Peiyu Yu, Suraj Kothawade, Sirui Xie +2
Most post-training methods for text-to-image samplers focus on model weights: either fine-tuning the backbone for alignment or distilling it for few-step efficiency. We take a diff…