arXivAI Research
DMAD: Recasting Distribution Matching as Adversarial Distillation for Fast Visual Generation
DMAD：將分佈匹配重塑為對抗蒸餾，實現超快速視覺生成
DMAD simplifies Distribution Matching Distillation by recasting it as a classification task, eliminating costly auxiliary models and achieving state-of-the-art few-step visual generation.
2 min read