AI TrackerFrontier desk · AEST
Back to news
normalai

DMAD: Distribution Matching as Adversarial Distillation for Fast Visual Generation

arXiv cs.AI · 2 Oct 2026, 03:59 AEST

Distribution Matching Distillation (DMD) trains a few-step student from the difference between separately estimated target and student scores, so it must keep an auxiliary diffusion model fitted to the student's. We introduce DMAD, Distribution Matching as Adversarial Distillation, which recasts distribution matching as classification and learns the required log-density ratios directly.

Key facts

Read aloud

Distribution Matching Distillation DMD trains a few-step student from the difference between separately estimated target and student scores, so it must keep an auxiliary diffusion model fitted to the student's. We introduce DMAD, Distribution Matching as Adversarial Distillation, which recasts distribution matching as classification and learns the required log-density ratios directly. Distribution Matching Distillation DMD trains a few-step student from the difference between separately estimated target and student scores, so it must keep an auxiliary diffusion model fitted to the student's. We introduce DMAD, Distribution Matching as Adversarial Distillation, which recasts distribution matching as classification and learns the required log-density ratios directly. Two discriminator heads on a shared backbone distinguish real data and teacher samples from the student's, and linear losses on their logits train the student without auxiliary score fitting. We prove that at the discriminator optimum these losses recover the distribution-matching gradient underlying DMD, through the classical identity linking discriminator logits to log-density ratios. We further introduce gap-based reweighting, which adapts teacher supervision across noise levels from the real-data head's empirical logit gap between real and teacher samples.

Sources in this cluster

DMAD: Distribution Matching as Adversarial Distillation for Fast Visual Generation · AI Tracker