DMD2: regression-free distribution matching with critic and GAN updates

What changed

Removed paired-trajectory regression from DMD, stabilized synthetic-score learning with two-timescale updates, added real-data adversarial loss and simulated multi-step inference inputs during training.

Scientific scope

Removed paired-trajectory regression from DMD, stabilized synthetic-score learning with two-timescale updates, added real-data adversarial loss and simulated multi-step inference inputs during training. Inherits DMD, GANs and score matching. Removing paired teacher trajectories is not removing real training data, the teacher or critic costs. Source-specific improved FID does not guarantee a student dominates its teacher for every metric.

Evidence

The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. Inherits DMD, GANs and score matching. Removing paired teacher trajectories is not removing real training data, the teacher or critic costs. Source-specific improved FID does not guarantee a student dominates its teacher for every metric.

Contributors

• Source-listed authors and technical contributors · SCIENCE: Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.

• Improved Distribution Matching Distillation for Fast Image Synthesis

Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.

Full scientific record · Browse contribution grades · Request a correction