Improves few-step text-to-image distillation by matching feature means and per-coordinate standard deviations. It retains DMD’s full score correction and uses the generator’s EMA copy for features, avoiding an external encoder.