AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

Scale-wise Distillation of Diffusion Models

Nikita Starodubcev, Ilya Drobyshevskiy, Denis Kuznedelev, Artem Babenko, Dmitry Baranchuk

ICLR 2026 · Image generation › Fast diffusion sampling · approach: Training-free diffusion acceleration · 10 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
upsample x0 prediction then renoise (B) instead of noise then upscale (C) (ablation)SD3.5 noisy latents · COCO, 64->128, t=400FID-5K: 122 -> 30.9Improved
scale-wise sampling instead of full-resolution at 4 steps (ablation)SD3.5-M distilled · COCO 30KHPSv3: 11.3 -> 11.7Improved
scale-wise instead of full-resolution at 6 steps (ablation)SD3.5-M distilled · COCO 30KFID: 20.4 -> 23.1Hurt
scale-wise instead of full-resolution at 2 steps (ablation)SD3.5-M distilled · COCO 30KHPSv3: 10.8 -> 10.6Improved
scale-wise 4 steps vs full-scale 2 steps at matched time (ablation)SD3.5-M · COCO 30KHPSv3: 10.8 -> 11.7Improved
scale-wise vs full-scale sampling time (4 steps) (ablation)SD3.5-M · sampling latencysec/image: 0.29 -> 0.17Improved
scale-wise vs full-scale training iteration (full objective) (ablation)Wan2.1 · training timesec/iteration: 70.6 -> 23.9Improved
add temporal scaling to spatial-only SwD (ablation)SwD on Wan2.1-1.3B · MovieGenBenchlatency s/video: 2.1 -> 1.8Improved
add temporal scaling to spatial-only SwD (ablation)SwD on Wan2.1-1.3B · MovieGenBenchVBench2 Overall: 52.8 -> 53.2Improved
use MMD loss alone instead of full objective (ablation)SwD · training timesec/iteration (SD3.5-M scale-wise): 3.2 -> 0.4Improved
remove MMD loss from L_SwD (ablation)SwD SD3.5-M · MJHQautomatic metricsImproved
average feature tokens across batch (B) or extract features from clean samples (C) (ablation)L_MMD · SD3.5-M MJHQautomatic metricsHurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy