AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

Fewer Denoising Steps or Cheaper Per-Step Inference: Towards Compute-Optimal Diffusion Model Deployment

Zhenbang Du, Yonggan Fu, Lifu Wang, Jiayi Qian, Xiao Luo, Yingyan Celine Lin

ICCV 2025 · Image generation › Fast diffusion sampling · approach: Training-free diffusion acceleration · 3 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
add mixed-resolution denoising (low resolution in early steps) (ablation)SD V1.5, 20 steps · MS-COCO 2014 valFID: 18.42 -> 15.69Improved
add mixed-resolution denoising (ablation)PixArt-alpha, 20 steps · MS-COCO 2014 valFID: 29.16 -> 25.99Improved
add mixed-resolution denoising (ablation)SDXL, 20 steps · MS-COCO 2014 valFID: 14.10 -> 14.37No clear effect
add hybrid module caching on top of mixed-resolution (ablation)SD V1.5, 20 steps · MS-COCO 2014 valClip Score: 30.78 -> 30.25Hurt
add hybrid module caching on top of mixed-resolution (ablation)LCM, 8 steps · MS-COCO 2014 valFID: 21.01 -> 22.76Hurt
reduce denoising steps from 20 to 8SD V1.5 · MS-COCO 2014 valFID: 18.42 -> 20.60Hurt
reduce denoising steps from 20 to 8PixArt-alpha · MS-COCO 2014 valFID: 29.16 -> 33.09Hurt
abandon CFG entirely (ablation)SD V1.5 · MS-COCO 2014 valFID: 18.42 -> 31.98Hurt
cache conditional (Cond) cross-attention instead of CFG-combined map (ablation)DC+CA m=5 · MS-COCO 2014 valFID: 103.71 -> 18.82Improved
increase cache interval k from 2 to 3 (ablation)PostDiff · MS-COCO 2014 valFID: 16.65 -> 26.96Hurt
abandon CFG earlier (m=0.45 instead of 0.75) (ablation)PostDiff · MS-COCO 2014 valClip Score: 30.25 -> 28.82Hurt
raise resolution scale beta from 0.5 to 0.875 (ablation)mixed-resolution denoising · MS-COCO 2014 valFIDHurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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