AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

LogCD: Local-to-global Consistency Distillation for Few-step Image Generation

Qingsong Xie, Zhenyi Liao, Chen Chen, Zhijie Deng, Haonan Lu

CVPR 2026 · Image generation › Fast diffusion sampling · approach: Distillation for few-step generation

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
replace MSE with latent LPIPS loss in local CD (ablation)LogCD (SDXL) · MSCOCO-2017 5KImageReward: 0.52 -> 0.62Improved
add global consistency distillation after local CD (ablation)LogCD (SDXL) · MSCOCO-2017 5KImageReward: 0.62 -> 0.76Improved
add GAN loss to global CD (ablation)LogCD (SDXL) · MSCOCO-2017 5KImageReward: 0.76 -> 0.82Improved
add DMD loss to global CD (ablation)LogCD (SDXL) · MSCOCO-2017 5KImageReward: 0.82 -> 0.95Improved
increase local interval number M from 4 to 8 (ablation)LogCD (SDXL) · MSCOCO-2017 5KImageReward: 0.89 -> 0.95Improved
increase local interval number M beyond 8 (ablation)LogCD (SDXL) · MSCOCO-2017 5KImageReward: 0.95 -> 0.98No clear effect
use multi-step teacher solver (p=3) in global CD instead of one step (ablation)LogCD (SDXL) · MSCOCO-2017 5KFID: 27.1 -> 25.9Improved
increase student denoising steps q from 2 to 3 (ablation)LogCD (SDXL) · MSCOCO-2017 5KImageReward: 0.88 -> 0.94Improved
increase student denoising steps q beyond 3 (ablation)LogCD (SDXL) · MSCOCO-2017 5KCLIP score: 33.5 -> 33.6No clear effect
increase LogCM sampling steps from 2 to 4LogCM (SDXL) · MSCOCO-2017 5KImageReward: 0.93 -> 0.95Improved
distill FLUX.1-dev with LogCD instead of LCM (4 steps)FLUX.1-dev · MSCOCO-2017 5KCLIP score: 31.6 -> 32.6Improved
LogCM 4-step versus 25-step teacher on FLUX.1-devFLUX.1-dev · MSCOCO-2017 5KCLIP score: 32.6 -> 32.6No clear effect

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