AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

Consistent Diffusion Models: Mitigating Sampling Drift by Learning to be Consistent

Giannis Daras, Yuval Dagan, Alex Dimakis, Constantinos Costis Daskalakis

NeurIPS 2023 · Image generation › Fast diffusion sampling · approach: Diffusion theory and design space · 6 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 consistency regularization to EDM-VP trainingEDM-VP · CIFAR-10 conditionalFID (best): 1.82 -> 1.77Improved
add consistency regularization to EDM-VP trainingEDM-VP · CIFAR-10 unconditionalFID (best): 2.01 -> 1.95Improved
add consistency regularization to EDM-VP trainingEDM-VP · AFHQ 64x64FID (best): 2.26 -> 2.21Improved
add consistency regularizationEDM-VP · AFHQ 64x64FID at 100k iterations: 2.61 -> 2.30Improved
add consistency regularizationEDM-VP · AFHQ 64x64FID at 70k iterations: 2.41 -> 2.44No clear effect
add consistency regularizationEDM-VP · CIFAR-10 conditionalFID at 150k iterations: 1.85 -> 1.88Hurt
finetune pretrained EDM-VP for 5k steps with consistency regularizationEDM-VP · FFHQ 64x64FID: 2.68 -> 2.61Improved
reduce batch size from 512 to 256 to fit memoryEDM-VP · AFHQ 64x64FID: 1.96 -> 2.29Hurt
add consistency regularization with DSM at all t (ablation)EDM · CIFAR-10 conditional (10k steps)FID: 5.81 -> 5.45Improved
drop DSM loss for 20% of t on clean side, keep consistency everywhere (ablation)CDM · CIFAR-10 conditional (10k steps)FID: 5.45 -> 6.59Hurt
stop sampling early at t_threshold instead of sampling through untrained times (ablation)CDM trained without DSM for some t · CIFAR-10 conditional (10k steps)FID: 6.59 -> 14.52Hurt
use 6 sampling steps in consistency lossCDM · training costtraining time: 1.5xHurt

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