AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

Optimal Stopping in Latent Diffusion Models

Yu-Han Wu, Quentin Berthet, Gérard Biau, Claire Boyer, Romuald Elie, Pierre Marion

ICML 2026 · Image generation › Fast diffusion sampling · approach: Diffusion theory and design space · 0 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
stop the backward diffusion early instead of at t = Tlatent diffusion model · ImageNet-256 (32x32x4 latents)FID: minimum at t=0.956 rather than at the endImproved
use a larger latent dimensionlatent diffusion model · ImageNet-256optimal stopping timeMixed
use a noisy autoencoder as proxy for the full LDMLDM model selection · ImageNet-256FID curve crossing and optimal stopping time: LDM min-time 0.956, AE min-time 0.956Improved
use the exact score and very fine discretization on GMM data (ablation)latent diffusion · GMM dataFID trajectoryNo clear effect
project onto the data subspace dimension for low-rank datalatent diffusion · Gaussian with low-rank covarianceFrechet distanceImproved
use a larger latent dimension for linear AEs on MNISTLDM with linear AE · MNIST (latent 64, 256, 400)FIDMixed
cap score weight norm C on exponentially decaying spectrumclipped linear score · 10-dim Gaussianoptimal projection dimension: d_min in {5, 6} predicted and observedMixed
apply noisy AE proxy to pre-trained SD-VAE and FLUX-AEnoisy AE proxy · ImageNet-256FIDImproved

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