AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

Skipping the Zeros in Diffusion Models for Sparse Data Generation

Phil Ostheimer, Mayank Nagda, Andriy Balinskyy, Gabriel Vicente Rodrigues, Jean Radig, Carl Herrmann, Stephan Mandt, Marius Kloft, Sophie Fellenz

ICML 2026 · Image generation › Fast diffusion sampling · approach: Fast diffusion ODE solvers · 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
use SED instead of dense DDPM for sparse calorimeter imagesSED · Muon SignalWasserstein PT: 220.32 -> 16.31Improved
use SED instead of domain-specific SARMSED · Muon BackgroundWasserstein PT: 12.61 -> 9.63Improved
use SED instead of latent DM with same parameter countSED · Muon SignalWasserstein PT: 28.20 -> 16.31Improved
add post-hoc thresholding to dense DDPM (ablation)DDPM · Muon SignalWasserstein PT: 220.32 -> 24.22Improved
add post-hoc thresholding to LDM (ablation)LDM · Muon SignalWasserstein PT: 28.20 -> 43.66Hurt
add post-hoc thresholding to LDM (ablation)LDM · Tabula MurisSCC: 0.87 -> 0.26Hurt
use SED instead of scDiffusionSED · Tabula MurisMMD: 1.53 -> 0.54Improved
use SED instead of dense LDM on scRNASED · Tabula MurisSCC: 0.87 -> 0.76 SCC; 5.82 -> 0.54 MMDMixed
skip zeros so compute scales with non-zeros as dimensionality growsSED · scRNA with added zero genesrelative FLOPsImproved
use SED instead of DDPM sampling on highly sparse dataSED · Muon Signalsampling time (ms per batch): 453.1 -> 23.5Improved
use SED on less sparse Fashion-MNISTSED · Fashion-MNIST (50% sparse)sampling time (ms per batch): LDM 77.7 vs SED 354.6Hurt
mean pooling vs special [CLS]-like token in encoder (ablation)SAVAE encoder · latent representationperformanceNo 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