Research map / Image generation / Fast diffusion sampling
Fast Sampling of Diffusion Models with Exponential Integrator
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 changed | Where | Measured | Effect |
|---|---|---|---|
| Switch Euler discretization to exponential integration while retaining the score parameterization (ablation) | Probability flow ODE sampler · CIFAR10 | FID | Hurt |
| Use εθ parameterization instead of the score sθ in the exponential integrator (ablation) | DEIS · CIFAR10 | FID | Helped |
| Add polynomial extrapolation of εθ instead of zero-order approximation (ablation) | DEIS · CIFAR10 | FID | Helped |
| Optimize the sampling timestamps (ablation) | DEIS · CIFAR10 | FID | Helped |
| Use higher-order ρ-coordinate Runge-Kutta methods at small NFE (ablation) | ρRK-DEIS · CIFAR10 | FID | Hurt |
| Increase polynomial extrapolation order (ablation) | DEIS · CIFAR10 | FID: 15.37 vs 16.31 | Helped |
| Use multistep tAB-DEIS rather than ρ-coordinate midpoint/RK samplers at low NFE | Diffusion sampler · Class-conditioned 64×64 ImageNet | FID: 6.65 vs 9.12 | Helped |
| Use deterministic ODE sampling rather than an SDE sampler with additional noise | Diffusion sampler · Diffusion model sampling | sample quality | Helped |
Closest papers by meaning
- On Accelerating Diffusion-Based Sampling Processes via Improved Integration Approximation (ICLR 2024 · similarity 0.93)
- Efficient Integrators for Diffusion Generative Models (ICLR 2024 · similarity 0.92)
- Distilling ODE Solvers of Diffusion Models into Smaller Steps (CVPR 2024 · similarity 0.91)
- Fast ODE-based Sampling for Diffusion Models in Around 5 Steps (CVPR 2024 · similarity 0.90)
- Enhanced Diffusion Sampling via Extrapolation with Multiple ODE Solutions (ICLR 2025 · similarity 0.90)
- A Unified Sampling Framework for Solver Searching of Diffusion Probabilistic Models (ICLR 2024 · similarity 0.90)
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.