Research map / Image generation / Fast diffusion sampling
SEEDS: Exponential SDE Solvers for Fast High-Quality Sampling from Diffusion Models
Martin Gonzalez, Nelson Fernandez, Thuy Vinh Dinh Tran, Elies Gherbi, Hatem Hajri, Nader Masmoudi
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 changed | Where | Measured | Effect |
|---|---|---|---|
| Use SEEDS-3 instead of DDPM | CIFAR-10 discrete VP unconditional sampling · CIFAR-10 discrete VP unconditional | FID: 3.16 -> 3.08 | Improved |
| Use SEEDS-3 instead of EDM (S_churn=0) | CIFAR-10 continuous VP conditional sampling · CIFAR-10 continuous VP conditional | FID: 2.48 -> 2.08 | Improved |
| Use SEEDS-3 instead of optimized EDM | CIFAR-10 continuous VP unconditional sampling · CIFAR-10 continuous VP unconditional | FID: 2.27 -> 2.39 | Hurt |
| Use SEEDS-3 instead of F-PNDM (linear) | CelebA-64 discrete VP unconditional sampling · CelebA-64 discrete VP unconditional | FID: 2.71 -> 1.88 | Improved |
| Use SEEDS-3 instead of EDM (S_churn=0) | FFHQ-64 continuous VP unconditional sampling · FFHQ-64 continuous VP unconditional | FID: 3.39 -> 3.40 | Hurt |
| Use SEEDS-3 instead of optimized EDM | ImageNet-64 EDM conditional sampling · ImageNet-64 EDM conditional | FID: 1.36 -> 1.38 | Hurt |
| Use SEEDS-3 instead of optimized EDM | ImageNet-64 EDM conditional sampling · ImageNet-64 EDM conditional | NFE: 511 -> 270 | Improved |
| Use noise prediction rather than data prediction | SEEDS · Reported image-generation experiments | FID | Improved |
| Use Markov-preserving noise combinations rather than naive noise combinations (ablation) | SEEDS-2 and SEEDS-3 · CIFAR-10 VP conditional | FID | Improved |
Closest papers by meaning
- Adaptive Stochastic Coefficients for Accelerating Diffusion Sampling (NeurIPS 2025 · similarity 0.92)
- PFDiff: Training-Free Acceleration of Diffusion Models Combining Past and Future Scores (ICLR 2025 · similarity 0.91)
- TADA: Improved Diffusion Sampling with Training-free Augmented DynAmics (NeurIPS 2025 · similarity 0.90)
- A Unified Sampling Framework for Solver Searching of Diffusion Probabilistic Models (ICLR 2024 · similarity 0.90)
- Distilling ODE Solvers of Diffusion Models into Smaller Steps (CVPR 2024 · similarity 0.90)
- Fast Solvers for Discrete Diffusion Models: Theory and Applications of High-Order Algorithms (NeurIPS 2025 · 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.
