Research map / Image generation / Fast diffusion sampling
STORK: Faster Diffusion and Flow Matching Sampling by Resolving both Stiffness and Structure-Dependence
Zheng Tan, Weizhen Wang, Andrea L. Bertozzi, Ernest K. Ryu
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 |
|---|---|---|---|
| add Taylor-approximated virtual NFEs to stabilized RK4 (ablation) | SRK4 · CIFAR-10 NFE 10 | FID: 443.812 -> 5.497 | Improved |
| use STORK-4 instead of classical RK4 (ablation) | RK4 · CIFAR-10 NFE 20 | FID: 33.662 -> 4.167 | Improved |
| use sub-steps s=9 instead of s=5 (ablation) | STORK-4 SANA-0.6B · MJHQ-30K NFE 9 | FID: 8.537 -> 6.526 | Improved |
| increase sub-steps from s=9 to s=104 (ablation) | STORK-4 SANA-0.6B · MJHQ-30K NFE 9 | FID: 6.526 -> 7.412 | Hurt |
| raise Taylor order from 1st to 2nd (ablation) | STORK-4 · MJHQ-30K NFE 7 | FID: 7.659 -> 54.179 | Hurt |
| raise Taylor order from 1st to 3rd (ablation) | STORK-4 · MJHQ-30K NFE 9 | FID: 6.526 -> 12.556 | Hurt |
| raise solver order from STORK-1 to STORK-4 (1st-order Taylor) (ablation) | STORK · MJHQ-30K NFE 9 | FID: 20.799 -> 6.526 | Improved |
| raise solver order from STORK-2 to STORK-4 (1st-order Taylor) (ablation) | STORK · MJHQ-30K NFE 9 | FID: 7.417 -> 6.526 | Improved |
| use STORK-4 instead of DPM-Solver++ | DPM-Solver++ · CIFAR-10 NFE 10 | FID: 6.471 -> 5.497 | Improved |
| use STORK-4 instead of UniPC | UniPC · LSUN-Bedroom NFE 10 | FID: 12.146 -> 9.594 | Improved |
| use STORK instead of Flow-UniPC | Flow-UniPC Hunyuan · EvalCrafter NFE 4 | final score: 206 -> 210 | Improved |
| use STORK instead of Flow-DPM-Solver++ | Flow-DPM-Solver++ Hunyuan · EvalCrafter NFE 8 | final score: 216 -> 216 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- UniPC — UniPC: A Unified Predictor-Corrector Framework for Fast Sampling of Diffusion Models (NeurIPS 2023) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- TADA: Improved Diffusion Sampling with Training-free Augmented DynAmics (NeurIPS 2025 · similarity 0.89)
- Accelerating Diffusion Sampling with Optimized Time Steps (CVPR 2024 · similarity 0.89)
- PFDiff: Training-Free Acceleration of Diffusion Models Combining Past and Future Scores (ICLR 2025 · similarity 0.89)
- Improving Progressive Generation with Decomposable Flow Matching (NeurIPS 2025 · similarity 0.89)
- Elucidating Flow Matching ODE Dynamics via Data Geometry and Denoisers (ICML 2025 · similarity 0.89)
- Shortcutting Pre-trained Flow Matching Diffusion Models is Almost Free Lunch (NeurIPS 2025 · similarity 0.88)
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.
