Research map / Image generation / Fast diffusion sampling
SA-Solver: Stochastic Adams Solver for Fast Sampling of Diffusion Models
Shuchen Xue, Mingyang Yi, Weijian Luo, Shifeng Zhang, Jiacheng Sun, Zhenguo Li, Zhi-Ming Ma
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 data-prediction instead of noise-prediction (ablation) | SA-Solver · ImageNet 256x256 LDM, NFE=20 | FID: 310.5 -> 3.88 | Improved |
| use data-prediction instead of noise-prediction (ablation) | SA-Solver · ImageNet 256x256 LDM, NFE=80 | FID: 3.41 -> 3.38 | Improved |
| predictor 3 steps instead of 1 step (no corrector) (ablation) | SA-Solver · CIFAR10, NFE 15 | FID: 13.76 -> 5.30 | Improved |
| predictor 3 steps instead of 1 step (no corrector) (ablation) | SA-Solver · CIFAR10, NFE 47 | FID: 14.67 -> 2.98 | Improved |
| add 1-step corrector to 1-step predictor (ablation) | SA-Solver · CIFAR10, NFE 15 | FID: 13.76 -> 8.49 | Improved |
| add 3-step corrector to 3-step predictor (ablation) | SA-Solver · CIFAR10, NFE 15 | FID: 5.30 -> 4.91 | Improved |
| add 3-step corrector to 3-step predictor (ablation) | SA-Solver · CIFAR10, NFE 47 | FID: 2.98 -> 2.92 | Improved |
| use smaller nonzero tau under few NFEs | SA-Solver · CIFAR10/ImageNet/LSUN, small NFE | FID | Improved |
| use larger tau under 20-100 NFEs | SA-Solver · CIFAR10/ImageNet/LSUN, 20-100 NFE | FID | Improved |
| SA-Solver with NFE=60 replaces DDPM with NFE=250 | DiT-XL/2-G · ImageNet 256x256 | FID: 2.27 -> 2.02 | Improved |
| SA-Solver with NFE=20 replaces Heun with NFE=50 | Min-SNR DiT-XL/2-G · ImageNet 256x256 | FID: 2.06 -> 1.93 | Improved |
| SA-Solver with NFE=60 replaces DDPM with NFE=250 | DiT-XL/2-G · ImageNet 512x512 | FID: 3.04 -> 2.80 | Improved |
Later papers that compared with this method
1 later accepted paper reports SA-Solver as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Efficient Integrators for Diffusion Generative Models (ICLR 2024 · better on 2 of 2 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
- A Unified Sampling Framework for Solver Searching of Diffusion Probabilistic Models (ICLR 2024 · similarity 0.92)
- S4S: Solving for a Fast Diffusion Model Solver (ICML 2025 · similarity 0.90)
- Adaptive Stochastic Coefficients for Accelerating Diffusion Sampling (NeurIPS 2025 · similarity 0.90)
- Diffusion Sampling Correction via Approximately 10 Parameters (ICML 2025 · similarity 0.89)
- Enhanced Diffusion Sampling via Extrapolation with Multiple ODE Solutions (ICLR 2025 · similarity 0.89)
- Distilling ODE Solvers of Diffusion Models into Smaller Steps (CVPR 2024 · 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.
