Research map / Image generation / Fast diffusion sampling
Restart Sampling for Improving Generative Processes
Yilun Xu, Mingyang Deng, Xiang Cheng, Yonglong Tian, Ziming Liu, Tommi S. Jaakkola
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 Restart instead of the previous best SDE sampler | VP diffusion sampler · CIFAR-10 | NFE at the same FID: 10-fold fewer steps | Improved |
| Use Restart instead of the previous best SDE sampler | EDM sampler · ImageNet 64×64 | NFE at the same FID: 2-fold fewer steps | Improved |
| Use Restart instead of Heun | CIFAR-10 generation · CIFAR-10, VP and stronger-model settings | FID | Improved |
| Use Improved SDE instead of the ODE baseline | EDM sampler · CIFAR-10 | FID | Hurt |
| Use Restart instead of DPM-solver | VP sampler · CIFAR-10, NFE 16–36 | FID: 2.11 FID at NFE 30 | Improved |
| Add repeated Restart cycles, increasing K (ablation) | Restart sampler · CIFAR-10 | FID | Mixed |
| Use a smaller turning-point K for the stronger EDM model (ablation) | Restart sampler · CIFAR-10 with EDM | FID | Mixed |
| Add multiple Restart intervals instead of using one level (ablation) | Restart sampler · ImageNet 64×64 | FID | Improved |
| Use Heun instead of DDIM in the Restart backward process (ablation) | Stable Diffusion Restart sampler · Text-to-image generation | sample quality | Improved |
Closest papers by meaning
- Adaptive Stochastic Coefficients for Accelerating Diffusion Sampling (NeurIPS 2025 · similarity 0.90)
- Distilling ODE Solvers of Diffusion Models into Smaller Steps (CVPR 2024 · similarity 0.89)
- On the Error Correcting Effects of Stochasticity in Discrete Diffusion (NeurIPS 2026 · similarity 0.88)
- CCS: Controllable and Constrained Sampling with Diffusion Models via Initial Noise Perturbation (NeurIPS 2025 · similarity 0.88)
- Fast ODE-based Sampling for Diffusion Models in Around 5 Steps (CVPR 2024 · similarity 0.88)
- On the Trajectory Regularity of ODE-based Diffusion Sampling (ICML 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.
