Research map / Image generation / Fast diffusion sampling
Fast Sampling of Diffusion Models via Operator Learning
Hongkai Zheng, Weili Nie, Arash Vahdat, Kamyar Azizzadenesheli, Anima Anandkumar
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 DSNO instead of one-evaluation progressive distillation | CIFAR-10 diffusion sampling · CIFAR-10 | FID: 9.12 -> 3.78 | Improved |
| Use DSNO instead of one-evaluation progressive distillation | ImageNet-64 diffusion sampling · ImageNet-64 | FID: 15.99 -> 7.83 | Improved |
| Add temporal convolution blocks to the U-Net (ablation) | DSNO on CIFAR-10 · CIFAR-10 | FID: 8.09 -> 4.23; 7.85 -> 4.12 | Improved |
| Use SNR^0.5 loss weighting instead of uniform weighting (ablation) | DSNO on CIFAR-10 · CIFAR-10 | FID: 4.56 -> 4.21 | Improved |
| Use quadratic instead of uniform time discretization (ablation) | DSNO on CIFAR-10 · CIFAR-10 | FID: 4.33 -> 4.21 | Improved |
| Increase temporal resolution from 2 to 4 (ablation) | DSNO on CIFAR-10 · CIFAR-10 | FID: 5.01 -> 4.21 | Improved |
| Increase temporal resolution from 4 to 8 (ablation) | DSNO on CIFAR-10 · CIFAR-10 | FID: 4.21 -> 3.98 | Improved |
| Use LPIPS instead of L1 loss (ablation) | DSNO on CIFAR-10 · CIFAR-10 | FID: 4.12 -> 3.78 | Improved |
Closest papers by meaning
- A Simple Early Exiting Framework for Accelerated Sampling in Diffusion Models (ICML 2024 · similarity 0.90)
- Neural Diffusion Models (ICML 2024 · similarity 0.90)
- TADA: Improved Diffusion Sampling with Training-free Augmented DynAmics (NeurIPS 2025 · similarity 0.90)
- Faster Diffusion: Rethinking the Role of the Encoder for Diffusion Model Inference (NeurIPS 2024 · similarity 0.89)
- One Step Diffusion via Shortcut Models (ICLR 2025 · similarity 0.89)
- Accelerating Diffusion Models with Parallel Sampling: Inference at Sub-Linear Time Complexity (NeurIPS 2024 · similarity 0.89)
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.
