Research map / Image generation / Fast diffusion sampling
WILD-Diffusion: A WDRO Inspired Training Method for Diffusion Models under Limited Data
Xianglu Wang, Wanlin Zhang, Hu Ding
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 WILD-Diffusion | EDM-DDPM++ · CIFAR-10, 20% data | FID: 13.91 -> 12.14 (-12.72%) | Improved |
| add WILD-Diffusion | EDM-DDPM++ · CIFAR-10, 100% data | FID: 1.97 -> 1.93 (-2.03%) | Improved |
| add WILD-Diffusion | EDM-NCSN++ · FFHQ 64x64, 20% data | FID: 9.38 -> 7.89 (-15.88%) | Improved |
| add WILD-Diffusion | EDM-NCSN++ · FFHQ 64x64, 100% data | FID: 2.57 -> 2.54 (-1.16%) | Improved |
| add WILD-Diffusion | Patch Diffusion · CIFAR-10, 20% data | FID: 12.53 -> 11.78 (-5.99%) | Improved |
| add WILD-Diffusion | DeepCache · CIFAR-10, 20% data | FID: 15.33 -> 13.96 (-8.94%) | Improved |
| add WILD-Diffusion | EDM-ADM · LSUN-Church 256x256, 50% data | FID: 5.79 -> 5.13 (-11.40%) | Improved |
| add WILD-Diffusion without pretraining | EDM-NCSN++ · 100-shot Obama | FID: 37.10 -> 34.52 | Improved |
| add WILD-Diffusion with pretraining | LD-Diffusion · 100-shot Obama | FID: 13.00 -> 12.54 | Improved |
| add WILD-Diffusion with pretraining | LD-Diffusion · AnimalFace Cat | FID: 12.77 -> 12.93 | No clear effect |
| increase interval m between WDRO sample updates (ablation) | WILD-Diffusion · FFHQ 50% data | FID | Hurt |
| increase number of inner steps K up to 5 (ablation) | WILD-Diffusion · FFHQ 50% data | FID | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- DeepCache — DeepCache: Accelerating Diffusion Models for Free (CVPR 2024) · better on 1 of 1 shared result
- Patch Diffusion — Patch Diffusion: Faster and More Data-Efficient Training of Diffusion Models (NeurIPS 2023) · better on 1 of 1 shared result
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
- EVODiff: Entropy-aware Variance Optimized Diffusion Inference (NeurIPS 2025 · similarity 0.89)
- Rethinking Peculiar Images by Diffusion Models: Revealing Local Minima’s Role (AAAI 2024 · similarity 0.89)
- Stochastic Forward–Backward Deconvolution: Training Diffusion Models with Finite Noisy Datasets (ICML 2025 · similarity 0.89)
- Assessing the quality of denoising diffusion models in Wasserstein distance: noisy score and optimal bounds (NeurIPS 2025 · similarity 0.89)
- Efficient Dataset Distillation via Minimax Diffusion (CVPR 2024 · similarity 0.89)
- DEFT: Efficient Fine-tuning of Diffusion Models by Learning the Generalised $h$-transform (NeurIPS 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.
