Research map / Image generation / Fast diffusion sampling
Diffusion Models as Dataset Distillation Priors
Duo Su, Huyu Wu, Huanran Chen, Yiming Shi, Yuzhu Wang, Xi Ye, Jun Zhu
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 representativeness guidance from the diffusion backbone | DiT · ImageNet-1K IPC10 (soft-label) | Top-1 accuracy: 39.6 -> 49.1 | Improved |
| add representativeness guidance from the diffusion backbone | DiT · ImageNet-1K IPC50 (soft-label) | Top-1 accuracy: 52.9 -> 62.7 | Improved |
| add DAP guidance on top of Minimax | Minimax · ImageNette ConvNet-6 IPC10 | Top-1 accuracy: 58.2 -> 64.2 | Improved |
| add DAP guidance on top of Minimax | Minimax · ImageWoof ConvNet-6 IPC10 | Top-1 accuracy: 33.5 -> 38.2 | Improved |
| subsample IPC10 from IPC100 instead of generating it for IPC10 (ablation) | MGD3 · ImageNette ResNet-18 IPC10 | Top-1 accuracy: 56.0 versus DAP 63.7 | Hurt |
| use DAP rather than IGD under dataset scale reduction | DAP · ImageWoof ResNet-18 IPC10 | Top-1 accuracy: 35.2 -> 39.4 | Improved |
| use DAP rather than IGD on unseen architectures | DAP · ImageNet-1K IPC10 MobileNet-V2 | Top-1 accuracy: 39.2 -> 43.1 | Improved |
| use hard-label instead of soft-label supervision for DAP | DAP on Stable Diffusion · ImageNet subsets | Top-1 accuracy | No clear effect |
| use RBF kernel instead of linear kernel (ablation) | DAP · ImageNette/ImageWoof | Top-1 accuracy | No clear effect |
| use the Mid U-Net layer as the guidance feature extractor rather than final layers (ablation) | DAP on Stable Diffusion · ImageNette ResNet-18 | Top-1 accuracy | Improved |
| raise the guidance scale gamma beyond a moderate value (ablation) | DAP · ImageNette ResNet-18 | Top-1 accuracy | Hurt |
| stop guidance at t_stop=25 (early stop) (ablation) | DAP · ImageNette ResNet-18 | Top-1 accuracy | Improved |
Closest papers by meaning
- Taming Diffusion for Dataset Distillation with High Representativeness (ICML 2025 · similarity 0.92)
- Unlocking Dataset Distillation with Diffusion Models (NeurIPS 2025 · similarity 0.92)
- MGD$^3$ : Mode-Guided Dataset Distillation using Diffusion Models (ICML 2025 · similarity 0.91)
- D^4: Dataset Distillation via Disentangled Diffusion Model (CVPR 2024 · similarity 0.91)
- Efficient Dataset Distillation via Minimax Diffusion (CVPR 2024 · similarity 0.91)
- Multistep Distillation of Diffusion Models via Moment Matching (NeurIPS 2024 · similarity 0.90)
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.
