Research map / Architectures and efficiency / Pruning and efficient tuning
Condensing Large-Scale Datasets Directly with Minimal Information Loss
Xinyi Shang, Peng Sun, Bei Shi, Zixuan Wang, Tao Lin
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 Relabel (ablation) | CIM · ImageNet-1k (IPC=10, ResNet-18) | top-1 accuracy: 22.0 -> 48.7 | Improved |
| add Relabel (ablation) | CIM · CIFAR-10 (IPC=10, ResNet-18) | accuracy: 51.6 -> 66.2 | Improved |
| add Relabel (ablation) | SRe2L · ImageNet-1k (IPC=10, ResNet-18) | top-1 accuracy: 1.1 -> 31.1 | Improved |
| add Relabel (ablation) | G-VBSM · ImageNet-1k (IPC=10, ResNet-18) | top-1 accuracy: 0.8 -> 35.7 | Improved |
| add Relabel (ablation) | RDED · ImageNet-1k (IPC=10, ResNet-18) | top-1 accuracy: 19.7 -> 41.1 | Improved |
| use RDED selection instead of random selection (ablation) | CIM · ImageNet-1k (IPC=10) | top-1 accuracy: 45.3 -> 48.6 | Improved |
| use RDED selection instead of K-means (ablation) | CIM · ImageNet-1k (IPC=10) | top-1 accuracy: 44.8 -> 48.6 | Improved |
| use RDED selection instead of herding (ablation) | CIM · Tiny-ImageNet (IPC=10) | top-1 accuracy: 51.3 -> 53.2 | Improved |
| use RDED selection instead of random selection (ablation) | CIM · CIFAR-10 (IPC=10) | accuracy: 64.5 -> 66.6 | Improved |
| apply Relabel to non-relabel methods ADD and DataDAM as distillation proceeds (ablation) | ADD / DataDAM · ImageNet (IPC=10) | accuracy | Mixed |
| increase compression iterations M beyond 200 (ablation) | CIM · ConvNet IPC=10 | accuracy | No clear effect |
| set number of squeezed images N=4 (ablation) | CIM · four datasets (ConvNet IPC=10) | accuracy | Improved |
Closest papers by meaning
- Are Large-scale Soft Labels Necessary for Large-scale Dataset Distillation? (NeurIPS 2024 · similarity 0.87)
- GIFT: Unlocking Full Potential of Labels in Distilled Dataset at Near-zero Cost (ICLR 2025 · similarity 0.86)
- Large Scale Dataset Distillation with Domain Shift (ICML 2024 · similarity 0.86)
- Enhancing Dataset Distillation via Non-Critical Region Refinement (CVPR 2025 · similarity 0.86)
- Breaking Class Barriers: Efficient Dataset Distillation via Inter-Class Feature Compensator (ICLR 2025 · similarity 0.86)
- Rectifying Soft-Label Entangled Bias in Long-Tailed Dataset Distillation (NeurIPS 2025 · similarity 0.86)
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.
