Research map / Label-efficient and robust learning / Knowledge distillation
DREAM: Efficient Dataset Distillation by Representative Matching
Yanqing Liu, Jianyang Gu, Kai Wang, Zheng Zhu, Wei Jiang, Yang You
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 representative matching (sub-cluster sampling of original images) (ablation) | IDC · CIFAR10 IPC 10 | top-1 accuracy (%): 67.5 -> 68.9 | Improved |
| Add clustering-based initialization (ablation) | IDC · CIFAR10 IPC 10 | top-1 accuracy (%): 67.5 -> 68.1 | Improved |
| Combine representative matching and clustering initialization (ablation) | IDC · CIFAR10 IPC 10 | top-1 accuracy (%) and iterations: 67.5 -> 69.4; 1000 -> 150 iterations | Improved |
| Plug DREAM into DC | DC · CIFAR10 IPC 10 | top-1 accuracy (%): 44.9 -> 45.9 | Improved |
| Plug DREAM into DSA | DSA · CIFAR10 IPC 10 | top-1 accuracy (%): 52.1 -> 53.1 | Improved |
| Use DREAM-distilled data on unseen architecture ResNet-10 | IDC distilled on ConvNet-3 · CIFAR10 IPC 10 cross-architecture | top-1 accuracy (%): 63.5 -> 66.3 | Improved |
| Increase sub-cluster number from 32 to 128 (ablation) | DREAM · CIFAR10 IPC 10 | top-1 accuracy (%): 67.2 -> 69.4 | Improved |
| Increase sub-cluster number from 128 to 256 (ablation) | DREAM · CIFAR10 IPC 10 | top-1 accuracy (%): 69.4 -> 68.9 | Hurt |
| Draw more than one sample per sub-cluster (ablation) | DREAM · CIFAR10 IPC 10, N=32 | top-1 accuracy (%): 67.2 -> 67.7 | No clear effect |
| Use DREAM instead of DSA at larger IPC on FashionMNIST | DREAM · FashionMNIST IPC 50 | top-1 accuracy (%): 88.7 -> 86.8 | Hurt |
Later papers that compared with this method
4 later accepted papers report DREAM as a baseline in their results tables; 4 of them report a better number on most of the results they share with it.
- Going Beyond Feature Similarity: Effective Dataset distillation based on Class-aware Conditional Mutual Information (ICLR 2025 · better on 6 of 7 shared results)
- Rectifying Soft-Label Entangled Bias in Long-Tailed Dataset Distillation (NeurIPS 2025 · better on 1 of 1 shared result)
- Distill Gold from Massive Ores: Bi-level Data Pruning towards Efficient Dataset Distillation (ECCV 2024 · better on 1 of 1 shared result)
- CEMA – Cost-Efficient Machine-Assisted Document Annotations (AAAI 2023 · better on 2 of 2 shared results)
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
- DELT: A Simple Diversity-driven EarlyLate Training for Dataset Distillation (CVPR 2025 · similarity 0.90)
- Dataset Distillation for Pre-Trained Self-Supervised Vision Models (NeurIPS 2025 · similarity 0.90)
- SelMatch: Effectively Scaling Up Dataset Distillation via Selection-Based Initialization and Partial Updates by Trajectory Matching (ICML 2024 · similarity 0.90)
- D^4: Dataset Distillation via Disentangled Diffusion Model (CVPR 2024 · similarity 0.90)
- Efficient Dataset Distillation for Pre-Trained Self-Supervised Models via Statistical Flow Matching (NeurIPS 2026 · similarity 0.90)
- Towards Lossless Dataset Distillation via Difficulty-Aligned Trajectory Matching (ICLR 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.
