Research map / Label-efficient and robust learning / Recognition and long-tail
Rethinking Long-tailed Dataset Distillation: A Uni-Level Framework with Unbiased Recovery and Relabeling
Xiao Cui, Yulei Qin, Xinyue Li, Wengang Zhou, Hongsheng Li, Houqiang Li
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 model debiasing (mixture consistency + class-wise debias loss) (ablation) | Ours · CIFAR-100-LT (IF=50, IPC=10) | accuracy (%): 31.7 -> 42.1 | Improved |
| add fair BN statistics recalibration (ablation) | Ours · CIFAR-100-LT (IF=50, IPC=10) | accuracy (%): 40.9 -> 42.1 | Improved |
| add confidence-guided multi-round initialization (ablation) | Ours · CIFAR-100-LT (IF=50, IPC=10) | accuracy (%): 40.8 -> 42.1 | Improved |
| replace bi-level trajectory matching (DAMED) with uni-level BN statistic alignment | Ours · CIFAR-100-LT (IF=10, IPC=10) | accuracy (%): 31.5 -> 47.1 | Improved |
| replace DAMED with ours on Tiny-ImageNet-LT | Ours · Tiny-ImageNet-LT (IF=10, IPC=10) | accuracy (%): 26.0 -> 37.8 | Improved |
| replace DAMED with ours under extreme imbalance IF=256 with ResNet-50 | Ours · ImageNet-LT (IF=256, IPC=10) | accuracy (%): 17.2 -> 48.2 | Improved |
| replace DAMED with ours at IPC=1 | Ours · CIFAR-10-LT (IF=100, IPC=1) | accuracy (%): 24.1 -> 44.8 | Improved |
| use ours vs DAMED peak memory | Ours · CIFAR-100-LT | peak GPU memory (GB): 10.2 -> 3.1 (IPC=10) | Improved |
Closest papers by meaning
- Rectifying Soft-Label Entangled Bias in Long-Tailed Dataset Distillation (NeurIPS 2025 · similarity 0.93)
- Distilling Long-tailed Datasets (CVPR 2025 · similarity 0.91)
- A Label is Worth A Thousand Images in Dataset Distillation (NeurIPS 2024 · similarity 0.88)
- Long-tailed Recognition with Model Rebalancing (NeurIPS 2025 · similarity 0.88)
- Rethinking Dataset Distillation for Classification: Do Distilled Sets Outperform Coresets? (NeurIPS 2026 · similarity 0.88)
- Optimizing Distributional Geometry Alignment with Optimal Transport for Generative Dataset Distillation (NeurIPS 2025 · 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.
