Research map / Label-efficient and robust learning / Knowledge distillation
Is Retain Set All You Need in Machine Unlearning? Restoring Performance of Unlearned Models with Out-Of-Distribution Images
Jacopo Bonato, Marco Cotogni, Luigi Sabetta
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 distillation-trick on out-of-distribution surrogate images to the Mahalanobis forget loss (ablation) | SCAR · CIFAR100 class removal | AUS: 0.871 -> 0.935 | Improved |
| add Mahalanobis metric-learning forget loss to distillation-trick (ablation) | SCAR · CIFAR100 class removal | AUS: 0.675 -> 0.935 | Improved |
| add distillation-trick in homogeneous removal (ablation) | SCAR · CIFAR100 homogeneous removal | AUS: 0.409 -> 0.934 | Improved |
| use Mahalanobis distance instead of L2 distance in forget loss (ablation) | SCAR · TinyImageNet class removal | AUS: 0.881 -> 0.940 | Improved |
| use Mahalanobis distance instead of cosine similarity (ablation) | SCAR · CIFAR100 class removal | AUS: 0.933 -> 0.935 | No clear effect |
| use Gaussian noise as surrogate dataset instead of ImageNet subset (ablation) | SCAR · CIFAR100 class removal | AUS: 0.935 -> 0.388 | Hurt |
| use COCO subset instead of ImageNet subset as surrogate (ablation) | SCAR · CIFAR100 class removal | AUS: 0.935 -> 0.933 | No clear effect |
| apply SCAR to ViT-B16 instead of ResNet18 | SCAR · CIFAR100 class removal | AUS: 0.935 -> 0.945 | No clear effect |
| use self-forget variant without forget set | SCAR · CIFAR100 class removal | AUS: 0.935 -> 0.929 | No clear effect |
| SCAR versus retain-set methods in homogeneous removal | SCAR · CIFAR100 homogeneous removal | AUS: 0.934 vs 0.965 (DUCK) | Hurt |
Closest papers by meaning
- Machine Unlearning for Image-to-Image Generative Models (ICLR 2024 · similarity 0.89)
- What makes unlearning hard and what to do about it (NeurIPS 2024 · similarity 0.88)
- When to Forget? Complexity Trade-offs in Machine Unlearning (ICML 2025 · similarity 0.88)
- Label-Agnostic Forgetting: A Supervision-Free Unlearning in Deep Models (ICLR 2024 · similarity 0.88)
- From Dormant to Deleted: Tamper-Resistant Unlearning Through Weight-Space Regularization (NeurIPS 2025 · similarity 0.88)
- Decoupled Distillation to Erase: A General Unlearning Method for Any Class-centric Tasks (CVPR 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.
