Machine Unlearning on CIFAR-10 (Utility and Specific Privacy Metrics)
87.92Accuracy (Forget)Original Model
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| Original ModelModel=DeiT-S2025.06 | 87.92 | 84.78 | 76.1 | 71.12 | — | 92.94 | 84.2 | |
| Original ModelModel=ViT-S2025.06 | 86.62 | 85.1 | 75.3 | 71.03 | — | 91.48 | 84.12 | |
| SpotterModel=DeiT-S2025.06 | 0.5 | 85.35 | 0.1 | 72.46 | 0.0047 | 3.2 | 85.26 | |
| SpotterModel=ViT-S2025.06 | 0.3 | 85.19 | 0.4 | 71.69 | 0.0029 | 5.52 | 85.37 | |
| Retrain ModelModel=ViT-S2025.06 | 0 | 87.99 | 0 | 75.48 | — | 14 | 87.56 | |
| Retrain ModelModel=DeiT-S2025.06 | 0 | 87.15 | 0 | 73.59 | — | 27.12 | 86.52 |