Class-wise Machine Unlearning on CIFAR-20 veg (test)
95.88Ar (Accuracy Retention)baseline
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| baselinemodel=ViT2026.02 | 95.88 | 97.67 | 91.48 | |
| MaGAmodel=ViT2026.02 | 95.64 | 0 | 0 | |
| SSDmodel=ViT2026.02 | 95.61 | 0 | 1.44 | |
| retrainmodel=ViT2026.02 | 94.95 | 0 | 4.16 | |
| UNSIRmodel=ViT2026.02 | 93.75 | 86.57 | 63.6 | |
| AMNCmodel=ViT2026.02 | 93.31 | 0 | 1.05 | |
| FTmodel=ViT2026.02 | 88.52 | 0.59 | 14.44 | |
| SSDmodel=RN182026.02 | 82.64 | 45.34 | 2.08 | |
| baselinemodel=RN182026.02 | 82.59 | 88.94 | 93.28 | |
| MaGAmodel=RN182026.02 | 82.53 | 0 | 0 | |
| retrainmodel=RN182026.02 | 82.24 | 0 | 9.24 | |
| AMNCmodel=RN182026.02 | 81.66 | 0 | 3.04 | |
| UNSIRmodel=RN182026.02 | 81.23 | 70.24 | 43.27 | |
| FTmodel=RN182026.02 | 72.56 | 0 | 29.16 |