Image Classification Unlearning on Tiny ImageNet (30% Class-wise Forgetting)
92.7Accuracy (Train Retained)Retraining
Evaluation results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| RetrainingBackbone=ResNet2026.04 | 92.7 | 59.3 | 0 | 0 | 0.909 | 0 | — | — | — | |
| IEU w/GABackbone=ResNet2026.04 | 89 | 56.7 | 0 | 0 | 0.908 | 0.013 | — | — | — | |
| IEU w/NoisyBackbone=ResNet2026.04 | 88.2 | 57.7 | 0 | 0 | 0.908 | 0.012 | — | — | — | |
| FTBackbone=ResNet2026.04 | 88.1 | 57.7 | 0 | 0 | 0.909 | 0.012 | — | — | — | |
| IEU w/GA+NoisyBackbone=ResNet2026.04 | 87.3 | 57.8 | 0 | 0 | 0.909 | 0.014 | — | — | — | |
| RLBackbone=ResNet2026.04 | 81.7 | 58 | 0 | 0 | 0.909 | 0.025 | — | — | — | |
| SCRUBBackbone=ResNet2026.04 | 75.3 | 58.6 | 0 | 0 | 0.906 | 0.037 | — | — | — | |
| IEU w/NoisyBackbone=ViT2026.04 | 62.9 | 42.6 | — | — | — | 0.013 | 0 | 0 | 0.957 | |
| IEU w/GABackbone=ViT2026.04 | 61.7 | 42.9 | — | — | — | 0.011 | 0 | 0 | 0.962 | |
| IEU w/GA+NoisyBackbone=ViT2026.04 | 60.4 | 42.5 | — | — | — | 0.008 | 0 | 0 | 0.956 | |
| FTBackbone=ViT2026.04 | 59.2 | 43 | — | — | — | 0.005 | 0 | 0 | 0.958 | |
| RetrainingBackbone=ViT2026.04 | 59 | 42.8 | — | — | — | 0 | 0 | 0 | 0.936 | |
| SALUNBackbone=ResNet2026.04 | 56.9 | 48.6 | 0.2 | 0.4 | 0.911 | 0.095 | — | — | — | |
| SCRUBBackbone=ViT2026.04 | 56 | 44.6 | — | — | — | 0.017 | 0 | 0 | 0.973 | |
| RLBackbone=ViT2026.04 | 55.4 | 44.6 | — | — | — | 0.013 | 0 | 0 | 0.946 | |
| SALUNBackbone=ViT2026.04 | 36 | 35.7 | — | — | — | 0.062 | 0.001 | 0.001 | 0.944 |