Image Classification on Fashion-MNIST Class Unlearning (Dft)
92.5AccuracyOriginal
Evaluation results
| Method | Links | |
|---|---|---|
| OriginalBackbone=ViT2024.11 | 92.5 | |
| OriginalBackbone=ResNet182024.11 | 90.6 | |
| FinetuneBackbone=ResNet182024.11 | 76.8 | |
| FinetuneBackbone=ViT2024.11 | 62.6 | |
| NegGradBackbone=ResNet182024.11 | 9.7 | |
| SCRUBBackbone=ResNet182024.11 | 5.3 | |
| BadTBackbone=ViT2024.11 | 0.2 | |
| SCRUBBackbone=ViT2024.11 | 0.2 | |
| RetrainBackbone=ResNet182024.11 | 0 | |
| BadTBackbone=ResNet182024.11 | 0 | |
| Residual Feature Alignment UnlearningBackbone=ResNet182024.11 | 0 | |
| RetrainBackbone=ViT2024.11 | 0 | |
| NegGradBackbone=ViT2024.11 | 0 | |
| Residual Feature Alignment UnlearningBackbone=ViT2024.11 | 0 |