Multi-class unlearning on CIFAR-100 20 Classes
79.93Forgotten Accuracy (Acc_ft)Gradient Ascent
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Gradient AscentBackbone=ViT-Small, Rank=322026.06 | 79.93 | 82.79 | 95 | |
| L2ULBackbone=ViT-Small, Rank=322026.06 | 79.1 | 83.11 | 94.5 | |
| Boundary ExpandBackbone=ViT-Small, Rank=322026.06 | 73.48 | 87.67 | 1.2 | |
| SalUnBackbone=ViT-Small, Rank=322026.06 | 71.84 | 97.17 | 12.4 | |
| FinetuneBackbone=ViT-Small, Rank=322026.06 | 56.49 | 88.5 | 73.9 | |
| Boundary ShrinkBackbone=ViT-Small, Rank=322026.06 | 17.73 | 32.35 | 90.6 | |
| DeleteBackbone=ViT-Small, Rank=322026.06 | 13.51 | 96.92 | 3.6 | |
| Random LabelBackbone=ViT-Small, Rank=322026.06 | 12.55 | 29.25 | 94.1 | |
| Bad TeacherBackbone=ViT-Small, Rank=322026.06 | 3.23 | 4.44 | 83 | |
| TrustEraseBackbone=ViT-Small, Rank=322026.06 | 0 | 90.59 | 32 |