Multi-class unlearning on CIFAR-100 2 Classes
86.55Forgotten Accuracy (Acc_ft)Gradient Ascent
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Gradient AscentBackbone=ViT-Small, Rank=322026.06 | 86.55 | 83 | 98.1 | |
| L2ULBackbone=ViT-Small, Rank=322026.06 | 85.55 | 83.03 | 97.7 | |
| SalUnBackbone=ViT-Small, Rank=322026.06 | 75.84 | 97.44 | 18 | |
| FinetuneBackbone=ViT-Small, Rank=322026.06 | 59.54 | 86.21 | 78 | |
| Boundary ShrinkBackbone=ViT-Small, Rank=322026.06 | 46.51 | 83.02 | 70.2 | |
| Boundary ExpandBackbone=ViT-Small, Rank=322026.06 | 19.69 | 96.74 | 99.6 | |
| Bad TeacherBackbone=ViT-Small, Rank=322026.06 | 4.82 | 96.62 | 0 | |
| Random LabelBackbone=ViT-Small, Rank=322026.06 | 4.58 | 82.73 | 8.9 | |
| DeleteBackbone=ViT-Small, Rank=322026.06 | 1.85 | 97.26 | 12 | |
| TrustEraseBackbone=ViT-Small, Rank=322026.06 | 0 | 88.5 | 39 |