Machine Unlearning on ImageNet Forget set
98.02AuFISHER
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| FISHERBackbone=ViT-b162025.03 | 98.02 | 0.605 | 0.83 | |
| OriginalBackbone=ViT-b162025.03 | 94.26 | — | — | |
| FTBackbone=ViT-b162025.03 | 91.97 | — | — | |
| SCRUBBackbone=ViT-b162025.03 | 45.82 | 0.314 | 0.84 | |
| InfluenceBackbone=ViT-b162025.03 | 15.28 | 0.112 | 0.04 | |
| NegGradBackbone=ViT-b162025.03 | 0.41 | 0.03 | 0.23 | |
| GDR-GMABackbone=ViT-b162025.03 | 0.15 | 0.025 | 0.54 | |
| OrthoGradBackbone=ViT-b162025.03 | 0.1 | 0.002 | 0.3 | |
| NegGrad+Backbone=ViT-b162025.03 | 0.05 | 0.024 | 0.38 | |
| RetrainBackbone=ViT-b162025.03 | 0 | — | — | |
| DUCKBackbone=ViT-b162025.03 | 0 | 0.064 | 0.11 |