Machine Unlearning on ImageNet-100 (forgetting test)
100UART
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| RTBackbone=ViT, Forgetting Scenario=class-wise2025.05 | 100 | 100 | 92.01 | 88.17 | 100 | 0 | 606.93 | |
| GABackbone=ViT, Forgetting Scenario=class-wise2025.05 | 100 | 100 | 81.42 | 78.11 | 100 | 4.13 | 0.76 | |
| SFRonBackbone=ViT, Forgetting Scenario=class-wise2025.05 | 100 | 100 | 81.38 | 80.97 | 100 | 3.57 | 87.96 | |
| MCUBackbone=ViT, Forgetting Scenario=class-wise2025.05 | 100 | 100 | 92.32 | 87.92 | 100 | 0.09 | 105.49 | |
| MCUβBackbone=ViT, Forgetting Scenario=class-wise2025.05 | 100 | 100 | 92.18 | 88 | 100 | 0.07 | 98.12 | |
| NegTVBackbone=ViT, Forgetting Scenario=class-wise2025.05 | 97.85 | 100 | 91.39 | 87.6 | 99.15 | 0.84 | 1.24 | |
| NegGrad+Backbone=ViT, Forgetting Scenario=class-wise2025.05 | 97.46 | 99 | 92.17 | 87.9 | 96.58 | 1.48 | 69.14 | |
| RLBackbone=ViT, Forgetting Scenario=class-wise2025.05 | 96.15 | 100 | 92.21 | 88.1 | 100 | 0.82 | 200.23 | |
| SalUnBackbone=ViT, Forgetting Scenario=class-wise2025.05 | 95.35 | 100 | 92.06 | 88.01 | 100 | 0.97 | 174.67 | |
| FTBackbone=ViT, Forgetting Scenario=class-wise2025.05 | 80.69 | 83 | 92.33 | 87.82 | 83.27 | 10.74 | 100.68 |