Machine Unlearning on CIFAR-100 10% Random Forgetting VGG16-BN
99.51Retention Accuracy (RA)RELOAD
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| RELOADBackbone=VGG16-BN2026.04 | 99.51 | 3.37 | 0.4 | 2 | 0.24 | 0.11 | 0.51 | |
| SCRUBBackbone=VGG16-BN2026.04 | 98.43 | 26.62 | 1.66 | 14 | 0.02 | 0.06 | 0.66 | |
| GABackbone=VGG16-BN2026.04 | 98.41 | 26.4 | 1.64 | 14 | 0 | 0.06 | 0.66 | |
| EU-kBackbone=VGG16-BN2026.04 | 98.35 | 26.16 | 1.67 | 15 | 0.27 | 0.06 | 0.57 | |
| CF-kBackbone=VGG16-BN2026.04 | 98.3 | 26.26 | 1.68 | 15 | 0.27 | 0.06 | 0.56 | |
| FTBackbone=VGG16-BN2026.04 | 98.27 | 12.65 | 1.16 | 8 | 0.25 | 0.06 | 0.5 | |
| Retrained (Baseline)Backbone=VGG16-BN2026.04 | 97.8 | 68.25 | 1.82 | 50 | — | — | — | |
| SSDBackbone=VGG16-BN2026.04 | 22.86 | 61.38 | 2.57 | 2 | 0 | 8.01 | 7.57 |