Image Classification Unlearning on CIFAR-10 50% Random Forgetting
0.0112MIABARRIER
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| BARRIERBackbone=ResNet-18, TParams=0.05%2026.05 | 0.0112 | — | 0.56 | 99.48 | 94.86 | |
| GABackbone=ResNet-18, TParams=100%2026.05 | 0.0122 | — | 0.4 | 99.61 | 94.34 | |
| FTBackbone=ResNet-18, TParams=100%2026.05 | 0.0215 | — | 0.44 | 99.96 | 94.23 | |
| ℓ1-sparseBackbone=ResNet-18, TParams=100%2026.05 | 0.0476 | — | 1.44 | 99.52 | 93.13 | |
| SEMUBackbone=ResNet-18, TParams=0.64%2026.05 | 0.072 | — | 1.77 | 98.12 | 91.8 | |
| IUBackbone=ResNet-18, TParams=100%2026.05 | 0.0729 | — | 3.97 | 96.21 | 90 | |
| SEMU-remainBackbone=ResNet-18, TParams=0.72%2026.05 | 0.0754 | — | 1.82 | 98.12 | 91.72 | |
| SalUnBackbone=ResNet-18, TParams=100%2026.05 | 0.1699 | — | 7.75 | 94.28 | 89.29 | |
| RetrainBackbone=ResNet-18, TParams=100%2026.05 | 0.1929 | — | 7.91 | 100 | 91.72 | |
| BEBackbone=ResNet-18, TParams=100%2026.05 | 0.2487 | — | 3.08 | 96.84 | 90.41 | |
| SalUn-softBackbone=ResNet-18, TParams=100%2026.05 | 0.315 | — | 3.41 | 99.62 | 91.82 | |
| BSBackbone=ResNet-18, TParams=100%2026.05 | 0.3215 | — | 9.76 | 90.19 | 83.71 | |
| RLBackbone=ResNet-18, TParams=100%2026.05 | 0.4195 | — | 4.8 | 99.55 | 91.31 | |
| RLBackbone=ViT2026.04 | 71.8 | 0.013 | — | — | — | |
| IEU w/GA+NoisyBackbone=ViT2026.04 | 72.3 | 0.016 | — | — | — | |
| SCRUBBackbone=ViT2026.04 | 72.8 | 0.065 | — | — | — | |
| IEU w/GABackbone=ViT2026.04 | 73 | 0.054 | — | — | — | |
| FTBackbone=ViT2026.04 | 73.6 | 0.065 | — | — | — | |
| IEU w/NoisyBackbone=ViT2026.04 | 74 | 0.012 | — | — | — | |
| SALUNBackbone=ViT2026.04 | 74.2 | 0.168 | — | — | — | |
| RetrainingBackbone=ViT2026.04 | 75.7 | 0 | — | — | — |