Machine Unlearning on CIFAR-10 SUBCLASS (deer)
1.9KL Divergence Target (nats)GA
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| GABackbone=resnet18, Scenario=SUBCLASS (deer)2026.05 | 1.9 | 4.8 | 2.6 | 72.4 | 56.3 | 8.3 | |
| SalUnBackbone=resnet18, Scenario=SUBCLASS (deer)2026.05 | 0.37 | 0.91 | 0.37 | 88.7 | 97.2 | 17 | |
| SCRUBBackbone=resnet18, Scenario=SUBCLASS (deer)2026.05 | 0.31 | 0.28 | 0.28 | 95 | 97.8 | 22 | |
| DIRBackbone=resnet18, Scenario=SUBCLASS (deer)2026.05 | 0.23 | 0.83 | 0.3 | 96.3 | 100 | 22 | |
| RL+FTBackbone=resnet18, Scenario=SUBCLASS (deer)2026.05 | 0.22 | 0.79 | 0.31 | 97.1 | 100 | 17 | |
| FTBackbone=resnet18, Scenario=SUBCLASS (deer)2026.05 | 0.13 | 0.17 | 0.44 | 97.9 | 100 | 75 | |
| GA+FTBackbone=resnet18, Scenario=SUBCLASS (deer)2026.05 | 0.12 | 0.14 | 0.42 | 97.8 | 100 | 75 | |
| pθBackbone=resnet18, Scenario=SUBCLASS (deer)2026.05 | 0.1 | — | 0.38 | 98 | 100 | — | |
| LDA-MixBackbone=resnet18, Scenario=SUBCLASS (deer)2026.05 | 0.1 | 0.09 | 0.17 | 96.6 | 99.8 | 24 | |
| LDA-2CBackbone=resnet18, Scenario=SUBCLASS (deer)2026.05 | 0.1 | 0.1 | 0.17 | 96.5 | 99.7 | 24 | |
| DIR-2CBackbone=resnet18, Scenario=SUBCLASS (deer)2026.05 | 0.1 | 0.1 | 0.15 | 96.9 | 99.8 | 22 | |
| LDABackbone=resnet18, Scenario=SUBCLASS (deer)2026.05 | 0.09 | 0.1 | 0.18 | 97 | 99.8 | 24 | |
| pθrBackbone=resnet18, Scenario=SUBCLASS (deer)2026.05 | 0 | — | 0 | 97.7 | 95.7 | 100 |