Machine Unlearning on CIFAR100 scenario SUBCLASS (apple) (test)
0KL Divergence (Test Set)pθr
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| pθrBackbone=dinov2-mlp1, Scenario=SUBCLASS (apple)2026.05 | 0 | — | 0 | 92.3 | 96.2 | 100 | |
| pθBackbone=dinov2-mlp1, Scenario=SUBCLASS (apple)2026.05 | 0.08 | — | 0.35 | 92.4 | 100 | — | |
| LDA-MixBackbone=dinov2-mlp1, Scenario=SUBCLASS (apple)2026.05 | 0.09 | 0.09 | 0.12 | 92.2 | 94.9 | 1.5 | |
| DIRBackbone=dinov2-mlp1, Scenario=SUBCLASS (apple)2026.05 | 0.09 | 0.15 | 0.31 | 92.2 | 84.5 | 1.1 | |
| DIR-2CBackbone=dinov2-mlp1, Scenario=SUBCLASS (apple)2026.05 | 0.09 | 0.09 | 0.18 | 92.2 | 98.7 | 1.2 | |
| SCRUBBackbone=dinov2-mlp1, Scenario=SUBCLASS (apple)2026.05 | 0.09 | 0.09 | 0.28 | 92.3 | 98.2 | 0.63 | |
| SalUnBackbone=dinov2-mlp1, Scenario=SUBCLASS (apple)2026.05 | 0.09 | 0.52 | 0.1 | 92.3 | 96.7 | 0.45 | |
| RL+FTBackbone=dinov2-mlp1, Scenario=SUBCLASS (apple)2026.05 | 0.09 | 0.26 | 0.16 | 92.2 | 93.2 | 0.6 | |
| GABackbone=dinov2-mlp1, Scenario=SUBCLASS (apple)2026.05 | 0.09 | 4.2 | 0.13 | 92.2 | 93.3 | 0.21 | |
| LDA-2CBackbone=dinov2-mlp1, Scenario=SUBCLASS (apple)2026.05 | 0.1 | 0.09 | 0.32 | 92.1 | 100 | 1.2 | |
| LDABackbone=dinov2-mlp1, Scenario=SUBCLASS (apple)2026.05 | 0.11 | 0.09 | 0.32 | 91.9 | 100 | 1.2 | |
| FTBackbone=dinov2-mlp1, Scenario=SUBCLASS (apple)2026.05 | 0.12 | 0.13 | 0.29 | 91.9 | 100 | 17 | |
| GA+FTBackbone=dinov2-mlp1, Scenario=SUBCLASS (apple)2026.05 | 0.13 | 0.13 | 0.19 | 91.6 | 98.8 | 17 |