Machine Unlearning on CIFAR10 frog
0KL Divergence (t)pθr
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| pθrBackbone=dinov2-mlp12026.05 | 0 | — | 0 | 99.4 | 97.1 | 100 | |
| DIR-2CBackbone=dinov2-mlp12026.05 | 0.02 | 0.02 | 0.17 | 99.6 | 99.9 | 5 | |
| SCRUBBackbone=dinov2-mlp12026.05 | 0.02 | 4.7 | 0.12 | 99.2 | 95.9 | 6.6 | |
| FTBackbone=dinov2-mlp12026.05 | 0.02 | 0.02 | 0.17 | 99.6 | 99.7 | 19 | |
| pθBackbone=dinov2-mlp12026.05 | 0.03 | — | 0.31 | 99.7 | 100 | — | |
| LDABackbone=dinov2-mlp12026.05 | 0.03 | 0.03 | 0.25 | 99.6 | 100 | 5 | |
| LDA-MixBackbone=dinov2-mlp12026.05 | 0.03 | 0.03 | 0.24 | 99.6 | 100 | 5.5 | |
| LDA-2CBackbone=dinov2-mlp12026.05 | 0.03 | 0.03 | 0.24 | 99.6 | 99.9 | 5 | |
| GA+FTBackbone=dinov2-mlp12026.05 | 0.2 | 0.2 | 2.1 | 97 | 71.8 | 19 | |
| DIRBackbone=dinov2-mlp12026.05 | 1 | 1.3 | 10 | 89.8 | 6.7 | 5 | |
| RL+FTBackbone=dinov2-mlp12026.05 | 1 | 1.3 | 10 | 89.7 | 6 | 4.4 | |
| SalUnBackbone=dinov2-mlp12026.05 | 1.5 | 1.5 | 14 | 89 | 0.2 | 4.5 | |
| GABackbone=dinov2-mlp12026.05 | 11 | 1,230 | 43 | 49.4 | 0 | 2.1 |