Machine Unlearning on CIFAR-100 10+10 classes (test)
74.3Retain AccuracyOriginal
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Originalstrategy=Original2026.02 | 74.3 | 84.2 | — | |
| GradAscstrategy=GradAsc2026.02 | 74.2 | 68.7 | 62.2 | |
| TFER + Orthogonalstrategy=Modular Orthogonal, parametric_isolation=LoRA modules2026.02 | 74.1 | 41.2 | 42.1 | |
| Naive Continualstrategy=Naive Continual2026.02 | 73.3 | 73.2 | 84 | |
| Random Labelstrategy=Random Label2026.02 | 73.1 | 46 | 46.6 | |
| Retrainstrategy=Retrain2026.02 | 41.8 | 95.4 | — |