Machine Unlearning Evaluation on TOFU
61.4Model UtilityGround Truth
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Ground Truth2025.12 | 61.4 | 100 | |
| WISECategory=Model Editing, Target=Avoidant2025.12 | 61.2 | 0 | |
| WISECategory=Model Editing, Target=Dummy2025.12 | 61 | 0 | |
| WISECategory=Model Editing, Target=Incorrect2025.12 | 60.3 | 0.0001 | |
| GDCategory=Model Unlearning2025.12 | 58.9 | 0 | |
| POCategory=Model Unlearning2025.12 | 53.8 | 0 | |
| IKECategory=Model Editing, Target=Incorrect2025.12 | 47.3 | 24.9 | |
| IKECategory=Model Editing, Target=Avoidant2025.12 | 45.2 | 0 | |
| IKECategory=Model Editing, Target=Dummy2025.12 | 33.8 | 0 | |
| ROMECategory=Model Editing, Target=Dummy2025.12 | 15.3 | 12.1 | |
| KLCategory=Model Unlearning2025.12 | 0 | 0 | |
| GACategory=Model Unlearning2025.12 | 0 | 0 |