Machine Unlearning Evaluation on TOFU Real World
89.7ROUGEGround Truth
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Ground Truth2025.12 | 89.7 | 0.414 | 54.4 | |
| GDCategory=Model Unlearning2025.12 | 88.9 | 0.465 | 60.7 | |
| WISECategory=Model Editing, Target=Dummy2025.12 | 87.5 | 0.424 | 56.4 | |
| WISECategory=Model Editing, Target=Incorrect2025.12 | 87.5 | 0.426 | 57.5 | |
| WISECategory=Model Editing, Target=Avoidant2025.12 | 87.5 | 0.426 | 55.9 | |
| POCategory=Model Unlearning2025.12 | 85.6 | 0.391 | 48.4 | |
| IKECategory=Model Editing, Target=Incorrect2025.12 | 82.1 | 0.438 | 56.7 | |
| IKECategory=Model Editing, Target=Avoidant2025.12 | 72.9 | 0.481 | 61.7 | |
| ROMECategory=Model Editing, Target=Dummy2025.12 | 62.6 | 0.43 | 58.3 | |
| IKECategory=Model Editing, Target=Dummy2025.12 | 22.6 | 0.456 | 58 | |
| ROMECategory=Model Editing, Target=Incorrect2025.12 | 2.7 | 0.258 | 23.9 | |
| ROMECategory=Model Editing, Target=Avoidant2025.12 | 0.3 | 0.235 | 19 | |
| KLCategory=Model Unlearning2025.12 | 0 | 0.251 | 38.1 | |
| GACategory=Model Unlearning2025.12 | 0 | 0.239 | 38.2 |