Machine Unlearning Evaluation on TOFU (Forget)
0.986ROUGEWISE
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| WISECategory=Model Editing, Target=Avoidant2025.12 | 0.986 | 0.899 | 51.1 | |
| WISECategory=Model Editing, Target=Incorrect2025.12 | 0.982 | 0.866 | 60.7 | |
| WISECategory=Model Editing, Target=Dummy2025.12 | 0.965 | 0.815 | 51.8 | |
| IKECategory=Model Editing, Target=Dummy2025.12 | 0.466 | 0.677 | 50.8 | |
| IKECategory=Model Editing, Target=Incorrect2025.12 | 0.424 | 0.386 | 71.2 | |
| Ground Truth2025.12 | 0.408 | 0.148 | 67.4 | |
| IKECategory=Model Editing, Target=Avoidant2025.12 | 0.383 | 0.567 | 53.9 | |
| ROMECategory=Model Editing, Target=Incorrect2025.12 | 0.099 | — | — | |
| POCategory=Model Unlearning2025.12 | 0.053 | 0.848 | 54.7 | |
| ROMECategory=Model Editing, Target=Dummy2025.12 | 0.041 | 0.012 | 69.7 | |
| ROMECategory=Model Editing, Target=Avoidant2025.12 | 0.036 | — | — | |
| GDCategory=Model Unlearning2025.12 | 0.0024 | 0 | 82.2 | |
| GACategory=Model Unlearning2025.12 | 0.0017 | 0 | 84.5 | |
| KLCategory=Model Unlearning2025.12 | 0 | 0 | 86.6 |