Unlearning on WMDP
0.489WMDP ScoreTask Vector
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Task VectorBase Model=Ministral-8B-Instruct-24102025.09 | 0.489 | 281.883 | 0 | 0 | 8 | |
| DOORBase Model=Ministral-8B-Instruct-24102025.09 | 0.481 | 99.667 | 8.5 | 0.289 | 5 | |
| GAGDRBase Model=Ministral-8B-Instruct-24102025.09 | 0.424 | 77.667 | 30.3 | 0.528 | 4 | |
| OriginalBase Model=Ministral-8B-Instruct-24102025.09 | 0.393 | — | — | — | — | |
| NPOGDRBase Model=Ministral-8B-Instruct-24102025.09 | 0.393 | 180.189 | 38.4 | 0.556 | 3 | |
| PRISMBase Model=Ministral-8B-Instruct-24102025.09 | 0.308 | 110.033 | 79.2 | 0.761 | 1 | |
| SAM+NPOBase Model=Ministral-8B-Instruct-24102025.09 | 0.261 | 123.952 | 98.1 | 0.721 | 2 | |
| GABase Model=Ministral-8B-Instruct-24102025.09 | 0.256 | 50.7 | 100 | 0 | 7 | |
| NPOBase Model=Ministral-8B-Instruct-24102025.09 | 0.256 | 65.989 | 100 | 0.055 | 6 |