Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Unlearning on WMDP (Accuracy and Refusal Rates)

79.6Bio Accuracy

Original

23.33637.94352.5567.157May 26, 2026
Updated 7d ago

Evaluation Results

MethodLinks
2026.05
79.657.959.1---
2026.05
69.74251.5---
2026.05
46.142.539.9---
2026.05
29.327.725.5---
2026.05
28.928.726.291.188.896.6
2026.05
28.825.529.9---
2026.05
27.126.526.69694.395.3
2026.05
25.825.425.597.797.799
2026.05
25.626.723---
2026.05
25.626.723.2---
2026.05
25.625.427.596.897.193.4
2026.05
25.527.823.7---