WMDP
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
WMDP Cyber
7.19Rel
9
WMDP Bio
6.72Rel Score
9
WMDP multiple-choice QA
65.1Bio Accuracy
9
WMDP
64.62Accuracy (Bio)
8
WMDP Cyber
68.5Forget Rate
8
WMDP
34.6Score
8
WMDP bio
0.522Unlearn Score
8
WMDP
41.2RAP
8
WMDP
0.029MIA
8
WMDP Cyber
44.69WMDP Accuracy
7
WMDP-Bio
63.55WMDP Acc
7
WMDP Bio
64.31RR
7
WMDP-Cyber
41.5Accuracy (Pre-Unlearning)
7
WMDP Bio
67.6Accuracy (Pre-Unlearning)
7
WMDP Bio
72.43Accuracy (All)
7
WMDP Cyber
-1.1Forget Accuracy
6
WMDP Cyber
46.7ES
6
WMDP Bio
62.3ES
6
WMDP Chem
95.3Refusal Rate
4
WMDP Cyber
59.6Refusal Rate
4
WMDP Bio
39.5Refusal Rate
4
WMDP-Cyber
75.8MMLU Score
4
WMDP
36.77Average VRAM (GB)
4
WMDP Cyber
44.7Accuracy
4
WMDP Bio (test)
65Accuracy
3