Our new X account is live! Follow @wizwand_team for updates
Home
/
Benchmarks
Jailbreak evaluation on Synthetic dataset (held-out)
Loading...
100
Good Bot Rate
AIM
-2.96
23.77
50.5
77.23
Jul 5, 2023
Good Bot Rate
Bad Bot Rate
Unclear Rate
Updated 4d ago
Evaluation Results
Method
Method
Links
Good Bot Rate
Bad Bot Rate
Unclear Rate
AIM
Model=Claude v1.3
2023.07
100
-
0
AIM
Model=GPT-4
2023.07
13
-
1
combination_3
Model=Claude v1.3
2023.07
12
-
2
Adaptive attack
Model=GPT-4
2023.07
4
96
-
combination_2
Model=GPT-4
2023.07
3
-
10
combination_2
Model=Claude v1.3
2023.07
3
-
8
combination_3
Model=GPT-4
2023.07
2
-
5
Adaptive attack
Model=Claude v1.3
2023.07
1
99
-
Feedback
Search any
task
Search any
task