Our new X account is live! Follow @wizwand_team for updates
Home
/
Benchmarks
Illicit Task Completion on AgentHarm English prompts
Loading...
72.7
AgentHarm Score (AHS)
STING
-0.62
18.415
37.45
56.485
Feb 18, 2026
AgentHarm Score (AHS)
Updated 4d ago
Evaluation Results
Method
Method
Links
AgentHarm Score (AHS)
STING
Target Agent Model=Qwe...
2026.02
72.7
STING
Target Agent Model=Qwe...
2026.02
67.8
STING
Target Agent Model=Dee...
2026.02
61.8
STING
Target Agent Model=Dee...
2026.02
57.2
STING
Target Agent Model=Gem...
2026.02
50.9
STING
Target Agent Model=Gem...
2026.02
47.6
Single-turn prompting
Target Agent Model=Gem...
2026.02
45.9
Single-turn prompting
Target Agent Model=Qwe...
2026.02
35.1
STING
Target Agent Model=GPT...
2026.02
34.1
STING
Target Agent Model=Cla...
2026.02
32.3
Single-turn prompting
Target Agent Model=Dee...
2026.02
31.2
STING
Target Agent Model=GPT...
2026.02
29.7
STING
Target Agent Model=Cla...
2026.02
28
X-Teaming
Target Agent Model=Qwe...
2026.02
27
Single-turn prompting
Target Agent Model=GPT...
2026.02
24.3
Single-turn prompting
Target Agent Model=Cla...
2026.02
16
X-Teaming
Target Agent Model=Dee...
2026.02
15.1
X-Teaming
Target Agent Model=Gem...
2026.02
13.8
X-Teaming
Target Agent Model=GPT...
2026.02
5
X-Teaming
Target Agent Model=Cla...
2026.02
2.2
Feedback
Search any
task
Search any
task