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

Controllability on HarmBench

87.5HarmBench Score

AAS

-3.520.12543.7567.375Jan 27, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.01
87.5
2026.01
87.5
2026.01
86.54
2026.01
86.54
2026.01
85.58
2026.01
84.62
2026.01
84.62
2026.01
82.69
2026.01
82.69
2026.01
82.69
2026.01
79.81
2026.01
77.88
2026.01
77.88
2026.01
76.92
2026.01
74.04
2026.01
74.04
2026.01
74.04
2026.01
74.04
2026.01
70.19
2026.01
70.19
2026.01
70.19
2026.01
68.27
2026.01
67.31
2026.01
55.77
2026.01
54.81
2026.01
53.85
2026.01
52.88
2026.01
50.96
2026.01
50.96
2026.01
39.42
2026.01
32.69
2026.01
25
2026.01
25
2026.01
13.46
2026.01
13.46
2026.01
11.54
2026.01
0
2026.01
0
2026.01
0
2026.01
0