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

Jailbreak Defense on HarmBench and AdvBench (test)

91.2GCG Score

No Defense

-2.08822.13146.3570.569Jan 15, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.01
91.285.379.594.841.295.6
2026.01
85.280.475.192.438.694.5
2026.01
84.375.773.589.236.490.5
2026.01
82.676.370.216.212.411.5
2026.01
79.266.845.312.48.59.6
2026.01
78.468.264.582.132.484.6
2026.01
78.364.742.615.410.29.8
2026.01
72.559.438.210.57.28.4
2026.01
35.541.633.221.415.817.6
2026.01
32.438.635.218.412.615.2
2026.01
29.316.619.49.212.27.1
2026.01
28.633.430.216.511.213.6
2026.01
26.812.113.48.510.46.2
2026.01
26.231.428.614.89.511.4
2026.01
25.311.412.38.810.26.5
2026.01
24.332.826.317.212.414.6
2026.01
22.528.124.713.49.611.8
2026.01
22.410.211.57.28.55.1
2026.01
19.181.573.186.435.221.4
2026.01
18.431.229.510.26.45.2
2026.01
16.229.727.412.87.26.1
2026.01
15.228.525.48.55.44.1
2026.01
14.568.458.285.632.418.2
2026.01
14.412.610.29.87.58.4
2026.01
14.331.22813.58.16.8
2026.01
13.416.518.95.13.21.8
2026.01
13.113.711.28.66.87.9
2026.01
12.415.217.75.63.32.1
2026.01
12.212.610.46.84.12.6
2026.01
11.613.413.25.43.52.2
2026.01
11.213.415.24.12.82.4
2026.01
11.267.756.781.230.514.8
2026.01
10.213.416.24.22.51.5
2026.01
10.258.449.275.828.512.5
2026.01
9.312.615.44.92.93.7
2026.01
8.86.77.42.63.45.1
2026.01
8.59.87.36.15.44.2
2026.01
8.511.212.43.42.13.2
2026.01
8.56.57.32.83.72.7
2026.01
8.19.19.57.46.85.1
2026.01
4.53.84.25.24.53.2
2026.01
2.53.12.95.24.13.1
2026.01
1.72.41.41.32.13
2026.01
1.52.21.31.52.43.5