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

Adversarial Attack Transfer on 388 harmful behaviors

81.19Attack Success Rate (ASR)

I-GCG

-3.247618.673740.59562.5163Jun 4, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.06
81.19
2026.06
79.9
2026.06
76.8
2026.06
70.1
2026.06
64.43
2026.06
63.66
2026.06
63.4
2026.06
61.34
2026.06
50.77
2026.06
43.04
2026.06
41.49
2026.06
40.46
2026.06
37.89
2026.06
25.52
2026.06
24.83
2026.06
24.73
2026.06
24.68
2026.06
22.38
2026.06
20.62
2026.06
18.92
2026.06
18.2
2026.06
14.9
2026.06
12.89
2026.06
6.7
2026.06
4.12
2026.06
3.61
2026.06
3.09
2026.06
2.06
2026.06
1.8
2026.06
1.55
2026.06
1.55
2026.06
1.29
2026.06
0.77
2026.06
0.77
2026.06
0.52
2026.06
0.52
2026.06
0.26
2026.06
0.26
2026.06
0.26
2026.06
0.26
2026.06
0
2026.06
0
2026.06
0
2026.06
0
2026.06
0
2026.06
0
2026.06
0
2026.06
0