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

Jailbreak Defense on MiniGPT-4

7.32Attack Success Rate (ASR)

NullSteer

5.467217.973630.4842.9864Mar 23, 2026
Updated 2mo ago

Evaluation Results

MethodLinks
2026.03
7.32
2026.03
7.62
2026.03
9.09
2026.03
9.09
2026.03
10.72
2026.03
13.18
2026.03
13.58
2026.03
15.46
2026.03
16.36
2026.03
17.27
2026.03
20
2026.03
22.73
2026.03
24.55
2026.03
24.55
2026.03
24.55
2026.03
24.55
2026.03
27.27
2026.03
28.18
2026.03
29.09
2026.03
30
2026.03
32.73
2026.03
33.64
2026.03
35.45
2026.03
36.36
2026.03
37.27
2026.03
38.15
2026.03
38.18
2026.03
40.91
2026.03
41.82
2026.03
42.73
2026.03
42.73
2026.03
43.64
2026.03
44.55
2026.03
47.27
2026.03
49.09
2026.03
53.64