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

Visual Jailbreak Defense on Visual Adversarial Attacks epsilon = 16/255

12.45Attack Success Rate

IMMUNE

10.07426.11242.1558.188Nov 27, 2024
Updated 1mo ago

Evaluation Results

MethodLinks
2024.11
12.45
2024.11
13.22
2024.11
13.32
2024.11
17.32
2024.11
18.42
2024.11
23.41
2024.11
27
2024.11
34.25
2024.11
35.21
2024.11
38.11
2024.11
38.15
2024.11
39
2024.11
44.23
2024.11
45.21
2024.11
47.09
2024.11
47.82
2024.11
49.1
2024.11
53.19
2024.11
53.4
2024.11
53.98
2024.11
59.81
2024.11
62.57
2024.11
66.12
2024.11
68
2024.11
71.85