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

Jailbreak Detection on MM-SafetyBench

99.18AUROC

Mahal-OOD

50.008862.774475.5488.3056Dec 12, 2025Jan 10, 2026Feb 8, 2026Mar 9, 2026Apr 7, 2026May 6, 2026Jun 5, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2025.12
99.1899.32
2025.12
97.0197.62
2026.06
96.8-
2025.12
95.2395.91
2025.12
91.2698.04
2026.06
91.1-
2025.12
85.4790.65
2025.12
85.1487.52
2026.06
84.8-
2025.12
82.6993.53
2026.06
81.8-
2026.06
78.1-
2026.06
77.4-
2026.06
75-
2025.12
74.2787.29
2026.06
68.8-
2026.06
68.6-
2026.06
67.3-
2026.06
67.3-
2026.06
61.2-
2025.12
61.0680.2
2026.06
53.3-
2026.06
51.9-