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

Policy Violation Localization on AgentDojo, MSB, and AgentDAM Average

88AUROC @5

GIF

6268.7575.582.25Jun 22, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.06
8894959697859394959616.38
2026.06
86888895968285859394160.91
2026.06
849091919376808282843.92
2026.06
8087889091707780828315.98
2026.06
78767894966260628488103.37
2026.06
7881808383576060626336.25
2026.06
78828186886874748083162.54
2026.06
778890929262778083826.75
2026.06
7785888990737984868723.72
2026.06
748182899167747784868.89
2026.06
73747682876264677681147.9
2026.06
70707275866263687484157.38
2026.06
67687275815456616773161.71
2026.06
63636368764545485667141.47