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

Trace-level safety monitoring on ImpossibleBench

95.7ROC-AUC

Meerkat

48.17260.51172.8585.189Apr 13, 2026
Updated 4d ago

Evaluation Results

MethodLinks
2026.04
95.7
2026.04
95.1
2026.04
94.9
2026.04
94.8
2026.04
94.4
2026.04
94.3
2026.04
93.5
2026.04
93.5
2026.04
93.3
2026.04
92.8
2026.04
92.4
2026.04
92.4
2026.04
91.5
2026.04
91.4
2026.04
90
2026.04
89.8
2026.04
89.8
2026.04
89.3
2026.04
88
2026.04
88
2026.04
87.8
2026.04
87.8
2026.04
87.6
2026.04
87.3
2026.04
87.1
2026.04
87
2026.04
85.5
2026.04
85.5
2026.04
84.8
2026.04
84.8
2026.04
80
2026.04
78.9
2026.04
78
2026.04
74.2
2026.04
73.8
2026.04
71.8
2026.04
70
2026.04
61.9
2026.04
50
2026.04
50