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

Trace-level detection on ImpossibleBench

88.7AP

Meerkat

0.61223.48146.3569.219Apr 13, 2026
Updated 5d ago

Evaluation Results

MethodLinks
2026.04
88.70.07
2026.04
84.50.056
2026.04
83.2-
2026.04
820.032
2026.04
82-0.012
2026.04
81.7-
2026.04
81.7-
2026.04
81.5-
2026.04
79.3-
2026.04
78.9-
2026.04
78.9-
2026.04
78.8-
2026.04
78.8-
2026.04
78.1-
2026.04
75.50.017
2026.04
75.3-0.036
2026.04
750.012
2026.04
74.8-
2026.04
73.8-
2026.04
73.8-
2026.04
73.1-0.085
2026.04
72.6-
2026.04
70-
2026.04
68.1-
2026.04
67.4-
2026.04
65.9-
2026.04
64.1-
2026.04
64-
2026.04
63.6-
2026.04
62-
2026.04
58-
2026.04
57.8-
2026.04
57.3-
2026.04
53.2-
2026.04
49.7-
2026.04
46.2-
2026.04
41.9-
2026.04
26.4-
2026.04
10-
2026.04
4-