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

ImpossibleBench

Benchmarks

Task NameDataset NameSOTA ResultTrend
Trace-level safety monitoringImpossibleBench
ROC-AUC95.7
40
Trace-level detectionImpossibleBench
AP88.7
40
Sabotage DetectionImpossibleBench
Average Precision100
40
Case-level safety detectionImpossibleBench
ROC AUC100
40
Showing 4 of 4 rows