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

Misalignment Detection on Covert code sabotage Bloom behaviors (in-distribution)

0.959AUROC

18-indicator ensemble

0.90180.916650.93150.94635Jun 23, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.06
0.959
2026.06
0.956
2026.06
0.943
2026.06
0.904