List your company

Collusion Detection on NARCBENCH Research Pipeline

0.94AUROCDeception Split
0.54480.64740.750.8526Apr 1, 2026

Evaluation results

MethodLinks
Deception SplitTraining Split=CORE, Layers=26–30, Evaluation Protocol=Zero-shot2026.040.94
Asymmetry ProbeTraining Split=CORE, Layers=26–30, Evaluation Protocol=Zero-shot2026.040.92
Peak SuspicionTraining Split=CORE, Layers=26–30, Evaluation Protocol=Zero-shot2026.040.83
Sorted ConcatenationTraining Split=CORE, Layers=26–30, Evaluation Protocol=Zero-shot2026.040.78
Activation AlignmentTraining Split=CORE, Layers=26–30, Evaluation Protocol=Zero-shot2026.040.56