Collusion Detection on NARCBENCH CI/CD
0.86AUROCAsymmetry Probe
Evaluation Results
| Method | Links | |
|---|---|---|
| Asymmetry ProbeTraining Split=CORE, Layers=26–30, Evaluation Protocol=Zero-shot2026.04 | 0.86 | |
| Deception SplitTraining Split=CORE, Layers=26–30, Evaluation Protocol=Zero-shot2026.04 | 0.82 | |
| Peak SuspicionTraining Split=CORE, Layers=26–30, Evaluation Protocol=Zero-shot2026.04 | 0.77 | |
| Activation AlignmentTraining Split=CORE, Layers=26–30, Evaluation Protocol=Zero-shot2026.04 | 0.76 | |
| Sorted ConcatenationTraining Split=CORE, Layers=26–30, Evaluation Protocol=Zero-shot2026.04 | 0.37 |