Collusion Detection on NARCBENCH News Framing
0.88AUROCAsymmetry Probe
Evaluation Results
| Method | Links | |
|---|---|---|
| Asymmetry ProbeTraining Split=CORE, Layers=26–30, Evaluation Protocol=Zero-shot2026.04 | 0.88 | |
| Peak SuspicionTraining Split=CORE, Layers=26–30, Evaluation Protocol=Zero-shot2026.04 | 0.87 | |
| Deception SplitTraining Split=CORE, Layers=26–30, Evaluation Protocol=Zero-shot2026.04 | 0.85 | |
| Activation AlignmentTraining Split=CORE, Layers=26–30, Evaluation Protocol=Zero-shot2026.04 | 0.79 | |
| Sorted ConcatenationTraining Split=CORE, Layers=26–30, Evaluation Protocol=Zero-shot2026.04 | 0.43 |