Machine-generated text detection on MAGE roct (held-out)
0.907AUROCconfidence-weighted ensemble
Evaluation results
| Method | Links | ||
|---|---|---|---|
| confidence-weighted ensembleEnsemble Strategy=Confidence-weighted (Eq. 1)2026.07 | 0.907 | 53.2 | |
| FOMAML+LoRAK-shot=102026.07 | 0.867 | 44.6 | |
| Vanilla RoBERTa baselineBackbone=RoBERTa-large2026.07 | 0.865 | 37.4 |