AI-generated text detection on M4GT-Bench Wikipedia (test)
60.04PrecisionXLM-R
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| XLM-RTraining Protocol=Leave-one-domain-out2026.02 | 60.04 | 42.55 | 38.8 | 41.95 | |
| Stylistic-SVMTraining Protocol=Leave-one-domain-out2026.02 | 47.94 | 21.21 | 27.77 | 16.06 | |
| NELA-SVMTraining Protocol=Leave-one-domain-out2026.02 | 46.38 | 20.74 | 25.06 | 18.76 | |
| RoBERTaTraining Protocol=Leave-one-domain-out2026.02 | 46.37 | 52.22 | 39.37 | 51.91 | |
| DependencyAITraining Protocol=Leave-one-domain-out2026.02 | 44.45 | 34.03 | 28.9 | 34.3 | |
| GLTR-LRTraining Protocol=Leave-one-domain-out2026.02 | 41.62 | 36.95 | 34.62 | 35.18 | |
| GLTR-SVMTraining Protocol=Leave-one-domain-out2026.02 | 34.86 | 30.95 | 26.72 | 29.81 |