Machine-Generated Text Detection on Reuters StableLM 1.0 (test)
34.84TPR @ FPR=1%Log-Rank
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Log-RankCalibration=Markov-informed score calibration, Training Source=Claude2026.02 | 34.84 | 90.36 | |
| LikelihoodCalibration=Markov-informed score calibration, Training Source=Claude2026.02 | 34.58 | 90.57 | |
| Log-RankCalibration=None, Training Source=Claude2026.02 | 24.58 | 85.68 | |
| LikelihoodCalibration=None, Training Source=Claude2026.02 | 19.42 | 86.45 | |
| EntropyCalibration=Markov-informed score calibration, Training Source=Claude2026.02 | 18 | 83.03 | |
| FastGPTCalibration=Markov-informed score calibration, Training Source=Claude2026.02 | 8.62 | 71.74 | |
| DetectGPTCalibration=Markov-informed score calibration, Training Source=Claude2026.02 | 6 | 70.99 | |
| DNA-GPTCalibration=Markov-informed score calibration, Training Source=Claude2026.02 | 5.38 | 72.95 | |
| EntropyCalibration=None, Training Source=Claude2026.02 | 4.36 | 76.24 | |
| DNA-GPTCalibration=None, Training Source=Claude2026.02 | 2.44 | 65.3 | |
| FastGPTCalibration=None, Training Source=Claude2026.02 | 1.42 | 58.69 | |
| DetectGPTCalibration=None, Training Source=Claude2026.02 | 0.44 | 58.01 |