AI-generated text detection on Peer-Review Gemini generated
43.96TPR @ FPR=0.1%Calibrated DMAP
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Calibrated DMAPVariant=Calibrated2026.05 | 43.96 | 71.25 | 0.9788 | |
| Calibrated Log-SurprisalVariant=Calibrated2026.05 | 38.08 | 67.73 | 0.9686 | |
| DALD2026.05 | 12.08 | 33.75 | 0.8646 | |
| DMAP2026.05 | 5.76 | 37.81 | 0.9523 | |
| Calibrated Token Fast-DetectVariant=Calibrated2026.05 | 2.73 | 24.4 | 0.8609 | |
| Calibrated Log-RankVariant=Calibrated2026.05 | 0.92 | 18.5 | 0.8346 | |
| Calibrated Token NPRVariant=Calibrated2026.05 | 0.81 | 15.3 | 0.8129 | |
| Baseline BinocularsVariant=Baseline2026.05 | 0.74 | 4.04 | 0.6255 | |
| Baseline Fast-DetectGPTVariant=Baseline2026.05 | 0.23 | 1.89 | 0.6063 | |
| Token Fast-DetectVariant=Token2026.05 | 0.22 | 1.86 | 0.6067 | |
| Token NPRVariant=Token2026.05 | 0.18 | 1.52 | 0.5843 | |
| Baseline NPRVariant=Baseline2026.05 | 0.11 | 1.28 | 0.5816 | |
| Baseline Log-SurprisalVariant=Baseline2026.05 | 0 | 0.29 | 0.559 | |
| Baseline Log-RankVariant=Baseline2026.05 | 0 | 0.25 | 0.5269 |