AI-generated text detection on Peer-Review Claude generated
64.02TPR@FPR=0.1%Calibrated DMAP
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Calibrated DMAPVariant=Calibrated2026.05 | 64.02 | 88.6 | 0.9943 | |
| Calibrated Log-SurprisalVariant=Calibrated2026.05 | 59.15 | 83.86 | 0.9897 | |
| DALD2026.05 | 38.11 | 68.19 | 0.9646 | |
| Calibrated Token Fast-DetectVariant=Calibrated2026.05 | 11.02 | 46.25 | 0.9322 | |
| Calibrated Token NPRVariant=Calibrated2026.05 | 7.48 | 31.01 | 0.8996 | |
| Calibrated Log-RankVariant=Calibrated2026.05 | 3.74 | 26.59 | 0.9104 | |
| DMAP2026.05 | 0.87 | 13.52 | 0.8613 | |
| Baseline BinocularsVariant=Baseline2026.05 | 0.25 | 1.66 | 0.5077 | |
| Baseline NPRVariant=Baseline2026.05 | 0.11 | 0.96 | 0.4958 | |
| Token NPRVariant=Token2026.05 | 0.07 | 0.63 | 0.5372 | |
| Baseline Fast-DetectGPTVariant=Baseline2026.05 | 0.04 | 0.56 | 0.5433 | |
| Token Fast-DetectVariant=Token2026.05 | 0.02 | 0.6 | 0.5453 | |
| Baseline Log-SurprisalVariant=Baseline2026.05 | 0 | 0.02 | 0.5144 | |
| Baseline Log-RankVariant=Baseline2026.05 | 0 | 0.05 | 0.4791 |