AI-generated text detection on Peer-Review GPT-4o generated benchmark leaderboard · SOTA2 Research

AI-generated text detection on Peer-Review GPT-4o generated

33.04TPR @ 0.1% FPRCalibrated DMAP

Evaluation Results

Method
Calibrated DMAP2026.05	33.04	63.43	0.9734
Calibrated Log-Surprisal2026.05	32.71	64.35	0.9544
DALD2026.05	8.51	27.42	0.8523
Calibrated Token Fast-Detect2026.05	1.48	24.35	0.8623
Calibrated Token NPR2026.05	1.21	14.84	0.8089
DMAP2026.05	0.49	7.95	0.8337
Calibrated Log-Rank2026.05	0.22	13.72	0.8262
Baseline NPR2026.05	0.16	1.82	0.5996
Baseline Fast-DetectGPT2026.05	0.14	2.06	0.6397
Token Fast-Detect2026.05	0.13	1.97	0.6403
Token NPR2026.05	0.13	1.8	0.6157
Baseline Binoculars2026.05	0.02	0.31	0.2887
Baseline Log-Surprisal2026.05	0	0.52	0.6354
Baseline Log-Rank2026.05	0	0.32	0.6021