AI-generated text detection on RAID ChatGPT Classic (test)
70.77TPR@0.1%DALD
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| DALDtokens=2002026.05 | 70.77 | 89.2 | 0.9917 | |
| Calibrated DMAPtokens=2002026.05 | 67.62 | 88.66 | 0.991 | |
| DMAPtokens=2002026.05 | 47.54 | 79.06 | 0.9824 | |
| Calibrated Log-Surprisaltokens=2002026.05 | 47.31 | 79.96 | 0.9852 | |
| Baseline Fast-DetectGPTtokens=2002026.05 | 37.7 | 65.7 | 0.9642 | |
| Token Fast-Detecttokens=2002026.05 | 32.58 | 69.16 | 0.9666 | |
| Token NPRtokens=2002026.05 | 25.31 | 64.14 | 0.9602 | |
| Calibrated Token Fast-Detecttokens=2002026.05 | 23.73 | 63.94 | 0.9491 | |
| Baseline Binocularstokens=2002026.05 | 20.08 | 43.83 | 0.8933 | |
| Baseline log-ranktokens=2002026.05 | 3.51 | 19.62 | 0.8396 | |
| Baseline log-surprisaltokens=2002026.05 | 3.04 | 18.7 | 0.8519 | |
| Calibrated Log-Ranktokens=2002026.05 | 0.99 | 62.17 | 0.9412 | |
| Baseline NPRtokens=2002026.05 | 0.52 | 5.06 | 0.7053 | |
| Calibrated Token NPRtokens=2002026.05 | 0.18 | 61.07 | 0.9243 |