Machine-Generated Text Detection on Reuters Dolly
38.62TPR @ FPR=1%Likelihood-M
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Likelihood-MTraining Source=ChatGPT-turbo2026.02 | 38.62 | 79.93 | |
| Log-Rank-MTraining Source=ChatGPT-turbo2026.02 | 33.07 | 76.27 | |
| Log-RankTraining Source=ChatGPT-turbo2026.02 | 15.2 | 58.22 | |
| LikelihoodTraining Source=ChatGPT-turbo2026.02 | 12.53 | 58.77 | |
| Entropy-MTraining Source=ChatGPT-turbo2026.02 | 12.27 | 55.94 | |
| DNA-GPT-MTraining Source=ChatGPT-turbo2026.02 | 3.56 | 64.13 | |
| DetectGPT-MTraining Source=ChatGPT-turbo2026.02 | 2.49 | 59.5 | |
| EntropyTraining Source=ChatGPT-turbo2026.02 | 2.04 | 38.08 | |
| DNA-GPTTraining Source=ChatGPT-turbo2026.02 | 1.69 | 35.96 | |
| FastGPT-MTraining Data=ChatGPT, Variant=Markov-informed score calibration2026.02 | 0.6151 | 0.9604 | |
| DNA-GPT-MTraining Data=ChatGPT, Variant=Markov-informed score calibration2026.02 | 0.5933 | 0.9731 | |
| Likelihood-MTraining Data=ChatGPT, Variant=Markov-informed score calibration2026.02 | 0.3591 | 0.7954 | |
| Log-Rank-MTraining Data=ChatGPT, Variant=Markov-informed score calibration2026.02 | 0.3458 | 0.7676 | |
| DNA-GPTTraining Data=ChatGPT2026.02 | 0.1564 | 0.9582 | |
| Log-RankTraining Data=ChatGPT2026.02 | 0.152 | 0.5822 | |
| Entropy-MTraining Data=ChatGPT, Variant=Markov-informed score calibration2026.02 | 0.136 | 0.5654 | |
| LikelihoodTraining Data=ChatGPT2026.02 | 0.1253 | 0.5877 | |
| DetectGPT-MTraining Data=ChatGPT, Variant=Markov-informed score calibration2026.02 | 0.1169 | 0.6947 | |
| FastGPTTraining Source=ChatGPT-turbo2026.02 | 0.04 | 17.55 | |
| FastGPT-MTraining Source=ChatGPT-turbo2026.02 | 0.04 | 9.43 | |
| EntropyTraining Data=ChatGPT2026.02 | 0.0204 | 0.3808 | |
| FastGPTTraining Data=ChatGPT2026.02 | 0.0004 | 0.1755 | |
| DetectGPTTraining Data=ChatGPT2026.02 | 0 | 0.7725 | |
| DetectGPTTraining Source=ChatGPT-turbo2026.02 | 0 | 22.75 |