Machine-Generated Text Detection on Reuters ChatGPT-turbo
97.51TPR@FPR=1%DNA-GPT-M
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DNA-GPT-MTraining Data=ChatGPT, Variant=Markov-informed score calibration2026.02 | 97.51 | 0.9982 | |
| Log-Rank-MTraining Data=ChatGPT, Variant=Markov-informed score calibration2026.02 | 94.53 | 0.9903 | |
| Log-Rank-MTraining Source=ChatGPT-turbo2026.02 | 94 | 0.9899 | |
| Likelihood-MTraining Source=ChatGPT-turbo2026.02 | 93.91 | 0.9896 | |
| Likelihood-MTraining Data=ChatGPT, Variant=Markov-informed score calibration2026.02 | 93.24 | 0.9894 | |
| Log-RankTraining Data=ChatGPT2026.02 | 93.2 | 0.9867 | |
| Log-RankTraining Source=ChatGPT-turbo2026.02 | 93.2 | 0.9867 | |
| LikelihoodTraining Data=ChatGPT2026.02 | 90.31 | 0.9842 | |
| LikelihoodTraining Source=ChatGPT-turbo2026.02 | 90.27 | 0.9842 | |
| DNA-GPTTraining Data=ChatGPT2026.02 | 82.22 | 0.9938 | |
| DetectGPT-MTraining Data=ChatGPT, Variant=Markov-informed score calibration2026.02 | 71.24 | 0.972 | |
| DetectGPT-MTraining Source=ChatGPT-turbo2026.02 | 60.67 | 0.9653 | |
| Entropy-MTraining Data=ChatGPT, Variant=Markov-informed score calibration2026.02 | 33.29 | 0.9046 | |
| Entropy-MTraining Source=ChatGPT-turbo2026.02 | 31.69 | 0.8991 | |
| FastGPTTraining Data=ChatGPT2026.02 | 23.51 | 0.8871 | |
| FastGPTTraining Source=ChatGPT-turbo2026.02 | 23.51 | 0.8871 | |
| EntropyTraining Source=ChatGPT-turbo2026.02 | 9.6 | 0.8245 | |
| EntropyTraining Data=ChatGPT2026.02 | 9.56 | 0.8245 | |
| DetectGPTTraining Source=ChatGPT-turbo2026.02 | 9.24 | 0.7958 | |
| FastGPT-MTraining Source=ChatGPT-turbo2026.02 | 7.38 | 0.7628 | |
| DNA-GPT-MTraining Source=ChatGPT-turbo2026.02 | 6.67 | 0.8937 | |
| DNA-GPTTraining Source=ChatGPT-turbo2026.02 | 5.24 | 0.7063 | |
| FastGPT-MTraining Data=ChatGPT, Variant=Markov-informed score calibration2026.02 | 1.29 | 0.5604 | |
| DetectGPTTraining Data=ChatGPT2026.02 | 0.04 | 0.2042 |