Machine-Generated Text Detection on Reuters ChatGLM
99.29TPR@FPR=1%DNA-GPT-M
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DNA-GPT-MTraining Data=ChatGPT, Variant=Markov-informed score calibration2026.02 | 99.29 | 99.89 | |
| Log-Rank-MTraining Source=ChatGPT-turbo2026.02 | 98.22 | 99.73 | |
| Log-Rank-MTraining Data=ChatGPT, Variant=Markov-informed score calibration2026.02 | 98.13 | 99.75 | |
| Log-RankTraining Data=ChatGPT2026.02 | 97.33 | 99.61 | |
| Log-RankTraining Source=ChatGPT-turbo2026.02 | 97.33 | 99.61 | |
| Likelihood-MTraining Data=ChatGPT, Variant=Markov-informed score calibration2026.02 | 96 | 99.7 | |
| Likelihood-MTraining Source=ChatGPT-turbo2026.02 | 95.96 | 99.72 | |
| LikelihoodTraining Source=ChatGPT-turbo2026.02 | 94.44 | 99.62 | |
| LikelihoodTraining Data=ChatGPT2026.02 | 94.36 | 99.62 | |
| DNA-GPTTraining Data=ChatGPT2026.02 | 83.47 | 99.45 | |
| FastGPT-MTraining Data=ChatGPT, Variant=Markov-informed score calibration2026.02 | 26.67 | 92.12 | |
| DetectGPT-MTraining Data=ChatGPT, Variant=Markov-informed score calibration2026.02 | 23.64 | 80.47 | |
| Entropy-MTraining Data=ChatGPT, Variant=Markov-informed score calibration2026.02 | 9.02 | 69.75 | |
| Entropy-MTraining Source=ChatGPT-turbo2026.02 | 7.69 | 68.88 | |
| DNA-GPT-MTraining Source=ChatGPT-turbo2026.02 | 6.49 | 91.42 | |
| DNA-GPTTraining Source=ChatGPT-turbo2026.02 | 5.51 | 71.24 | |
| DetectGPT-MTraining Source=ChatGPT-turbo2026.02 | 4.04 | 71.41 | |
| FastGPTTraining Data=ChatGPT2026.02 | 3.16 | 66.79 | |
| FastGPTTraining Source=ChatGPT-turbo2026.02 | 3.16 | 66.79 | |
| EntropyTraining Data=ChatGPT2026.02 | 1.69 | 49.04 | |
| EntropyTraining Source=ChatGPT-turbo2026.02 | 1.69 | 49.04 | |
| FastGPT-MTraining Source=ChatGPT-turbo2026.02 | 0.53 | 36.46 | |
| DetectGPTTraining Data=ChatGPT2026.02 | 0 | 61.97 | |
| DetectGPTTraining Source=ChatGPT-turbo2026.02 | 0 | 38.03 |