Machine-generated text detection on Wikipedia M4 benchmark (test)
100AUROCRoBERTa
Evaluation results
| Method | Links | ||
|---|---|---|---|
| RoBERTaGenerator=ChatGPT2025.02 | 100 | 77.1 | |
| RoBERTaGenerator=GPT-42025.02 | 100 | 87.8 | |
| RoBERTaGenerator=Dolly-v22025.02 | 100 | 61.8 | |
| ExaGPTGenerator=ChatGPT2025.02 | 98.6 | 92.3 | |
| ExaGPTGenerator=GPT-42025.02 | 98.3 | 87.3 | |
| LR-GLTRGenerator=GPT-42025.02 | 97.8 | 85.7 | |
| LR-GLTRGenerator=ChatGPT2025.02 | 95 | 60 | |
| ExaGPTGenerator=Dolly-v22025.02 | 85.8 | 63.8 | |
| DNA-GPTGenerator=ChatGPT2025.02 | 84.8 | 49.4 | |
| LR-GLTRGenerator=Dolly-v22025.02 | 79.7 | 57.7 | |
| DNA-GPTGenerator=Dolly-v22025.02 | 68 | 61.5 | |
| DNA-GPTGenerator=GPT-42025.02 | 40.3 | 48.1 |