Machine-generated text detection on WikiHow M4 benchmark (test)
100AUROCRoBERTa
Evaluation results
| Method | Links | ||
|---|---|---|---|
| RoBERTaGenerator=ChatGPT2025.02 | 100 | 50 | |
| RoBERTaGenerator=GPT-42025.02 | 100 | 77.4 | |
| RoBERTaGenerator=Dolly-v22025.02 | 100 | 70.8 | |
| ExaGPTGenerator=ChatGPT2025.02 | 99.5 | 96 | |
| DNA-GPTGenerator=ChatGPT2025.02 | 99.4 | 93.5 | |
| ExaGPTGenerator=GPT-42025.02 | 98.8 | 92.2 | |
| LR-GLTRGenerator=ChatGPT2025.02 | 97.5 | 85.8 | |
| LR-GLTRGenerator=GPT-42025.02 | 94.8 | 77.8 | |
| ExaGPTGenerator=Dolly-v22025.02 | 94.3 | 75.6 | |
| DNA-GPTGenerator=Dolly-v22025.02 | 87.7 | 82.3 | |
| LR-GLTRGenerator=Dolly-v22025.02 | 72.4 | 55 | |
| DNA-GPTGenerator=GPT-42025.02 | 44.6 | 49.9 |