Table 6. Overall performance metrics from Scenario 2: AI-assisted abstracts.
| Metric | GPTZero (%) | ZeroGPT (%) | DetectGPT (%) |
|---|---|---|---|
| Under-Detection Rate (UDR) | 3.47 | 58.33 | 44.44 |
| • By author categories | |||
| • Native | 4.17 | 52.78 | 41.67 |
| • Non-Native | 2.78 | 63.89 | 47.22 |
| • By disciplines | |||
| • Technology & Engineering | 4.17 | 75.00 | 43.75 |
| • Social Sciences | 4.17 | 66.67 | 43.75 |
| • Interdisciplinary | 2.08 | 33.33 | 45.83 |
| • By LLMs | |||
| • ChatGPT o1 | 5.56 | 76.39 | 69.44 |
| • Gemini 2.0 Pro | 1.39 | 40.28 | 19.44 |
| Over-Detection Rate (ODR) | 18.06 | 2.08 | 34.03 |
| • By author categories | |||
| • Native | 11.11 | 1.39 | 33.33 |
| • Non-Native | 25.00 | 2.78 | 34.72 |
| • By disciplines | |||
| • Technology & Engineering | 16.67 | 0.00 | 41.67 |
| • Social Sciences | 12.50 | 6.25 | 29.17 |
| • Interdisciplinary | 25.00 | 0.00 | 31.25 |
| • By LLMs | |||
| • ChatGPT o1 | 8.33 | 0.00 | 22.22 |
| • Gemini 2.0 Pro | 27.78 | 4.17 | 45.83 |