AI-generated text detection on MIRAGE six task subsets
0.963AUROCVanilla(large)
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Vanilla(large)training=fine-tuned on 500 paired GPT-3.5 polish examples, seeds=32026.07 | 0.963 | 90.9 | 0.82 | 85.9 | |
| DetectAnyLLM2026.07 | 0.934 | 88.1 | 0.764 | 77.3 | |
| Vanilla(base)training=fine-tuned on 500 paired GPT-3.5 polish examples, seeds=32026.07 | 0.934 | 87.1 | 0.746 | 76.7 | |
| RoBERTa-largesource=OpenAI 20192026.07 | 0.515 | 52.6 | 0.053 | 8 | |
| RoBERTa-basesource=OpenAI 20192026.07 | 0.51 | 51.6 | 0.044 | 7.4 |