AI-generated text detection (Three-way classification) on FAID Unseen Domain+Generator
88.2AccuracyVanilla RoBERTa baseline
Evaluation results
| Method | Links | |
|---|---|---|
| Vanilla RoBERTa baselinefine-tuning on unseen split validation set=true2026.07 | 88.2 | |
| Vanilla RoBERTa baselinefine-tuning on unseen split validation set=false2026.07 | 67.7 | |
| FAID2026.07 | 66.6 | |
| LLM-DetectAIve2026.07 | 62.9 |