BERT-base-uncased — AI-Generated Text Detector (HC3)
Fine-tuned bert-base-uncased for binary classification of human-written vs AI-generated text.
Trained as part of an MSc AI dissertation at the University of the West of Scotland (UWS), 2025/26.
Supervisor: Dr Tahir Mahmood | Student: Abdul Hannaan Mohammed (B00409227)
Performance
| Condition | F1 | Accuracy | Recall |
|---|
| Clean HC3 (test) | 0.9845 | 0.9895 | 0.9997 |
| Pegasus paraphrase attack | — | — | ASR 1.8% |
| QuillBot-style attack | — | — | ASR 12.2% |
| ChatGPT rewrite attack | — | — | ASR 3.2% |
| M4 cross-dataset | 0.5999 | 0.7045 | — |
Training
- Base model: (Devlin et al., 2019)
- Dataset: HC3 (Hello-SimpleAI/HC3) — 55,156 training samples, 70/15/15 split
- Epochs: 3 | LR: 2e-5 | Batch: 16
- Hardware: Local GPU (6 GB VRAM)
Usage
Citation
