ai-text-detector-roberta-base

Fine-tuned roberta-base binary classifier for detecting AI-generated vs. human-written text.

• Labels: HUMAN (0), AI (1)
• Max sequence length: 256
• Test Accuracy: 0.9740
• Test F1 (weighted): 0.9740
• Trained on: Kaggle, single GPU

Usage

This model uses a custom classification head on top of roberta-base (not the standard RobertaForSequenceClassification head), so load it with the matching architecture rather than AutoModelForSequenceClassification. See the training notebook for the AITextClassifier class definition.