Evaluated on COVID-19 Twitter context (200K samples) ยท Fine-Tuned BERT vs BiLSTM
Or try one of these test cases:
| Model Architecture | Test Accuracy | F1-Score | ROC-AUC | Parameters |
|---|---|---|---|---|
| BERT-base-uncased (Fine-Tuned) | 89.1% | 0.891 | 0.950 | 109M |
| DistilBERT Baseline | 87.3% | 0.874 | 0.940 | 66M |
| BiLSTM + Embeddings | 78.8% | 0.800 | 0.870 | 2M |