Table 8 Comparison with baseline models.
From: Clickbait detection in news headlines using RoBERTa-Large language model and deep embeddings
Model | Type | Parameters | Accuracy (%) | F1-score (%) | Training time (per epoch) | Inference speed | Resource demand |
|---|---|---|---|---|---|---|---|
SVM | ML (Linear) | – | 92 | 92 | ~ 15s | Fast | Low (CPU) |
LSTM | RNN-based DL | 10 M | 95 | 96 | ~ 60s | Medium | Moderate (GPU/CPU) |
BiLSTM | RNN-based DL | 18 M | 95 | 96 | ~ 75s | Medium | Moderate |
T5 | Seq2Seq Transformer | 60 M | 87 | 88 | ~ 85s | Slower | High (GPU/RAM) |
DistilBERT | Transformer (distilled) | 66 M | 91 | 92 | ~ 40s | Fast | Moderate (GPU) |
RoBERTa-Large | Transformer (large) | 355 M | 97 | 98 | ~ 120s | Slower | Very High (GPU ~ 12GB) |
