VulDistilBERT: A CPS Vulnerability Severity Prediction Method Based on Distillation Model

Table 6

Results of evaluating the raw dataset with the pretrained model.

Models	Best epoch	Loss	Accuracy	Precision	Recall	F1 scores
BERT	26	1.038	0.54	0.42	0.33	0.32
XLNET	48	0.978	0.56	0.41	0.36	0.35
ROBERTA	74	1.043	0.54	0.34	0.32	0.29
DistilBERT	24	1.015	0.56	0.42	0.35	0.34