Table 4.

Medication mention extraction performance on the CMED test set.

Model	Span level		Token level
	P	R	F1	P	R	F1
FABLE	0.874	0.950	0.910	0.961	0.861	0.909
BioBERT	0.952	0.951	0.951	0.960	0.958	0.959 †
Clinical BERT	0.934	0.957	0.945	0.952	0.961	0.956†
BlueBERT	0.925	0.938	0.931	0.951	0.947	0.949†
PubMedBERT	0.925	0.958	0.941	0.941	0.963	0.952†

The highest F1 is bolded.

^†indicates performance significance compared to FABLE. $\left(p<0.05\right)$.