Table 5.

Performance of PICO sentence classification by BERT with different pre-trained weights on the validation set

	F1	Recall	Precision
BERT-base	86.6	87.7	87.2
BioBERT	87.7	89.6	88.1
PubMedBERT-abs	89.3	91.3	89.1
PubMedBERT-full	85.8	89.3	84.6