Table 1.

Paratope prediction performance by different metrics of the fine-tuned protein language model on the Development (Dev.) and Test sets via 10-fold cross-validation

Set	ROC AUC	F1	MCC
Dev.	$0.895\pm 0.010$	$0.702\pm 0.016$	$0.589\pm 0.021$
Test	$0.890\pm 0.007$	$0.686\pm 0.015$	$0.569\pm 0.019$

Average $\pm$ standard deviations among validation sets shown.