Precision, recall, and F1 score on the test set as functions of the confidence threshold. The F1 score is defined as F1 = 2PR/(P + R). The vertical dashed line marks confidence = 0.902, which maximizes the F1 score on the test set and was adopted for wide-field inference. The inset shows a zoomed-in view around the adopted threshold, with an expanded vertical scale to make the local trends in precision, recall, and F1 more visible.