#delayed-labels

Monitoring Models When Ground Truth Is Late or Never Arrives

Delayed labels are the defining hard problem of ML monitoring. Strategies for the blind period between prediction and ground truth — proxy signals