Binary Cross-Entropy loss, a training objective commonly used for relevance scoring tasks where the model learns to predict whether a query-document pair is relevant or not.