Adaptive batch selection improves multi-label classification by focusing on hard and imbalanced samples.
