Zero-shot incremental learning aims to enable a model to generalize to new classes without forgetting previously learned classes. However, the semantic gap between old and new sample classes can lead ...