DS-AL proposes a novel approach for exemplar-free class-incremental learning, achieving competitive performance across various datasets.
