Class-Incremental Learning with CLIP: Adaptive Representation Adjustment and Parameter Fusion