To address these challenges, we propose an unsupervised multi-modal enhancement method based on a high-order adaptive curve, named CLIP-AE. Specifically, we introduce a multi-modal recurrent ...