Tag
The paper addresses Incomplete Multimodal Sentiment Analysis with Unseen Modality Combinations and proposes the ContrastiveMixedPromptLearning (CMPL) model, which uses label-guided contrastive learning and modality-combination prompts to enhance generalization, achieving over 5% accuracy improvement on benchmark datasets.