MagXCL: enhanced multimodal adaptation gate and cross-modal contrastive learning for multimodal sentiment analysis
MagXCL, a unified framework designed to improve multimodal integration through more effective interaction between verbal and non-verbal modalities, is proposed, demonstrating the effectiveness of combining AMag with CrossCL to produce more accurate and robust multimodal sentiment predictions.