We propose a vision-language multimodal pretrained model called multiway-fuzzy-experts bidirectional retention network (VL-MFER), which is designed to effectively interpret model decision processes ...