The model is structured around an Encoder-Decoder framework ... integrating a novel multi-head cross-modal attention mechanism and a Region-Specific Dynamics (RSD) layer. This layer is ...