Recent deep learning advancements have improved spectral and spatial details in pan-sharpening. Yet, neural networks still struggle with frequency bias and adapting to diverse remote sensing content.