RevColV2: Exploring Disentangled Representations in Masked Image Modeling | Read Paper on Bytez