Yuhai Zhao

We propose a self-supervised coloring learning framework for heterophilic graph representation, which effectively captures both local and global structures without relying on delicate augmentation strategies.