Recent years have witnessed fast developments of graph neural networks (GNNs)
that have benefited myriads of graph analytic tasks and applications. In
general, most GNNs depend on the homophily assumption that nodes belonging to
the same class are more likely to be connected. However, as a ubiquitous graph
property in numerous real-world scenarios, heterophily, i.e., nodes with
different labels tend to be linked, significantly limits the performance of
tailor-made homophilic GNNs. Hence, \textit{GNNs for heterophilic graphs} are
gaining increasing attention in this community. To the best of our knowledge,
in this paper, we provide a comprehensive review of GNNs for heterophilic
graphs for the first time. Specifically, we propose a systematic taxonomy that
essentially governs existing heterophilic GNN models, along with a general
summary and detailed analysis. Furthermore, we summarize the mainstream
heterophilic graph benchmarks to facilitate robust and fair evaluations. In the
end, we point out the potential directions to advance and stimulate future
research and applications on heterophilic graphs.

本文首次全面回顾了异态图神经网络，提出了系统的分类法，分析了主要异态图神经网络模型，总结了主流异态图基准，并指出了未来研究和应用方向。