Unraveling and Mitigating Endogenous Task-oriented Spurious Correlations in Ego-graphs via Automated Counterfactual Contrastive Learning

Tianqianjin Lin, Yangyang Kang, Zhuoren Jiang, Kaisong Song, Kun Kuang, Changlong Sun, Cui Huang, Xiaozhong Liu

Abstract: Highlights•Uncover an endogenous mechanism of spurious correlations in node-level graph tasks.•Propose a novel counterfactual contrastive learning framework to mitigate the issue.•Experiments on 13 real-world datasets validate the superiority of the framework.