LR-EXP

该数据集是一个旨在衡量模型对于非自同构链接分配不同输出的能力的综合数据集，包含了1400个图，这些图被设计用来孤立并探究表达性。该数据集包括那些端点节点具有相同1-WL颜色的非自同构链接，这使得区分任务颇具挑战性。其规模被定义为小型，任务则是评估图神经网络在链接级别的表达性能力。

This comprehensive dataset is designed to evaluate a model's ability to assign distinct outputs to non-isomorphic links. It consists of 1400 graphs that are intentionally engineered to isolate and probe model expressive power. The dataset includes non-isomorphic links whose endpoint nodes share identical 1-dimensional Weisfeiler-Lehman (1-WL) colors, which renders the discrimination task highly challenging. It is categorized as a small-scale dataset, and the core task is to assess the link-level expressive capabilities of graph neural networks.