新基准数据集

新基准数据集是专为公平图学习方法评估设计的一系列数据集，包括合成、半合成和真实世界数据集。这些数据集精心设计，包含关键的图结构和偏差信息，旨在测试模型在利用图结构提升预测准确性的同时，有效处理和减轻数据中固有的偏差。数据集的创建过程考虑了图结构的效用和通过图结构放大偏差的可能性，确保只有既能有效利用图信息又能中和嵌入偏差的模型才能脱颖而出。这些数据集为开发和评估公平图学习算法提供了挑战性的基准，推动了该领域的发展。

This novel benchmark dataset suite is specifically designed for evaluating fair graph learning methods, covering synthetic, semi-synthetic, and real-world datasets. These datasets are carefully crafted to include key graph structures and bias information, with the goal of testing models' capacity to improve prediction accuracy by leveraging graph structures while effectively handling and mitigating inherent biases in the data. The development process of these datasets takes into account both the utility of graph structures and the possibility of bias amplification through graph structures, ensuring that only models that can effectively utilize graph information and neutralize embedded biases can stand out. These datasets provide challenging benchmarks for the development and evaluation of fair graph learning algorithms, promoting the advancement of this research field.