GraphLog documentation

GraphLog is a multi-purpose, multi-relational graph dataset built using rules grounded in first-order logic. GraphLog can be used to benchmark existing Graph Neural Network (GNN) family of models on relation prediction task.

GraphLog image


You can read more about GraphLog at our blog post.


GraphLog is introduced in the paper “Evaluating Logical Generalization in Graph Neural Networks”. If you find our dataset useful, consider citing our work.

    Author = {Koustuv Sinha and Shagun Sodhani and Joelle Pineau and William L. Hamilton},
    Title = {Evaluating Logical Generalization in Graph Neural Networks},
    Year = {2020},
    arxiv = {}