# Evaluating Logical Generalization in Graph Neural Networks

arXiv: https://arxiv.org/abs/2003.06560 · HTML: https://mlsanity.com/paper/2003.06560/

## Key facts

- Cited by 12 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 14 Mar 2020.

## Built on by (most cited first, 12 of 12)

- [Out-Of-Distribution Generalization on Graphs: A Survey](https://mlsanity.com/paper/2202.07987/) (2022)
- [Modular Lifelong Reinforcement Learning via Neural Composition](https://mlsanity.com/paper/2207.00429/) (2022)
- [The Neural Race Reduction: Dynamics of Abstraction in Gated Networks](https://mlsanity.com/paper/2207.10430/) (2022)
- [How to Reuse and Compose Knowledge for a Lifetime of Tasks: A Survey on Continual Learning and Functional Composition](https://mlsanity.com/paper/2207.07730/) (2022)
- [CompoSuite: A Compositional Reinforcement Learning Benchmark](https://mlsanity.com/paper/2207.04136/) (2022)
- [Neural Compositional Rule Learning for Knowledge Graph Reasoning](https://mlsanity.com/paper/2303.03581/) (2023)
- [Knowledge Hypergraph Embedding Meets Relational Algebra](https://mlsanity.com/paper/2102.09557/) (2021)
- [Relational Graph Convolutional Networks: A Closer Look](https://mlsanity.com/paper/2107.10015/) (2021)
- [R5: Rule Discovery with Reinforced and Recurrent Relational Reasoning](https://mlsanity.com/paper/2205.06454/) (2022)
- [State of the Art and Potentialities of Graph-level Learning](https://mlsanity.com/paper/2301.05860/) (2023)
- [Systematic Relational Reasoning With Epistemic Graph Neural Networks](https://mlsanity.com/paper/2407.17396/) (2024)
- [Relational Weight Priors in Neural Networks for Abstract Pattern Learning and Language Modelling](https://mlsanity.com/paper/2103.06198/) (2021)

## Abstract (arXiv)

Recent research has highlighted the role of relational inductive biases in building learning agents that can generalize and reason in a compositional manner. However, while relational learning algorithms such as graph neural networks (GNNs) show promise, we do not understand how effectively these approaches can adapt to new tasks. In this work, we study the task of logical generalization using GNNs by designing a benchmark suite grounded in first-order logic. Our benchmark suite, GraphLog, requires that learning algorithms perform rule induction in different synthetic logics, represented as knowledge graphs. GraphLog consists of relation prediction tasks on 57 distinct logical domains. We use GraphLog to evaluate GNNs in three different setups: single-task supervised learning, multi-task pretraining, and continual learning. Unlike previous benchmarks, our approach allows us to precisely control the logical relationship between the different tasks. We find that the ability for models to generalize and adapt is strongly determined by the diversity of the logical rules they encounter during training, and our results highlight new challenges for the design of GNN models. We publicly release the dataset and code used to generate and interact with the dataset at https://www.cs.mcgill.ca/~ksinha4/graphlog.