Measuring abstract reasoning in neural networks

TL;DR

This paper introduces a dataset and challenge inspired by IQ tests to evaluate and improve the abstract reasoning capabilities of neural networks, revealing their current limitations and potential for enhancement.

Contribution

It presents a new dataset and challenge for probing neural network reasoning, along with a novel architecture that outperforms standard models in abstract reasoning tasks.

Findings

Standard models like ResNets perform poorly on reasoning tasks

The proposed architecture shows significant improvement in reasoning ability

Training models to predict symbolic explanations enhances generalisation

Abstract

Whether neural networks can learn abstract reasoning or whether they merely rely on superficial statistics is a topic of recent debate. Here, we propose a dataset and challenge designed to probe abstract reasoning, inspired by a well-known human IQ test. To succeed at this challenge, models must cope with various generalisation `regimes' in which the training and test data differ in clearly-defined ways. We show that popular models such as ResNets perform poorly, even when the training and test sets differ only minimally, and we present a novel architecture, with a structure designed to encourage reasoning, that does significantly better. When we vary the way in which the test questions and training data differ, we find that our model is notably proficient at certain forms of generalisation, but notably weak at others. We further show that the model's ability to generalise improves markedly if it is trained to predict symbolic explanations for its answers. Altogether, we introduce and explore ways to both measure and induce stronger abstract reasoning in neural networks. Our freely-available dataset should motivate further progress in this direction.