Exploring The Spatial Reasoning Ability of Neural Models in Human IQ Tests
This work addresses the gap in measuring neural models' spatial reasoning, which is incremental as it applies existing methods to new datasets for benchmarking.
The study tackled the problem of evaluating neural models' spatial reasoning abilities by constructing datasets based on spatial IQ tests (rotation and shape composition) and testing six baseline models across various complexity levels. The results provided analysis on generalization abilities and how models solve these tests, offering insights into machine vs. human reasoning.
Although neural models have performed impressively well on various tasks such as image recognition and question answering, their reasoning ability has been measured in only few studies. In this work, we focus on spatial reasoning and explore the spatial understanding of neural models. First, we describe the following two spatial reasoning IQ tests: rotation and shape composition. Using well-defined rules, we constructed datasets that consist of various complexity levels. We designed a variety of experiments in terms of generalization, and evaluated six different baseline models on the newly generated datasets. We provide an analysis of the results and factors that affect the generalization abilities of models. Also, we analyze how neural models solve spatial reasoning tests with visual aids. Our findings would provide valuable insights into understanding a machine and the difference between a machine and human.