A Benchmark for Compositional Visual Reasoning

TL;DR

This paper introduces a new benchmark, CVR, to evaluate and improve the data efficiency of AI systems in compositional visual reasoning, inspired by human learning and reasoning tests.

Contribution

The paper presents a novel benchmark for compositional visual reasoning, including a scalable method for creating abstract rule compositions and datasets, to assess and enhance AI sample efficiency.

Findings

Convolutional architectures outperform transformers in most data regimes.

All models are less data efficient than humans even with self-supervised learning.

The benchmark measures generalization, transfer, and compositionality in visual reasoning.

Abstract

A fundamental component of human vision is our ability to parse complex visual scenes and judge the relations between their constituent objects. AI benchmarks for visual reasoning have driven rapid progress in recent years with state-of-the-art systems now reaching human accuracy on some of these benchmarks. Yet, a major gap remains in terms of the sample efficiency with which humans and AI systems learn new visual reasoning tasks. Humans' remarkable efficiency at learning has been at least partially attributed to their ability to harness compositionality -- such that they can efficiently take advantage of previously gained knowledge when learning new tasks. Here, we introduce a novel visual reasoning benchmark, Compositional Visual Relations (CVR), to drive progress towards the development of more data-efficient learning algorithms. We take inspiration from fluidic intelligence and non-verbal reasoning tests and describe a novel method for creating compositions of abstract rules and associated image datasets at scale. Our proposed benchmark includes measures of sample efficiency, generalization and transfer across task rules, as well as the ability to leverage compositionality. We systematically evaluate modern neural architectures and find that, surprisingly, convolutional architectures surpass transformer-based architectures across all performance measures in most data regimes. However, all computational models are a lot less data efficient compared to humans even after learning informative visual representations using self-supervision. Overall, we hope that our challenge will spur interest in the development of neural architectures that can learn to harness compositionality toward more efficient learning.