ShapeWorld - A new test methodology for multimodal language understanding

TL;DR

ShapeWorld presents a new framework for evaluating multimodal language understanding by automatically generating controlled artificial data, enabling detailed assessment of models' generalization and reasoning abilities.

Contribution

It introduces a novel, controllable data generation methodology for testing multimodal models, emphasizing their generalization capabilities beyond existing benchmarks.

Findings

Models show varying generalization abilities across tasks

Framework provides detailed insights into model strengths and weaknesses

Open-sourcing encourages further research in multimodal understanding

Abstract

We introduce a novel framework for evaluating multimodal deep learning models with respect to their language understanding and generalization abilities. In this approach, artificial data is automatically generated according to the experimenter's specifications. The content of the data, both during training and evaluation, can be controlled in detail, which enables tasks to be created that require true generalization abilities, in particular the combination of previously introduced concepts in novel ways. We demonstrate the potential of our methodology by evaluating various visual question answering models on four different tasks, and show how our framework gives us detailed insights into their capabilities and limitations. By open-sourcing our framework, we hope to stimulate progress in the field of multimodal language understanding.