Transparency by Design: Closing the Gap Between Performance and Interpretability in Visual Reasoning

TL;DR

This paper introduces visual reasoning primitives that enhance interpretability and achieve state-of-the-art accuracy in visual question answering, bridging the gap between model transparency and high performance.

Contribution

The authors propose a set of composable visual reasoning primitives that improve interpretability while maintaining high accuracy in complex visual reasoning tasks.

Findings

Achieved 99.1% accuracy on CLEVR dataset.

Significantly improved generalization on CoGenT with over 20 percentage points.

Enabled diagnosis of model strengths and weaknesses through primitive outputs.

Abstract

Visual question answering requires high-order reasoning about an image, which is a fundamental capability needed by machine systems to follow complex directives. Recently, modular networks have been shown to be an effective framework for performing visual reasoning tasks. While modular networks were initially designed with a degree of model transparency, their performance on complex visual reasoning benchmarks was lacking. Current state-of-the-art approaches do not provide an effective mechanism for understanding the reasoning process. In this paper, we close the performance gap between interpretable models and state-of-the-art visual reasoning methods. We propose a set of visual-reasoning primitives which, when composed, manifest as a model capable of performing complex reasoning tasks in an explicitly-interpretable manner. The fidelity and interpretability of the primitives' outputs enable an unparalleled ability to diagnose the strengths and weaknesses of the resulting model. Critically, we show that these primitives are highly performant, achieving state-of-the-art accuracy of 99.1% on the CLEVR dataset. We also show that our model is able to effectively learn generalized representations when provided a small amount of data containing novel object attributes. Using the CoGenT generalization task, we show more than a 20 percentage point improvement over the current state of the art.