Visual Set Program Synthesizer

TL;DR

This paper introduces a visual program synthesis approach for complex set-based reasoning in visual question answering, outperforming existing methods by generating explicit symbolic programs for transparent and accurate reasoning.

Contribution

It proposes a novel framework that treats visual reasoning as program synthesis, along with a new benchmark for evaluating set-based visual reasoning tasks.

Findings

Significantly outperforms state-of-the-art baselines on complex reasoning tasks.

Produces more systematic and transparent reasoning behavior.

Improves answer accuracy in visual question answering.

Abstract

A user pointing their phone at a supermarket shelf and asking "Which soda has the least sugar?" poses a difficult challenge for current visual Al assistants. Such queries require not only object recognition, but explicit set-based reasoning such as filtering, comparison, and aggregation. Standard endto-end MLLMs often fail at these tasks because they lack an explicit mechanism for compositional logic. We propose treating visual reasoning as Visual Program Synthesis, where the model first generates a symbolic program that is executed by a separate engine grounded in visual scenes. We also introduce Set-VQA, a new benchmark designed specifically for evaluating set-based visual reasoning. Experiments show that our approach significantly outperforms state-of-the-art baselines on complex reasoning tasks, producing more systematic and transparent behavior while substantially improving answer accuracy. These results demonstrate that program-driven reasoning provides a principled alternative to black-box visual-language inference.