KAT: A Knowledge Augmented Transformer for Vision-and-Language

TL;DR

This paper introduces KAT, a transformer model that effectively integrates explicit knowledge with implicit information for improved reasoning in vision-and-language tasks, achieving state-of-the-art results and enhanced interpretability.

Contribution

KAT is a novel end-to-end model that combines explicit knowledge retrieval with implicit reasoning in a unified architecture for multimodal understanding.

Findings

Achieves +6 points absolute improvement on OK-VQA.

Enhances interpretability of model predictions.

Successfully integrates explicit and implicit knowledge during reasoning.

Abstract

The primary focus of recent work with largescale transformers has been on optimizing the amount of information packed into the model's parameters. In this work, we ask a different question: Can multimodal transformers leverage explicit knowledge in their reasoning? Existing, primarily unimodal, methods have explored approaches under the paradigm of knowledge retrieval followed by answer prediction, but leave open questions about the quality and relevance of the retrieved knowledge used, and how the reasoning processes over implicit and explicit knowledge should be integrated. To address these challenges, we propose a novel model - Knowledge Augmented Transformer (KAT) - which achieves a strong state-of-the-art result (+6 points absolute) on the open-domain multimodal task of OK-VQA. Our approach integrates implicit and explicit knowledge in an end to end encoder-decoder architecture, while still jointly reasoning over both knowledge sources during answer generation. An additional benefit of explicit knowledge integration is seen in improved interpretability of model predictions in our analysis.