Visually-Augmented Language Modeling

TL;DR

VaLM is a novel pre-training framework that enhances language models with relevant visual information by retrieving images and fusing visual knowledge, significantly improving performance on visual knowledge reasoning tasks.

Contribution

Introduces VaLM, a new method for visually augmenting language models through image retrieval and multimodal fusion, enabling better reasoning on visual knowledge tasks.

Findings

VaLM outperforms language-only models on visual knowledge reasoning.

VaLM surpasses existing vision-language baselines in object commonsense tasks.

Visual augmentation improves understanding of colors, sizes, and shapes.

Abstract

Human language is grounded on multimodal knowledge including visual knowledge like colors, sizes, and shapes. However, current large-scale pre-trained language models rely on text-only self-supervised training with massive text data, which precludes them from utilizing relevant visual information when necessary. To address this, we propose a novel pre-training framework, named VaLM, to Visually-augment text tokens with retrieved relevant images for Language Modeling. Specifically, VaLM builds on a novel latent text-image alignment method via an image retrieval module to fetch corresponding images given a textual context. With the visually-augmented context, VaLM uses a visual knowledge fusion layer to enable multimodal grounded language modeling by attending to both text context and visual knowledge in images. We evaluate VaLM on various visual knowledge-intensive commonsense reasoning tasks, which require visual information to excel. The experimental results illustrate that VaLM outperforms all strong language-only and vision-language baselines with substantial gains in reasoning object commonsense including color, size, and shape. Our code is available at https://github.com/Victorwz/VaLM.