KM-BART: Knowledge Enhanced Multimodal BART for Visual Commonsense Generation

TL;DR

KM-BART is a novel multimodal transformer model that integrates external commonsense knowledge to enhance visual commonsense generation, achieving state-of-the-art results through specialized pretraining tasks.

Contribution

The paper introduces KM-BART, the first dedicated model for visual commonsense generation that incorporates external knowledge via novel pretraining tasks.

Findings

Achieves state-of-the-art performance on VCG task.

Effective integration of external commonsense knowledge.

Pretraining with KCG boosts model accuracy.

Abstract

We present Knowledge Enhanced Multimodal BART (KM-BART), which is a Transformer-based sequence-to-sequence model capable of reasoning about commonsense knowledge from multimodal inputs of images and texts. We adapt the generative BART architecture to a multimodal model with visual and textual inputs. We further develop novel pretraining tasks to improve the model performance on the Visual Commonsense Generation (VCG) task. In particular, our pretraining task of Knowledge-based Commonsense Generation (KCG) boosts model performance on the VCG task by leveraging commonsense knowledge from a large language model pretrained on external commonsense knowledge graphs. To the best of our knowledge, we are the first to propose a dedicated task for improving model performance on the VCG task. Experimental results show that our model reaches state-of-the-art performance on the VCG task by applying these novel pretraining tasks.