OFA: Unifying Architectures, Tasks, and Modalities Through a Simple Sequence-to-Sequence Learning Framework

TL;DR

OFA introduces a simple, unified sequence-to-sequence framework for multimodal pretraining that supports diverse tasks and modalities without task-specific customization, achieving state-of-the-art results with relatively small data.

Contribution

It proposes OFA, a task-agnostic, modality-agnostic model that unifies multiple cross-modal and unimodal tasks in a single sequence-to-sequence framework, requiring no extra task-specific layers.

Findings

Achieves new SOTA in various cross-modal tasks.

Performs competitively on unimodal tasks.

Effectively transfers to unseen tasks and domains.

Abstract

In this work, we pursue a unified paradigm for multimodal pretraining to break the scaffolds of complex task/modality-specific customization. We propose OFA, a Task-Agnostic and Modality-Agnostic framework that supports Task Comprehensiveness. OFA unifies a diverse set of cross-modal and unimodal tasks, including image generation, visual grounding, image captioning, image classification, language modeling, etc., in a simple sequence-to-sequence learning framework. OFA follows the instruction-based learning in both pretraining and finetuning stages, requiring no extra task-specific layers for downstream tasks. In comparison with the recent state-of-the-art vision & language models that rely on extremely large cross-modal datasets, OFA is pretrained on only 20M publicly available image-text pairs. Despite its simplicity and relatively small-scale training data, OFA achieves new SOTAs in a series of cross-modal tasks while attaining highly competitive performances on uni-modal tasks. Our further analysis indicates that OFA can also effectively transfer to unseen tasks and unseen domains. Our code and models are publicly available at https://github.com/OFA-Sys/OFA.