Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer

TL;DR

This paper introduces a unified text-to-text framework for transfer learning in NLP, systematically comparing various approaches and achieving state-of-the-art results across multiple language understanding tasks.

Contribution

It presents a comprehensive, unified approach to transfer learning in NLP, including a new data set and models, and demonstrates improved performance on diverse benchmarks.

Findings

Unified text-to-text framework outperforms previous methods

Achieved state-of-the-art results on multiple NLP benchmarks

Released new data set, models, and code for future research

Abstract

Transfer learning, where a model is first pre-trained on a data-rich task before being fine-tuned on a downstream task, has emerged as a powerful technique in natural language processing (NLP). The effectiveness of transfer learning has given rise to a diversity of approaches, methodology, and practice. In this paper, we explore the landscape of transfer learning techniques for NLP by introducing a unified framework that converts all text-based language problems into a text-to-text format. Our systematic study compares pre-training objectives, architectures, unlabeled data sets, transfer approaches, and other factors on dozens of language understanding tasks. By combining the insights from our exploration with scale and our new ``Colossal Clean Crawled Corpus'', we achieve state-of-the-art results on many benchmarks covering summarization, question answering, text classification, and more. To facilitate future work on transfer learning for NLP, we release our data set, pre-trained models, and code.