Adapting Pretrained Text-to-Text Models for Long Text Sequences

TL;DR

This paper explores adapting pretrained text-to-text models for long sequences by modifying architecture, training objectives, and data, resulting in state-of-the-art performance on long-text tasks.

Contribution

It introduces a novel approach combining pooling-augmented attention and span prediction pretraining, improving long-text model performance from existing short-text models.

Findings

Achieved state-of-the-art results on five long-text summarization datasets.

Demonstrated the effectiveness of concatenating short documents for pretraining.

Replaced full attention with pooling-augmented blockwise attention for efficiency.

Abstract

We present an empirical study of adapting an existing pretrained text-to-text model for long-sequence inputs. Through a comprehensive study along three axes of the pretraining pipeline -- model architecture, optimization objective, and pretraining corpus, we propose an effective recipe to build long-context models from existing short-context models. Specifically, we replace the full attention in transformers with pooling-augmented blockwise attention, and pretrain the model with a masked-span prediction task with spans of varying length. In terms of the pretraining corpus, we find that using randomly concatenated short-documents from a large open-domain corpus results in better performance than using existing long document corpora which are typically limited in their domain coverage. With these findings, we build a long-context model that achieves competitive performance on long-text QA tasks and establishes the new state of the art on five long-text summarization datasets, often outperforming previous methods with larger model sizes. Our code has been released at https://github.com/facebookresearch/bart_ls.