Adapting Pretrained Text-to-Text Models for Long Text Sequences
This work addresses the challenge of handling long-context inputs in NLP, which is important for applications like summarization and QA, but it is incremental as it builds on existing short-context models.
The authors tackled the problem of adapting pretrained text-to-text models for long text sequences by proposing a recipe involving model architecture, optimization, and corpus changes, resulting in a model that achieves competitive performance on long-text QA and sets new state-of-the-art results on five long-text summarization datasets.
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.