Simple Local Attentions Remain Competitive for Long-Context Tasks

Wenhan Xiong, Barlas Oguz, Anchit Gupta, Xilun Chen, Diana Liskovich, Omer Levy, Scott Yih, Yashar Mehdad


Abstract
Many NLP tasks require processing long contexts beyond the length limit of pretrained models. In order to scale these models to longer text sequences, many efficient long-range attention variants have been proposed. Despite the abundance of research along this direction, it is still difficult to gauge the relative effectiveness of these models in practical use cases, e.g., if we apply these models following the pretrain-and-finetune paradigm. In this work, we aim to conduct a thorough analysis of these emerging models with large-scale and controlled experiments. For each attention variant, we pretrain large-size models using the same long-doc corpus and then finetune these models for real-world long-context tasks. Our findings reveal pitfalls of an existing widely-used long-range benchmark and show none of the tested efficient attentions can beat a simple local window attention under standard pretraining paradigms. Further analysis on local attention variants suggests that even the commonly used attention-window overlap is not necessary to achieve good downstream results — using disjoint local attentions, we are able to build a simpler and more efficient long-doc QA model that matches the performance of Longformer with half of its pretraining compute.
Anthology ID:
2022.naacl-main.144
Volume:
Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies
Month:
July
Year:
2022
Address:
Seattle, United States
Venue:
NAACL
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
1975–1986
Language:
URL:
https://aclanthology.org/2022.naacl-main.144
DOI:
10.18653/v1/2022.naacl-main.144
Bibkey:
Cite (ACL):
Wenhan Xiong, Barlas Oguz, Anchit Gupta, Xilun Chen, Diana Liskovich, Omer Levy, Scott Yih, and Yashar Mehdad. 2022. Simple Local Attentions Remain Competitive for Long-Context Tasks. In Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 1975–1986, Seattle, United States. Association for Computational Linguistics.
Cite (Informal):
Simple Local Attentions Remain Competitive for Long-Context Tasks (Xiong et al., NAACL 2022)
Copy Citation:
PDF:
https://aclanthology.org/2022.naacl-main.144.pdf
Code
 pytorch/fairseq
Data
BookCorpusLRA