Back to Search Start Over

Difference-Masking: Choosing What to Mask in Continued Pretraining

Authors :
Wilf, Alex
Akter, Syeda Nahida
Mathur, Leena
Liang, Paul Pu
Mathew, Sheryl
Shou, Mengrou
Nyberg, Eric
Morency, Louis-Philippe
Publication Year :
2023

Abstract

The self-supervised objective of masking-and-predicting has led to promising performance gains on a variety of downstream tasks. However, while most approaches randomly mask tokens, there is strong intuition that deciding what to mask can substantially improve learning outcomes. We investigate this in continued pretraining setting in which pretrained models continue to pretrain on domain-specific data before performing some downstream task. We introduce Difference-Masking, a masking strategy that automatically chooses what to mask during continued pretraining by considering what makes a task domain different from the pretraining domain. Empirically, we find that Difference-Masking outperforms baselines on continued pretraining settings across four diverse language-only and multimodal video tasks.

Details

Database :
arXiv
Publication Type :
Report
Accession number :
edsarx.2305.14577
Document Type :
Working Paper