HiddenCut: Simple Data Augmentation for Natural Language Understanding with Better Generalization

TL;DR

HiddenCut is a simple data augmentation method that improves the generalization of NLP models by dropping contiguous spans in hidden space during training, outperforming existing methods on benchmarks and out-of-distribution data.

Contribution

We introduce HiddenCut, a novel data augmentation technique that enhances model regularization and generalization in NLP by strategically dropping hidden representations.

Findings

Outperforms state-of-the-art augmentation methods on GLUE benchmark

Improves out-of-distribution generalization

Enhances robustness to challenging counterexamples

Abstract

Fine-tuning large pre-trained models with task-specific data has achieved great success in NLP. However, it has been demonstrated that the majority of information within the self-attention networks is redundant and not utilized effectively during the fine-tuning stage. This leads to inferior results when generalizing the obtained models to out-of-domain distributions. To this end, we propose a simple yet effective data augmentation technique, HiddenCut, to better regularize the model and encourage it to learn more generalizable features. Specifically, contiguous spans within the hidden space are dynamically and strategically dropped during training. Experiments show that our HiddenCut method outperforms the state-of-the-art augmentation methods on the GLUE benchmark, and consistently exhibits superior generalization performances on out-of-distribution and challenging counterexamples. We have publicly released our code at https://github.com/GT-SALT/HiddenCut.