An Empirical Study on Robustness to Spurious Correlations using Pre-trained Language Models

Lifu Tu, Garima Lalwani, Spandana Gella, He He


Abstract
Recent work has shown that pre-trained language models such as BERT improve robustness to spurious correlations in the dataset. Intrigued by these results, we find that the key to their success is generalization from a small amount of counterexamples where the spurious correlations do not hold. When such minority examples are scarce, pre-trained models perform as poorly as models trained from scratch. In the case of extreme minority, we propose to use multi-task learning (MTL) to improve generalization. Our experiments on natural language inference and paraphrase identification show that MTL with the right auxiliary tasks significantly improves performance on challenging examples without hurting the in-distribution performance. Further, we show that the gain from MTL mainly comes from improved generalization from the minority examples. Our results highlight the importance of data diversity for overcoming spurious correlations.1
Anthology ID:
2020.tacl-1.40
Volume:
Transactions of the Association for Computational Linguistics, Volume 8
Month:
Year:
2020
Address:
Cambridge, MA
Editors:
Mark Johnson, Brian Roark, Ani Nenkova
Venue:
TACL
SIG:
Publisher:
MIT Press
Note:
Pages:
621–633
Language:
URL:
https://aclanthology.org/2020.tacl-1.40
DOI:
10.1162/tacl_a_00335
Bibkey:
Cite (ACL):
Lifu Tu, Garima Lalwani, Spandana Gella, and He He. 2020. An Empirical Study on Robustness to Spurious Correlations using Pre-trained Language Models. Transactions of the Association for Computational Linguistics, 8:621–633.
Cite (Informal):
An Empirical Study on Robustness to Spurious Correlations using Pre-trained Language Models (Tu et al., TACL 2020)
Copy Citation:
PDF:
https://aclanthology.org/2020.tacl-1.40.pdf