Towards Interpreting and Mitigating Shortcut Learning Behavior of NLU Models

Recent studies indicate that NLU models are prone to rely on shortcut features for prediction, without achieving true language understanding. As a result, these models fail to generalize to real-world out-of-distribution data. In this work, we show that the words in the NLU training set can be model...

Full description

Saved in:
Bibliographic Details
Published inarXiv.org
Main Authors Du, Mengnan, Manjunatha, Varun, Jain, Rajiv, Deshpande, Ruchi, Dernoncourt, Franck, Gu, Jiuxiang, Sun, Tong, Hu, Xia
Format Paper
LanguageEnglish
Published Ithaca Cornell University Library, arXiv.org 13.04.2021
Subjects
Learning
Training
Online AccessGet full text

Cover

More Information
Summary:Recent studies indicate that NLU models are prone to rely on shortcut features for prediction, without achieving true language understanding. As a result, these models fail to generalize to real-world out-of-distribution data. In this work, we show that the words in the NLU training set can be modeled as a long-tailed distribution. There are two findings: 1) NLU models have strong preference for features located at the head of the long-tailed distribution, and 2) Shortcut features are picked up during very early few iterations of the model training. These two observations are further employed to formulate a measurement which can quantify the shortcut degree of each training sample. Based on this shortcut measurement, we propose a shortcut mitigation framework LTGR, to suppress the model from making overconfident predictions for samples with large shortcut degree. Experimental results on three NLU benchmarks demonstrate that our long-tailed distribution explanation accurately reflects the shortcut learning behavior of NLU models. Experimental analysis further indicates that LTGR can improve the generalization accuracy on OOD data, while preserving the accuracy on in-distribution data.
ISSN:2331-8422