Understanding and Mitigating Spurious Correlations in Text Classification
Preprint
Abstract
Recent work has shown that deep learning models are prone to exploit spurious correlations that are present in the training set, yet may not hold true in general. A sentiment classifier may erroneously learn that the token spielberg is always tied to positive movie reviews. Relying on spurious correlations may lead to significant degradation in generalizability and should be avoided. In this paper, we propose a neighborhood analysis framework to explain how exactly language models exploit spurious correlations. Driven by the analysis, we propose a family of regularization methods, NFL (do Not Forget your Language) to prevent the situation. Experiments on two text classification tasks show that NFL brings a significant improvement over standard fine-tuning in terms of robustness without sacrificing in-distribution accuracy.Keywords
All Related Versions
This publication has 17 references indexed in Scilit:
- Model-tuning Via Prompts Makes NLP Models Adversarially Robust2023
- Identifying and Mitigating Spurious Correlations for Improving Robustness in NLP Models2022
- P-Tuning: Prompt Tuning Can Be Comparable to Fine-tuning Across Scales and Tasks2022
- A Closer Look at How Fine-tuning Changes BERT2022
- MedNLI Is Not Immune: Natural Language Inference Artifacts in the Clinical Domain2021
- An Empirical Study on Robustness to Spurious Correlations using Pre-trained Language ModelsTransactions of the Association for Computational Linguistics, 2020
- Learning to Model and Ignore Dataset Bias with Mixed Capacity Ensembles2020
- Identifying Spurious Correlations for Robust Text Classification2020
- Nuanced Metrics for Measuring Unintended Bias with Real Data for Text Classification2019
- Unlearn Dataset Bias in Natural Language Inference by Fitting the Residual2019