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

By Lifu Tu, Garima Lalwani, Spandana Gella, He He. 2020

207 citations recorded by the source corpus.

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

Subfields: Artificial Intelligence, Large Language Models, Safety, Robustness & Interpretability

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Yejin Choi
2. Mohit Bansal
3. Noah A. Smith
4. Yonatan Belinkov
5. Luke S. Zettlemoyer
6. Douwe Kiela
7. Jianfeng Gao
8. Percy Liang

In Large Language Models

1. Lu Hou
2. Qun Liu
3. Xin Jiang
4. Lifeng Shang
5. Yejin Choi
6. Noah A. Smith
7. Yonatan Belinkov
8. Swabha Swayamdipta

In Safety, Robustness & Interpretability

1. Yejin Choi
2. Noah A. Smith
3. Swabha Swayamdipta
4. Yonatan Belinkov
5. Alisa Liu
6. Divyansh Kaushik
7. Douwe Kiela
8. Joe Stacey

Citations in the corpus

• WANLI: Worker and AI Collaboration for Natural Language Inference Dataset Creation (2022, background): “This is important since Tu et al. (2020)”
• An Empirical Investigation of the Role of Pre-training in Lifelong Learning (2021, neutral): “, 2020), and out-of-distribution robustness (Hendrycks et al., 2020; Tu et al., 2020).”
• On the Efficacy of Adversarial Data Collection for Question Answering: Results from a Large-Scale Randomized Study (2021, result): “Prior works have hypothesized that the bigger size and increased diversity of the pretraining corpus of RoBERTa (compared to those of BERT and ELECTRA) might somehow be responsible for RoBERTa’s better out-of-domain generalization, (Baevski et al., 2019; Hendrycks et al., 2020; Tu et al., 2020).”
• Natural Language Inference with a Human Touch: Using Human Explanations to Guide Model Attention (2021, methodology): “…2019a; Stacey et al., 2020), using data-augmentation (Min et al., 2020; Minervini and Riedel, 2018), fine-tuning on minority examples (Yaghoobzadeh et al., 2021), multi-task learning (Tu et al., 2020) or creating compressed representations to remove irrelevant information (Mahabadi et al., 2021).”
• Improved OOD Generalization via Adversarial Training and Pre-training (2021, background): “OOD generalization measures a model’s ability to extrapolate beyond the training distribution (Hendrycks & Dietterich, 2018), and has been widely explored in both CV (Recht et al., 2019; Schneider et al., 2020; Salman et al., 2020b) and NLP domains (Tu et al., 2020; Lohn, 2020).”
• Supervising Model Attention with Human Explanations for Robust Natural Language Inference (2021, background): “(Tu et al. 2020) or creating compressed representations to remove irrelevant information (Mahabadi, Belinkov, and Henderson 2021).”