Increasing Robustness to Spurious Correlations using Forgettable Examples
This addresses the issue of spurious correlations in NLP models for improving generalization, but it is incremental as it builds on existing minority example techniques.
The paper tackles the problem of neural NLP models relying on spurious correlations by proposing a method to identify minority examples using example forgetting, without prior knowledge of spurious correlations, and robustifies models through double fine-tuning, resulting in substantial improvements in out-of-distribution generalization on MNLI, QQP, and FEVER datasets.
Neural NLP models tend to rely on spurious correlations between labels and input features to perform their tasks. Minority examples, i.e., examples that contradict the spurious correlations present in the majority of data points, have been shown to increase the out-of-distribution generalization of pre-trained language models. In this paper, we first propose using example forgetting to find minority examples without prior knowledge of the spurious correlations present in the dataset. Forgettable examples are instances either learned and then forgotten during training or never learned. We empirically show how these examples are related to minorities in our training sets. Then, we introduce a new approach to robustify models by fine-tuning our models twice, first on the full training data and second on the minorities only. We obtain substantial improvements in out-of-distribution generalization when applying our approach to the MNLI, QQP, and FEVER datasets.