Avoiding the Hypothesis-Only Bias in Natural Language Inference via Ensemble Adversarial Training
This addresses unwanted biases in NLI models for researchers and practitioners, though it is incremental as it builds on prior adversarial training approaches.
The paper tackles the problem of hypothesis-only bias in Natural Language Inference datasets, where neural networks exploit spurious correlations, by proposing an ensemble adversarial training method that reduces bias in sentence representations and outperforms previous de-biasing efforts when generalized to 12 other datasets.
Natural Language Inference (NLI) datasets contain annotation artefacts resulting in spurious correlations between the natural language utterances and their respective entailment classes. These artefacts are exploited by neural networks even when only considering the hypothesis and ignoring the premise, leading to unwanted biases. Belinkov et al. (2019b) proposed tackling this problem via adversarial training, but this can lead to learned sentence representations that still suffer from the same biases. We show that the bias can be reduced in the sentence representations by using an ensemble of adversaries, encouraging the model to jointly decrease the accuracy of these different adversaries while fitting the data. This approach produces more robust NLI models, outperforming previous de-biasing efforts when generalised to 12 other datasets (Belinkov et al., 2019a; Mahabadi et al., 2020). In addition, we find that the optimal number of adversarial classifiers depends on the dimensionality of the sentence representations, with larger sentence representations being more difficult to de-bias while benefiting from using a greater number of adversaries.