Partial-input baselines show that NLI models can ignore context, but they dont.

When strong partial-input baselines reveal artifacts in crowdsourced NLI datasets, the performance of full-input models trained on such datasets is often dismissed as reliance on spurious correlations. We investigate whether state-of-the-art NLI models are capable of overriding default inferences made by a partial-input baseline. We introduce an evaluation set of 600 examples consisting of perturbed premises to examine a RoBERTa models sensitivity to edited contexts. Our results indicate that NLI models are still capable of learning to condition on contexta necessary component of inferential reasoningdespite being trained on artifact-ridden datasets.