While recent progress on abstractive summarization has led to remarkably fluent summaries, factual errors in generated summaries still severely limit their use in practice. In this paper, we evaluate summaries produced by state-of-the-art models via crowdsourcing and show that such errors occur frequently, in particular with more abstractive models. We study whether textual entailment predictions can be used to detect such errors and if they can be reduced by reranking alternative predicted summaries. That leads to an interesting downstream application for entailment models. In our experiments, we find that outof-the-box entailment models trained on NLI datasets do not yet offer the desired performance for the downstream task and we therefore release our annotations as additional test data for future extrinsic evaluations of NLI.

Models for natural language understanding (NLU) tasks often rely on the idiosyncratic biases of the dataset, which make them brittle against test cases outside the training distribution. Recently, several proposed debiasing methods are shown to be very effective in improving out-of-distribution performance. However, their improvements come at the expense of performance drop when models are evaluated on the in-distribution data, which contain examples with higher diversity. This seemingly inevitable trade-off may not tell us much about the changes in the reasoning and understanding capabilities of the resulting models on broader types of examples beyond the small subset represented in the outof-distribution data. In this paper, we address this trade-off by introducing a novel debiasing method, called confidence regularization, which discourage models from exploiting biases while enabling them to receive enough incentive to learn from all the training examples. We evaluate our method on three NLU tasks and show that, in contrast to its predecessors, it improves the performance on out-of-distribution datasets (e.g., 7pp gain on HANS dataset) while maintaining the original in-distribution accuracy. 1

NLU models often exploit biases to achieve high dataset-specific performance without properly learning the intended task. Recently proposed debiasing methods are shown to be effective in mitigating this tendency. However, these methods rely on a major assumption that the types of bias should be known a-priori, which limits their application to many NLU tasks and datasets. In this work, we present the first step to bridge this gap by introducing a self-debiasing framework that prevents models from mainly utilizing biases without knowing them in advance. The proposed framework is general and complementary to the existing debiasing methods. We show that it allows these existing methods to retain the improvement on the challenge datasets (i.e., sets of examples designed to expose models' reliance on biases) without specifically targeting certain biases. Furthermore, the evaluation suggests that applying the framework results in improved overall robustness. 1

No abstract

Recent prompt-based approaches allow pretrained language models to achieve strong performances on few-shot finetuning by reformulating downstream tasks as a language modeling problem. In this work, we demonstrate that, despite its advantages on low data regimes, finetuned prompt-based models for sentence pair classification tasks still suffer from a common pitfall of adopting inference heuristics based on lexical overlap, e.g., models incorrectly assuming a sentence pair is of the same meaning because they consist of the same set of words. Interestingly, we find that this particular inference heuristic is significantly less present in the zero-shot evaluation of the prompt-based model, indicating how finetuning can be destructive to useful knowledge learned during the pretraining. We then show that adding a regularization that preserves pretraining weights is effective in mitigating this destructive tendency of few-shot finetuning. Our evaluation on three datasets demonstrates promising improvements on the three corresponding challenge datasets used to diagnose the inference heuristics. 1