后戒断期：针对 QA 中已弃用实例的可靠重新尝试（英文版）

英文标题：Post-Abstention: Towards Reliably Re-Attempting the Abstained Instances in QA中文摘要：本文提出一种 “Post-Abstention” 任务来增加自然语言处理系统的覆盖率并保持准确度，通过对 11 个 QA 数据集的全面实验证明，该方法可以显著提高性能指标。英文摘要：Despite remarkable progress made in natural language processing, even thestate-of-the-art models often make incorrect predictions. Such predictionshamper the reliability of systems and limit their widespread adoption inreal-world applications. 'Selective prediction' partly addresses the aboveconcern by enabling models t