Date Understanding

此任务旨在测试预先训练的语言模型通过询问一天的日期来理解日期的能力。上下文是一个或两个带有日期信息的句子，它们要么是显式的 (例如，今天是...)，要么是隐式的 (例如，昨天是...)。给定上下文，模型应该回答以下问题: MM/DD/YYYY中的 [日] 是什么日期？，[日] 可能是今天，明天等。虽然这项任务对人类来说很容易，但对许多语言模型来说却很难。This task is designed to evaluate the date comprehension capabilities of pre-trained language models. The task setup involves supplying context comprising one or two sentences containing date information, which can be either explicit (e.g., "Today is...") or implicit (e.g., "Yesterday was..."). Given the provided context, the model is required to respond to the question: "What is the date of [day] in the format MM/DD/YYYY?", where [day] may refer to today, tomorrow, or other similar temporal references. While this task is trivial for human beings, it presents substantial challenges for numerous pre-trained language models.

搜集汇总

背景与挑战

背景概述

该数据集旨在评估预训练语言模型对日期信息的理解能力，通过提供包含显式或隐式日期信息的句子，要求模型以MM/DD/YYYY格式回答特定日期问题。数据集适用于测试模型在简单但常见任务上的表现。