CLOTH(CLOze test by TeacHers)

CLOTH 是一个大规模的完形填空测试数据集，包含 7,131 个段落和 99,433 个问题。该数据集来自中国的中学和高中英语考试。 CLOTH 评估机器对自然语言多个方面的理解，包括词汇、推理和语法。此外，CLOTH 可用于评估语言模型对长上下文建模的能力。

CLOTH is a large-scale cloze test dataset containing 7,131 passages and 99,433 questions. It is sourced from Chinese middle school and high school English examinations. CLOTH is designed to evaluate multiple aspects of machine natural language understanding, including vocabulary, reasoning and grammar. Additionally, it can be utilized to assess the long-context modeling capabilities of language models.

搜集汇总

背景与挑战

背景概述

CLOTH是一个基于中国中学和高中英语考试构建的大规模完形填空测试数据集，包含7,131个段落和99,433个问题。它用于评估机器对自然语言多方面（如词汇、推理和语法）的理解能力，以及长上下文建模的性能。