The Children’s Book Test 问答数据集

CBT 数据集由文字段落和相应问题构建，问答数据均来自古腾堡项目免费提供的书籍，该数据集用于直接测量语言模型、更广泛的语言环境用于问答和仿真查找。

The CBT dataset is constructed from textual passages and their corresponding questions. All question-answering pairs within the dataset are sourced from books freely provided by the Project Gutenberg. This dataset is primarily used to directly evaluate language models, and also supports a broader range of language-centric applications including question answering and simulated information retrieval.

搜集汇总

背景与挑战

背景概述

该数据集是Facebook于2016年发布的The Children's Book Test问答数据集，基于古腾堡项目的书籍构建，用于评估语言模型和问答系统。