SummScreen

该数据集包含了大约5600个单词的电视节目剧集转录文本，而每个参考输出则是由人类编写的约100词的剧情概述。这些数据集被用于评估长上下文摘要的能力，旨在解决长文本语境下的摘要任务。

This dataset comprises television episode transcripts with a total length of approximately 5,600 words, and each corresponding reference output is a human-written plot summary of roughly 100 words. This dataset is utilized to evaluate long-context summarization capabilities, aiming to address summarization tasks in long-text contexts.