SeriesBench

该数据集是一个包含105部精心策划的以叙事驱动的系列作品的标准基准，覆盖了28个需要深入理解叙事的专业任务。数据集通过分层随机抽样的方式，按照8:1:1的比例划分为训练集、验证集和测试集，重点强调对叙事的理解，并包含了多种任务以评估模型的表现。该数据集的规模为105部系列作品，其任务旨在理解戏剧系列中的复杂叙事。

This dataset is a standardized benchmark containing 105 carefully curated narrative-driven series, covering 28 specialized tasks that demand in-depth narrative comprehension. It is partitioned into training, validation, and test sets at an 8:1:1 ratio via stratified random sampling, with a core emphasis on narrative comprehension and incorporating diverse tasks to assess model performance. Composed of 105 series works, this dataset features tasks aimed at understanding complex narratives within dramatic series.