StorySumm

该数据集名为StorySumm，包含了从业余写作论坛收集的32篇英文故事，每篇故事都由三种GPT或Claude系列的大型语言模型进行了摘要，总计96个摘要和568个句子。每个摘要和句子都被标注为忠实、不忠实或不适用，并附有多重标注者的标签和书面解释。此外，数据集还包括了对主观性的标注，用于分析摘要中陈述的模糊性。在规模上，该数据集包含32个故事、96个摘要和568个句子。该数据集的任务是评估叙事摘要中的主观性和陈述忠实度。

This dataset, named StorySumm, consists of 32 English stories collected from amateur writing forums. Each story has been summarized by three large language models (LLMs) from the GPT and Claude series, yielding a total of 96 summaries and 568 sentences. Every summary and sentence is annotated as faithful, unfaithful, or not applicable, accompanied by multiple annotator labels and written explanations. Additionally, the dataset includes subjectivity annotations for analyzing the ambiguity of statements within the summaries. In terms of scale, the dataset contains 32 stories, 96 summaries, and 568 sentences. The core task of this dataset is to evaluate the subjectivity and statement faithfulness in narrative summaries.