Twenty Questions dataset

该数据集包含了78,890个实体，用于评估大型语言模型在BrainKing游戏中的问题解决能力，该游戏要求参与者通过是非问答来识别实体，即便可能遇到误导性的答案。此外，该数据集使用GPT3.5进行了常见性筛选，并为每个实体特征提供了一个分层的概念列表。在筛选后，数据集包含了10,000个常见实体。该任务的目标是通过是非问题来进行实体识别。

This dataset contains 78,890 entities, intended to evaluate the problem-solving capabilities of Large Language Models (LLMs) in the BrainKing game. The BrainKing game requires participants to identify entities through yes-no questions, even when encountering misleading answers. Additionally, the dataset underwent commonness filtering using GPT-3.5, and a hierarchical list of concepts is provided for each entity's features. After filtering, the dataset includes 10,000 common entities. The objective of this task is entity identification via yes-no questions.