Twenty Questions Dataset

该数据集是从THINGS数据库中精心挑选的一组对象构建而成的，每个对象都作为一个潜在的隐藏实体，配备了由高质量的大型语言模型生成的多样化候选问题及其对应答案。该数据集有助于评估自适应引导方法，并为与大型语言模型交互式提问设计、决策制定和规划的研究提供了便利。该数据集的规模包括800个对象，每个对象配有1200个问题的答案。这项任务旨在通过策略性查询实现自适应信息收集。

This dataset is constructed using a carefully curated set of objects sourced from the THINGS database. Each object functions as a latent hidden entity, paired with diverse candidate questions and their corresponding answers generated by high-quality large language models. This dataset supports the evaluation of adaptive guidance methods and facilitates research on interactive question design, decision-making and planning for interactions with large language models. The dataset comprises 800 objects, with each object equipped with answers to 1200 questions. This task aims to accomplish adaptive information gathering through strategic queries.