stair-lab/information-gathering

信息收集数据集支持大型语言模型在自适应提问和决策方面的研究。该数据集能够用于训练和评估系统在不确定性下，通过互动获取相关信息以进行准确决策的能力。数据集包括三个关键领域：1D-ARC符号推理任务、GSM8K-Q口头数学推理基准和Fermi现实世界估计问题。

The Information Gathering dataset supports research in adaptive question-asking and decision-making with large language models (LLMs). It enables training and evaluation of systems that must interactively acquire relevant information before making accurate decisions under uncertainty. The dataset includes three key domains: 1D-ARC symbolic reasoning task, GSM8K-Q verbal math reasoning benchmark, and Fermi real-world estimation problems.