ChildPlay Benchmark Suite

该数据集包含了一系列游戏，如井字棋、四子棋、战舰和乐高组装等，以及空间推理任务，旨在评估大型语言模型在战略思考和决策能力方面的表现。此外，该数据集还包括了800张图片用于有效性测试，其中一半是有效的构造，另一半是无效的。模型根据提示生成额外的构造，并对其有效性进行评估。该数据集涵盖了多个游戏以及生成的构造和棋盘状态，其规模可用来评估大型语言模型在非语言领域的推理和决策能力。

This dataset includes a series of games such as Tic-Tac-Toe, Connect Four, Battleship, Lego assembly, and spatial reasoning tasks, designed to evaluate the performance of large language models (LLMs) in strategic thinking and decision-making abilities. Additionally, the dataset contains 800 images for validity testing, with half being valid constructions and the other half invalid. Models are required to generate additional constructions based on given prompts and assess their validity. This dataset covers multiple games, along with generated constructions and board states, and its scale can be used to evaluate the reasoning and decision-making capabilities of LLMs in non-linguistic domains.