Procedurally Generated Game Levels
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/BKhaleque/Evaluating-Environments-using-Exploratory-Agents
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含了5个吸引人程度较高的关卡和5个吸引人程度较低的关卡，这些关卡是由两种不同的生成器生成的。其目的是评估游戏环境的探索潜力。在三个不同的出生点，利用探索性代理对这些关卡进行了3分钟的探索，同时测量了包括覆盖度、新颖性、熵值和物体检查在内的多种指标。该数据集规模为10个关卡（其中5个具有较高吸引力，5个具有较低吸引力），任务是对游戏关卡的探索潜力进行评估。
This dataset comprises 10 game levels, including 5 highly engaging levels and 5 low-engagement levels, which were generated by two distinct level generators. The core task of this dataset is to evaluate the exploration potential of game environments. For each level, exploratory AI agents were deployed to conduct 3-minute exploration sessions from three distinct spawn points, while a suite of metrics including coverage, novelty, entropy, and object inspection were measured simultaneously.
