Agar.io
收藏arXiv2025-09-30 收录
下载链接：
https://sites.google.com/view/staghuntrpg
下载链接
链接失效反馈官方服务：
资源简介：
该数据集来源于一个现实世界的游戏，在这个游戏中存在多个均衡状态，并且利用奖励随机化来探索多样化的策略行为。与其他标准的多代理政策梯度方法相比，该数据集显示角色扮演游戏（RPG）一致性地发现具有更高收益的策略。该研究在1000个剧集上进行评估，任务是探索多代理的战略行为。
This dataset is derived from a real-world game that features multiple equilibrium states, and employs reward randomization to explore diverse strategic behaviors. Compared with other standard multi-agent policy gradient methods, this dataset demonstrates that the role-playing game (RPG) consistently discovers strategies with higher returns. This study was evaluated over 1000 episodes, with the core task of exploring multi-agent strategic behaviors.
