Hanabi Learning Environment (HLE)

该数据集是关于汉纳比（Hanabi）的，这是一种完全合作的纸牌游戏，玩家需要共同协作，完成被称为“烟花”的一堆堆纸牌。在游戏中，玩家无法看到自己的牌，必须与队友有效沟通才能玩好游戏。汉纳比游戏的最大得分是25分，条件是完成所有五个烟花，每个烟花由五张牌组成。此游戏具有独特的沟通要素，玩家必须分享信息才能成功。数据集中的规模包括50张牌的牌组，以及多名玩家（2至5人）。任务涉及多代理强化学习。

This dataset focuses on Hanabi, a fully cooperative card game where players collaborate to build stacks of cards referred to as "fireworks". In the game, players cannot see their own hands and must communicate effectively with their teammates to achieve strong performance. The maximum score in a Hanabi game is 25 points, which is achieved by completing all five fireworks, with each firework stack consisting of five cards. This game features a unique communication mechanism, as players must share information to succeed. The dataset includes a 50-card deck and supports 2 to 5 players. The tasks involved in this dataset relate to multi-agent reinforcement learning.