We present Meta MMO, a collection of many-agent minigames for use as a
reinforcement learning benchmark. Meta MMO is built on top of Neural MMO, a
massively multiagent environment that has been the subject of two previous
NeurIPS competitions. Our work expands Neural MMO with several computationally
efficient minigames. We explore generalization across Meta MMO by learning to
play several minigames with a single set of weights. We release the
environment, baselines, and training code under the MIT license. We hope that
Meta MMO will spur additional progress on Neural MMO and, more generally, will
serve as a useful benchmark for many-agent generalization.

我们提出了 Meta MMO，这是一个用于强化学习基准的多代理小游戏集合，它建立在之前两次 NeurIPS 竞赛的研究基础 Neural MMO 之上，并扩展了该环境的多个计算效率高的小游戏。我们通过使用一组权重学习玩多个小游戏来探索 Meta MMO 的泛化能力。我们以 MIT 许可证发布了环境、基线和训练代码。我们希望 Meta MMO 能够推动 Neural MMO 和更广泛的多代理泛化研究取得额外进展，并且成为一个有用的基准。