Massively Multiagent Minigames for Training Generalist Agents

Kyoung Whan Choe, Ryan Sullivan, Joseph Suárez
6/7/2024
cs.AIcs.LGcs.MA

No official code found. Best implementation: kywch/meta-mmo (7 stars, MIT, last commit 2024-08-30).

Abstract

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.

View on arXivView PDF

Code Implementations(1)

kywch/meta-mmoCommunity100%
MIT

Massively Multiagent Minigames for Training Generalist Agents

70Shell, CythonApr 13, 20242 years agoMIT

Similar Papers

The NeurIPS 2022 Neural MMO Challenge: A Massively Multiagent Competition with Specialization and Trade

85% match2026

Minimax Exploiter: A Data Efficient Approach for Competitive Self-Play

84% match2023

Generative Evolutionary Meta-Solver (GEMS): Scalable Surrogate-Free Multi-Agent Learning

83% match2025

A Comprehensive Review of Multi-Agent Reinforcement Learning in Video Games

83% match2025

Artificial Generals Intelligence: Mastering Generals.io with Reinforcement Learning

83% match1 repo1112025

Learning Controllable and Diverse Player Behaviors in Multi-Agent Environments

82% match2025

Cite this paper

@article{choe2024massively,
  title  = {Massively Multiagent Minigames for Training Generalist Agents},
  author = {Kyoung Whan Choe and Ryan Sullivan and Joseph Suárez},
  year   = {2024},
  eprint = {2406.05071},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2406.05071v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2406.05071.svg)](https://codeofpaper.com/paper/2406.05071)

Discussion