SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2406.05071 · NeurIPS · 2024

Massively Multiagent Minigames for Training Generalist Agents

Whan Kyoung, Ryan Choe, Joseph Sullivan, Suárez

arXiv · PDF · Open in the Atlas

Code that ran

1 of 1 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 1 failure of our instrument; 0 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2406.05071.json · badge: /api/ran/2406.05071.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
kywch/meta-mmo canonical (classifier label)
classifier basis: classifier_2025
1 ran · 0 no recorded run
FunctionStatusWhere it lives
get_eval_config Ran
ran; our contract draft was wrong, not the code
kywch/meta-mmo/evaluate.py
code served (permissive licence) · get_code("d3590a66327201cb")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned dict. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

ARGS = {
    'debug': False
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block

Also present in other papers' code

Of 1 samples lifted from this paper's linked repositories, 1 carry a function identity and 0 do not. Of the 1 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 1 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

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. Of these, XLand has 2 agents and is not publicly available. The rest are single-agent. In contrast, Neural MMO features 100+ agents in a multi-task, open-source environment that allows us to study generalization across tasks, opponents, and maps [40,41]. In Neural MMO, agents are presented with diverse challenges including collecting resources, engaging in combat, training professions, and trading on a player-controlled market. Most of the progress on Neural MMO has been driven by competitions at NeurIPS and IJCAI totaling 1300+ participants. In the most recent NeurIPS 2023 competition, participants trained goal-conditioned agents capable of completing a variety of tasks. Despite years of sustained interest, the best Neural MMO agents are only proficient at a few tasks and cannot reach high levels or play effectively as a team. Meta MMO extends Neural MMO with a diverse set of minigames. Our main contributions are: 1. Meta MMO as a benchmark for many-agent generalization. Minigames feature free-for-all and team settings, built-in domain randomization, and adaptive difficulty. 2. Optimized training up to 3x faster with minigames. Each of our experiments is run on a commercial off-the-shelf desktop with a single RTX 4090. 3. A generalist agent capable of playing several minigames with a single set of weights. It is trained using PPO and a simple curriculum learning method. Neural MMO evaluates generalization over tasks, opponents, and maps. Meta MMO enables further evaluation over variations in gameplay mechanics and runs up to 10 times faster than Neural MMO 2. We demonstrate that a RL can learn sophisticated behaviors on multiple individual and team-based Preprint. Under review.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2406.05071")
get_code_for_paper("2406.05071")
have("2406.05071")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2406.05071.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2406.05071.svg)](https://syntology.ai/paper/2406.05071)

Connect an agent — have() is free.