MASS: Multiplayer World Models with Authoritative Shared State

arXiv 2026 46.6 method

TLDR

Proposes MAS, a multiplayer world model that separates authoritative shared state from view rendering, achieving scalable and consistent multi-agent simulation.

Reasoning

The paper introduces a novel architecture that disentangles world dynamics from view rendering, addressing scalability and consistency issues in multiplayer environments. Strengths include clear methodology and strong empirical results on a benchmark, but limitations include evaluation only on a simple Snake environment and lack of real-world complexity.

Read-first score

Read-first score 46.6, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 55.

Recency 6%

100

Uses a gentle age decay so recent papers surface without erasing older foundations. 2026

Topical relevance 29%

78.6

Uses existing LLM keyword relevance scores normalized to 0-100. world model,world simulator,generative world model,interactive world model,video world model,world dynamics prediction,model-based reinforcement learning world model

Methodology quality 18%

Screens visible abstract and analysis fields for experiment, dataset, baseline, metric, and limitation evidence. markers=baseline,benchmark,result

Reproducibility 18%

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=none

Citation impact 18%

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. cited_by_count=0

Citation velocity 12%

Citation velocity estimates citations per publication-year to reduce old-paper bias. velocity=0.00

Field roles

FrontierMethodology anchor

Rank sensitivity

Stability: volatile; rank range: 515.

Keyword Scores

world model

world simulator

interactive world model

world dynamics prediction

generative world model

video world model

model-based reinforcement learning world model

Deep Analysis

Innovations

Disentanglement of world dynamics and view rendering into a learned Logic Engine and Rendering Engine
Authoritative typed state as sole recurrent memory and synchronization reference, eliminating hand-crafted transition functions
Scalable multi-agent simulation supporting 1,024 concurrent players and 10,000 recurrent steps

Methodology

MAS architecture with a Logic Engine that updates a global authoritative typed state from joint actions and a Rendering Engine that generates independent views on demand. The model is trained on a multiplayer Snake benchmark and compared against state-of-the-art multi-view world model baselines.

Key Results

MAS achieves higher state accuracy and lower cross-view inconsistency than baselines, and successfully scales to 1,024 players over 10,000 steps.

Resources

Paper