StatePlay: State-Aware Game World Models for Mechanics-Consistent Generation

arXiv 2026 40.2 method

TLDR

StatePlay jointly predicts visual content and game states via a mixture-of-transformers architecture to ensure mechanics-consistent game world generation.

Reasoning

The paper addresses a clear gap in game world models by explicitly modeling state-dependent game mechanics, with strong quantitative results (state prediction L1 < 0.06, 18.6% improvement in mechanics fidelity). However, the evaluation is limited to game domains, and the claim of 'complete' generation is not fully supported by the abstract alone.

Read-first score

Read-first score 40.2, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 48.

Recency 6%

100

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

Topical relevance 29%

68.6

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=experiment

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

Frontier

Rank sensitivity

Stability: volatile; rank range: 312.

Keyword Scores

world model

generative world model

interactive world model

world dynamics prediction

world simulator

video world model

model-based reinforcement learning world model

Deep Analysis

Innovations

State-aware game world model that jointly predicts visual frames and game states for mechanics-consistent generation.
Mixture-of-transformers (MoT) architecture with specialized visual and state branches, cross-modal interaction, and modality-specific optimization objectives.
Explicit state prediction to enforce game mechanics, going beyond pixel-level realism.

Methodology

StatePlay is a mixture-of-transformers model that jointly predicts visual content and game states conditioned on player actions. It preserves specialized visual and state representations with cross-modal interaction, allowing predicted states to guide frame generation, and each branch is optimized with a distinct objective suitable for its modality.

Key Results

StatePlay achieves an average normalized L1 distance below 0.06 for state prediction and improves mechanics fidelity by 18.6% over models without explicit state modeling.

Resources

Paper