Gamma-World: Generative Multi-Agent World Modeling Beyond Two Players

arXiv 2026 60.8 method

TLDR

A generative multi-agent world model for interactive simulation with novel agent encoding and sparse attention, enabling real-time rollout.

Reasoning

The paper introduces innovative techniques (Simplex Rotary Agent Encoding, Sparse Hub Attention) for scalable multi-agent world modeling, with real-time generation via distillation. However, it lacks real-world validation and the abstract cuts off before presenting experimental results.

Read-first score

Read-first score 60.8, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 53.

Recency 6%

100

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

Citation impact 18%

85.1

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. citation_normalized_percentile=0.85116031

Topical relevance 29%

75.7

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,experiment,metric

Reproducibility 18%

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

Citation velocity 12%

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

Field roles

FoundationFrontierBridgeMethodology anchor

Rank sensitivity

Stability: volatile; rank range: 443.

Keyword Scores

world model

generative world model

interactive world model

video world model

world simulator

world dynamics prediction

model-based reinforcement learning world model

Deep Analysis

Innovations

Simplex Rotary Agent Encoding: a parameter-free extension of 3D RoPE that represents agents as vertices of a regular simplex in rotary angle space, enabling permutation-equivalent agent identity without learned per-slot identities or fixed ordering.
Sparse Hub Attention: learnable hub tokens mediate token interaction across agents, reducing cross-agent attention cost from quadratic to linear in the number of agents.
Distillation of a full-context diffusion teacher into a causal student with KV caching for real-time rollout at 24 FPS.

Methodology

The model uses Simplex Rotary Agent Encoding to assign distinct phases to agents while maintaining permutation symmetry, and Sparse Hub Attention to efficiently handle cross-agent interactions. A full-context diffusion teacher is distilled into a causal student that generates temporal blocks sequentially with KV caching, enabling action-responsive generation at 24 FPS. Experiments are conducted in multiplayer virtual environments, comparing against slot-based and dense-attention baselines.

Key Results

The proposed model improves video fidelity, action controllability, and inter-agent consistency over slot-based and dense-attention baselines, and generalizes from two to four players without additional training.

Resources

Paper