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

2026-05-27

AI Summary

Existing world models struggle to achieve scalable, consistent, and efficiently controllable video generation in multi-agent interactive environments. This work proposes a generative multi-agent world model that employs a parameter-free Simplex Rotary Agent Encoding to yield permutation-equivariant yet independently controllable agent identity representations. It introduces Sparse Hub Attention to reduce cross-agent attention complexity from quadratic to linear and integrates KV cache–based causal diffusion distillation with 3D RoPE extensions to enable action-responsive real-time video synthesis. The method substantially improves video fidelity, action controllability, and agent consistency in multi-character virtual scenes, achieving real-time inference at 24 FPS and demonstrating zero-shot generalization to four-agent scenarios without additional training.

Abstract

World models for interactive video generation have largely focused on single-agent settings, where future observations are generated from a single control signal. However, many generated environments require multi-agent interaction: multiple players, robots, or embodied agents act simultaneously within a shared space. Scaling world models to such settings requires a principled multi-agent design: agents should remain independently controllable, permutation-symmetric, and support efficient inference while maintaining consistency across time and perspectives. In this paper, we present our generative multi-agent world model for interactive simulation. It introduces Simplex Rotary Agent Encoding, a parameter-free extension of 3D RoPE that represents agents as vertices of a regular simplex in rotary angle space. This gives each agent a distinct phase while making all agents permutation-equivalent, enabling scalable agent identity without learned per-slot identities or a fixed agent ordering. To avoid dense all-to-all attention across agents, we further propose Sparse Hub Attention, where learnable hub tokens mediate token interaction across agents, reducing cross-agent attention cost from quadratic to linear in the number of agents. For real-time rollout, we distill a full-context diffusion teacher into a causal student that generates temporal blocks sequentially with KV caching, enabling action-responsive generation at 24 FPS. Experiments in multiplayer virtual environments show that our model improves video fidelity, action controllability, and inter-agent consistency over slot-based and dense-attention baselines, while generalizing from two to four players without additional training.

Problem

• multi-agent world modeling
• interactive video generation
• permutation symmetry
• scalable agent identity
• cross-agent consistency

Innovation

• Simplex Rotary Agent Encoding
• Sparse Hub Attention
• Generative Multi-Agent World Model
• Permutation-Equivariant Representation
• Causal Diffusion Distillation