Adnan KhanAdnanProduct & AI
San Francisco Bay Area
Get in touch

Source note3 min read

Multiplayer Interactive World Models with Representation Autoencoders

Source

arXiv

arxiv.org/abs/2607.05352v1

Read the source

Source

  • Source name: arXiv
  • Source kind: paper
  • Authors / organization: Anthony Hu, Václav Volhejn, Adrien Ramanana Rahary, Chris Mulder, Aditya Makkar, Amélie Royer, Manu Orsini, Alyx Liao, Adam Jelley, Eloi Alonso, Florian Laurent, Fredrik Norén, James Swingos, Jan Hünermann, Kent Rollins, Lucas Hosseini, Matthieu Le Cauchois, Maxim Peter, Pim de Witte, Tim Brown, Vincent Micheli, Moritz Böhle, Gabriel de Marmiesse, Viktoriia Sharmanska, Lucia Specia, Michael Black, Patrick Pérez
  • Venue: arXiv: cs.CV
  • Published: 2026-07-06
  • Collected: 2026-07-07
  • URL: http://arxiv.org/abs/2607.05352v1
  • PDF: https://arxiv.org/pdf/2607.05352v1
  • DOI: UNKNOWN

Plain-Language Brief

Researchers have developed a new multiplayer world model for complex games like Rocket League, which can generate four-player matches in real time using a latent diffusion model trained on 10,000 hours of gameplay. The model can maintain stability far beyond the training horizon and has been tested with various design choices.

What the Document Says

  • The paper introduces a multiplayer world model for complex games with fast dynamics.
  • The model is trained on Rocket League gameplay data collected using publicly available bots.
  • It generates four-player matches in real time and produces 20 frames per second on a single Nvidia B200 GPU.
  • The model can maintain stability even beyond the training horizon.
  • They investigate central design choices, including the video codec, generative objective, and multiplayer conditioning scheme.
  • They characterize how behavior changes with model and data scale.
  • They develop targeted evaluations to probe the model's physical understanding rather than just its visual appearance.
  • The authors release their dataset, training and inference codebase, and a live demo for continued research.

Key Concepts Mentioned

  • multiplayer world model
  • Rocket League
  • latent diffusion model
  • training horizon
  • central design choices
  • targeted evaluations

Claims Made by the Document

  • The introduced multiplayer world model can generate four-player matches in real time.
  • The model maintains stability far beyond the training horizon.
  • They investigate central design choices and characterize behavior changes with model and data scale.

Evidence Used

  • The paper describes the development of a multiplayer world model for complex games.
  • It mentions that the model is trained on 10,000 hours of Rocket League gameplay data.
  • They state that the model generates four-player matches in real time and produces 20 frames per second.
  • The paper mentions that the model can maintain stability beyond the training horizon.
  • It is stated that they investigate central design choices and characterize behavior changes with model and data scale.

Methods / Approach

UNKNOWN

Limitations Stated by the Authors

UNKNOWN

Limitations Noted by the Agent

Agent interpretation — not from the source.

UNKNOWN

Open Questions

UNKNOWN

Notes for Future Q&A

May include interpretation. Labeled where it does.

UNKNOWN

Raw Abstract / Excerpt

We introduce the first multiplayer world model for highly dynamic environments governed by complex physical interactions. Whereas single-player world models treat the other agents as part of the environment, ours conditions on the action streams of multiple agents, learning to attribute changes in the scene to the correct player and to stay coherent under arbitrary combinations of their actions. We study this problem in the game of Rocket League, where players compete and cooperate under fast, tightly coupled dynamics. Trained on 10,000 hours of gameplay collected with publicly available bots, our 5-billion-parameter latent diffusion model generates four-player matches in real time, producing 20 frames per second on a single Nvidia B200 GPU. Although trained only on short clips, its rollouts stay stable far beyond the training horizon: distributional quality holds steady out to five minutes, the longest horizon we measure, and in practice we observe rollouts continuing for hours with no sign of collapse. We systematically investigate the central design choices: the video codec, the generative objective, and the multiplayer conditioning scheme. In addition, we characterize how behavior changes with model and data scale, including the capabilities that emerge and the failure modes that persist. We further develop targeted evaluations that probe the model's physical understanding rather than visual appearance alone. To support continued research on multiplayer world models, we release our dataset, our full training and inference codebase, and a live demo.