MultiGen: Level-Design for Editable Multiplayer Worlds in Diffusion Game Engines
Ryan Po, David Junhao Zhang, Amir Hertz, Gordon Wetzstein*, Neal Wadhwa*, Nataniel Ruiz* · 2026
Read MultiGen: Level-Design for Editable Multiplayer Worlds in Diffusion Game Engines by Ryan Po, David Junhao Zhang, Amir Hertz, Gordon Wetzstein*, Neal Wadhwa*, Nataniel Ruiz* (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-06679.txt.
Description
Po and team address play limits of video world models: user control for editable runs, and shared inference where players influence a world. They bring an external memory, a lasting state free of the context window, updated by user actions and queried throughout roll-out. Unlike diffusion engines acting as next-frame guessers, their approach splits generation into Memory, Observation, and Dynamics modules. Users gain control over environment structure via memory shape, and the design extends to multi-user rollouts, keeping views coherent and cross-player interactions consistent.
Catalog record
| CREATOR | Ryan Po, David Junhao Zhang, Amir Hertz, Gordon Wetzstein*, Neal Wadhwa*, Nataniel Ruiz* |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 34403 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.06679