Recent advances in video diffusion have enabled the development of "world models" capable of simulating interactive environments. However, these models are largely restricted to single-agent settings, failing to control multiple agents simultaneously in a scene. In this work, we tackle a fundamental issue of action binding in existing video diffusion models, which struggle to associate specific actions with their corresponding subjects. For this purpose, we propose ActionParty, an action controllable multi-subject world model for generative video games. It introduces subject state tokens, i.e. latent variables that persistently capture the state of each subject in the scene. By jointly modeling state tokens and video latents with a spatial biasing mechanism, we disentangle global video frame rendering from individual action-controlled subject updates. We evaluate ActionParty on the Melting Pot benchmark, demonstrating the first video world model capable of controlling up to seven players simultaneously across 46 diverse environments. Our results show significant improvements in action-following accuracy and identity consistency, while enabling robust autoregressive tracking of subjects through complex interactions.

翻译：近期视频扩散技术的进展催生了能够模拟交互式环境的“世界模型”。然而，这类模型大多局限于单智能体场景，无法在同一个画面中同时控制多个智能体。本文旨在解决现有视频扩散模型中动作绑定的基础性问题——这些模型难以将特定动作与对应主体准确关联。为此，我们提出ActionParty——一种面向生成式电子游戏的动作可控多主体世界模型。该模型引入主体状态令牌（subject state tokens），即能持续捕获场景中各主体状态的隐变量。通过联合建模状态令牌与视频隐变量并采用空间偏置机制，我们实现了全局视频帧渲染与个体动作驱动主体更新的解耦。我们在Melting Pot基准上评估ActionParty，首次展示了能在46种不同环境中同时控制多达七个玩家的视频世界模型。实验结果表明，该方法在动作跟随准确率与主体身份一致性上取得显著提升，同时能够通过复杂交互过程对主体实现稳健的自回归跟踪。