LiveAct: a method for real-time multi-agent world models
What happened
Xinyu Wang and colleagues propose LiveAct, a video world model for real-time multi-agent settings. Its core is the Heterogeneous Action Interface (HAI), which lets the model jointly generate visual futures for multiple controlled agents and handle diverse control inputs, from game button presses to robot instructions, in a unified way. According to the study, LiveAct reaches agent attribution accuracy of 0.898 on games and 0.942 on multi-robot manipulation tasks, without relying on agent position or identity cropping. Its causal generator runs at 17.70 FPS at 512×448 resolution. These figures come from the authors' self-reported experiments, and so far only this single arXiv paper reports them.
AI-written from the coverage · updated 2026-10-22
Coverage timeline
Follow the coverage to see the story from different sides.
- arXiv Game World Models & GenerationPickLiveAct: heterogeneous action interfaces for real-time multi-agent world models
LiveAct introduces a heterogeneous action interface that lets a multi-agent video world model generate joint visual futures for multiple controlled agents, handling control inputs ranging from game buttons to robot commands. The method reaches agent attribution accuracy of 0.898 on games and 0.942 on multi-robot manipulation, without relying on agent position or identity cropping. Its causal generator runs at 17.70 FPS at 512×448 resolution.