Rolling-WAM: World Action Models with Rolling Imagination
Rolling-WAM speeds up robot replanning by spreading video-action denoising across steps instead of restarting the full horizon each cycle.
The paper proposes a sliding window of future video-action chunks held at staggered noise levels. Each new observation advances the window, fully denoising the next action for execution while continuing to refine later chunks. The authors report competitive results on LIBERO, RoboTwin, and a real Unitree G1 humanoid, with a 4.5x steady-state replanning speedup over standard joint WAMs. HF Daily Papers' note
The paper proposes a sliding window of future video-action chunks held at staggered noise levels. Each new observation advances the window, fully denoising the next action for execution while continuing to refine later chunks. The authors report competitive results on LIBERO, RoboTwin, and a real Unitree G1 humanoid, with a 4.5x steady-state replanning speedup over standard joint WAMs. HF Daily Papers' note
score 5