AffordanceWAM: Spatiotemporal Affordance for Joint World-Action Modeling
October 17, 2026
AffordanceWAM integrates object-centric spatiotemporal affordance—via scalar affordance and heatmaps—into generative World Action Models. This allows for shared interaction targets between human and robot videos, grounding visual prediction in task-relevant regions.
HOW THIS AFFECTS YOU
●
builderYou can leverage this to bridge the gap between human demonstration data and robot execution.
●
researcherThis enables more effective cross-embodiment learning by using shared affordance representations.