UMO: Unified In-Context Learning Unlocks Motion Foundation Model Priors | ScienceToStartup | ScienceToStartup