Rectifying Action Inequality in Robotic Foundation Models

Existing robotic foundation models, while powerful, are predicated on an implicit assumption of temporal homogeneity: treating all actions as equally informative during optimization. This "flat" training paradigm remains indifferent to the underlying physical hierarchy of manipulation. In reality, robot trajectories are fundamentally heterogeneous — low-velocity segments often dictate task success through precision-demanding interactions, while high-velocity motions serve as error-tolerant transitions. To rectify this misalignment, we introduce AttenA+, an architecture-agnostic framework that prioritizes kinematically critical segments via velocity-driven action attention. By reweighting the training objective based on the inverse velocity field, AttenA+ naturally aligns the model's learning capacity with the physical demands of manipulation. As a plug-and-play enhancement, it integrates into existing backbones without structural modifications or additional parameters. Extensive experiments demonstrate significant improvements across state-of-the-art models, with real-world validation on a Franka manipulator further showcasing its robustness and cross-task generalization.