paper
Learning Action Priors for Cross-embodiment Robot Manipulation
paperactiveprovisional
learning-action-priors-for-cross-embodiment-robot-manipulation-7d20f405·1 events·first seen 17h agoAliases: Learning Action Priors for Cross-embodiment Robot Manipulation
Co-occurring entities
More like this (12)
Geometric Action Model for Robot Policy Learningbi-manual robotic manipulationOperator LearningOpen X-EmbodimentPose-ICL: 3D-Aware In-Context Learning for Pose-Controllable Subject CustomizationOpenAI Dexterous HandLearning Red Agent Policy from Observations for Neurosymbolic Autonomous Cyber AgentsModal-Aware Rotary Positional EmbeddingA Taxonomy of Conceptual Alignment in Human-Robot DialogueEmbodied Minds Labembodied agentsUMass Embodied AGI
Recent events (1)
Two-stage action prior pretraining improves cross-embodiment VLA robot manipulation
Researchers propose a two-stage training framework for Vision-Language-Action (VLA) models that pretrains the action module with motion priors before cross-modal alignment begins. Stage 1 uses a flow-matching-based encoder-decoder to learn temporal motion structure from unconditioned action trajectories alone; Stage 2 transfers this prior to VLA training via decoder reuse and latent distillation. Evaluated across 13 cross-embodiment tasks in simulation and real-world settings, the approach achieves faster convergence, higher success rates, and notably better performance in data-scarce real-world scenarios compared to VLA training without action priors.