Search papers, labs, and topics across Lattice.
This paper introduces LIFT, a novel post-training framework that enhances pretrained vision-language-action (VLA) policies by incorporating reactive force injection, allowing for improved performance in contact-rich manipulation tasks. By integrating a reactive action expert and utilizing causal force memory alongside an online DAgger loop, LIFT effectively addresses the challenges posed by occluded scenes and depth ambiguities during execution. Experimental results demonstrate that LIFT significantly accelerates learning and improves task performance in scenarios such as towel folding and book insertion compared to traditional vision-only approaches.
LIFT accelerates learning in vision-language-action policies by injecting reactive force feedback, achieving superior performance in contact-rich tasks.
Pretrained vision-language-action (VLA) policies provide strong language-conditioned manipulation knowledge, but they remain largely vision-driven and can struggle once manipulation enters contact states where the scene is occluded, depth is ambiguous, or small force errors push execution off the offline demonstration distribution. We present LIFT (Late Reactive Injection of Force for VLA Post-Training), a force-aware post-training framework that adds contact reactivity to a pretrained VLA policy while preserving its general manipulation knowledge. LIFT grafts a reactive action expert beside the original action expert, initializes it from pretrained action weights, and injects recent 6D end-effector force through causal force memory and zero-initialized cross attention, enabling actions to be refreshed during execution. To address the policy-dependent distribution shift of contact feedback, LIFT further couples reactive force injection with an online DAgger loop that trains on a mixture of offline task-alignment data and human-corrected online rollouts. Across towel folding, book insertion, and Hanoi ring placement, LIFT learns faster and reaches higher performance than vision-only post-training, while ablations show that reactive force memory and online corrective data are both important for robust contact-rich manipulation. Our code and data will be publicly available.
