This work introduces AVP (Action with Visual Primitives), an end-to-end architecture that implements this visual-primitive-centric interface: the VLM infers the next-stage target and emits visual-primitive tokens that condition a flow-matching action expert, with supervision derived from end-effector kinematics.
T2T(Thickening-to-Thinning), a dynamic reward framework inspired by human learning processes, implements a dual-phase mechanism that significantly outperforms standard GRPO and recent baselines, achieving superior performance.
Wen-Ze Lin, Zhen Yang, Xi-Tai Jiang et al.· 1 citation
This work reformulates the implicit reward of sampled-token OPD based on trajectory correctness, then applies a ReLU gating mechanism to ensure that correct trajectories receive non-negative rewards and incorrect ones receive non-positive rewards, making it readily combinable with any policy gradient algorithm, such as...
