Search papers, labs, and topics across Lattice.
2
0
3
0
Mid-generation intervention during LRLM inference can preserve valid robotic plans and correct errors, boosting success rates and efficiency beyond one-shot or iterative refinement alone.
Safe RL policies, designed to avoid unsafe actions, can be effectively attacked using a novel framework that learns safety constraints from demonstrations and then crafts adversarial perturbations, even without access to the target policy's gradients.
