RARM: Confidence-Gated Progress Reward Modeling for RL in Manipulation
Industry
Source: arXiv cs.ROPublish time unverified
arXiv:2606.22027v3 Announce Type: replace Abstract: Reinforcement learning for robot manipulation is often bottlenecked by reward design, especially in long-horizon tasks: sparse success rewards provide weak supervision, while hand-crafted dense rewards are tedious to design and generalize poorly across tasks.