AI RESEARCH

RubricEM: Meta-RL with Rubric-guided Policy Decomposition beyond Verifiable Rewards

arXiv CS.LG

ArXi:2605.10899v1 Announce Type: cross