POW3R: Policy-Aware Rubric Rewards for RLVR

A new arXiv preprint (2605.20164) introduces POW3R, a policy-aware rubric reward framework for reinforcement learning with verifiable rewards (RLVR). The authors argue that standard static rubric aggregations conflate human-assigned importance with optimization usefulness, as criteria may be saturated or unreachable. POW3R preserves human weights and category balance while adapting criterion-level reward weights during training using rollout-level contrast. This addresses the limitation that criteria distinguishing rollouts are not necessarily those with largest human weights.