MDP-GRPO: Stabilized Group Relative Policy Optimization for Multi-Constraint Instruction Following 事件

MDP-GRPO: Stabilized Group Relative Policy Optimization for Multi-Constraint Instruction Following arXiv:2606.06058v1 Announce Type: cross Abstract: Reinforcement learning with verifiable rewards is ideal for multi-constraint instruction following, yet standard group-relative policy optimization (GRPO) becomes unstable under discrete, low-dispersion rewards, where within-group reward distributions are frequently homogeneous. We identify and formalize three pathologies of z-score group normaliza