OPD & PRM: On-policy Distillation and Process Reward Models