"""Reward functions for verifiable-reward RL training."""