Mixture-of-Minds: Multi-Agent Reinforcement Learning for Table Understanding Written on October 24, 2025