Training Language Models for Social Deduction with Multi-Agent Reinforcement Learning Paper • 2502.06060 • Published Feb 9 • 34