Learning Equilibria from Data: Provably Efficient Multi-Agent Imitation Learning | Read Paper on Bytez