[2210.01050] Faster Last-iterate Convergence of Policy Optimization in Zero-Sum Markov Games