The authors demonstrate the fast convergence of no-regret learning algorithms to correlated equilibria in Markov games, closing a gap in existing research by achieving eO(T^-1) rates.
