Learning from A Single Markovian Trajectory: Optimality and Variance Reduction | Read Paper on Bytez