Provably Efficient Reinforcement Learning with Multinomial Logit Function Approximation | Read Paper on Bytez