Provably Efficient Interactive-Grounded Learning with Personalized Reward | Read Paper on Bytez