Hybrid Reinforcement Learning Breaks Sample Size Barriers in Linear MDPs Kevin Tan, Wei Fan, Y uting Wei Department of Statistics and Data Science The Wharton School, University of Pennsylvania