Optimal Single-Policy Sample Complexity and Transient Coverage for Average-Reward Offline RL | Read Paper on Bytez