Distribution Learning with Valid Outputs Beyond the Worst-Case | Read Paper on Bytez