Citadel interview question

can neural networks with appropriate regularization replicate rl training