Uses reinforcement learning to solve for the optimal tax policy in an agent-based world. Tax policy ends up with a tax policy that isn't progressive but has substantial transfers. Outperforms other frameworks
link