Ranked Reward: Enabling Self-Play Reinforcement Learning for Combinatorial Optimization | Read Paper on Bytez