ModelsUS vs ChinaOfficialSelected85°00:21Chinese contentOpenAI Dota 2：自我对弈实现超人类表现展示了自我对弈在复杂策略游戏中突破监督学习上限的潜力，对强化学习和AI博弈研究具有重要参考价值。#self-play#reinforcement-learning#dota-2#superhuman2 sourcesEventOPrimary sourceOpenAI BlogSourceRead laterReadWorth followingUsefulFollow self-play