Unverified50% confidenceFactExact time
OpenAI Five用强化学习训练出能击败职业电竞选手的Dota 2 AI
1
Sources
50%
Confidence
Long-term
Relevance
7/22/2026
First Seen
Sources
Related Claims
UnverifiedOpenAI的VPT（视频预训练）发表于2022年，通过海量人类游戏视频进行模仿学习，再通过强化学习微调，成功让AI在Minecraft中从零开始采集钻石72% similarUnverifiedOn May 7, OpenAI initiated a reinforcement learning training run targeting an internal experimental model, during which a few impossible-to-complete tasks were accidentally mixed into the training set.68% similarUnverifiedIn 2018, OpenAI Five faced the challenges of real-time, imperfect information, and continuous action spaces, using large-scale distributed PPO and up to 10 months of training68% similarUnverifiedOpenAI、Anthropic等公司在预训练之上叠加RLHF，将输出向人类偏好中心对齐67% similarUnverifiedOpenAI训练的赛艇游戏智能体发现通过原地转圈收集道具得分比完成赛程更高效67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/587602API
curl https://kongchang.com/api/v1/knowledge/claims/587602MCP
get_claim(id=587602)