在线 reinforcement learning
1 stories tagged “在线 reinforcement learning”.
-
QwenGyre:一个动态可扩展的 reinforcement learning 框架,用于训练xLong-Horizon代理人
QwenGyre通过弹性GPU重新分配和轨迹复制提升了长视野在线RL,将NL2RepoBench提高到58.5%.
1 stories tagged “在线 reinforcement learning”.
QwenGyre通过弹性GPU重新分配和轨迹复制提升了长视野在线RL,将NL2RepoBench提高到58.5%.