One of the coolest at-scale RL resources made public yet! You love to see it.
Nearly half a year of silence. We spent it studying one problem: how far RL can scale. MiMo-V2.6 is in the middle of its RL run right now. Three things we scale...
小米MiMo-V2.6正进行大规模强化学习训练,并公开了训练细节。其RL扩展了三个维度:计算(每步约2B tokens,1568 prompts × 16 rollouts,全异步)、环境与harness(多任务智能体RL,单次运行混合多种harness)以及评分计算(智能体组内信用分配,基于测试用例和评分标准的奖励)。相关细节将在未来几周逐步开源。
小米MiMo-V2.6正进行大规模强化学习训练,并公开了训练细节。其RL扩展了三个维度:计算(每步约2B tokens,1568 prompts × 16 rollouts,全异步)、环境与harness(多任务智能体RL,单次运行混合多种harness)以及评分计算(智能体组内信用分配,基于测试用例和评分标准的奖励)。相关细节将在未来几周逐步开源。
One of the coolest at-scale RL resources made public yet! You love to see it.
来源:Nathan Lambert· x.com