mimo-vl-vision-language

【效率与协作】Build state-of-the-art 7B multimodal models by combining four-stage vision-language pretraining with mixed on-policy RL integrating verifiable and human feedback rewards.

独立 GitHub 快照 · 自动分类 · 以原作者文档为准