mimo-vl-vision-language

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.

Independent GitHub snapshot · Automated classification · Original documentation governs