This model is the stage 1 checkpoint of one of the thirteen settings, CLIP@336, used in the [Law of Vision Representation in MLLMs](https://arxiv.org/abs/2408.16357).