Question about experts used in this model

[Lorenz5622]

What are the specific models of the four experts used in the MoE mentioned in this paper? Do they have their own specialized fields (such as text processing and image analysis), or are each expert a multimodal large model?
这篇论文中提到的MoE中使用了四个专家，这四个专家具体是哪些模型？它们是有各自专精领域（如有的专门处理文本，有的处理图像分析）还是每个专家都是多模态大模型？