Comment by ddevnyc

Does MoE help with multimodality? Can it in general enable reasoning in imagery (technical drawings, diagrams, schematics) rather than text-based?