exactly, that is the current objective. To proove that generation for a specific domain is on-par with causal attention models