I wonder, is this exponential relation specific to multi-modal models? From my admittedly naïve view it seems to make sense that "...what is rare is not proper

I wonder, is this exponential relation specific to multi-modal models? From my admittedly naïve view it seems to make sense that "...what is rare is not properly learned" would apply generally?