Comment by purple-leafy

Dumb question: can you train a model to predict the next byte of ANOTHER MODEL

So apply this same logic to compressing a bigger model within a smaller model

I know this is absolutely regarded, but humour me please