> this proves that all llm models converge to a certain point when trained on the same data They are also all trained to do well on the same evals, right? So d

> this proves that all llm models converge to a certain point when trained on the same data They are also all trained to do well on the same evals, right? So doesn't it just boil down to neural nets being universal function approximators?