Comment by mcv

Hardly surprising that models specialised at language aren't as good at a highly specialist task as a model trained specifically at that task, isn't it?

I know some people want to believe that LLMs are good at everything, but they're not. They're good at language.