LLMs are known to do bad calculations because all it does is to predict the next token. It’s pretty annoying, and I don’t understand why calculation (eg via Pyt

LLMs are known to do bad calculations because all it does is to predict the next token. It’s pretty annoying, and I don’t understand why calculation (eg via Python) is still not a core functionality