6 ms·
Language models are known to be a bit crap at maths though in comparison to natural language prompts
by 0xFF0123 2y ago
Language models are known to be a bit crap at maths though in comparison to natural language prompts
- margalabargala 2y agoNah, for complex problems maybe, not for single digit addition that should be in the training corpus directly. Regardless, the intention here is to highlight a difference between Gemini and ChatGPT/Claude, neither of which will agree to simple mathberrors.