top | item 42326282 (no title) 0xFF0123 | 1 year ago Language models are known to be a bit crap at maths though in comparison to natural language prompts discuss order hn newest margalabargala|1 year ago Nah, 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.
margalabargala|1 year ago Nah, 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.
margalabargala|1 year ago
Regardless, the intention here is to highlight a difference between Gemini and ChatGPT/Claude, neither of which will agree to simple mathberrors.