Bug in tackling mathematical questions

The AI sometimes gives inaccurate feedbacks when answering mathematical question.
some of the errors are really basic and stupid like addition and subtraction errors.

its not designed for solving stuff like that. if you need to do addition use a calculator xD

Large Language Models are Zero-Shot Reasoners | Simply adding “Let’s think step by step” before each answer increases the accuracy on MultiArith from 17.7% to 78.7% and GSM8K from 10.4% to 40.7% with GPT-3.


Notes on Teaching GPT-3 Adding Numbers by Ekin Akyürek and Afra Feyza Akyürek