14 ms·

You should repeat this experiment by feeding the answer of a GPT instance to another one as an input and let it judge the correctness.

• jonnycomputer 3y ago
  That's a good idea.