Merge pull request #196 from openai/ted/undo-incorrect-fix

restores correct split between prompt and completion
2 years ago · dc757f8997
parent 3a1c3af2df f9ee6221a5
commit dc757f8997
1 changed files with 5 additions and 1 deletions
--- a/techniques_to_improve_reliability.md
+++ b/techniques_to_improve_reliability.md
@ -18,7 +18,11 @@ As an example, if you ask `text-davinci-002` the following math problem about ju

 ```text-davinci-002
 Q: A juggler has 16 balls. Half of the balls are golf balls and half of the golf balls are blue. How many blue golf balls are there?
-A: There are 8 blue golf balls.
+A:
+```
+
+```text-davinci-002
+There are 8 blue golf balls.
 ```

 Does this mean that GPT-3 cannot do simple math problems? No; in fact, it turns out that by prompting the model with `Let's think step by step`, the model solves the problem reliably: