Inconsistent responses between playground and API

Unlike other topics about inconsistency, I really use exactly the same options and prompts. (See vscode and playground code preview side-by-side below.)

I’m trying to parse some unstructured data from pubmed.

Prompt:

Extract only author's names from the following data and print them as a comma-separated list (no spaces after commas). From each section delimited by "@@", extract a author's name only once.

Stamey T A,Meares E M,Meares E M,Winningham D G@@de Ruiter P,Vlietstra H P

Playground output:
Stamey T A,Meares E M,Winningham D G,de Ruiter P,Vlietstra H P

API output:
Stamey T A,Meares E M,Meares E M,Winningham D G@@de Ruiter P,Vlietstra H P

You have “temperature: 1”.

That’s like asking why the game of Yahtzee is not the same every time, even though you are using the exact same dice.

1 Like
  1. If that was the root cause, results should be inconsistent between runs regardless of platform. Yet, I’ve tried this example about 5 times with each playground and API and both are perfectly consistent in themselves.
  2. I just tried both with a temperature of 0 and nothing changed.

Never mind - a really stupid mistake on my part. I was accidentally looking at the input instead of the output… :person_facepalming:
It works perfectly fine.