Break lines characters seems to be semantically relevant

Hi does anyone detected prompts results may be affected on how break lines are represented? I mean sometimes same messages sent to open AI api with different line breaks representation (\r\n vs \n) produce very different results.

1 Like

Welcome to our dev community forums!

Yes, those tokens are taken into account.

1 Like

Ok thanks for your quick answer. We’ll keep an eye on breaklines then :slight_smile:

1 Like

No problem. Hope you stick around. We’ve got a wealth of hidden gems and a lot of great people. Enjoy your stay.