Review & Rate Outputs

Our output results are ready in the Evaluation section:

Now, we need to rate each one.

We need to think about what makes a good email subject line for our use case.

On the one hand, they are email subject lines (that's a start), and the content seems relevant to the input. On the other hand, they have double quotes around them, which is not what we want. We'll subtract 2 stars for improper formatting, and rank each one from 1 to 3 stars based on how much we like the result.

You might notice that these results from GPT-4 Omni are eerily similar to the ones in our validation examples. That's because when writing this guide, we used GPT-4 Omni to come up with those email subject lines in the first place! If you selected a different model, you would get more diverse outputs.

To fix the formatting, we need to update our prompt.