I fine-tuned Davinci on 9 prompt styles excluding prompt 5 on Chapter 1, but the performance significantly decreased, with style0 (included) as 0.0, style3 (included) 0.17, style5 (excluded). I included low-token style0 this time in training for easier comparison with Aly's results.
For chapter 5 (the previous results), it was 0.67 for included and 0.47 for excluded.
Hi Mike,
I fine-tuned Davinci on 9 prompt styles excluding prompt 5 on Chapter 1, but the performance significantly decreased, with style0 (included) as 0.0, style3 (included) 0.17, style5 (excluded). I included low-token style0 this time in training for easier comparison with Aly's results. For chapter 5 (the previous results), it was 0.67 for included and 0.47 for excluded.
I'm wondering what I should do for the next step.