It's Like Davinci-IS and Curie-IS Generate Better Completions Through Using Their Own Completions. And a New Question

Prompting
1

Hello everyone,

So recently I’ve been testing how I can get my prompt to be better effective in giving me what I want, and I, now at this point youre thinking this is absurd, but I got better completions after using ones outputted by the engines themselves, with minor editing.

For a few weeks I have been struggling to the engines to give me proper results, but I realized throught the experience, the “best of” setting is your friend, the frequency setting is your best friend, and temperature is your BESTEST FRIEND.

Literally the one digit above or below in temp made the difference between a proper output or garbage over creative nonsense. A few examples (two to three) are enough, and anything more would seem to confuse the engines. All three settings are extremely important in getting a proper output. I am now able to use Curie IS after being absolutely incapable of getting any decent output! But still more to learn! I have not totally mastered the outputs.

Which brings me to a new question, is there such a thing as a perfect prompt and perfect settings?

2 Likes
2

Further findings:

If you find the completion has ended with “and they are:” when you prompted the engine to list a points, continue the prompt by submitting “Please continue” so that it lists the point. Keep prompting it to “Please continue” until you get the number of points desired.

3

I’m sorry but, WHAT HAPPENED WITH DAVINCI IS? IT IS NOW LITERALLY COMPLETING LIKE REGULAR DAVINCI!!!

5

Interesting. Only after numerous submissions, did it finally snap back into gear. Havent used it for over a week, and came back surprised. Settings and everything are exactly the same, It’s like it needed to wake up or something. Would love an explanation, or to be called a moron, thanks

1 Like