Hi all! Recently created a fine-tuned version of Gemini 1.0 Pro tuned on a proprietary dataset. For now, our goal is to be able to get an input sentence converted into a certain step-by-step process as text output. We want about 10-20 of these sentences converted in a single prompt. I’ve just discovered Structured Prompting, and want to explore that in hopes of getting these batch results better than in a chat environment. However, it seems Structured Prompts rely on you putting in your own example input-output-pairs for the model to follow.
- Do I still need to put in these Examples for a Structured Prompt if the model has already been fine-tuned on the data I want it to mimic?
- If so, is it a good idea to reuse the tuning data as the Examples, or would I want new data? Since I didn’t know about the Examples feature, I used all the data I had available (barring testing reserves) for the Tuning process.
Please also let me know if I’ve got the wrong picture about something, or Structured Prompts don’t sound like what I’m looking for. Thanks in advance!