Generating Questions in Rise with AI - Model Refinements

Typically I find myself copying and pasting the question and answers into a quiz block in Rise from other source material. To make the process more efficient, I typically copy the entire question with all of the answers, and after each answer I include the specific feedback related to each answer. Then I select Add Question > Generate with AI. I use this method because if I am manually importing many questions I don't want to have to copy the question, answers, and each feedback separately. That would be a lot of unnecessary work.

When I use the "Generate Question" feature in Rise, I have found that things start changing that you didn't ask for:

• The wording of the question changes
• The order of the answers change
• The specific feedback I provide for each answer also changes, particularly when I am trying to give the "Correct/Incorrect / Why what you selected was wrong or why it was right / When Incorrect, give the correct answer"

To counteract this, I started including very specific instructions in my prompt such as "do not change the order of the answers" and "do not modify the original question". And what I have found is that...it does not listen very often. I have to repeat it, and then sometimes it does catch on, and other times it just doesn't understand.

It feels defeating prompting with absolute specificity only to then have very inconsistent results.

My suggestion: train the underlying LLM better or allow for a copy/paste method that would work without the need to click into each individual field separately.

Thanks! 

- Frankie