Best Practices regarding defining references for AI

I’m having consistency problems with the AI not repeating itself 100% of the time.

Is option A or B better to get more consistency???.

OPTION A
To define all your contexts at the beginning of the prompt then ask all the questions the AI must answer? (ie requiring the AI to have good memory)

Example
CONTEXT
Here is article 1.
Here is article 2.
Here is article 3.

QUESTIONS
Summarize article 1.
Summarize article 2.
Summarize article 3.

OPTION B
Define the context, then immediately ask the question, then define additional context, then ask the next question?

CONTEXT
Here is article 1.
Summarize article 1.

Here is article 2.
Summarize article 2.

Here is article 3.
Summarize article 3.

Thanks…

Hi @VFX

I’d love to see papers about this!

From my personal experience it will depend case-by-case and when you are writing a prompt that will be used in an app, you need to do a lot of “unit” testing. That gives you the answers you need.