Long contextMultiple choiceMediumCopy prompt

Without changing the retriever's choices, what would you change to raise accuracy?

1. ARetrieve 30 documents instead of 20 so the answer is more likely there.
2. BSwitch to a version of the model with a longer context window.
3. CShuffle the order of the documents randomly for every question.
4. DPlace the top-ranked documents at the beginning and end of the prompt.