Ground answers in sources

Lower hallucination risk by giving source material, asking the model to use only that source, requesting uncertainty when needed, and checking important claims elsewhere. You will learn what prompting can improve and what still needs verification.