Retrieval before reasoning
The instinct with language models is to make them smarter. Usually the bigger lever is making sure they're looking at the right thing in the first place. Why grounding, not raw intelligence, is what makes AI trustworthy enough for real work.
Read in full →