Choose a prompt-iteration sequence that uses zero-shot baseline results before adding few-shot examples.
The queue lead needs a classifier by lunch. Marcus has label definitions but no evidence that examples will help yet. What should he do next? Run the label definitions zero-shot on a fixed test slice, record misses, then add few-shot examples only for repeated confusions. Right. This follows an iterate-and-evaluate loop: baseline first, diagnose failure, then use examples to improve consistency where the model actually needs guidance. Add ten polished examples immediately so the model sees what good tickets look like. Weaker. Examples can help, but adding them before a baseline may encode accidental style and leaves Marcus unable to…
Sign up free: one personalized lesson every day, matched to your role and goals.
Already have an account? Sign in