Key points
- Few-shot examples are the most effective technique for consistently formatted, actionable output when instructions alone fall short.
- Examples demonstrate ambiguous-case handling, such as tool selection for unclear requests or whether a branch counts as a coverage gap.
- Good examples let the model generalize judgment to novel patterns rather than only matching the cases shown.
- In extraction, examples reduce hallucination on informal measurements and varied document structures.
- Two to four targeted examples, each with the reasoning for the chosen action, are usually enough.
Skills the exam expects
- Create 2-4 examples for ambiguous scenarios that show why one action beat plausible alternatives.
- Demonstrate the exact output format (location, issue, severity, suggested fix).
- Pair acceptable code patterns with genuine issues to cut false positives while keeping generalization.
- Show correct extraction across document layouts (inline citations vs bibliographies, methodology sections vs embedded details).
- Use examples to fix empty or null required fields in extraction.
Common misconceptions
- "More examples are always better." A handful of targeted, ambiguous cases beats a pile of easy ones.
- "Telling the model not to hallucinate works as well as showing it." Demonstration teaches handling; prohibition does not.
- "Examples only teach the exact cases shown." They are chosen so the model generalizes.
Watch for
- A third page of instructions when two pages have already failed.
- Examples that only cover obvious cases and leave the ambiguity unaddressed.
- Fixing empty fields with retries when the real gap is unseen document layouts.
Practice task 4.2
56 practice items cover this task in CCA Prep Studio: flashcards, matching, scenario questions and explain-it-back exercises, each tracked toward your mastery score.