Golden Examples Set: Collecting and Maintaining Reference Answers
Build a small set of reference interactions and keep it current.
Added October 15, 2026. Written with AI assistance and checked for clarity and placeholders; not reviewed by a lawyer, accountant, clinician or other professional. How these prompts are made
The Open buttons send the prompt text as currently shown below to that site only when you click. Fill in the placeholders first.
Help me build a golden set for an assistant that [ASSISTANT_PURPOSE]. Sources for realistic inputs: [INPUT_SOURCES]. Experts who can review answers: [EXPERTS]. How often the domain changes: [CHANGE_FREQUENCY]. Define what a good golden example contains: the input, the context supplied, the ideal answer, the sources behind it, the date last verified, and the scoring notes. Provide a target mix (common, hard, borderline, refusal) and the number needed to start (for example 40). Write the collection routine: sample real inputs, remove personal data, draft an answer, have an expert verify, record disagreement, and store. Include a review calendar with triggers for re-verification (policy change, product change, a failure in production). Add a table template and a naming convention, and instructions for adding new failures as regression cases. Mention how to avoid overfitting prompts to the set. Return the template and a sample of three entries with invented content. Add a worked entry with invented content: input 'Can I return a sale item after 20 days?', context policy v3 (30 days, original packaging), ideal answer, source line, date verified 3 March, and scoring note 'must mention packaging'. Then list what to do when the policy changes to v4.
Fill in before use
Optional: type your details and the prompt above updates. Nothing is saved or sent.
Replace each bracketed placeholder with your own details, then paste into Any AI or another assistant.