Prompt A/B testing and evaluation

Prompt A/B testing and evaluation

Why it matters: A/B testing prompts reveals which changes improve quality reliably.

Goals

• Define evaluation criteria.
• Run a small A/B test.
• Pick a winner with evidence.

Key definitions

• Variant: A version of a prompt with one intentional change.
• Evaluation set: A set of inputs used to compare variants fairly.
• Metric: A measurable signal of quality (rubric score, error rate, time to edit).

Workflow (step-by-step)

1. Define 1–3 metrics (rubric score, factual accuracy, edit time).
2. Create an evaluation set (5–10 representative inputs).
3. Write variant A and variant B with one change.
4. Run both on the same set and score with a rubric.
5. Adopt the winner and document the change.

Example (good vs bad)

✅ Good: You compare prompts using the same inputs and a rubric, then adopt the winner.

❌ Bad: You pick a prompt based on a single run and vibes.

Checklist

• One change per variant.
• Same inputs used.
• Rubric scoring used.
• Winner documented.

Metrics / criteria

• Rubric score improves; edit time decreases.
• Measurable: You can say whether the output is correct/complete (pass/fail or a score).
• Effort: You can produce the result in 15–30 minutes using the workflow.
• Clarity: Another person can run your prompt and get a similar outcome.

Common mistakes + fixes

• Pitfall: Changing multiple variables at once. Fix: Only change one thing per variant.
• Pitfall: Vague instructions. Fix: Add constraints, examples, and a success checklist.
• Pitfall: One-shot prompting. Fix: Iterate: draft → critique → revise → verify.
• Pitfall: No validation. Fix: Add QA steps, tests, and explicit “what counts as done”.

Student tasks

1. Apply the workflow to one real work task related to prompt a/b testing and evaluation.
2. Create a small artifact: prompt, checklist, rubric, table, draft, or decision note.
3. Run one quality check and record what changed after the check.
4. Write a 3-sentence reflection: what worked, what failed, and what you will reuse.

Useful external sources

• OpenAI prompt engineering guide: https://platform.openai.com/docs/guides/prompt-engineering — Prompt structure, iteration, constraints, and evaluation habits.
• NIST AI Risk Management Framework: https://www.nist.gov/itl/ai-risk-management-framework — Risk framing, governance language, and verification mindset.
• OWASP Top 10 for Large Language Model Applications: https://owasp.org/www-project-top-10-for-large-language-model-applications/ — Practical AI safety risks such as prompt injection, leakage, and insecure output handling.
• Google People + AI Guidebook: https://pair.withgoogle.com/guidebook/ — Human-centered AI design, user control, feedback, and failure modes.

Bibliography

• OpenAI prompt engineering guide. https://platform.openai.com/docs/guides/prompt-engineering
• NIST AI Risk Management Framework. https://www.nist.gov/itl/ai-risk-management-framework
• OWASP Top 10 for Large Language Model Applications. https://owasp.org/www-project-top-10-for-large-language-model-applications/
• Google People + AI Guidebook. https://pair.withgoogle.com/guidebook/