SESSION 10
Session 10 — Prompt Engineering
Improve eval scores with disciplined prompt iteration: specify, structure, constrain, align.
1.5–2 hours•6 exercises · 2 phases
What you'll be able to do by the end
- ✓ Follow the measure-compare-keep discipline for prompt changes
- ✓ Apply the 6 prompt engineering stages: Specify, Structure, Demonstrate, Constrain, Align, Iterate
- ✓ Run A/B tests comparing two prompt variants on the same eval dataset
- ✓ Align generator and judge vocabulary to avoid false low scores
- ✓ Use a pre-run checklist before trusting any score improvement
Prerequisites
The arc
First learn the discipline and the engineering levers. Then apply them with real A/B comparisons.
Discipline
Loop + Stages
Measure
A/B + Checklist
Exercises
The Discipline
The improvement loop and the core engineering stages: specify, constrain, structure, demonstrate.
Measure & Compare
Align prompts with rubrics, run A/B tests, and validate results with a checklist.