Prompt experiment: instructions alone versus one worked example

Here is an experiment idea for a repeated task: compare a prompt containing instructions alone with a version that also includes one worked example. No outcome is assumed.

1. Choose a narrow task, such as assigning a label from a fixed list.
2. Prepare shareable test inputs and an expected result or scoring rubric.
3. Write prompt A with clear instructions.
4. Create prompt B by adding one example while keeping the other instructions the same.
5. Use the same model and settings, record each run, and compare the outputs.

Keep the worked example separate from your test inputs. Record failures as well as successes, and explain how you scored ambiguous cases.

Post both prompts, the model identifier, test inputs, outputs, and your interpretation. What changed, and would the result still be useful on a different set of inputs?