Teach the model by example: supply 2-3 input to output samples, then have it apply the pattern to your task.
Prompts / Techniques / Few-Shot Exemplar Set Builder With Edge Cases
Few-Shot Exemplar Set Builder With Edge Cases
Designs a balanced, labeled few-shot example bank that teaches a task by demonstration.
ROLE: You are a prompt-engineering instructor who specializes in in-context learning. CONTEXT: I need few-shot examples that teach a model to perform [TASK_DESCRIPTION] on inputs shaped like [INPUT_FORMAT], producing [OUTPUT_FORMAT]. TASK: 1) Restate the task as a one-line rule the examples must demonstrate. 2) Write [N=6] input/output pairs ordered from prototypical to ambiguous. 3) Include at least two near-miss edge cases (tempting wrong answer plus the correct one). 4) Add one counterexample labeled as such, showing what NOT to produce. 5) For each pair, note in one phrase the specific pattern it teaches. CONSTRAINTS: Keep every example self-contained; vary surface wording so the model learns the rule, not the phrasing; never reuse the same entity twice; do not include explanations inside the examples themselves. OUTPUT FORMAT: A numbered list of pairs (Input: / Output: / Teaches:), then a final 'Ordering rationale' paragraph of 2-3 sentences explaining the chosen sequence.