Being specific

Prompt Engineering: The Concept of Specificity

• Goal of Specificity: To direct the language model (LLM) toward a desired output, preventing infinite variations in length, structure, or content.
• Two Primary Techniques for Specificity:

• Type A: Output Attributes/Guidelines
• Lists specific qualities or constraints the final output must possess.
• Used to control the characteristics of the response (e. g. , length, structure, required elements).
• Example: Keep the story under one thousand words; include at least one supporting character.

• Type B: Sequential Steps
• Provides a defined process or workflow the model must follow to generate the response.
• Used to control how the model arrives at the final product, forcing it to consider specific stages.
• Example: First, brainstorm three special talents; second, select the most interesting one; third, outline the scene.

• Best Practices:
• Both Type A and Type B techniques can and often should be mixed in professional prompts.
• General Recommendation: Almost always use Type A (listing output qualities/attributes).
• When to use Type B (Steps): Use sequential steps for complex problems where you need to force the model to consider a wider view or topics beyond its natural inclination.

Takeaways

• Specificity directs the language model (LLM) toward a desired output, preventing infinite variations in length, structure, or content.
• Type A (Output Attributes/Guidelines) controls the characteristics of the final response, such as length or required elements.
• Type B (Sequential Steps) controls the process or workflow the model must follow to generate the response.
• Type A is the general recommendation and should almost always be used in professional prompts.
• Type B is best used for complex problems to force the model to consider a wider view or specific stages.