| created |
modified |
tags |
type |
status |
|
2025-02-12 09:09 |
llm |
large-language-model |
gpt |
chat-gpt |
chat-interface |
nlp |
natural-language-processing |
prompt |
prompting |
prompt-engineering |
|
|
|
This is a simple prompt engineering approach I learned from Gianluca Mauro.
The idea is to separate your prompt into Context, Instructions, Details, Input:
- Set the scene for the model - important background information framing the task.
Context is designed to give the LLM a clear role and style.- Example: “You are a copywriter at an ice cream company specialising in healthy ice creams, targeting affluent customers”
Instruction(s) specify the task that you want the LLM to achieve- Be assertive and crystal clear. The LLM doesn’t have feelings, so clarity trumps tact.
- Example: “I am going to give you a message and you are going to write an instagram post containing this message”
- Think about what an ideal output would look like, and describe it here. Provide some model examples.
Input is information/data directly relevant to this specific request/task.Input differs from context in that context is the kind of information that may be reused with different inputs, whereas input is information specific to only this request/task.- Example: “The message I want you to write an Instagram post about is 'we are releasing a new product called choco-early, which contains no dairy or cane sugar. The target audience for this post is middle-aged men.'”
- Links to references (source material) go here