Open AI Prompt Engeneering Guide
Open AI Prompt Engeneering Guide
Write Clear Instructions:
Be Specific: Clearly articulate your request, as if guiding someone unfamiliar with the task.
Format Matters: Choose a format (like bullet points, paragraphs, etc.) and ask the model to follow it.
Length Control: Specify whether you want a brief summary or a detailed explanation.
Provide Examples: Offer examples of the type of response you're looking for.
Use Delimiters: Clearly separate different parts of your input with symbols or headings.
Provide Reference Text:
Contextual Information: Give background information or specific details to guide the response.
Citation Requirement: If needed, instruct the model to include citations from the reference text.
Split Complex Tasks into Simpler Subtasks:
Break Down Tasks: Divide a complex request into smaller, more manageable parts.
Sequential Approach: Tackle each part one at a time for better accuracy and coherence.
Summarize in Stages: For long inputs, summarize parts individually before creating an overall summary.
Give the Model Time to 'Think':
Chain of Thought: Ask the model to explain its reasoning process step by step.
Sequential Queries: Build up to the final answer with a series of related questions.
Review and Reflect: Instruct the model to reconsider its previous responses for potential improvements.
Use External Tools:
Leverage Technology: Utilize specific functions or tools that the model can access for better results.
Knowledge Retrieval: Use tools for efficient information gathering, like embeddings-based search.
Code Execution: For precise calculations or executing code snippets, use a code execution engine.
Test Changes Systematically:
Evaluate Performance: Regularly compare the model's responses to a set of high-quality reference answers.
Adjust and Improve: Continuously refine your prompts based on the evaluation to enhance performance.
Categories
