User Guide: Advanced Prompting

⚙️ Core Concepts

Every AI model has its own unique training approach, which means standard prompting techniques might not always give you the best results. Let's explore how to optimize your prompts for maximum effectiveness.

🛠️ Essential Techniques

1️⃣ Using XML Tags for Structure

XML tags are powerful tools for organizing your prompts. They help separate different types of information and make your instructions crystal clear:

• Instructions vs. context
• Examples vs. input data
• Questions vs. requirements

<text>
Your input content here
</text>

<example>
Sample interaction here
</example>

💡 Pro tip: You can use any tag names you prefer - what matters is the consistent structure with opening and closing tags.

2️⃣ Being Direct and Specific

Clarity is key when working with AI. Here's how to make your prompts more effective:

• Use affirmative instructions ("do this") rather than negative ones ("don't do that")
• Provide detailed context
• Specify exactly where to find information using tags
• Be explicit about expected outputs

Before:

Summarize this text.

After:

Summarize the main ideas from the provided article within the <text> tags, focusing on key conclusions and actionable insights.

3️⃣ Controlling Output Format

AI models can sometimes be chatty or inconsistent in their responses. Here's how to maintain consistent outputs:

• Provide explicit format templates
• Use example outputs
• Start the response with a specific token or format

Example:

Follow this format for your response:
<format>
→ Key finding 1
→ Key finding 2
→ Key finding 3
</format>

A: →

💡 Important: Starting with "A: →" helps ensure the AI begins directly with the requested format.

4️⃣ Assigning Roles

Role assignment helps the AI understand the context and style needed:

• Be specific about the expertise level
• Define the role's responsibilities
• Set clear expectations

Example:

You're an expert technical writer with 15 years of experience in API documentation.

5️⃣ Encouraging Thoughtful Responses

For complex tasks, guide the AI through a structured thinking process:

When generating your response:
1. First analyze the problem in <thinking></thinking> tags
2. Then provide your solution in <answer></answer> tags

💡 Advanced Strategies

Using Examples Effectively

Few-shot prompting can dramatically improve response quality:

• Include 2-3 relevant examples
• Show both input and expected output
• Cover different scenarios
• Use consistent formatting

Managing Long Documents

When working with extensive content:

• Place important instructions at the end of your prompt
• Use clear section markers
• Break complex tasks into steps
• Reference specific sections using tags

Step-by-Step Processing

For complex tasks:

1. Break down the main task into subtasks
2. Provide clear instructions for each step
3. Request intermediate outputs if needed
4. Specify how to combine results

Example:

Please follow these steps:
1. Extract key metrics from the data
2. Analyze trends and patterns
3. Generate insights
4. Format the final report

💡 Remember: Effective prompt engineering is an iterative process. Keep testing and refining your approaches based on results.