Deeplearning.ai - ChatGPT Prompt Engineering for Developers

Last updated Jul 16, 2024

# Introduction

Instrustion Tuned LLM

• the model tries to follow given instruction
• clear and specific than Base LLM
• more helpful, honest, harmless
• RLHF: Reinforcement Learning with Human Feedback

vs. Base LLMInstruction Tuned LLM

• Q: What is the capital of France?
• Base LLM: search also “what is France’s largest city” or “what is population in France”
• Instruction Tuned LLM: The capital city of France is Paris.

# Guidelines

Prompt Principles

1. write clear and specific
2. give the model time to thick