Course Link

ChatGPT Prompt Engineering for Developers - DeepLearning.AI

Types of LLMs

General LLMS vs ChatGPT like “instruction tuned LLM” ( built using RLHF - Reinforcement Learning with Human Feedback)

image.png

key takeaways

Foundational Principles

  1. Clear and specific prompts lead to better results. Avoid vague instructions—be direct about what you want.
  2. Delimiters help with clarity. Use things like triple backticks to separate instructions from content or context.
  3. Ask the model to think step-by-step. You’ll get more accurate outputs when the model is guided to reason.
  4. Use examples to guide behavior. Few-shot prompting—giving one or two examples—often improves performance.
  5. Iterate and refine. Good prompts don’t always come on the first try; test and tweak them to improve results.

Working with Different Use Cases

  1. Summarization works well with clear guidance. Specify word count, tone, or summary type (e.g., bulleted list or paragraph).