Prompt Overview and Tips

table of contents

1. What is a prompt?
2. The role of prompts
3. The importance of prompts
4. Tips for writing prompts
5. Examples
6. summary
7. References

---

1. What is a prompt?

A prompt is a piece of text or instructions that is fed into an LLM (large-scale language model) to produce a response that is generated for a specific purpose. Prompts are designed based on the information or task a user is looking for and guide how the LLM should operate to produce that response.

In essence, it is a set of instructions for the AI.

2. The role of prompts

The prompt serves the following purposes:

• Question format : The text used when a user asks a specific question.
• Form of instruction : A piece of text instructing the LLM to perform a specific task.
• Providing Context : Text that provides background information or facts that the LLM needs to generate a more accurate response.
• Output format specification : Instructions for generating the LLM output in a specific format (e.g. JSON, Markdown).

3. The importance of prompts

Prompts have a significant impact on the quality and accuracy of LLM responses, so it is important that they are designed properly. The importance of prompts is evident from the emergence of the job title "prompt engineer" and the high demand for prompts.

4. Tips for writing prompts

Here are some tips for writing effective prompts:

1. Write concisely : Even if it is the same command, a shorter sentence will make it easier to convey your intention to the LLM and will also be less costly.
2. Write clearly : Do not write in a way that leaves room for interpretation.
3. Be informative : Including necessary background information and examples will elicit more accurate responses.
4. Use delimiters : If you have multiple instructions or elements, it is a good idea to use delimiters to clearly separate them.
5. Assign roles : By assigning specific roles to LLMs, you can control the style and tone of the response you expect.

5. Specific Examples

Here are some specific examples of writing prompts:

Keep it concise

Bad Example :

Good example :

Write clearly

Bad Example :

Good example :

Give sufficient information

Bad Example :

Good example :

Please write a report of no more than 1000 words on the following topic:

Topic: The future of AI and its social impact

Use delimiters

Bad Example :

Good example :

Please answer the following three questions:

question: """

1. Today's Weather

2. Tomorrow's weather

3. This weekend's weather

"""

Assigning roles

Bad Example :

Good example :

6. Summary

Prompts are an important factor that greatly affects the quality and accuracy of LLM responses. To write an effective prompt, it is important to keep in mind the following tips: conciseness, clarity, sufficient information, use of punctuation marks, and role assignment. Please refer to the concrete examples to help you create your own prompts.

7. References

• OpenAI Guide to Prompt Engineering
• Best Practices for Prompt Engineering with the OpenAI API