Prompt Basics

Prompts are a technique for effectively communicating with LLM (Large Language Model). The quality of the results greatly depends on how you structure the prompt (instructions, context, and sometimes examples).

Default prompt

Even simple prompts can be powerful. Let's look at an example:

Basic prompt example

Propmt : 고양이는

Output : 고양이는 사랑스럽고 독특한 동물 중 하나로, 수많은 사람들에게 사랑을 받고 있습니다.

This example shows how the model follows a given prompt, but it also highlights the importance of context and clear instructions, as simple prompts like this may not always match the intended task.

Propmt : 문장을 완성하세요: 고양이는

  Output : 고양이는 우아한 모습과 호기심 많은 성격으로 많은 사람들의 마음을 사로잡는 동물입니다.
이들은 그 특유의 미모와 유연한 몸뚱아리로 우리를 매료시키며, 독립적인 성격과 함께 때로는 애정을 
드러내어 주인들과 강한 유대감을 형성합니다. 또한 고양이는 수많은 역사적인 순간과 신화에 등장하여 
그들의 중요성을 상징적으로 보여주었으며, 현대에는 애완 동물로서 우리 가정에 행복과 안정을 가져다 
주고 있습니다. 고양이는 집안을 깨끗하게 유지하고 스트레스를 줄여주는 등 우리의 삶에 긍정적인 영향을 
끊임없이 미치며, 정말로 특별한 존재입니다.

Here, asking the model to complete sentences yields more consistent and relevant output. These improvements illustrate the essence of prompt engineering: designing prompts that guide the model to perform a specific task.

If you try the 'cat' example I just showed you, you will never get the same answer. The answer will usually be similar or similar in meaning, but occasionally you will get a somewhat bizarre answer. The example above shows that even the same question can get completely different answers depending on how you ask it.

When I simply typed "cat", LLM assumed that was the starting point and wrote a sentence starting with "cat is~". When I typed the prompt "Complete the sentence: cat is", I got a much longer and more elaborate answer. This is because I saw how LLM handled each question.

When the cat provided the question intent as identified by the language model:

[질문의도]: 사용자의 질문을 이해하고 고양이에 관한 정보를 제공합니다.

Complete the sentence: If the cat provided , the language model figured out the question intent:

[질문의도]: 사용자가 주어진 문장을 완성하여 고양이에 대한 정보를 요청한 것으로 추정됩니다.

Prompt, which is considered a similar concept from a human perspective, is interpreted as a different intention by AI and generates an answer. This is the core reason why Prompting is important. Even if you are curious about the same thing and want to do it, there are cases where AI has sufficient capabilities but cannot do it because it does not know how to use it. The biggest purpose of this guide is to overcome that.

It may be used for commercial purposes with permission from the copyright holder, provided the source is cited.

Made with SlashPage