Prompt Engineering

Zero-shot Learning

Basically using LLM without any examples. If, i have to review sentiment of the sentense "I love programming.", I can just give the prompt below.

Prompt : Please classify the sentiment of the following sentence as 'positive', 'negative', or 'neutral'.
The sentence is: "I love programming."
Answer :

Few-shot Learning

Prompt : Pailindrome is a word, phrase, number, or other sequence of characters which reads the same backward as forward. 
Examples of palindromes include "madam", "racecar"
Examples of non-palindromes include "hello", "world"
Answer 'yes' or 'no' please.
The word 'steve' is a palindrome?
Answer : 

Chain-of-Thought Prompting

현 시점으로 LLM들에는 Thinking이라는 능력이 탑재되어 있고, 우리가 단계별로 생각해보세요!라고 유도하지 않아도 "Instruncion-tuning"되었기에 무언가 입력했을 때, 생각하는 과정을 거쳐 답변을 도출해낸다.

Prompt Chaning, Adversarial Prompting ...

등등해서 많은 기법들이 있지만, 여기에 다 작성하는건 부질없기에 그떄 그때 필요하면 찾아보고 사용할 것을 권장한다.