"Engenharia de Prompt"
ou
"Se fazes perguntas pobres vais receber respostas pobres que te fazem pensar que o ChatGPT não presta!"
“ Prompt engineering is the art of communicating with a generative large language model. ” ChatGPT, 2023
Principled Instructions Are All You Need for Questioning LLaMA-1/2, GPT-3.5/4
Artigo de referência, EN, dez.2023, 24 pág.
Artigo original disponivel em:
Este texto foi criado pelo seguinte GPT disponivel na loja de GPTs com o nome: "Prompt Professor"
Ver esta interação em chat.openai.com/share/228fad1b-8fb9-4e9c-92d4-29e87b58ee1e
[Os 26 principios em Português] Tradução automática com erros!
A "roda" dos Principios de P1 a P26
Prompt Structure and Clarity (cor verde)
P2: Audience integration / Intended Audience: Integrate the intended audience in the prompt.
P4: Use Affirmative Directives: Use affirmative directives like ‘do’, avoiding negative language like ‘don’t’.
P12: Think Step-by-Step: Use leading words like “think step by step.”
P20: Output Primers: Use output primers by starting your response with the beginning of the desired output.
P17: Use Delimiters: Employ delimiters in the prompt.
P8: Formatted Prompt / Structured Formatting: Start prompts with ‘###Instruction###’, followed by ‘###Example###’ or ‘###Question###’ if relevant.
Exemplos dos 26 Principios: https://github.com/VILA-Lab/ATLAS/blob/main/data/README.md#overview-of-26-prompt-principles