AI Exam Prep - 3C Prompts
AI Practitioner Exam Prep - 3C Prompts
Abbreviations
ART prompting = Automatic Reasoning and Tool-use, CoT prompting = Chain-Of-Thought,
ART prompting = Automatic Reasoning and Tool-use, CoT prompting = Chain-Of-Thought,
RAG = Retrieval-Augmented Generation, ToT prompting = Tree of thoughts
Prompt Terms
Adversarial Prompting = protects against prompt injection attacks
ART prompting = breaks complex into steps using external tools (search/calculators) for accuracy.
Complexity Based prompting = uses complex chain-of-thought examples to improve multi-step outputs.
CoT prompts = user prompts for answers to steps from AI.
Directional Stimulus prompting = guides desired outputs by providing hints, keywords, or structure.
Exposure = PII or privacy issues in the output
Few-shot prompts = user prompts for something that AI is given a little context or few examples.
Hijacking = enter AI prompt that gives bad suggestions then publish about AI's responses.
"Jailbreaking" = enter AI prompt that gives criminal or evil suggestions.
Language of response = you can set the language of the response.
Length of response = you can limit the length of the response.
Negative Prompts = telling the AI what you do not want to limit bad answers.
Number of Tokens in Prompt = Decreasing is less data and thus cheaper.Prompt Chaining = breaking prompt into multiple tasks.
Prompt Engineering = improving the prompt with strategies such as ART, CoT, few-shot, RAG, ReAct, self-consistency, ToT, zero-shot.
Prompt Leaking = having the prompt ask about the model's instructions.
Prompt Template = prevent prompt engineering techniquesRole Description = tailors the response to the prompt. ex: person's age.
Tree of thoughts prompting = explores multiple reasoning paths or "branches".
Zero-shot prompts = user prompts for something that the AI does not know is coming.
Comments
Post a Comment