prompt engineering basics quiz

Prompt engineering is best described as:

• Training neural networks from scratch

• Writing software code for AI

• Crafting input instructions to optimize AI model outputs

• Evaluating AI systems automatically

What core architecture do most LLMs use?

• Recurrent Neural Network

• Convolutional Neural Network

• Transformer

• Decision Tree

During LLM “pre‑training,” the model predominantly learns to:

• Classify images

• Predict next tokens in text

• Cluster topics

• Remove bias

Which LLM prompting technique uses no examples and directly provides instructions?

• Few-shot

• Zero-shot

• Chain-of-thought

• Retrieval-augmented

Few-shot prompting differs from one-shot in that it includes:

• No examples

• Exactly one example

• Multiple examples (≥ 2)

• Numeric reasoning cues

"Chain‑of‑thought" prompting is especially effective for tasks requiring:

• Single-word translation

• Step-by-step reasoning

• Image classification

• Noise reduction

A well‑structured prompt typically includes all EXCEPT:

• Role/persona setting

• Tone or style directive

• Task objective

• GPU memory specs

In the context of LLMs, the context window refers to:

• Maximum length for everything model can consider simultaneously

• Number of layers in the model

• Amount of GPU memory

• Latency for token generation

Fine-tuning differs from pre-training because it:

• Uses unsupervised data only

• Trains with smaller, task-specific datasets

• Always uses the full original dataset

• Only affects tokenization strategy

Embedding layers in LLMs help by:

• Converting words/tokens into dense numerical vectors

• Masking parts of the data

• Filtering training noise

• Automatically fine-tuning the model