wayground logo

Free Printable Worksheets

NEW

Font size

S
M
L
XL
Worksheets

AI Concepts Quiz: Questions 1-10

Total questions: 25

Worksheet time: 13mins

Name
Class
Date
1.

Transformers introduced which core concept enabling models to focus on relevant tokens?

a)

Pooling

b)

Self-Attention

c)

Activation

2.

Which transformer block is responsible for generating outputs token-by-token?

a)

Encoder

b)

Decoder

c)

Feed-Forward

3.

Which model is encoder-only?

a)

T5

b)

BERT

c)

GPT

4.

Large Language Models are primarily trained on:

a)

Transaction logs

b)

Massive text datasets

c)

Image datasets

5.

In Generative AI, inference refers to:

a)

Model initialization

b)

Generating responses to new inputs

c)

Loss minimization

6.

Which term best describes Generative AI models creating new content?

a)

Creative generation

b)

Reconstructive

c)

Predictive

7.

The main goal of RAG is to solve:

a)

High training costs

b)

Hallucination and outdated knowledge

c)

Tokenization errors

8.

In RAG, which component retrieves supporting documents?

a)

Generator

b)

Tokenizer

c)

Retriever

9.

Embeddings convert text into:

a)

JSON

b)

HTML

c)

Numerical vectors

10.

Similar vectors in an embedding space indicate:

a)

Incorrect tokenization

b)

Semantically related meanings

c)

Randomly distributed tokens

11.

Vector databases are mainly used for:

a)

Semantic similarity search

b)

Storing raw PDFs

c)

Grammar correction

12.

In training, the objective is to:

a)

Remove embeddings

b)

Minimize loss

c)

Increase number of tokens

13.

Backpropagation adjusts:

a)

Model weights

b)

Dataset labels

c)

Token IDs

14.

Which optimizer is adaptive and widely used?

a)

Newton

b)

SGD

c)

Adam

15.

Overfitting means:

a)

Model generalizes well

b)

Model memorizes training data

c)

Model does not learn enough

16.

Function calling allows AI to:

a)

Improve batch size

b)

Execute real-world actions via APIs

c)

Increase token length

17.

First step in function calling flow:

a)

Model formats response

b)

User sends a query

c)

API executes

18.

Multi-agent systems rely on:

a)

Collaboration among multiple agents

b)

Single model decision

c)

Static prompts

19.

Which is an example of agent roles?

a)

Deployer agent

b)

Retriever & Summarizer agents

c)

Billing agent

20.

MAS communication commonly uses:

a)

Binary encodings

b)

Structured messages / JSON

c)

Images

21.

Positional encoding in transformers helps:

a)

Train without embeddings

b)

Add token order information

c)

Increase sequence length

22.

Granite LLM series focuses mainly on:

a)

Gaming applications

b)

Enterprise-grade, trusted AI

c)

Crypto analytics

23.

In RAG architecture, the knowledge base stores:

a)

Temporary token scores

b)

Real factual documents

c)

Random synthetic prompts

24.

Chunking documents before embedding helps:

a)

Increase token repetition

b)

Maintain context and avoid truncation errors

c)

Reduce inference cost fully

25.

Evaluation metrics like BLEU and ROUGE apply to:

a)

Image recognition

b)

Password encryption

c)

Text-based models