NEW
Font size
WorksheetsGenerative AI and Transformers Quiz
Total questions: 25
Worksheet time: 13mins
What best defines Generative AI?
AI that only classifies data
AI models that generate new, coherent, context-aware content
Rule-based automation systems
AI for numerical optimization
Which component enables attention-based contextual understanding?
CNN
RNN
Transformer
Decision Tree
What is the purpose of foundation models?
Task-specific training only
Training on massive, diverse datasets
Real-time data retrieval
Model evaluation
Which stage produces responses to user queries?
Pre-training
Fine-tuning
Inference
Feedback
Why is ethical and responsible AI important in GenAI?
To reduce training cost
To avoid plagiarism, bias, and misuse
To increase creativity
To speed up inference
What key limitation of RNNs did Transformers solve?
Overfitting
Sequential processing speed
Vocabulary size
Tokenization issues
Why do Transformers support faster training?
Smaller models
Parallel token processing
No attention layers
Fewer parameters
Which are the two core blocks of a Transformer?
Input & Output
Encoder & Decoder
CNN & RNN
Tokenizer & Classifier
What is the primary role of the encoder?
Generate next tokens
Convert input into contextual representations
Mask future tokens
Apply softmax
Why is positional encoding required?
To reduce parameters
Transformers lack inherent word-order awareness
To remove attention
To improve embeddings size
What is the benefit of multi-head attention?
Reduces memory usage
Captures multiple relationships in parallel
Removes feed-forward layers
Eliminates positional encoding
What best describes a Large Language Model (LLM)?
Rule-based NLP system
Deep neural network trained on massive text data
Database query engine
Symbolic reasoning system
What enabled the rise of LLMs?
Smaller datasets
Manual feature engineering
Transformers + massive data + compute power
Keyword matching
Which pre-training objective is used by BERT?
Next-word prediction
Masked language modeling
Reinforcement learning
Text-to-text learning
GPT models are primarily designed for:
Bidirectional understanding
Step-by-step text generation
Token masking
Classification tasks only
What makes T5 unique?
Decoder-only architecture
Frames all NLP tasks as text-to-text
Vision-only model
No fine-tuning required
IBM Granite models are mainly built for:
Entertainment chatbots
Academic research only
Enterprise-grade AI workflows
Gaming applications
A key focus of Granite models is:
Creativity over accuracy
Trust, transparency, and reliability
Image generation
Social media content
Why are Granite models suitable for enterprise use?
They are closed-source
They emphasize governed, factual outputs
They avoid fine-tuning
They remove evaluation
What primary problem in LLMs does RAG address?
Slow inference
Hallucination and outdated information
Token limits
Poor embeddings
What happens during Context Injection in RAG?
Model training
Retrieved documents added to the prompt
Tokenization
Model evaluation
Why does RAG reduce retraining needs?
It updates model weights automatically
It changes data instead of the model
It removes embeddings
It avoids inference
In a text summarization project, the primary goal is to:
Translate text into another language
Generate a shorter version while preserving key meaning
Classify documents into labels
Detect grammatical errors
Which task best represents a text classification project?
Generating code from prompts
Assigning labels like sentiment or topic to text
Fixing syntax errors in code
Summarizing research papers
In code generation and bug-fixing projects, the model is expected to:
Only explain code
Generate, analyze, and correct faulty code logic
Compile programs
Replace developers
