wayground logo

Free Printable Worksheets

Font size

S
M
L
XL
Worksheets

Machine Learning and Generative AI — MCQ Worksheet Extraction

Total questions: 111

Worksheet time: 56mins

Name
Class
Date
1.

MCQ — LEC 1: Which statement best describes supervised learning in machine learning?

a)

The algorithm is trained using only unlabeled data and finds patterns without any guidance.

b)

The algorithm tries to generate new data similar to the training data without any specific labels

c)

The algorithm learns by receiving rewards or penalties for its actions in a simulated environment.

d)

The algorithm is given input data paired with the correct output labels and learns to predict new outputs based on this labeled data.

2.

MCQ — LEC 1: ………………… deals with predicting categorical target variables, which represent discrete classes or labels.

a)

Classification

b)

Regression

c)

Clustering

d)

Association

3.

MCQ — LEC 1: Which task is most appropriate for a regression algorithm?

a)

Sorting emails into “spam” or “not spam” folders.

b)

Estimating the selling price of a used car based on its characteristics.

c)

Grouping photos by the people in them without labels.

d)

Identifying faces in crowded images and marking their positions.

4.

MCQ — LEC 1: The primary goal of Unsupervised learning is…………………………

a)

To predict future outcomes based on historical labeled data and maximize prediction accuracy

b)

To discover hidden patterns, similarities, or clusters within data for purposes such as data exploration, visualization, and dimensionality reduction

c)

To classify data into predefined categories using training examples with known correct answers

d)

To optimize decision-making processes through trial-and-error interactions with an environment

5.

MCQ — LEC 1: What is association rule learning?

a)

A supervised learning technique that predicts customer purchasing behavior based on demographic data and past transaction labels

b)

A clustering algorithm that groups similar customers together based on their shopping preferences and spending patterns

c)

A rule-based machine learning technique used to discover interesting relationships between variables in large datasets, particularly known for market basket analysis

d)

A reinforcement learning method that optimizes product recommendations by learning from customer feedback and purchase rewards.

6.

MCQ — LEC 1: What are the main characteristics that distinguish reinforcement learning from other machine learning approaches?

a)

Uses labeled datasets, requires human supervision, and focuses on finding patterns in static data without environmental interaction

b)

Discovers hidden structures in data, performs clustering and dimensionality reduction, and works with unlabeled datasets exclusively

c)

Learns through trial-and-error interactions with an environment, uses reward feedback to improve performance, and excels at autonomous decision-making

d)

Processes sequential text data, generates natural language responses, and relies on attention mechanisms for understanding context

7.

MCQ — LEC 1: What are Neural Networks (NNs) in the context of deep learning?

a)

The building blocks of deep learning that mimic the structure of the brain to identify patterns and make decisions

b)

Statistical models that use linear regression techniques to analyze data relationships and make predictions based on correlation coefficients

c)

Database management systems designed to store and retrieve large amounts of structured data efficiently using indexing algorithms

d)

Programming frameworks that provide pre-written code libraries for implementing machine learning algorithms and data processing tasks

8.

MCQ — LEC 1: What are the two main components of a Generative Adversarial Network (GAN)?

a)

Encoder and decoder networks

b)

Generator and discriminator networks

c)

Input and output layers

d)

Training and testing datasets

9.

MCQ — LEC 1: RNNs are specifically designed to process ……………… data and can maintain ……………… to remember information from previous steps in a sequence.

a)

static image; long-term memory

b)

tabular; permanent storage

c)

sequential; short-term memory

d)

categorical; external memory

10.

MCQ — LEC 1: Which statement best explains why transformers revolutionized NLP compared to previous architectures?

a)

They process words sequentially one at a time, maintaining better memory of previous contexts

b)

They require less computational power by processing only the most relevant words in a sentence

c)

They generate synthetic text by competing with discriminator networks through adversarial training

d)

They use attention mechanisms to consider all words simultaneously and weigh their importance

11.

MCQ — LEC 1: What is the main difference between traditional AI and on-device AI?

a)

Traditional AI uses smaller models while on-device AI requires large cloud servers for processing

b)

Traditional AI relies on cloud servers to process data from edge devices, while on-device AI operates directly on devices using integrated chipsets and smaller LLMs

c)

Traditional AI focuses on mobile applications while on-device AI is designed only for desktop computers

d)

Traditional AI processes data faster while on-device AI provides more accurate results through cloud computing

12.

MCQ — LEC 2: What does Generative AI refer to?

a)

A set of artificial intelligence methodologies that can predict future events

b)

A set of artificial intelligence methodologies that can produce novel content resembling their training data

c)

A set of artificial intelligence methodologies that only analyze data

d)

A set of artificial intelligence methodologies that create physical objects

13.

MCQ — LEC 2: Which scenario best illustrates the concept of a foundation model?

a)

A model trained specifically for medical diagnosis that only works with X-ray images and cannot be used for other medical tasks

b)

A small specialized model designed to perform one specific function like spam email detection with high accuracy

c)

A large pre-trained model like GPT that can be fine-tuned for various tasks such as writing, coding, translation, and summarization

d)

A model that requires training from scratch for each new application without leveraging any previous learning

14.

MCQ — LEC 2: What is the primary purpose of Gen AI Studio?

a)

To allow developers to quickly explore, customize, and deploy GenAI models on Google Cloud with collaborative tools and resources

b)

To provide cloud storage solutions for large datasets used in traditional machine learning applications

c)

To offer computational resources for training AI models from scratch without any pre-built components

d)

To serve as a marketplace for purchasing pre-trained AI models from third-party vendors

15.

MCQ — LEC 2: What is the main issue with LLM consistency?

a)

Always provide the same answer regardless of context

b)

They produce wildly different answers when asked the same question multiple times

c)

They only work with specific types of questions

d)

They require constant internet connection to function

16.

MCQ — LEC 2: Why do LLMs often exhibit bias in their responses?

a)

They inherit biases present in their internet-scraped training data

b)

They are designed to favor certain user groups

c)

They lack sufficient computational power

d)

They are programmed with specific political viewpoints

17.

MCQ — LEC 2: In the context of LLMs, what are hallucinations?

a)

Visual distortions that occur when the model processes image inputs incorrectly

b)

Words or phrases generated by the model that are often nonsensical or grammatically incorrect

c)

Temporary system failures that cause the model to stop responding to user queries

d)

Intentional creative outputs designed to enhance the model's storytelling capabilities

18.

MCQ — LEC 2: What makes LLM interpretability challenging?

a)

They process data too slowly for analysis

b)

They require specialized hardware to operate

c)

It's difficult to understand why they generate particular responses

d)

They only work with text-based inputs

19.

MCQ — LEC 2: The main memory challenge for LLMs is ……………………

a)

They cannot remember any previous conversations

b)

They have unlimited memory capacity

c)

They can only store numerical data

d)

Efficiently remembering important details over long conversations remains difficult

20.

MCQ — LEC 2: What limitation do LLMs have regarding current information?

a)

They can only process historical data from decades ago

b)

They lack information about changes that occurred after their training cutoff

c)

They cannot access any external data sources

d)

They require manual updates every few hours

21.

MCQ — LEC 2: What is the primary purpose of PaLM API in Google's generative AI ecosystem?

a)

To serve as a simple entry point for developers to access Google's LLMs optimized for various use cases

b)

To provide a complex multi-step authentication system for accessing Google's cloud services

c)

To offer data storage solutions for large-scale machine learning datasets

d)

To provide hardware acceleration for training custom neural networks from scratch

22.

MCQ — LEC 2: How does MakerSuite facilitate generative AI application development?

a)

By requiring extensive coding knowledge and manual model training processes

b)

By providing an approachable platform for prototyping, prompt iteration, dataset augmentation, and model customization

c)

By focusing exclusively on image generation and computer vision tasks

d)

By offering only pre-built applications without customization options

23.

MCQ — LEC 3: What is the primary function of sequence-to-sequence (seq2seq) models?

a)

To classify images into categories

b)

To map an input sequence to an output sequence

c)

To detect anomalies in large datasets

d)

To generate static numerical summaries of tabular data

24.

What is one of the main issues associated with processing long sequences in RNNs?

a)

Increased accuracy

b)

Faster training

c)

Reduced memory usage

d)

Slow computations

25.

Which problem in RNNs occurs when gradients become too large during training?

a)

Vanished memory

b)

Slow computations

c)

Exploding gradients

d)

Vanishing gradients

26.

What does the term 'vanished memory' refer to in the context of RNNs?

a)

The ability to store large datasets

b)

The loss of information from earlier time steps during training

c)

The increase in computational speed

d)

The improvement in gradient flow

27.

Which of the following is NOT listed as a problem with RNNs?

a)

Vanishing gradients

b)

Exploding gradients

c)

Perfect memory retention

d)

Slow computations

28.

What is the eventual size of the gradient when vanishing gradients occur?

a)

Infinitely large

b)

Infinitely small

c)

Moderately stable

d)

Equal to one

29.

What happens to the gradient size in each step due to vanishing gradients?

a)

It becomes smaller and smaller

b)

It remains constant

c)

It becomes larger and larger

d)

It fluctuates randomly

30.

What happens to the gradient when it is multiplied by a number larger than one in the context of exploding gradients?

a)

It keeps increasing

b)

It remains constant

c)

It becomes smaller

d)

It becomes zero

31.

What can happen to the parameters of a network when gradients become extremely large?

a)

They become negative

b)

They stabilize automatically

c)

They decrease to zero

d)

They result in NaN values

32.

What is a potential drawback of word-based tokenization?

a)

Short sequences

b)

Out-of-vocabulary (OOV) words

c)

Small vocabulary size

d)

Easy semantic capture

33.

Which tokenization method breaks words into smaller units and is used in GPT?

a)

Character-based tokenization

b)

Byte-Pair Encoding (BPE)

c)

Word-based tokenization

d)

Sentence Piece

34.

Which tokenization method is used in BERT and breaks words into smaller units?

a)

Sentence Piece

b)

Byte-Pair Encoding (BPE)

c)

Character-based tokenization

d)

Word Piece

35.

Why are positional embeddings used in Transformers?

a)

To increase the vocabulary size

b)

To encode the position or order of tokens

c)

To replace attention scores

d)

To process sequences sequentially like RNNs

36.

What does the attention scores matrix represent in Transformers?

a)

The difference between token positions

b)

The total number of tokens

c)

The similarity between query and key vectors of tokens

d)

The weight of the output sequence

37.

What is the primary role of the query vector (Q) for each token?

a)

To store the token's content

b)

To determine the attention it receives

c)

To measure its relevance to other tokens

d)

To mix data based on relevance scores

38.

What is the function of the key vector (K) in relation to other tokens?

a)

To store the actual information of a token

b)

To compare against query vectors to determine attention

c)

To measure relevance to other tokens

d)

To mix data based on relevance scores

39.

What does the value vector (V) represent for each token?

a)

The relevance score to other tokens

b)

The actual information or content used if deemed relevant

c)

The comparison metric for attention allocation

d)

The position of the token in the sequence

40.

What are the "heads" in multi-head attention?

a)

Sequential layers in the Transformer

b)

Output tokens

c)

Single query vectors

d)

Parallel attention mechanisms

41.

What is a key characteristic of multi-head attention in Transformers?

a)

It processes tokens sequentially

b)

It uses multiple parallel attention mechanisms

c)

It relies solely on a single attention score

d)

It does not use query, key, or value vectors

42.

What is a key requirement of a causal model?

a)

It must not be able to see future words

b)

It must predict future words accurately

c)

It must process all words simultaneously

d)

It must ignore past words

43.

What do the key and value vectors produced by the encoder represent?

a)

The order of tokens in the output sequence

b)

The contextual information and what the model has learned about the input

c)

The probabilities of the next token

d)

The similarity between input and output tokens

44.

How does the decoder generate the output sequence?

a)

By processing the entire sequence at once

b)

By ignoring the input sequence

c)

One token at a time

d)

By using only key vectors

45.

What is a potential drawback of using greedy search?

a)

It always leads to optimal results

b)

It can lead to suboptimal results due to its lack of exploration

c)

It increases the speed of decoding

d)

It eliminates the need for SoftMax

46.

What is the primary purpose of fine-tuning a pre-trained Large Language Model (LLM)?

a)

To erase all previous knowledge

b)

To adapt it for a specific task or domain

c)

To increase the size of the dataset

d)

To train it from scratch

47.

What does it mean when a model is described as pre-trained?

a)

It has been trained only on a small dataset

b)

It has been exposed to a massive dataset with general knowledge

c)

It has never been trained before

d)

It is designed for a single task

48.

What is a key characteristic of Self-Supervised Learning (SSL) in domain-specific fine-tuning?

a)

It requires fully labeled datasets

b)

It generates its own labels from the input data

c)

It only works with small datasets

d)

It ignores the surrounding context

49.

How does Self-Supervised Learning help models like LLMs?

a)

By predicting missing words based on context

b)

By limiting the use of unlabeled data

c)

By avoiding the use of training corpora

d)

By focusing only on labeled data

50.

How does supervised fine-tuning improve a model's performance?

a)

By mapping inputs to correct outputs based on explicit feedback

b)

By ignoring question-answer pairs

c)

By limiting the model to general knowledge

51.

What is the core of the Reinforcement Learning from Human Feedback (RLHF) procedure?

a)

Using only supervision and demonstration data

b)

Establishing and training a reward model (RM)

c)

Ignoring human feedback

d)

Fine-tuning without a reward model

52.

How does RLHF improve an LLM's behavior?

a)

By collecting feedback from humans to train a reward model

b)

By limiting the model to unsupervised learning

c)

By avoiding candidate response exploration

d)

By aligning with machine preferences only

53.

What is the main benefit of Low-Ranking Adaptation (LoRA) in fine-tuning LLMs?

a)

It increases the number of parameters

b)

It reduces computational and financial costs using low-rank approximation

c)

It requires full 16-bit finetuning

d)

It avoids specific task adaptation

54.

How does Parameter Efficient Fine Tuning (PEFT) enhance LLM performance?

a)

By fine-tuning all parameters

b)

By adjusting a small set of parameters to reduce costs and combat forgetting

c)

By limiting performance to general tasks

d)

By avoiding multimodal applications

55.

What is the main benefit of Low Ranking Adaptation (LoRA) in fine-tuning LLMs?

a)

It increases the number of parameters

b)

It reduces computational and financial costs using low-rank approximation

c)

It requires full 16-bit finetuning

d)

It avoids specific task adaptation

56.

During fine-tuning of a pre-trained model like DistilBERT for a specific task, what strategy can help reduce catastrophic forgetting?

a)

Fine-tuning with a larger batch size

b)

Using Parameter-Efficient Fine-Tuning (PEFT) methods like LoRA

c)

Increasing the maximum sequence length during tokenization

d)

Training on noisy, un-labelled data

57.

What is a primary cause of model drift in a fine-tuned model?

a)

Insufficient computational resources

b)

Changes in data distribution or environment

c)

Overfitting during initial training

d)

Lack of model complexity

58.

What is a potential consequence of bias amplification during fine-tuning?

a)

Improved model accuracy

b)

Increased computational efficiency

c)

Ethical concerns due to biased predictions

d)

Reduced need for re-fine-tuning

59.

What is the primary function of the reward model in the RLHF process?

a)

To generate initial text outputs for evaluation

b)

To assign a score to the base model's output based on human preferences

c)

To directly update the base model's parameters

d)

To reduce the computational resources needed for fine-tuning

60.

How does the reinforcement learning algorithm (e.g., PPO) utilize the reward model’s scores in RLHF?

a)

It uses the scores to generate new training data

b)

It adjusts the base model to favor high-reward outputs and avoid lowerward ones

c)

It replaces the base model with a new model

d)

It ignores low-reward outputs without updating the model

61.

What is the purpose of human annotators ranking responses in the reward model training process?

a)

To directly update the fine-tuned language model

b)

To provide preference data to train the reward model to predict humanpreferred scores

c)

To generate new prompts for the language model

d)

To reduce bias in the fine-tuned model's outputs

62.

Why is fine-tuning a pre-trained model more efficient than training a model from scratch?

a)

It requires no computational resources

b)

It skips initial training stages and converges faster with less data

c)

It eliminates the need for task-specific adaptation

d)

It avoids the use of pre-trained model knowledge

63.

What is a key benefit of fine-tuning in continuous learning scenarios?

a)

It eliminates the need for any initial training

b)

It allows periodic model updates to adapt to changing data without starting from scratch

c)

It ensures the model never requires additional data

d)

It completely removes biases from the pre-trained model

64.

What is a prompt in the context of NLP and LLMs?

a)

A pre-trained model output

b)

An input provided to the model to generate a response or prediction

c)

A type of fine-tuning technique

d)

A method to evaluate model performance

65.

What does prompt engineering involve?

a)

Only asking simple questions

b)

Designing high-quality prompts, tinkering for the best prompt, and optimizing length and style

c)

Ignoring creativity and logic

d)

Avoiding task-specific adjustments

66.

What is the primary function of an instruction in a prompt?

a)

To provide background context

b)

To tell the model what action to take, e.g., "Summarize the following text"

c)

To specify the output format

d)

To process input data

67.

How does an output indicator enhance a prompt?

a)

By providing additional context

b)

By reducing the prompt length

c)

By guiding the model on the desired response format, e.g., "In the style of Shakespeare"

d)

By limiting the input data

68.

How does context enhance a prompt?

a)

By giving the model a broader scenario or background, e.g., "Considering the economic downturn"

b)

By limiting the input data

c)

By defining the output style

d)

By reducing the prompt length

69.

What defines a zero-shot prompt?

a)

It includes multiple examples

b)

It provides a task description and some text without examples

c)

It relies on specific training data

d)

It requires prior model training on the task

70.

What is a key ability tested by the zero-shot technique?

a)

The model's ability to generalize and produce outputs without prior examples

b)

The model's ability to memorize training data

c)

The model's ability to follow complex instructions

d)

The model's ability to process large datasets

71.

What is the main difference between zero-shot and one-shot prompting?

a)

Zero-shot uses multiple examples, while one-shot uses none

b)

Zero-shot provides no examples, while one-shot uses a single example

c)

One-shot requires specific training data, while zero-shot does not

d)

Zero-shot is more complex than one-shot

72.

How does one-shot prompting help the model perform a task?

a)

By providing multiple demonstrations

b)

By avoiding any examples

c)

By relying on prior training data alone

d)

By showing the model a single example to understand and generate text accordingly

73.

How can the effectiveness of one-shot prompting be improved for difficult tasks?

a)

By reducing the number of examples

b)

By removing the prompt context

c)

By experimenting with increasing demonstrations (e.g., 3-shot, 5-shot, 10-shot)

d)

By limiting the input data

74.

What is the main benefit of chain-of-thought (CoT) prompting?

a)

It simplifies the task by avoiding examples

b)

It limits the model's output to simple responses

c)

It reduces the need for model training

d)

It enables complex reasoning through intermediate steps, improving language understanding

75.

What is the core idea behind Self-Consistency CoT prompting?

a)

It uses a single reasoning path for all tasks

b)

It calls the LM multiple times on the same prompt and selects the most consistent answer using Majority Voting

c)

It avoids intermediate reasoning steps

d)

It relies solely on standard prompting techniques

76.

How does Self-Consistency CoT prompt mimic human reasoning?

a)

By providing diverse perspectives and encouraging critical evaluation of its own reasoning

b)

By limiting the model to a single example

c)

By avoiding consultation of multiple resources

d)

By focusing only on arithmetic tasks

77.

What does RAG primarily enable a generative AI model to do?

a)

Generate images from text descriptions

b)

Access and use external information that was not part of its original training data/parameters

c)

Train itself in real time on user inputs

d)

Compress its model size while keeping performance

78.

Which of the following is NOT listed as one of the core RAG components on this slide?

a)

Retriever

b)

Ranker

c)

Generator

d)

Planner

79.

In a standard RAG pipeline, what is the role of the Retriever component?

a)

To write the final answer in natural language

b)

To fetch relevant passages or documents from a large corpus of external knowledge

c)

To re-rank the retrieved documents by relevance after the initial retrieval

d)

To store and update the external data corpus

80.

What is the primary purpose of a Vector Database in a RAG system?

a)

To store vector embeddings and enable similarity search for efficient retrieval

b)

To compress the base language model to reduce memory footprint

c)

To generate final answers after retrieval

d)

To train the Retriever component on-the-fly

81.

Before data can be stored and searched in a Vector DB for RAG, what mandatory preprocessing step must be?

a)

Data Chunking – dividing the data into manageable “chunks” or segments

b)

Translation into English

c)

Compression using gzip

d)

Removal of all stop-words

82.

Which of the following is NOT a real-world example of a Vector Database?

a)

Chroma

b)

Pinecone

c)

Weaviate

d)

Redis

83.

What is the main job of the RAG Ranker component?

a)

To fetch the initial set of documents from the vector database

b)

To refine the retrieved documents by assigning relevance and importance, assigning scores and prioritizing the most relevant ones

c)

To generate the final answer in natural language

d)

To split the original documents into chunks

84.

According to RAG, which component is responsible for efficiently pulling the initial relevant information from a huge corpus?

a)

The Ranker

b)

The Retriever

c)

The Generator

d)

The Augmentor

85.

What best describes Naive/Standard RAG?

a)

It uses multiple rounds of retrieval and reasoning

b)

It is the simplest form: chunk documents → embed → retrieve top chunks on query → directly send to LLM for answer

c)

It always rewrites the user query before retrieval

d)

It compresses all retrieved documents into a single sentence

86.

What is the key difference between Naive RAG and Advanced RAG?

a)

Advanced RAG removes the vector database entirely

b)

Advanced RAG only works with images, not text

c)

Advanced RAG uses smaller language models

d)

Advanced RAG adds extra processing steps both before and after the initial retrieval

87.

What is the defining characteristic of Modular RAG?

a)

It adds a loop that corrects factual mistakes after generation

b)

It breaks the traditional RAG pipeline into independent, swappable, re-orderable modules (e.g., separate search, re-ranking, summarization, generation modules)

c)

It only works with a single large vector database

d)

It removes the retrieval step completely

88.

Corrective RAG is specifically designed for scenarios where:

a)

Speed is more important than accuracy

b)

The knowledge base is very small

c)

A factual error in the answer could have serious negative consequences

d)

Users only ask creative questions

89.

What is the key new capability that Agentic RAG introduces compared to all previous RAG variants?

a)

It uses a larger vector database

b)

It removes the generator LLM completely

c)

It only works with images

d)

The system gains autonomy: the model actively decides if/when it needs more information and iteratively refines its own queries

90.

Fusion RAG is distinguished from standard RAG because it:

a)

Uses only one document at a time

b)

Pulls relevant information from multiple heterogeneous data sources and blends/synthesizes them into a single cohesive answer

c)

Relies solely on the LLM’s internal knowledge without any retrieval

d)

Runs everything in parallel on a single GPU

91.

What is currently one of the biggest real-world benefits of Self-RAG?

a)

It is the fastest RAG method

b)

It drives hallucination rates in RAG systems close to zero while keeping high accuracy

c)

It requires no vector database

d)

It works offline without any external data

92.

According to Anthropic, both Workflows and Agents belong to the same broad category. What is that category?

a)

Large Language Models

b)

Agentic Systems

c)

Tool-augmented Models

d)

Reasoning Engines

93.

What is the key characteristic of a Workflow according to Anthropic?

a)

The model dynamically decides the next step at runtime.

b)

The model can change its goals during execution.

c)

It operates without any tools or external functions

d)

The sequence of steps is predefined by the designer.

94.

In an Anthropic-style Agent, who or what decides what to do next during task execution?

a)

A hardcoded script written by the developer.

b)

The user in real time.

c)

The model itself dynamically

d)

A separate supervisor model

95.

Which of the following statements is true about the process flexibility of Agents compared to Workflows (per Anthropic’s distinction)?

a)

Workflows can adapt and replan, while Agents follow a fixed path

b)

Both Workflows and Agents have completely fixed processes

c)

Agents have a fixed process, while Workflows can react and adapt

d)

Agents can react, adapt, and plan as they go, whereas Workflows follow a fixed script

96.

What is Prompt Chaining according to Anthropic?

a)

Asking the same LLM multiple times in parallel for better answers

b)

Breaking a large task into a sequence of smaller, well-defined sub-tasks

c)

Letting the model freely decide the order of operations

d)

Combining multiple unrelated prompts into one giant prompt

97.

In Anthropic’s Prompt Chaining pattern, how are the individual sub-tasks typically executed?

a)

All sub-tasks run simultaneously on different models

b)

Each sub-task is handled by a separate LLM call (or sequence of calls), with results passed forward

c)

The entire chain is executed inside a single LLM call

d)

Sub-tasks are performed only by custom Python code without any LLM

98.

In the diagram shown, what is the purpose of the “Gate” component placed after LLM Call 1?

a)

To add new tools to the next LLM call

b)

To merge outputs from multiple parallel LLM calls

c)

To perform a quality check and decide whether to continue the chain (Pass) or stop early (Exit/Fail)

d)

To rewrite the original user query

99.

Which of the following is NOT listed as a benefit of Prompt Chaining?

a)

Precision

b)

Control

c)

Modularity

d)

Autonomous decision-making

100.

Which statement best reflects Anthropic’s view on Prompt Chaining as presented in this slid?

a)

It is a type of fully agentic system where the model freely decides every next action

b)

It puts the entire process on “guardrails” with every step defined in advance

c)

It is inferior to agentic approaches in every situation

d)

It removes the need for any LLM calls after the first one initial prompt

101.

What is the primary role of the LLM Router in Anthropic’s Routing design pattern?

a)

To generate the final answer itself

b)

To act as a “traffic controller” that decides which specialized (“expert”) model should handle the incoming task

c)

To combine outputs from all specialist models into one response

d)

To translate the user query into a different language before processing

102.

Which of the following is NOT listed as a benefit of using the Routing pattern?

a)

Specialization

b)

Efficiency

c)

Modularity

d)

Full autonomy (the router can change the user’s original request)

103.

What is the core idea of the Parallelization design pattern according to Anthropic?

a)

Running the same task on multiple identical LLMs and voting on the best answer

b)

Splitting a task into parts and having multiple specialist LLMs work on different parts simultaneously, then combining results

c)

Letting one LLM decide everything step-by-step without any parallel work

d)

Sending the entire task to the fastest available LLM

104.

According to Anthropic, Parallelization is classified as:

a)

A fully agentic system

b)

A workflow pattern (not an agent), because the splitting and aggregation logic is coded by humans

c)

Exactly the same as Prompt Chaining

105.

In orchestrator-work pattern, which component is responsible for deciding how to break down the original task?

a)

Human-written code

b)

The user in real time

c)

The Orchestrator LLM

d)

The Synthesizer LLM

106.

What is the key advantage highlighted for using LLMs as both Orchestrator and Synthesizer?

a)

It is cheaper because fewer API calls are needed

b)

It guarantees perfect parallelism every time

c)

It completely eliminates the need for any worker LLMs

d)

The process becomes more dynamic and can better adjust to the specifics of each individual input thanks to the LLM’s reasoning abilities

107.

What is the main purpose of the Evaluator-Optimizer (Validation Agent) pattern?

a)

To parallelize work across many LLMs

b)

To iteratively improve an answer by having one LLM generate solutions and another LLM critically check and reject/improve them

c)

To route the query to the single best specialist LLM

d)

To translate the user query into multiple languages simultaneously

108.

In Evaluator-Optimizer (Validation Agent) pattern, what happens when the Evaluator LLM rejects a proposed solution?

a)

The process stops immediately and returns an error

b)

The rejected solution (plus feedback) is sent back to the Generator LLM for another attempt

c)

The Evaluator automatically fixes the answer itself

d)

The input is forwarded to a completely different model

109.

What is the biggest quality-related benefit of the Evaluator-Optimizer pattern?

a)

It is faster than a single LLM call

b)

Solutions get progressively better each round as the Generator corrects its own mistakes using feedback

c)

It guarantees the answer is 100% factually correct

d)

It reduces token usage to the absolute minimum

110.

The Evaluator-Optimizer pattern delivers “more consistent, high-quality outputs” even when:

a)

Using very small models

b)

The Generator LLM is “messy” or inconsistent on its own

c)

There is no internet access

d)

The user provides very short prompts

111.

In the context of Anthropic’s overall workflow-vs-agent distinction, the Evaluator-Optimizer loop is still categorized as:

a)

A workflow (not an agent), because the loop structure, roles, and exit condition are all predefined by the human designer

b)

A true agent, because it can loop until satisfied

c)

The only pattern that is neither workflow nor agent

d)

The first real agentic pattern in the deck