AI0-001 Practice Test 19 — 15 Questions

Question 1

A team is designing a RAG system for a large collection of PDFs. They need to choose document chunking strategies. Which TWO strategies are considered best practices? (Choose two.)

Accepted Answer

Semantic chunking (e.g., sentence or paragraph boundaries). Semantic chunking preserves natural boundaries, and hierarchical chunking respects document structure, both improving retrieval quality.

Answer

Fixed-size chunking with no overlap

Answer

Single chunk per document

Answer

Random character-length chunks

Question 2

Which similarity search metric is BEST for comparing dense vector embeddings when the magnitude of the vectors is not important, only the direction?

Accepted Answer

Cosine similarity. Cosine similarity measures the cosine of the angle between vectors, focusing on direction and ignoring magnitude, which is ideal when only direction matters.

Answer

Euclidean distance

Answer

Dot product

Answer

Manhattan distance

Question 3

An AI team is deploying a fine-tuned LLM for a code generation assistant. They need to ensure the model outputs only syntactically valid JSON for integration with downstream systems. Which prompt engineering technique is MOST effective for enforcing structured output?

Accepted Answer

Enable JSON mode in the API call, specifying the desired JSON schema. JSON mode (or structured output) instructs the model to output valid JSON, often with schema enforcement. Few-shot examples may help but aren't as reliable as a dedicated mode. System prompts set the role but don't enforce syntax. Chain-of-thought improves reasoning but not format.

Answer

Provide a few-shot example of a valid JSON response in the prompt

Answer

Include a system prompt that says 'You are a helpful coding assistant.'

Answer

Use chain-of-thought prompting to have the model reason step-by-step before answering

Question 4

In the AI project lifecycle, after a model is trained and evaluated, it is deployed to a production environment. What is the NEXT critical step to ensure the model continues to perform well over time?

Accepted Answer

Monitoring the model's performance and data drift. Monitoring tracks model performance, data drift, and concept drift in production, triggering retraining or alerts when degradation is detected.

Answer

Collect more training data

Answer

Archive the model and start a new project

Answer

Re-train the model from scratch

Question 5

An AI developer is building an agent that can book flights and hotels by calling external APIs. The agent needs to decide which API to call and in what order based on user requests. Which pattern is BEST suited for this multi-step reasoning and tool use?

Accepted Answer

Apply the ReAct pattern (Reasoning and Acting). The ReAct pattern (Reasoning + Acting) combines chain-of-thought reasoning with tool use, allowing the agent to plan, call APIs, and incorporate results. Function calling alone doesn't provide multi-step reasoning. Simple RAG is for retrieval, not action. Fine-tuning doesn't give dynamic tool selection.

Answer

Implement a simple Retrieval-Augmented Generation (RAG) pipeline

Answer

Fine-tune a model to output API call sequences directly

Answer

Use function calling with a fixed sequence of API calls

Question 6

A company is deciding between fine-tuning and RAG for a domain-specific legal assistant that must provide accurate answers based on a changing set of statutes and regulations. The statutes are updated quarterly. Which approach is PREFERRED and why?

Accepted Answer

RAG, because it can retrieve the latest documents without retraining the model. RAG is preferred when the knowledge base changes frequently, as it retrieves the latest documents at inference time without requiring model retraining.

Answer

Fine-tuning, because it allows the model to internalize the statutes for faster inference

Answer

Fine-tuning with LoRA, because it is parameter-efficient and can be updated frequently

Answer

RAG with fine-tuning on the initial statutes, then update the index quarterly

Question 7

A team is training a image classification model. They split the dataset into training, validation, and test sets. After training, the model achieves 98% accuracy on the training set but only 72% on the test set. Which step in the AI project lifecycle should the team focus on?

Accepted Answer

Model selection – use regularization or reduce model complexity. The large gap indicates overfitting, which is a model selection/regularization issue. They need to apply techniques like dropout, data augmentation, or reduce model complexity.

Answer

Data acquisition – collect more data

Answer

Deployment – re-deploy with a different serving framework

Answer

Data preparation – check for train/test leakage

Question 8

A financial institution is deploying an AI system to approve personal loans. To comply with the EU AI Act's high-risk AI requirements, the bank must ensure meaningful human oversight. Which implementation BEST satisfies this requirement?

Accepted Answer

Require a human to review and approve every loan decision before it becomes final. The EU AI Act requires that high-risk AI systems allow for human oversight, including the ability to override or reverse the system's decisions. A mandatory human review before final approval ensures that the human can intervene. The other options either do not provide effective oversight or allow for rubber-stamping.

Answer

Use a separate AI model to audit the primary AI's decisions weekly

Answer

Allow applicants to appeal AI decisions through a customer service process

Answer

Provide a dashboard showing the AI's confidence score for each application

Question 9

A financial institution wants to use AI for loan approvals and must comply with fair lending laws. Which TWO practices should the institution adopt to mitigate bias and ensure compliance?

Accepted Answer

Apply fairness-aware machine learning techniques during model training. To mitigate bias, using fairness-aware algorithms and conducting disparate impact analysis are direct steps. Using only demographic data is illegal (redlining). Removing all features reduces model utility. A black-box model without explanation would hinder compliance.

Answer

Remove all features except credit score to avoid bias

Answer

Use a black-box model without explainability to protect intellectual property

Answer

Use only demographic features to ensure equal treatment

Question 10

A company using an AI-based hiring tool receives a candidate request for explanation of an automated rejection. Which GDPR principle is most directly relevant?

Accepted Answer

Right to explanation. The GDPR includes a right to explanation for automated individual decision-making, including profiling. The right to access is broader. The right to erasure is about deletion. The right to data portability is about data transfer.

Answer

Right to erasure

Answer

Right to data portability

Answer

Right to access

Question 11

A company is evaluating fairness metrics for a hiring model. They want to ensure that the model has similar true positive rates (TPR) across demographic groups. Which fairness metric should they use?

Accepted Answer

Equalized odds. Equalized odds requires that the true positive rate and false positive rate are equal across groups. Demographic parity requires equal selection rates. Individual fairness requires similar individuals to be treated similarly. Calibration ensures predicted probabilities match actual outcomes for each group. The scenario specifies TPR, which is part of equalized odds.

Answer

Calibration

Answer

Individual fairness

Answer

Demographic parity

Question 12

A hospital wants to train a diagnostic model using data from multiple hospitals without sharing raw patient data. Which technique allows model training across decentralised data while preserving privacy?

Accepted Answer

Federated learning. Federated learning trains a shared model by aggregating updates from local data without moving the data itself, preserving privacy. Differential privacy adds noise but doesn't decentralise data. Data centralisation violates privacy. Anonymisation alone doesn't allow collaborative training.

Answer

Differential privacy applied to the combined dataset

Answer

Centralising all data in one location and anonymising it

Answer

Using pseudonymisation and then pooling the data

Question 13

A company is forming an AI ethics board to oversee the development of a high-stakes AI system for bail decision recommendations. Which THREE responsibilities should the board primarily undertake?

Accepted Answer

Review model outputs for disparate impact across demographic groups. An AI ethics board should define fairness criteria, review models for bias, and establish a human oversight process. Designing the algorithm is a technical task for engineers. Marketing the system is a business function, not an ethics board duty.

Answer

Market the AI system to potential clients

Answer

Write the production code for the AI model

Question 14

A data scientist is evaluating a binary classifier for a hiring tool. They compute demographic parity and find that the selection rate for Group A is 0.2 and for Group B is 0.4. Which action would MOST directly address this disparity?

Accepted Answer

Retrain the model with a fairness constraint that enforces demographic parity. Demographic parity requires equal selection rates. Retraining with a fairness constraint that enforces demographic parity directly adjusts the model to achieve equal rates. Rebalancing the dataset (if the disparity stems from imbalanced labels) might help, but it does not guarantee parity. Modifying thresholds can also achieve parity, but post-processing without retraining may degrade other metrics; retraining with constraint is more direct.

Answer

Use a different evaluation metric such as equalized odds

Answer

Remove the sensitive attribute from the training data

Answer

Collect more data for Group A to increase its representation

Question 15

A hospital is implementing an AI system to analyze patient X-rays for potential fractures. The hospital must comply with HIPAA regulations. Which privacy-preserving technique allows the model to be trained on data from multiple hospitals without sharing raw patient data?

Accepted Answer

Federated learning. Federated learning trains model copies across multiple sites and aggregates only model updates (gradients) to a central server, keeping patient data local. Differential privacy adds noise to data but does not inherently prevent data sharing. Anonymisation and pseudonymisation still require sharing data, albeit de-identified.

Answer

Differential privacy

Answer

Data anonymisation

Answer

Data pseudonymisation