IBM Certified watsonx Generative AI Engineer - Associate

Last checked October 22, 2026 against the official IBM exam page.

1. Question 1Designing a Generative AI Solution

   A team wants a generative AI assistant that answers questions about its internal policy documents. Which requirement should be defined first?

   1. AThe number of GPUs that will be bought, so that the hardware budget is fixed before any requirements are gathered
   2. BThe colour scheme of the chat window, so that the interface matches the corporate brand guidelines used on the intranet
   3. CThe business question the assistant must answer, the users who will ask it, and the accuracy and risk tolerance for its answers
   4. DThe name of the foundation model, which should be chosen first because every other design decision depends on it
   Show answer and explanationHide answer

   Correct answer: C

   Defining the business question, users, and tolerance for error shapes every later decision, including model choice, retrieval design, and evaluation. Colour schemes, GPU purchases, and a pre-chosen model are downstream decisions that should follow from those requirements.

2. Question 2Selecting and Customizing Foundation Models

   What is the main difference between prompt tuning and fine-tuning a foundation model?

   1. AFine-tuning requires no training data at all, whereas prompt tuning requires a large labelled dataset of many thousands of examples
   2. BFine-tuning changes only the user interface of the application, whereas prompt tuning changes the weights of the base model
   3. CPrompt tuning deletes the base model and replaces it with a new one trained from random values, which fine-tuning does not do
   4. DPrompt tuning learns a small set of trainable prompt parameters while the base model stays frozen, whereas fine-tuning updates the model's weights
   Show answer and explanationHide answer

   Correct answer: D

   Prompt tuning keeps the base model frozen and learns a small set of prompt parameters, while fine-tuning updates weights of the model itself. Interface changes, replacing the model, and the data requirements in the other options reverse or misstate these differences.

3. Question 3Prompting and Retrieval-Augmented Generation

   What is retrieval-augmented generation (RAG)?

   1. ATraining a model from scratch on random data so that its outputs are not influenced by any particular source
   2. BA pipeline that retrieves relevant passages from a knowledge source and supplies them to the model to ground its answer
   3. CA cache of images that the interface loads so that the pages of the application are displayed more quickly to users
   4. DA printing method that places the generated text on the page in a layout chosen by the user before each print run
   Show answer and explanationHide answer

   Correct answer: B

   RAG retrieves relevant passages at query time and gives them to the model, which grounds the answer in a source. Training from scratch on random data, printing layouts, and image caches are unrelated to this pattern.

4. Question 4Deploying and Governing Generative AI Applications

   Which practice supports monitoring a deployed generative AI application?

   1. AChecking only that the server's power light is on, since the other aspects of the service are assumed to be working
   2. BTracking response quality, latency, cost, and user feedback over time, with alerts when they move outside agreed limits
   3. CPrinting the application logs once a month and filing them in a cabinet for a review that is held once a year
   4. DTurning off the application logs so that the storage cost stays low and the monitoring team has less data to read
   Show answer and explanationHide answer

   Correct answer: B

   Ongoing tracking of quality, latency, cost, and feedback, with alerts, shows whether the application is performing as intended in production. A power light, monthly paper logs, and disabled logs each leave problems undetected.

5. Question 5Designing a Generative AI Solution

   Which factor most influences whether a generative AI use case should use a general-purpose foundation model or a smaller specialized model?

   1. AThe required quality, latency, cost, and data residency constraints of the specific use case being designed
   2. BThe size of the company's office, since larger offices can justify larger models while smaller offices cannot
   3. CThe popularity of the model on social media, which indicates that the model is the best choice for every task
   4. DThe age of the employees who will use the assistant, since younger users need general models and older users need small ones
   Show answer and explanationHide answer

   Correct answer: A

   Quality, latency, cost, and data residency requirements decide whether a general model or a smaller specialized one fits the use case. Office size, social media popularity, and user age do not determine the technical fit.