A team fine-tunes a pre-trained model on a new task. After fine-tuning, they notice that the model's performance on the original task (the one it was pre-trained on) has degraded significantly. What is the most likely cause, and what is a potential mitigation?

Negative transfer; use a higher learning rate to quickly adapt to the new task.

Bias amplification; remove sensitive features from the new task data.

Catastrophic forgetting; use a lower learning rate and include some original task data during fine-tuning.

Overfitting; use more data for the new task.

Which of the following is a recommended mitigation strategy for bias in pre-trained models?

Using a pre-trained model without any fine-tuning to avoid introducing bias

Selecting a model with the most parameters to reduce bias

Evaluating the model for bias on relevant subgroups before deployment

Ignoring bias because it is not a real concern in practice

A company is selecting a pre-trained model for a legal document classification task. They are concerned about the model's performance on legal terminology. Which mitigation strategy is most appropriate to address this concern?

Evaluate the model on a general language benchmark to ensure it is robust.

Deploy the model as-is and rely on post-processing to correct errors.

Use a larger pre-trained model to ensure it has more general knowledge.

Fine-tune the model on a large corpus of legal documents.

Certified Tester AI Testing

Domain 1Objective 4

Pre-Trained Models and Transfer Learning CT-AI Practice Questions (Page 6)

Part of the Domain 1: Introduction to AI domain, which makes up ~7% of our current practice bank. ISTQB does not publish an official question count, but from its 60-minute exam (~25–40 total, ~2–3 in this domain), expect 1–1 from this objective — we provide 31 practice questions to prepare you well beyond it. (estimate)

31questions here

7free pages

8concepts

Questions 26–30

26expert · hard
A team fine-tunes a pre-trained model on a new task. After fine-tuning, they notice that the model's performance on the original task (the one it was pre-trained on) has degraded significantly. What is the most likely cause, and what is a potential mitigation?
Select an answer first
27expert · hard
A team fine-tunes a pre-trained model on a new task. After fine-tuning, they observe that the model's performance on the new task is worse than if they had trained a simple model from scratch. What is the most likely cause?
Select an answer first
28foundation · easy
Which of the following is a recommended mitigation strategy for bias in pre-trained models?
Select an answer first
29application · medium
A company is selecting a pre-trained model for a legal document classification task. They are concerned about the model's performance on legal terminology. Which mitigation strategy is most appropriate to address this concern?
Select an answer first
30expert · hard
A healthcare startup is using a pre-trained model to predict patient readmission risk. The model was trained on data from a different country with different healthcare practices. They are concerned about the model's applicability. Which limitation is most relevant?
Select an answer first

Finished these 5 questions?

Review the revealed explanations, or continue through the curriculum.

Next 5 questions →Back to CT-AI

Free Basic Practice is a study aid with revealable answers — not a scored exam. Examers.io is independent and not affiliated with or endorsed by ISTQB. “CT-AI” is a trademark of its owner, used for identification only.

Pre-Trained Models and Transfer Learning CT-AI Practice Questions (Page 6)

Questions 26–30

A team fine-tunes a pre-trained model on a new task. After fine-tuning, they notice that the model's performance on the original task (the one it was pre-trained on) has degraded significantly. What is the most likely cause, and what is a potential mitigation?

A team fine-tunes a pre-trained model on a new task. After fine-tuning, they observe that the model's performance on the new task is worse than if they had trained a simple model from scratch. What is the most likely cause?

Which of the following is a recommended mitigation strategy for bias in pre-trained models?

A company is selecting a pre-trained model for a legal document classification task. They are concerned about the model's performance on legal terminology. Which mitigation strategy is most appropriate to address this concern?

A healthcare startup is using a pre-trained model to predict patient readmission risk. The model was trained on data from a different country with different healthcare practices. They are concerned about the model's applicability. Which limitation is most relevant?

Keep studying — Domain 1: Introduction to AI