LLM Fundamentals →hardMultiple ChoiceObjective-mapped

1Z0-1127-25 LLM Fundamentals Practice Question

A team is deploying a chatbot that must never output harmful or biased statements. They plan to use a pre-trained LLM with in-context learning. Which additional measure is MOST effective at reducing harmful outputs without retraining?

Answer choices

Why each option matters

Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.

Correct answer & explanation

✓

Include few-shot examples in the system prompt that demonstrate appropriate responses

Providing few-shot examples of desired behavior in the system prompt (in-context learning) can guide the model toward safe responses. Fine-tuning would require retraining, and prompt engineering is broader; few-shot examples are a specific, effective technique.

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

✗
Apply a larger context window to include more safety instructions
Why it's wrong here
While a longer context can include more instructions, it does not guarantee the model follows them; few-shot examples are more effective.
✗
Fine-tune the model on a curated dataset of safe dialogues
Why it's wrong here
Fine-tuning requires retraining, which the team wants to avoid according to the scenario.
✗
Use beam search with a high beam width
Why it's wrong here
Beam search affects output quality but does not directly reduce harmful or biased content.
✓
Include few-shot examples in the system prompt that demonstrate appropriate responses
Why this is correct
In-context learning with few-shot examples can bias the model toward desired behavior without any model update.

Go deeper

1Z0-1127-25 LLM Fundamentals Practice Question

Why each option matters

Option-by-option breakdown

Related to this question