AIF-C01 Applications of Foundation Models Practice Question

A developer is using Amazon Bedrock with the Claude model for text summarization. The output sometimes includes inaccurate information. What is the best practice to reduce hallucinations?

⚠ Common exam trap

AWS often tests the misconception that model size or output length adjustments are the primary levers for accuracy, when in fact grounding techniques like RAG are the standard solution for reducing hallucinations in production systems.

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

✓

Use retrieval augmented generation

Retrieval Augmented Generation (RAG) grounds the model's output in external, authoritative knowledge sources by retrieving relevant documents and injecting them into the prompt context. This directly reduces hallucinations because the model generates summaries based on factual retrieved data rather than relying solely on its parametric memory, which is the primary source of inaccuracies in text summarization tasks.

Answer analysis

Option-by-option breakdown

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

✗
Use a larger model
Why it's wrong here
A larger model may still hallucinate because scale does not ground outputs in source text; only retrieval-augmented generation with citations ties each claim to supplied documents. Larger models are chosen for complex reasoning or broader knowledge, not for factual fidelity against a provided summarisation input.
✗
Increase temperature
Why it's wrong here
Raising temperature increases sampling randomness, producing more varied and creative tokens, which worsens fabrication in summaries. Higher temperature suits brainstorming, marketing copy or divergent ideation, where novelty matters and factual precision against source text does not.
✓
Use retrieval augmented generation
Why this is correct
Retrieval augmented generation grounds the model's response in documents fetched from a knowledge base, so summarisation draws on supplied source text rather than parametric memory alone. This constrains the model to verifiable content, directly reducing fabricated output in the summarisation task.
✗
Decrease max tokens
Why it's wrong here
Max tokens caps response length, not factual grounding; truncation can even cut qualifying context and increase inaccuracy. It is used to control cost and latency, or to fit outputs into downstream buffers, not to align generated claims with source documents.

Go deeper

AIF-C01 Applications of Foundation Models Practice Question

Why each option matters

Option-by-option breakdown

Related to this question