AWS Certified AI Practitioner (AIF-C01) — Question 133

A company’s large language model (LLM) is experiencing hallucinations.

How can the company decrease hallucinations?

Answer options

• A. Set up Agents for Amazon Bedrock to supervise the model training.
• B. Use data pre-processing and remove any data that causes hallucinations.
• C. Decrease the temperature inference parameter for the model.
• D. Use a foundation model (FM) that is trained to not hallucinate.

Correct answer: C

Explanation

The correct answer is C because reducing the temperature inference parameter can lead to more deterministic outputs from the model, thereby minimizing the chances of hallucinations. Option A is incorrect as supervising training does not directly address hallucinations. Option B may help but doesn't guarantee a reduction in hallucinations, and option D is not the best approach since it involves changing the model entirely rather than adjusting parameters.