Hallucination detection in Prompt Engineering / GenAI - Cheat Sheet & Quick Revision

Recall & Review

beginner

What is hallucination in AI language models?

Hallucination is when an AI model generates information that is false, made-up, or not supported by the input data.

beginner

Why is hallucination detection important in AI?

Detecting hallucinations helps ensure AI outputs are trustworthy and accurate, which is crucial for applications like healthcare, law, and education.

intermediate

Name one simple method to detect hallucinations in AI outputs.

One simple method is to check if the AI's output matches trusted external sources or databases.

intermediate

What role does human feedback play in hallucination detection?

Human feedback helps identify when AI outputs are incorrect or misleading, improving detection and guiding model improvements.

advanced

How can AI models reduce hallucinations during training?

By training on high-quality, verified data and using techniques like reinforcement learning with human feedback, models learn to produce more accurate outputs.

What does hallucination in AI usually mean?

A Compressing data

B Improving model accuracy

C Generating false or unsupported information

D Speeding up training

Hallucination means the AI creates information that is not true or supported by data.

Which method can help detect hallucinations?

A Random sampling

B Increasing model size only

C Ignoring user feedback

D Comparing output to trusted sources

Checking AI output against trusted sources helps find false information.

Why is human feedback useful in hallucination detection?

A It deletes data

B It identifies wrong AI outputs

C It slows down the model

D It trains the AI without errors

Humans can spot when AI says something incorrect, helping improve detection.

Which training approach helps reduce hallucinations?

A Using verified, high-quality data

B Using random data

C Ignoring errors

D Training without feedback

Training on good data helps the AI learn to avoid making things up.

Hallucination detection is most important for which AI use case?

A Healthcare advice

B Playing games

C Image compression

D Data storage

In healthcare, wrong AI info can be harmful, so detecting hallucinations is critical.

Explain what hallucination means in AI and why detecting it matters.

Think about when AI says something untrue and why that can be a problem.

Describe two ways to detect or reduce hallucinations in AI models.

Consider both checking AI answers and improving training.