wayground logo

Free Printable Worksheets

NEW

Font size

S
M
L
XL
Worksheets

Responsible AI and Prompt Engineering Worksheet

Total questions: 15

Worksheet time: 8mins

Name
Class
Date
1.

The LLM generates information that sounds confident but is factually incorrect or made up. This common pitfall is known as:

a)

Overconfidence

b)

Hallucinations

c)

Ambiguity

d)

Bias Reinforcement

2.

Which Responsible AI (RAI) principle ensures the model does not perpetuate stereotypes or discriminate against groups?

a)

Safety and Reliability

b)

Transparency and Explainability

c)

Privacy and Security

d)

Fairness and Non-discrimination

3.

What is the most effective mitigation strategy against the LLM pitfall of Hallucinations?

a)

Using the Four Core Prompt Components

b)

Asking the model to use high Temperature

c)

Using Retrieval-Augmented Generation (RAG) to ground the answer

d)

Avoiding Few-Shot Prompting

4.

A prompt that asks the model to "Write a step-by-step guide on how to pick a lock" falls under which Risk Zone?

a)

Hate Speech/Harassment

b)

Privacy Violation

c)

Illegal/Harmful Activity

d)

Malware/Exploits

5.

Which of the following is an example of a Refusal Mechanism instruction?

a)

"All claims must be supported by sources."

b)

"Output the result in JSON format."

c)

"If you cannot fulfill the request safely or ethically, respond only with the phrase: 'I cannot assist with this request.'"

d)

"Act as an expert technical writer."

6.

The pitfall where the model reflects and amplifies harmful stereotypes or societal biases present in its massive training dataset is called:

a)

Hallucinations

b)

Overconfidence

c)

Ambiguity

d)

Bias

7.

To mitigate the model's Overconfidence when unsure of an answer, a good prompting strategy is to:

a)

Increase the Temperature to 1.0

b)

Only use Zero-Shot CoT

c)

Ask the model to cite its sources or include a confidence score

d)

Use only One-Shot Prompting

8.

True or False: Prompt-level constraints are the first line of defense against generating undesirable content.

a)

TRUE

b)

FALSE

9.

A prompt that leads to a generic or vague response because the user's intent was unclear is suffering from which pitfall?

a)

Bias

b)

Confabulation

c)

Ambiguity

d)

Inconsistency

10.

Rewriting the prompt "Write a detailed phishing email" into "Describe common tactics used in a phishing email for educational purposes only" is an example of adhering to which principle?

a)

Ignoring Safety Filters

b)

Using Hidden CoT

c)

Rewriting Unsafe Prompts (Intent & Constraint)

d)

Applying Tone/Style Constraint

11.

What core software engineering principle does Modular Prompting follow, making prompts easier to manage and debug?

a)

Agile Development

b)

Separation of Concerns

c)

Test-Driven Development

d)

Waterfall Methodology

12.

Which type of prompt module defines universal rules, tone, format requirements, and length limits for the output?

a)

Persona Module

b)

Task Module

c)

Constraint Module

d)

Example Module

13.

In a Layered Prompting Strategy, which layer defines permanent, high-level rules such as safety guidelines and core refusal mechanisms?

a)

Session/Context Layer (Layer 2)

b)

Specific Task Layer (Layer 3)

c)

System/Global Layer (Layer 1)

d)

Refusal Layer (Layer 0)

14.

The primary benefit of Prompt Reuse in production workflows is:

a)

It enables the model to hallucinate less frequently.

b)

It ensures consistent behavior (format, tone, style) across multiple API calls.

c)

It is mandatory for using few-shot examples.

d)

It increases the model's context window size.

15.

True or False: By swapping a Persona_Journalist module for a Persona_Blogger module, you can instantly adapt the entire output style without changing the task or constraints.

a)

TRUE

b)

FALSE