Aim: To iteratively improve an ambiguous prompt and observe the effect of temperature, Top-P and maximum-token settings on generation.
Tools Required: An LLM interface/playground that allows generation-parameter control.
If a parameter is unavailable, students record it as Not Available.
Prompt / Activity:
PART A - PROMPT REFINEMENT

Initial prompt:
"Explain AI agents."

Refined prompt:
"Explain AI agents to first-year postgraduate IT students in 180-220 words. 
Define an AI agent, identify its four core components, distinguish an agent from a normal LLM application, 
and give one practical example. Use one heading and a 4-row comparison table. Avoid vendor-specific terminology."

Compare the two responses for clarity, completeness and relevance.

PART B - GENERATION-PARAMETER EXPERIMENT

Use this same prompt for every trial:
"Suggest six names for a university Generative AI laboratory. For each name, give a one-line rationale."

Trial 1: Low temperature (for example 0.1-0.2), moderate Top-P, sufficient max tokens.
Trial 2: Medium temperature (for example 0.5-0.7), similar Top-P and max tokens.
Trial 3: Higher temperature (for example 0.9-1.0), similar Top-P and max tokens.
Trial 4: Keep temperature moderate but reduce maximum tokens significantly.

Record how diversity, predictability and completeness change.
