Skip to content

amdemw206-coder/AI-ML-Assignment-6-Prompt-Engineering

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

10 Commits
 
 
 
 
 
 

Repository files navigation

Project Title: AD 331 Prompt Engineering for Performance Improvement

By: Amde Wubshet

Task:

Asking a LLM to generate a .CSV file using multiple prompt techniques and raw text data (Mainline Pokemon Games) as input.

LLM Used:

Google Gemini 3

Prompt Techniques Output Format Score (1 to 10) Brief Observation
Baseline 6 It looks like it takes all the raw text data and sets it as CSV entries, but it also makes it look very bloated. For example, "First Pokémon generation" Could just be "1st"
Role Prompt 6.2 Only notable difference would be the "Release date (Worldwide)" column compared to baseline
Chain-of-Thought 7.2 A huge jump in improvement, most notably at the "Generation" column.
Self Refine 6.8 Even when asking Gemini to critique itself, it looks like it was satisfied with the inital baseline prompt output and didn't change much. The column titles are much better.
Final (COT + Self Refine) 7.5 This two-step final prompt approach produced an output I was pretty satisfied with. Although it's missing elements that already existed with the previous prompt techniques.

Key Lessons:

I feel like these prompt techniques do make a huge difference for output generation from LLMs, but only using these techniques as a singular query leads to answers that are less than ideal (in my opinion). By using multiple techniques and multiple prompts for a task, you can get a pretty decent response from an LLM.

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors