AI Trivia Agent
This project was a group-built AI agent system designed to answer trivia questions using language model reasoning, prompt engineering, and collaborative decision-making.
The goal was to explore how LLM-powered agents can break down questions, reason through possible answers, and coordinate responses in a structured way.
This was a group project based on a forked repository. My contribution focused on understanding the agent workflow, improving prompts, testing responses, and helping evaluate the system’s reasoning behavior.
Project Overview
The trivia agent takes in a question, reasons through the possible answer choices, and produces a final response. The project demonstrates how AI agents can be structured to solve open-ended reasoning tasks rather than simply return direct completions.
What I Worked On
- Tested the agent across different trivia question types
- Reviewed model responses for reasoning quality and accuracy
- Helped improve prompt structure and response formatting
- Explored how agents handle uncertainty and competing answer choices
- Documented the project workflow for public presentation
Technical Focus
What This Project Demonstrates
This project helped me understand how AI systems can be designed around structured reasoning instead of one-shot prompting. It also gave me experience working with collaborative AI codebases, testing agent behavior, and thinking critically about the limitations of LLM-generated answers.