Prompt Analysis and Refinement for Large Language Models

Bachelor & Master Thesis

Description:
Prompt design plays a critical role in determining the behavior and output quality of large language models (LLMs). However, prompts are often treated as unstructured text, making it hard to analyze, debug, or optimize them systematically. This thesis investigates how different prompt patterns influence model behavior and explores methods to visualize, compare, and refine prompts for better performance. Students will build or extend tools for prompt pattern recognition, performance pattern visualization, and debugging to support developers working with LLMs.

Requirements:

• Python experience and familiarity with LLM APIs (OpenAI, Claude, etc.)

• Interest in human-AI interaction, developer tools, or NLP behavior analysis