Preference-Driven Refinement (PDR)

This repository contains code, experiments, and results for Preference-Driven Refinement (PDR) — a structured prompt refinement framework for improving LLM output quality using explicit preferences.

Repository Structure

pdr/
├── pdr-gpt_results/
├── pdr-gpt4o-Comparison/
├── pdr-gpt4-withcritic/
├── pdr-gpt4-withoutcritic/
├── pdr-gpt5/
├── pdr-gpt5-software-development/
├── pdr-gpt5-Comparison/
├── code/
├── graphs/
└── README.md