This repository contains the implementation and results for the project:
"Evaluating Node Ordering Effects on Large Language Model Graph Reasoning: A Systematic Analysis"
The project investigates how node ordering in graph representations impacts the performance of Large Language Models (LLMs) on algorithmic graph reasoning tasks.
Experiments are conducted with Llama 3.2 and Gemma 3:12B, using various prompt configurations (unordered, ordered, ordered with explanation).
├── output_poly_multi_table/ # Experimental outputs and accuracy tables
├── ordering_experiment_poly_multi_table.py # Main script for running experiments
├── plot_results.py # Script to generate performance plots
├── run_experiment.sh # Shell script to automate experiments
├── run_ordering_poly_multi_table.sh # Shell script for ordering-specific experiments
└── README.md # Project documentation