README.md

Let's Ask GNN: Empowering Large Language Models for Graph In-Context Learning

This repository contains the implementation of the paper "Let's Ask GNN: Empowering Large Language Models for Graph In-Context Learning". The goal of this work is to combine Graph Neural Networks (GNNs) with Large Language Models (LLMs) to enhance graph understanding and relational reasoning capabilities in a variety of graph-based tasks.

scripts

To train the GNN using the SAGE model on the OGBN-Arxiv dataset, use the following command:

python gen_result_local_llm.py  --stru sage   --dataset ogbn-arxiv --ratio 0.05

• --stru: The type of GNN architecture to use (e.g., sage for GraphSAGE).
• --dataset: The graph dataset to use (e.g., ogbn-arxiv).
• --ratio: The proportion of data to use (e.g., 0.05 represents 5%).

After training the GNN and generating embeddings, the next step is to perform inference with a Large Language Model (LLM). Use the following command for inference:

python llm_inference.py  --stru sage   --dataset ogbn-arxiv --ratio 0.05  --llm_model qwen

• --stru: Specifies the GNN model (e.g., sage).
• --dataset: The graph dataset to use (e.g., ogbn-arxiv).
• --ratio: The proportion of the dataset to use (same as in the training step).
• --llm_model: The LLM model to use (e.g., qwen).