GNN-RAG integrates Graph Neural Networks (GNNs) with Retrieval-Augmented Generation (RAG) to enhance Large Language Model (LLM) reasoning over knowledge graphs. It targets researchers and developers working on knowledge-based question answering and LLM reasoning, offering improved accuracy and explainability by grounding LLM responses in structured knowledge.

How It Works

The system employs GNNs to retrieve relevant entities and relations from a knowledge graph, which are then used as context for an LLM. This approach leverages the structural information within knowledge graphs to provide more precise and contextually relevant answers than traditional RAG methods, enabling more robust reasoning.

Quick Start & Requirements

• Install: pip install -r requirements.txt (within gnn and llm directories respectively).
• Prerequisites: Python 3.x, PyTorch, Transformers, RDFLib. Specific GNN training may require additional libraries.
• Precomputed GNN Output: Skip GNN training by using precomputed results found in llm/results/gnn.
• Results: Detailed results and LLM generations are available in the results/ directory.

Highlighted Details

• Provides implementations for various Knowledge Graph Question Answering (KGQA) GNNs.
• Includes RAG-based KGQA implementation for LLMs.
• Offers precomputed GNN outputs to bypass training.
• Detailed results and LLM generations are included for reproducibility.

Maintenance & Community

This project is maintained by cmavro. No specific community channels or roadmap details are provided in the README.

Licensing & Compatibility

The README does not specify a license. Users should verify licensing for commercial use or integration into closed-source projects.

Limitations & Caveats

The project's licensing is not specified, which may pose a barrier to commercial adoption. The README implies that GNN training might be complex, suggesting users may prefer to rely on precomputed results.