UniGFM-RAG: Model Checkpoint
This repository provides the model checkpoint for UniGFM-RAG, associated with our paper:
UniGFM-RAG: Unify Graph Foundation Models and Large Language Models for End-to-End Retrieval and Reasoning
IEEE Transactions on Knowledge and Data Engineering (TKDE), 2026.
π Paper: IEEE Xplore | arXiv
Model Description
UniGFM-RAG is an end-to-end Graph Retrieval-Augmented Generation (GraphRAG) framework that unifies Graph Foundation Models (GFMs) and Large Language Models (LLMs) for complex multi-hop retrieval and reasoning.
The framework integrates a GFM-based retriever with an LLM generator through a differentiable graph bridge, enabling joint optimization of retrieval and reasoning.
Intended Use
This checkpoint is released for research and reproducibility purposes, particularly for graph-based retrieval, multi-hop question answering, and end-to-end GraphRAG.
For implementation details, please refer to the original paper.
Citation
If you find this model useful, please cite our paper:
@ARTICLE{zhao2026unigfmrag,
author={Zhao, Zicheng and Luo, Linhao and Yu, Zongji and Dong, Junnan and Tan, Yanchao and Pan, Shirui and Gong, Chen},
journal={IEEE Transactions on Knowledge and Data Engineering},
title={UniGFM-RAG: Unify Graph Foundation Models and Large Language Models for End-to-End Retrieval and Reasoning},
year={2026}
}