KG-Retriever: Efficient Knowledge Indexing for Retrieval-Augmented Large Language Models

This repository contains the code for the paper "KG-Retriever: Efficient Knowledge Indexing for Retrieval-Augmented Large Language Models"

Environment Requirements

• Python Version: 3.10.13

Installation Steps

STEP 1: Prepare Datasets

1. Download Datasets and LLMs:

2. Preprocess Datasets:

   • Run src/dataset/preprocess/xxx.py for each datasets
   • Run extract_triples for triples extraction

STEP 2: Run the Models Demo

1. Run run_demo_EX_hotpop_v0.py to run models in base mode
2. Run run_demo_EX_hotpop_v1.py to run models in additional modes

Citation

If you use this project in your research, please cite the following paper:

@article{chen2024kg,
  title={Kg-retriever: Efficient knowledge indexing for retrieval-augmented large language models},
  author={Chen, Weijie and Bai, Ting and Su, Jinbo and Luan, Jian and Liu, Wei and Shi, Chuan},
  journal={arXiv preprint arXiv:2412.05547},
  year={2024}
}