From Retrieval to Generation: Unifying External and Parametric Knowledge for Medical Question Answering

🔭 Overview

From Retrieval to Generation: Unifying External and Parametric Knowledge for Medical Question Answering (Accepted in WWW2026)

In this work, we propose MedRGAG, a unified retrieval--generation augmented framework that seamlessly integrates external and parametric knowledge for medical QA.

MedRGAG comprises two key modules: Knowledge-Guided Context Completion (KGCC), which directs the generator to produce background documents that complement the missing knowledge revealed by retrieval; and Knowledge-Aware Document Selection (KADS), which adaptively selects an optimal combination of retrieved and generated documents to form concise yet comprehensive evidence for answer generation.

⚙️ Installation

Note that the code in this repo runs under Linux system. We have not tested whether it works under other OS.

1. Clone this repository:

   cd MedRGAG

2. Create and activate the conda environment:

   conda create -n medrgag python=3.10
   conda activate automir
   pip install torch==2.6.0
   pip install faiss-gpu==1.7.2
   pip install deepspeed==0.17.4
   pip install transformers==44.53.2
   pip install sentence-transformers==5.0.0
   pip install datasets==3.6.0
   pip install vllm==0.8.5
   pip install openai==1.86.0

💽 Evaluate

Run the following command to get results:

python main.py --reader_name qwen2.5 --data_name medqa
* `--reader_name`: the reader LLM.
* `--data_name`: the dataset name.

You will get the evaluation results in the outputs/ folder, which contains the evaluation results.

📜Reference

If this code or dataset contributes to your research, please kindly consider citing our paper and give this repo ⭐️ :)

@misc{li2025retrievalgenerationunifyingexternal,
      title={From Retrieval to Generation: Unifying External and Parametric Knowledge for Medical Question Answering}, 
      author={Lei Li and Xiao Zhou and Yingying Zhang and Xian Wu},
      year={2025},
      eprint={2510.18297},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2510.18297}, 
}