EditMem: Enhancing Multi-Hop Fact Verification via Editable Memory

Yuqing Lan¹, Zhenghao Liu¹, Yu Gu¹, Ge Yu¹

¹School of Computer Science and Engineering, Northeastern University, Shenyang 110819, China

📖 Overview

EditMem, a multi-hop fact verification framework built upon an editable memory.

⚙️ Setup

Create the environment and install the required packages.

conda create --name editmem python==3.10
conda activate editmem
git clone https://github.com/xxx.git
cd editmem
pip install -r requirement.txt

🔧 Reproduction Editmem

Prepare data index

• Build the corpus index of Hover and Feverous.

cd ./scripts
bash build_hover_index.sh
bash build_feverous_index.sh

Evaluate Prediction Effectiveness

• These experimental results are shown in Table 1 of our paper.
• Go to the scripts folder and evaluate model performance as follows:

cd ./scripts
bash inference.sh
bash eval.sh

🥰 Citation

Please cite the paper in the following format.

@inproceedings{lan2026editmem,
  title={EditMem: Enhancing Multi-Hop Fact Verification via Editable Memory},
  author={Lan, Yuqing and Liu, Zhenghao and Gu, Yu and Yu, Ge},
  booktitle    = {2026 {IEEE} International Conference on Acoustics, Speech and Signal
                  Processing, {ICASSP} 2026, Barcelona, Spain, May 4-8, 2026},
  year={2026}
}

📧 Contact

If you have questions, suggestions, and bug reports, please email:

lanyuqing@stumail.neu.edu.cn