ReCon: Improving the Accuracy of Community Detection on Signed Networks via Community Refinement and Contrastive Learning

Hyunuk Shin^{, 1} · Hojin Kim, 1 · Chanyoung Lee^{*, 1} · Yeon-Chang Lee² · David Yoon Suk Kang^{†, 1}

¹ Chungbuk National University · ² Ulsan National Institute of Science and Technology

Setup

We ran the experiments in the following environment:

- ubuntu: 24.04
- python: 3.10
- cuda: 12.5
- pytorch: 2.5.1
- GPU: RTX 3090

Installation

uv sync
uv pip install torch-scatter torch-sparse -f https://data.pyg.org/whl/torch-2.5.1+cu124.html

For our implementation, we based our code on SPONGE and SSSNET. Specifically, SPONGE and SSSNET are utilized for the initial clustering within our framework.

Usage Example

uv run python main.py --data-path data/SSBM/N1000_K5_p0.01_eta0.02.pt --initial-method SPONGE --K 5

Citation

@article{shin2026improving,
  title={Improving the Accuracy of Community Detection on Signed Networks via Community Refinement and Contrastive Learning},
  author={Shin, Hyunuk and Kim, Hojin and Lee, Chanyoung and Lee, Yeon-Chang and Kang, David Yoon Suk},
  journal={arXiv preprint arXiv:2601.16372},
  year={2026}
}