GLPN-LLM

Source code (PyTorch) and dataset of the paper "Synergizing LLMs with Global Label Propagation for Multimodal Fake News Detection", which is accepted by The 63rd Annual Meeting of the Association for Computational Linguistics (ACL 2025).

Dataset

You can download the dataset from here.

Unzip the dataset and put it in the script/dataset folder. We have three datasets: weibo, twitter and pheme. The dataset is in the form of csv files.

The file structure is as follows:

script/dataset/
    weibo/
    twitter/
    pheme/

Requirements

You can install the requirements by running the following command:

pip install -r requirements.txt

Note: It's recommended to install the CLIP package directly from the official GitHub repository.

Running

You can run the code by running the following command:

sh run.sh

Note: you can use the psesudo labels generated by GPT-4o to train the model, dataset/{args.dataset_name}/{args.dataset_name}_analysis_results.csv is the file that stores the psesudo labels.

Cite

If you use GLPN-LLM in a scientific publication, we would appreciate citations to the following paper:

@article{hu2025synergizing,
  title={Synergizing LLMs with Global Label Propagation for Multimodal Fake News Detection},
  author={Hu, Shuguo and Hu, Jun and Zhang, Huaiwen},
  journal={arXiv preprint arXiv:2506.00488},
  year={2025}
}

License: GPLv3

Copyright (c) 2024-2025 IMU, China & NUS, Singapore.