🚀 Exciting News!
✨ We are thrilled to announce that our paper, titled "Consistent and Invariant Generalization Learning for Short-video Misinformation Detection", has been accepted for presentation at ACM MM 2025! 🎉 🎉
├── README
├── requirements.txt
├── FakeTT_Domain_output
├── FakeSV_Domain_output
├── data
├── models
├── utils
├── provided_ckp
├── fea
├── main.py
├── run.sh
└── run.pyWe conduct experiments on two datasets: FakeSV and FakeTT.
- You can download preprocessed features in this repo.
conda create -n DOCTOR python=3.9
conda activate DOCTOR
pip install -r requirements.txt # You can classify the data into different domains by this commend
python domain_classify.py- To facilitate access to domain-specific data, we provide domain-partitioned datasets (fakesv and fakett) in the data folder FakeSV_Domain_output and FakeTT_Domain_output.
You can train and test by following code:
# First, select different training and testing domains based on your criteria in domain_split.py, then run it to generate the splits.
python domain_split.py
# Then you can fast run by this:
./run.sh
# Or you can run like this:
# FakeSV:
python main.py --dataset fakesv --mode train --inference_ckp ./provided_ckp/FakingRecipe_fakesv --dg --diffusion --alpha 0.1 --beta 3 --gamma 0.05 --lr 5e-5
# FakeTT:
python main.py --dataset fakett --mode train --inference_ckp ./provided_ckp/FakingRecipe_fakett --dg --diffusion --alpha 0.1 --beta 3 --gamma 0.05 --lr 1e-3@inproceedings{guo2025consistent,
title={Consistent and Invariant Generalization Learning for Short-video Misinformation Detection},
author={Guo, Hanghui and Shi, Weijie and Li, Mengze and Li, Juncheng and Chen, Hao and Cui, Yue and Xu, Jiajie and Zhu, Jia and Shen, Jiawei and Chen, Zhangze and others},
booktitle={Proceedings of the 33rd ACM International Conference on Multimedia},
pages={2254--2263},
year={2025}
}