Breaking domain barriers: mixture of experts for cross-domain fake news detection

This is the accompanying code for the paper:

Breaking domain barriers: mixture of experts for cross-domain fake news detection

Requirements

See requirements.txt

Folder Structure

./
│
├── train.py - main script to start training
│
├── configs/config*.json - holds configurations for training
├── parse_config.py - class to handle config file and cli options
│
├── base/ - abstract base classes
├── data_loader/ - anything about data loading
├── dataset/ - default directory for storing input data
├── model/ - models, losses, and metrics
├── trainer/ - trainers
├── logger/ - module for tensorboard visualization and logging
└── utils/ - small utility functions

Usage

The code in this repo could be executed with these commands:

# Run train of base model then MoE

python train.py -c configs/config_crossdomain_moe_base_models.json --seed $i --output_folder $BASE_EXPERT_FOLDER
  
python train.py -c configs/config_crossdomain_moe.json --seed $i

Reference

Liguori, A., Pisani, F.S., Comito, C. et al. Breaking domain barriers: mixture of experts for cross-domain fake news detection. Mach Learn 114, 188 (2025). https://doi.org/10.1007/s10994-025-06827-9