The official code for paper "Tutor-Student Reinforcement Learning: A Dynamic Curriculum for Robust Deepfake Detection" (CVPR 2026)
TSRL is implemented within the framework of DeepfakeBench. The provided code should be placed in the corresponding folders in DeepfakeBench.
A more comprehensive and refined version of the code, along with detailed usage instructions, will be made available as soon as I got spare time.
You may cite our paper by:
@article{lei2026tutor,
title={Tutor-Student Reinforcement Learning: A Dynamic Curriculum for Robust Deepfake Detection},
author={Zhanhe Lei and Zhongyuan Wang and Jikang Cheng and Baojin Huang and Yuhong Yang and Zhen Han and Chao Liang and Dengpan Ye},
year={2026},
journal={arXiv preprint arXiv:2603.24139},
}