RoGA: Towards Generalizable Deepfake Detection through Robust Gradient Alignment

Open in new window