RoGA: Towards Generalizable Deepfake Detection through Robust Gradient Alignment

Re-assign community