Circumventing shortcuts in audio-visual deepfake detection datasets with unsupervised learning

Re-assign community