Researchers have proposed a new deepfake detection method called DF-MoE that can generalize across different deepfake generation methods. The approach combines features from multiple sources, including audio and visual modalities, to identify fake content. In experiments on five benchmarks, DF-MoE outperformed existing methods in detecting deepfakes. The model's code has been made publicly available. --- Why it matters: This matters because it addresses a significant challenge in deepfake detection: generalizing across different methods. Engineers and researchers can benefit from this work by developing more robust and effective deepfake detection systems. Source: https://arxiv.org/abs/2608.23363