Social-MAE: A Transformer-Based Multimodal Autoencoder for Face and Voice Paper • 2508.17502 • Published Aug 24 • 1 • 2