Skip to content
Apple Machine Learning Research·· 2026-10-14AI score22

Apple proposes the first tri-modal masked diffusion model, systematically studying multimodal scaling laws

The Design Space of Tri-Modal Masked Diffusion Models

AI briefing

Apple researchers have proposed the first tri-modal masked diffusion model (MDM) pretrained from scratch, handling text, image-text, and audio-text data simultaneously, and systematically analyzed multimodal scaling laws, modality mixing ratios, noise schedules, and batch size effects.

Open original machinelearning.apple.com