Tag
Flash-WAM introduces a modality-aware distillation method for world-action models, achieving real-time inference by compressing diffusion to a single step per modality, resulting in 23x speedup.