An end-to-end, hierarchical VTLA model for fine manipulation, enabling native, anthropomorphic last-millimeter interaction.
CraftNet is the combination of System 0 and System 1, forming a VTLA (Vision-Tactile-Language-Action) model that translates high-level, multimodal inputs into continuous, fine-grained actions executed by dexterous humanoid robots — at a level of precision and speed previously only achievable by humans.