Hyper-networks for Unified Visual Representation (HUVR) use implicit neural representation to bridge the gap between understanding and generation in vision encoders.