LATIDIA · Robótica
$Z$-sight: Unificación de la visión, el lenguaje y la acción y modelado del mundo latente
arXiv:2611.24940v1 Tipo de anuncio: nuevo Resumen: Presentamos $Z$-sight, un modelo fundacional unificado para robots que combina visión-lenguaje-acción (VLA) y el modelado generativo del mundo latente. Modelos de acción en el mundo basados en vídeo (WA
WhatsApp ↗Telegram ↗
La noticia
arXiv:2611.24940v1 Announce Type: new Abstract: We introduce $Z$-sight, a unified robot foundation model that couples vision-language-action (VLA) and generative latent world modeling. Video-based world action models (WAMs) show the value of joint future-and-action generation for control, yet inherit the heavy cost of video generators. Recent latent-space policies avoid pixel generation, yet typically align intermediate features or rely on frozen target encoders, decoupling future prediction from the VLM's native feature space. $Z$-sight instead predicts the future directly in the visual token space in which its VLM perceives the present. Coupling a foresight expert and a continuous action expert with the VLM in a mixture-of-transformers, $Z$-sight jointly denoises future spatial tokens and