Episode Details

Back to Episodes

DreamZero: A 14B-Parameter Vision-to-Action World Model for Robotics

Published 6 months, 1 week ago
Description
UNVERIFIED: 14B parameter autoregressive diffusion model that jointly models video and actions for state-of-the-art generalization on robotics benchmarks like MolmoSpaces and RoboArena, targeting generalizable manipulation beyond fine-tuned VLAs.
Listen Now

Love PodBriefly?

If you like Podbriefly.com, please consider donating to support the ongoing development.

Support Us