Preprint
Aug 2026
DREAM: Deployment-Time Demonstration Generation via Real-to-Sim for Scalable Policy Adaptation
DREAM is presented, a framework that generates fine-tuning data for a pretrained VLA from a captured workspace and a language instruction, without requiring a task-specific human demonstration, and whether it can serve as a scalable data-collection system for the deployment workspace.
Makoto Sato, T. Matsushima, Yutaka Matsuo et al.
· 1 citation