paper/arxiv/2609.03681

Robotics Dataset

This paper introduces WISE, a framework for efficient post-training of Vision-Language-Action (VLA) models using world-model-guided imagination scheduling, demonstrating consistent improvements across manipulation tasks with reduced GPU computation time.

Catalog metadata

Repository owner

Chenhao Zhang, Hanyu Zhao, Hang Cheng et al.

Source

Research paper

Task

WISE: World-model-guided Imagination Scheduling for Efficient Post-training of Vision-Language-Action Models

Cameras

0

Browse the catalog to compare robots, tasks, sensors, and licenses. Sign in to open protected download and model links.