2025/12/03 by Yuhang Huang, Huang, Yuhang, Shilong Zou +9
Computer Science · #Adaptation (eye) #Computational model #Control (management) #Data-driven #FOS: Computer and information sciences #Foundation (evidence) #Generative Adversarial Networks and Image Synthesis #Multimodal Machine Learning Applications #Reinforcement Learning in Robotics #Robotics (cs.RO) #Task (project management)
paper · pdf · doi:10.48550/arxiv.2512.03538
published in arXiv (Cornell University) (Cornell University)
openalex publication_date 2025/12/03 · openalex created_date 2025/12/05 · openalex updated_date 2026/07/28
World Foundation Models (WFMs) offer remarkable visual dynamics simulation capabilities, yet their application to precise robotic control remains limited by the gap between generative realism and control-oriented precision. While existing approaches use WFMs as synthetic data generators, they suffer from high computational costs and underutilization of pre-trained VLA policies. We introduce AdaPower (Adapt and Empower), a lightweight adaptation framework that transforms general-purpose WFMs into specialist world models through two novel components: Temporal-Spatial Test-Time Training (TS-TTT) for inference-time adaptation and Memory Persistence (MP) for long-horizon consistency. Integrated within a Model Predictive Control framework, our adapted world model empowers pre-trained VLAs, achieving over 41% improvement in task success rates on LIBERO benchmarks without policy retraining, while preserving computational efficiency and generalist capabilities.