VLA ModelsarXiv preprint2025
Harness VLA: Steering Frozen VLAs into Reliable Manipulation Primitives via Memory-Guided Agents
Yixian Zhang, Huanming Zhang, Feng Gao, Xiao Li, Zhihao Liu, Chunyang Zhu, Jiaxing Qiu, Yuchen Yan, Jiyuan Liu, Wenhao Tang, Zhengru Fang, Yi Nie, Changxu Wei, Yu Wang, Wenbo Ding, Chao Yu et al.Stanford University
摘要
We propose Harness VLA, a method for steering frozen pretrained VLA models into reliable manipulation primitives through memory-guided agents. Our approach enables deployment of large VLAs without fine-tuning by using episodic memory and retrieval-augmented planning to guide action selection.
VLAmanipulation primitivesmemory-guidedfrozen modelrobot control
