arXiv Robotics — research abstracts· Ke Wu, Hanwen Huang, Bo Gu, Kaizhao Zhang, Xiangting Meng, Yupeng Zheng, Zijun Xu, Jieru Zhao, Wenchao Ding·· 1 days agoEditorial score60
ΔWAM: Distilling Action Tangent Fields into World Action Models
{\Delta}WAM: Distilling Action Tangent Fields into World Action Models
Summary
This research proposes Action Tangent Fields to enhance World Action Models (WAM) by focusing on action-dependent dynamics rather than appearance-based predictions. The method uses a Residual-VAE space and an action-conditioned world model to distill local first-order structures into WAM, improving robustness to environmental perturbations without large-scale embodied pretraining.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
Source：arXiv Robotics — research abstracts · arxiv.org