Physical Words: A Vocabulary for Structuring Latent Reasoning in Robot Manipulation

原文：Physical Words: A Vocabulary for Structuring Latent Reasoning in Robot Manipulation

作者：Jiaming Liu, Xinghao Zhu, Yiming Zhang, Shanghang Zhang, Yinxi Wang, Yichu Yang, Zhuoyang Liu, Dantong Niu, Soha Pouya, Juekun Li, Yan Chang

来源：arXiv cs.RO（机器人）

正文

Computer Science > Robotics

arXiv:2610.26854v1 (cs)

[Submitted on 19 Oct 2026 (this version), latest version 20 Oct 2026 (v2)]

Title:Physical Words: A Vocabulary for Structuring Latent Reasoning in Robot Manipulation

Authors:Jiaming Liu, Xinghao Zhu, Yiming Zhang, Shanghang Zhang, Yinxi Wang, Yichu Yang, Zhuoyang Liu, Dantong Niu, Soha Pouya, Juekun Li, Yan Chang

View a PDF of the paper titled Physical Words: A Vocabulary for Structuring Latent Reasoning in Robot Manipulation, by Jiaming Liu and 10 other authors

View PDF

HTML (experimental)

Abstract:Multimodal reasoning provides robotic foundation models with predictive context needed for reliable action generation. However, language-based reasoning captures task semantics but struggles to express fine-grained physical dynamics, while latent reasoning models future state evolution but offers limited structure for organizing these dynamics into reusable reasoning primitives. To bridge this gap, we introduce PhysWord, a reasoning-before-acting foundation model built on Physical Words, which use atomic action verbs as shared semantic anchors for continuous representations of physical dynamics. Specifically, PhysWord predicts upcoming atomic action words and directly aligns their contextual hidden states with latent targets derived from the corresponding future observations. Building on these primitives, we construct a temporally structured PhysWord space with visual and optional tactile variants of each action word. Autoregressively predicting Physical Words across future timesteps captures transitions between execution stages, providing structured multimodal context for action generation. PhysWord achieves state-of-the-art success rates of 90.3% on visuotactile UniVTAC and 99.6% on vision-based LIBERO-Long, and outperforms competing methods on real-world tasks using bimanual grippers and dexterous hands. Further evaluations demonstrate that our structured latent reasoning framework generalizes to unseen scenarios and enables long-horizon manipulation.

Subjects:

Robotics (cs.RO)

Cite as:

arXiv:2610.26854 [cs.RO]

(or

arXiv:2610.26854v1 [cs.RO] for this version)

https://doi.org/10.48550/arXiv.2610.26854

Focus to learn more

arXiv-issued DOI via DataCite

主题

机器人

由「前沿雷达」于 2026-10-20 采集。正文取自原文页面，已保留出处链接。标题与正文版权归原作者所有。