Saved in:
Bibliographic Details
Main Authors: Zhang, Ruicheng, Chen, Guangyu, Xu, Zunnan, Liu, Zihao, Zhong, Zhizhou, Zhang, Mingyang, Zhou, Jun, Li, Xiu
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2603.12639
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913025737359360
author Zhang, Ruicheng
Chen, Guangyu
Xu, Zunnan
Liu, Zihao
Zhong, Zhizhou
Zhang, Mingyang
Zhou, Jun
Li, Xiu
author_facet Zhang, Ruicheng
Chen, Guangyu
Xu, Zunnan
Liu, Zihao
Zhong, Zhizhou
Zhang, Mingyang
Zhou, Jun
Li, Xiu
contents Scalable Embodied AI faces fundamental constraints due to prohibitive costs and safety risks of real-world interaction. While Embodied World Models (EWMs) offer promise through imagined rollouts, existing approaches suffer from geometric hallucinations and lack unified optimization frameworks for practical policy improvement. We introduce RoboStereo, a symmetric dual-tower 4D world model that employs bidirectional cross-modal enhancement to ensure spatiotemporal geometric consistency and alleviate physics hallucinations. Building upon this high-fidelity 4D simulator, we present the first unified framework for world-model-based policy optimization: (1) Test-Time Policy Augmentation (TTPA) for pre-execution verification, (2) Imitative-Evolutionary Policy Learning (IEPL) leveraging visual perceptual rewards to learn from expert demonstrations, and (3) Open-Exploration Policy Learning (OEPL) enabling autonomous skill discovery and self-correction. Comprehensive experiments demonstrate RoboStereo achieves state-of-the-art generation quality, with our unified framework delivering >97% average relative improvement on fine-grained manipulation tasks.
format Preprint
id arxiv_https___arxiv_org_abs_2603_12639
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle RoboStereo: Dual-Tower 4D Embodied World Models for Unified Policy Optimization
Zhang, Ruicheng
Chen, Guangyu
Xu, Zunnan
Liu, Zihao
Zhong, Zhizhou
Zhang, Mingyang
Zhou, Jun
Li, Xiu
Computer Vision and Pattern Recognition
Scalable Embodied AI faces fundamental constraints due to prohibitive costs and safety risks of real-world interaction. While Embodied World Models (EWMs) offer promise through imagined rollouts, existing approaches suffer from geometric hallucinations and lack unified optimization frameworks for practical policy improvement. We introduce RoboStereo, a symmetric dual-tower 4D world model that employs bidirectional cross-modal enhancement to ensure spatiotemporal geometric consistency and alleviate physics hallucinations. Building upon this high-fidelity 4D simulator, we present the first unified framework for world-model-based policy optimization: (1) Test-Time Policy Augmentation (TTPA) for pre-execution verification, (2) Imitative-Evolutionary Policy Learning (IEPL) leveraging visual perceptual rewards to learn from expert demonstrations, and (3) Open-Exploration Policy Learning (OEPL) enabling autonomous skill discovery and self-correction. Comprehensive experiments demonstrate RoboStereo achieves state-of-the-art generation quality, with our unified framework delivering >97% average relative improvement on fine-grained manipulation tasks.
title RoboStereo: Dual-Tower 4D Embodied World Models for Unified Policy Optimization
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2603.12639