Research

Paper

AI LLM March 13, 2026

RoboStereo: Dual-Tower 4D Embodied World Models for Unified Policy Optimization

Authors

Ruicheng Zhang, Guangyu Chen, Zunnan Xu, Zihao Liu, Zhizhou Zhong, Mingyang Zhang, Jun Zhou, Xiu Li

Abstract

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.

Metadata

arXiv ID: 2603.12639
Provider: ARXIV
Primary Category: cs.CV
Published: 2026-03-13
Fetched: 2026-03-16 06:01

Related papers

Raw Data (Debug)
{
  "raw_xml": "<entry>\n    <id>http://arxiv.org/abs/2603.12639v1</id>\n    <title>RoboStereo: Dual-Tower 4D Embodied World Models for Unified Policy Optimization</title>\n    <updated>2026-03-13T04:16:19Z</updated>\n    <link href='https://arxiv.org/abs/2603.12639v1' rel='alternate' type='text/html'/>\n    <link href='https://arxiv.org/pdf/2603.12639v1' rel='related' title='pdf' type='application/pdf'/>\n    <summary>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 &gt;97% average relative improvement on fine-grained manipulation tasks.</summary>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.CV'/>\n    <published>2026-03-13T04:16:19Z</published>\n    <arxiv:primary_category term='cs.CV'/>\n    <author>\n      <name>Ruicheng Zhang</name>\n    </author>\n    <author>\n      <name>Guangyu Chen</name>\n    </author>\n    <author>\n      <name>Zunnan Xu</name>\n    </author>\n    <author>\n      <name>Zihao Liu</name>\n    </author>\n    <author>\n      <name>Zhizhou Zhong</name>\n    </author>\n    <author>\n      <name>Mingyang Zhang</name>\n    </author>\n    <author>\n      <name>Jun Zhou</name>\n    </author>\n    <author>\n      <name>Xiu Li</name>\n    </author>\n  </entry>"
}