Phys4D: Fine-Grained Physics-Consistent 4D Modeling from Video Diffusion

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Lu, Haoran, Wu, Shang, Zhang, Jianshu, Su, Maojiang, Ye, Guo, Xu, Chenwei, Lu, Lie, Maneriker, Pranav, Du, Fan, Li, Manling, Wang, Zhaoran, Liu, Han
Formato: Preprint
Publicado: 2026
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866915839064670208
author Lu, Haoran
Wu, Shang
Zhang, Jianshu
Su, Maojiang
Ye, Guo
Xu, Chenwei
Lu, Lie
Maneriker, Pranav
Du, Fan
Li, Manling
Wang, Zhaoran
Liu, Han
author_facet Lu, Haoran
Wu, Shang
Zhang, Jianshu
Su, Maojiang
Ye, Guo
Xu, Chenwei
Lu, Lie
Maneriker, Pranav
Du, Fan
Li, Manling
Wang, Zhaoran
Liu, Han
contents Recent video diffusion models have achieved impressive capabilities as large-scale generative world models. However, these models often struggle with fine-grained physical consistency, exhibiting physically implausible dynamics over time. In this work, we present \textbf{Phys4D}, a pipeline for learning physics-consistent 4D world representations from video diffusion models. Phys4D adopts \textbf{a three-stage training paradigm} that progressively lifts appearance-driven video diffusion models into physics-consistent 4D world representations. We first bootstrap robust geometry and motion representations through large-scale pseudo-supervised pretraining, establishing a foundation for 4D scene modeling. We then perform physics-grounded supervised fine-tuning using simulation-generated data, enforcing temporally consistent 4D dynamics. Finally, we apply simulation-grounded reinforcement learning to correct residual physical violations that are difficult to capture through explicit supervision. To evaluate fine-grained physical consistency beyond appearance-based metrics, we introduce a set of \textbf{4D world consistency evaluation} that probe geometric coherence, motion stability, and long-horizon physical plausibility. Experimental results demonstrate that Phys4D substantially improves fine-grained spatiotemporal and physical consistency compared to appearance-driven baselines, while maintaining strong generative performance. Our project page is available at https://sensational-brioche-7657e7.netlify.app/
format Preprint
id arxiv_https___arxiv_org_abs_2603_03485
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Phys4D: Fine-Grained Physics-Consistent 4D Modeling from Video Diffusion
Lu, Haoran
Wu, Shang
Zhang, Jianshu
Su, Maojiang
Ye, Guo
Xu, Chenwei
Lu, Lie
Maneriker, Pranav
Du, Fan
Li, Manling
Wang, Zhaoran
Liu, Han
Computer Vision and Pattern Recognition
Artificial Intelligence
Robotics
Recent video diffusion models have achieved impressive capabilities as large-scale generative world models. However, these models often struggle with fine-grained physical consistency, exhibiting physically implausible dynamics over time. In this work, we present \textbf{Phys4D}, a pipeline for learning physics-consistent 4D world representations from video diffusion models. Phys4D adopts \textbf{a three-stage training paradigm} that progressively lifts appearance-driven video diffusion models into physics-consistent 4D world representations. We first bootstrap robust geometry and motion representations through large-scale pseudo-supervised pretraining, establishing a foundation for 4D scene modeling. We then perform physics-grounded supervised fine-tuning using simulation-generated data, enforcing temporally consistent 4D dynamics. Finally, we apply simulation-grounded reinforcement learning to correct residual physical violations that are difficult to capture through explicit supervision. To evaluate fine-grained physical consistency beyond appearance-based metrics, we introduce a set of \textbf{4D world consistency evaluation} that probe geometric coherence, motion stability, and long-horizon physical plausibility. Experimental results demonstrate that Phys4D substantially improves fine-grained spatiotemporal and physical consistency compared to appearance-driven baselines, while maintaining strong generative performance. Our project page is available at https://sensational-brioche-7657e7.netlify.app/
title Phys4D: Fine-Grained Physics-Consistent 4D Modeling from Video Diffusion
topic Computer Vision and Pattern Recognition
Artificial Intelligence
Robotics
url https://arxiv.org/abs/2603.03485