PhyWorld: Physics-Faithful World Model for Video Generation
Quick Take
PhyWorld enhances video generation models for physically faithful world simulations through two-stage post-training.
Key Points
- Implements flow matching for stable visual attributes.
- Aligns dynamics with physics using Direct Preference Optimization.
- Achieves higher scores in video consistency and physical plausibility.
📖 Reader Mode
~2 min readAuthors:Pu Zhao, Juyi Lin, Timothy Rupprecht, Arash Akbari, Chence Yang, Rahul Chowdhury, Elaheh Motamedi, Arman Akbari, Yumei He, Chen Wang, Geng Yuan, Weiwei Chen, Yanzhi Wang
Abstract:World simulators can provide safe and scalable environments for training Physical AI systems before real-world deployment. Large video generation models are emerging as a promising basis for such simulators because they can generate diverse and realistic visual futures. However, using them as world simulators requires physically faithful video continuations, namely, generated videos that preserve the physical state implied by the conditioning input, and evolve in ways consistent with basic physical principles. We propose PhyWorld, a video generation world model designed to produce temporally coherent and physically faithful scene continuations through two-stage post-training. In the first stage, we improve video-to-video continuation with flow matching fine-tuning, encouraging stable visual attributes and coherent motion dynamics across frames. In the second stage, we align generated dynamics with physical principles using Direct Preference Optimization (DPO) over physics preference pairs, guiding the model toward outputs with higher physical plausibility. To evaluate PhyWorld, we use both standard video-quality benchmarks and a dedicated physical-faithfulness benchmark with per-law scoring. Experiments show that PhyWorld improves video consistency, achieving an average score of 0.769 on VBench compared with 0.756 or below for state-of-the-art baselines. PhyWorld also improves physical plausibility, reaching an average score of 3.09 on our physical-faithfulness benchmark compared with 2.99 for the strongest baseline. These results suggest that post-training large video generation models with continuation and physics-preference signals can make them more effective world simulators for Physical AI.
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Emerging Technologies (cs.ET); Machine Learning (cs.LG); Multimedia (cs.MM) |
| Cite as: | arXiv:2605.19242 [cs.CV] |
| (or arXiv:2605.19242v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2605.19242 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Pu Zhao [view email]
[v1]
Tue, 19 May 2026 01:28:52 UTC (8,901 KB)
— Originally published at arxiv.org
More from arXiv cs.CV
See more →GeoSym127K: Scalable Symbolically-verifiable Synthesis for Multimodal Geometric Reasoning
GeoSym127K introduces a scalable neuro-symbolic framework for enhanced geometric reasoning in multimodal models.