Liming Lu, Xianzheng Ma, Wenkun He, Guanqi Zhan, Yilin Zhao, Junyu Chen, Mengyao Xu, Jiaojiao Fan, Wenhang Ge, Yuchao Gu, Yunze Liu, Boyi Li, Zhen Dong, Victor Prisacariu, Ming-Yu Liu, Song Han, Han Cai
Physis-Lang is a self-evolving framework that represents physical processes through language, improving video world models' physical plausibility. It introduces PhysCapBench to evaluate and refine physical captions. Physis-Lang-enhanced models surpass leading proprietary models.
Physis-Lang's approach can generate more realistic videos by adhering to physical principles, which is crucial for applications like robotics, gaming, and simulation. This can lead to more accurate predictions and better decision-making. Improved physical plausibility can also enhance user experience in various industries.
Thank you to arXiv for use of its open access interoperability. arxiv.org. aibrief.ai is not affiliated with or endorsed by arXiv.