NVIDIA发布Physis-Lang视频物理推理框架
NVIDIA出了个Physis-Lang,能让AI视频模型更懂物理规律,不重新训练就能提升5.62分基准成绩。
NVIDIA研究人员推出Physis-Lang,这是一个开放的自进化框架,为视频字幕添加物理推理能力。该框架解释场景展开的原因和方式,用于微调世界模型并在生成视频时添加推理到提示中。
仅将物理推理添加到提示中,就使NVIDIA Cosmos 3的PhyGenBench分数提高了5.62分,无需重新训练。
A video world model can make a convincing clip and still get the physics wrong. Our researchers just released Physis-Lang, an open self-evolving framework that adds physics reasoning to video captions. The captions explain why and how a scene unfolds. We use them to fine-tune world models and add that reasoning to prompts when generating video. Adding physics reasoning to the prompt alone improved NVIDIA Cosmos 3’s PhyGenBench score by 5.62 points without retraining. Paper and project: physis-intelligence.github.io/physis-lang-we…
This report is categorized under AITOP's "Models" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows and US-China foundation model benchmarks and commercialization comparison. Key entities and technical topics under continuous tracking include Physis-Lang, NVIDIA, Cosmos 3, video generation.
Source attribution: curated from NVIDIA AI (@NVIDIAAI). Cross-source corroboration: 11 independent sources have covered this development (including TestingCatalog, Replicate, Artificial Analysis), confirming industry-wide signal strength.