PhysVid：面向生成式视频模型的物理感知局部条件控制
PhysVid: Physics Aware Local Conditioning for Generative Video Models
把物理约束从全局文本提示词下沉到时间连续帧块级的局部条件，并用块感知交叉注意力融合、推理时用负向物理提示词引导采样，这套「局部条件+负向提示」机制可直接迁移到任何以文本为条件的视频扩散模型上提升物理合理性。
原文依据 we introduce negative physics prompts (descriptions of locally relevant law violations) to steer generation away from implausible trajectories
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论生成式视频模型能够达到很高的视觉保真度，但常常违背基本物理规律，这限制了其在真实场景中的可靠性。此前注入物理知识的尝试依赖条件控制：帧级信号具有领域特定性和短时程性，而全局文本提示词又过于粗糙且噪声较大，会遗漏细粒度的动力学信息。我们提出 PhysVid，一种在时间上连续的帧块上运作的物理感知局部条件控制方案。每个帧块都标注了基于物理的状态、交互与约束描述，并在训练时通过块感知交叉注意力与全局提示词融合。在推理阶段，我们引入负向物理提示词（对局部相关违法物理规律情形的描述），以引导生成远离不合理的轨迹。在 VideoPhy 上，PhysVid 相较基线视频生成器将物理常识得分提升约 33%，在 VideoPhy2 上提升最高约 8%。这些结果表明，局部的、物理感知的引导能够显著提升生成视频的物理合理性，并朝着物理接地的视频模型迈出一步。
英文原文摘要 1109 字符
Generative video models achieve high visual fidelity but often violate basic physical principles, limiting reliability in real-world settings. Prior attempts to inject physics rely on conditioning: frame-level signals are domain-specific and short-horizon, while global text prompts are coarse and noisy, missing fine-grained dynamics. We present PhysVid, a physics-aware local conditioning scheme that operates over temporally contiguous chunks of frames. Each chunk is annotated with physics-grounded descriptions of states, interactions, and constraints, which are fused with the global prompt via chunk-aware cross-attention during training. At inference, we introduce negative physics prompts (descriptions of locally relevant law violations) to steer generation away from implausible trajectories. On VideoPhy, PhysVid improves physical commonsense scores by ~ 33% over baseline video generators, and by up to ~ 8% on VideoPhy2. These results show that local, physics-aware guidance substantially increases physical plausibility in generative video and marks a step toward physics-grounded video models.