GAUGE: A Measurement-Grounded Benchmark for Physical Fidelity in Simulation Engines and Video World Models
作者: Shuai Wang, Yaxin Feng, Xuekun Jiang, Shihan Tian, Ningyu Yan, Xing Shen, Chaoyang Lyu, Hui Wang, Yunsong Zhou, Hanqing Wang, Jiangmiao Pang, Yang Xiang, Xing Gao, Chunhua Shen, Weinan Zhang
分类: cs.AI, cs.CV, cs.RO
发布日期: 2026-08-06
💡 一句话要点
提出GAUGE基准以评估模拟引擎和视频世界模型的物理真实度
🎯 匹配领域: 支柱二:RL算法与架构 (RL & Architecture)
关键词: 物理引擎 视频世界模型 物理真实度 评估基准 具身智能 模拟器 生成模型 控制任务
📋 核心要点
- 现有的物理真实度评估方法往往孤立进行,缺乏对物理原理的深入分析,导致评估结果的局限性。
- 本文提出GAUGE基准,通过22个控制任务家族联合评估数值模拟器和生成视频世界模型的物理真实度。
- 实验结果表明,没有统一的物理引擎能够完全忠实于现实,尤其在冲击接触和快速变形方面存在显著差异。
📝 摘要(中文)
物理引擎为具身智能的大规模训练和评估提供了便利,而生成的视频世界模型则作为未来状态和交互的隐式模拟器逐渐兴起。然而,现有的物理真实度评估往往孤立进行,且过于依赖感知相似性或人类判断,无法深入了解哪些物理原理或参数被违反。为此,本文提出GAUGE,一个基于真实世界的诊断基准,用于联合评估数值模拟器和生成视频世界模型如何再现或偏离真实世界物理。GAUGE涵盖22个控制任务家族,涉及刚体、柔性电缆、纺织品和体积可变形物体等,基于真实轨迹并配有校准的物理元数据和不确定性注释,涵盖碰撞、摩擦、动量转移、振荡、自接触和变形等基本物理过程。
🔬 方法详解
问题定义:本文旨在解决现有物理真实度评估方法的不足,特别是孤立评估导致的对物理原理理解的缺乏。
核心思路:GAUGE基准通过结合真实世界的轨迹和物理元数据,提供了一种系统化的评估框架,能够更全面地分析模拟器和视频模型的物理表现。
技术框架:GAUGE包含22个任务家族,涵盖刚体、柔性物体等,配合真实轨迹和校准的物理元数据,评估物理过程的准确性。
关键创新:GAUGE的创新在于其综合性评估方法,能够同时分析数值模拟器和生成模型的物理真实度,填补了现有评估方法的空白。
关键设计:在任务设计中,考虑了碰撞、摩擦等多种物理现象,并通过不确定性注释和任务特定可观察量增强评估的全面性。实验中使用了广泛的基准数据集和多种模型进行对比。
🖼️ 关键图片
📊 实验亮点
实验结果显示,评估的物理引擎在冲击接触、快速纺织物运动和体积变形方面存在显著差异,未能完全忠实于现实。视频世界模型能够生成符合预期方程形式的轨迹,但在加速度和动量转移等方面表现不佳,揭示了当前技术的局限性。
🎯 应用场景
GAUGE基准的提出为物理引擎和视频世界模型的开发提供了重要的评估工具,能够帮助研究者和工程师更好地理解和改进模拟器的物理真实度。这一方法在机器人、虚拟现实和自动驾驶等领域具有广泛的应用潜力,能够推动具身智能的发展。
📄 摘要(原文)
Physics engines facilitate large-scale training and evaluation for embodied intelligence, while generative video world models are emerging as implicit simulators of future states and interactions. However, existing evaluations of physical fidelity are often conducted in isolation and rely heavily on perceptual similarity or human judgments, providing limited insight into which physical principles or parameters are violated. We introduce GAUGE, a real-world-grounded diagnostic benchmark for jointly evaluating how numerical simulators and generative video world models reproduce or deviate from real-world physics. It comprises 22 controlled task families covering rigid bodies, flexible cables, textiles, and volumetric deformable objects. Grounded in real-world trajectories and paired with calibrated physical metadata, uncertainty annotations, and task-specific observables, these tasks cover fundamental physical processes including collision, friction, momentum transfer, oscillation, self-contact, and deformation across diverse materials and conditions. We benchmark Isaac Sim, Genesis, and Newton on 14 task families using generalized trajectory errors, and evaluate 6 image-to-video models on 5 rigid-body tasks by testing physical-law consistency and the temporal stability of inferred parameters. Our results reveal no uniformly faithful physics engine, with the largest discrepancies arising in impulsive contact, rapid textile motion, and volumetric deformation. We further find that video world models can produce trajectories with the expected equation form while recovering incorrect accelerations, momentum transfer, and oscillation timing. GAUGE lays the groundwork for developing more physically faithful simulators and world models for embodied intelligence.