Video foundation models generate visually realistic and temporally coherent content, but their reliability as world simulators depends on whether they capture physical, logical, and spatial constraints. Existing metrics such as Frechet Video Distance (FVD) emphasize perceptual quality and overlook reasoning failures, including violations of causality, physics, and global consistency. We introduce MMGR (Multi-Modal Generative Reasoning Evaluation and Benchmark), a principled evaluation framework based on five reasoning abilities: Physical, Logical, 3D Spatial, 2D Spatial, and Temporal. MMGR evaluates generative reasoning across three domains: Abstract Reasoning (ARC-AGI, Sudoku), Embodied Navigation (real-world 3D navigation and localization), and Physical Commonsense (sports and compositional interactions). MMGR applies fine-grained metrics that require holistic correctness across both video and image generation. We benchmark leading video models (Veo-3, Sora-2, Wan-2.2) and image models (Nano-banana, Nano-banana Pro, GPT-4o-image, Qwen-image), revealing strong performance gaps across domains. Models show moderate success on Physical Commonsense tasks but perform poorly on Abstract Reasoning (below 10 percent accuracy on ARC-AGI) and struggle with long-horizon spatial planning in embodied settings. Our analysis highlights key limitations in current models, including overreliance on perceptual data, weak global state consistency, and objectives that reward visual plausibility over causal correctness. MMGR offers a unified diagnostic benchmark and a path toward reasoning-aware generative world models.

翻译：视频基础模型能够生成视觉逼真且时序连贯的内容，但其作为世界模拟器的可靠性取决于其是否捕捉了物理、逻辑和空间约束。现有指标如弗雷歇视频距离（FVD）侧重于感知质量，而忽视了推理失败，包括违反因果性、物理规律和全局一致性的问题。本文提出MMGR（多模态生成式推理评估与基准），这是一个基于五种推理能力（物理、逻辑、三维空间、二维空间和时序）的原则性评估框架。MMGR在三个领域评估生成式推理：抽象推理（ARC-AGI、数独）、具身导航（真实世界三维导航与定位）和物理常识（运动与组合交互）。MMGR采用细粒度指标，要求视频和图像生成在整体上均正确。我们对主流视频模型（Veo-3、Sora-2、Wan-2.2）和图像模型（Nano-banana、Nano-banana Pro、GPT-4o-image、Qwen-image）进行了基准测试，揭示了各领域间显著的性能差距。模型在物理常识任务上表现尚可，但在抽象推理上表现较差（ARC-AGI准确率低于10%），并在具身环境中的长程空间规划方面存在困难。我们的分析突显了当前模型的关键局限，包括过度依赖感知数据、全局状态一致性弱，以及目标函数更注重视觉合理性而非因果正确性。MMGR提供了一个统一的诊断基准，并为实现具备推理意识的生成式世界模型指明了路径。