4DWorldBench：面向3D/4D世界生成模型的综合评估框架
4DWorldBench: A Comprehensive Evaluation Framework for 3D/4D World Generation Models
为3D/4D世界生成模型提供统一评测框架，其将多模态条件映射到文本空间并用LLM/MLLM作为评判器的自适应评估策略，可复用于其他生成任务的跨模态对齐与物理合理性度量。
原文依据 we introduce the 4DWorldBench, which measures models across four key dimensions: Perceptual Quality, Condition-4D Alignment, Physical Realism, and 4D Consistency.
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论世界生成模型正逐渐成为下一代多模态智能系统的基石。与传统的2D视觉生成不同，世界模型旨在从图像、视频或文本中构建逼真、动态且物理一致的3D/4D世界。这些模型不仅需要生成高保真的视觉内容，还需在空间、时间、物理规律和指令控制上保持连贯性，从而赋能虚拟现实、自动驾驶、具身智能和内容创作等应用。然而，现有基准各自侧重不同的评估维度，缺乏对世界真实感能力的统一评估。为系统评估世界模型，我们提出4DWorldBench，从四个关键维度衡量模型：感知质量、条件-4D对齐、物理真实感和4D一致性。该基准涵盖图像到3D/4D、视频到4D、文本到3D/4D等任务。此外，我们创新性地引入跨多模态的自适应条件，不仅整合而且扩展了传统评估范式。为适应不同模态条件的输入，我们在评估过程中将所有模态条件映射到统一的文本空间，并进一步集成LLM-as-judge、MLLM-as-judge以及传统的基于网络的方法。这种统一且自适应的设计能够对对齐、物理真实感和跨模态连贯性进行更全面、一致的评估。初步的人类研究进一步表明，我们的自适应工具选择与主观人类判断具有更高的一致性。我们希望该基准能为客观比较和改进奠定基础，加速从“视觉生成”向“世界生成”的转变。项目地址：https://yeppp27.github.io/4DWorldBench.github.io/。
英文原文摘要 1864 字符
World Generation Models are emerging as a cornerstone of next-generation multimodal intelligence systems. Unlike traditional 2D visual generation, World Models aim to construct realistic, dynamic, and physically consistent 3D/4D worlds from images, videos, or text. These models not only need to produce high-fidelity visual content but also maintain coherence across space, time, physics, and instruction control, enabling applications in virtual reality, autonomous driving, Embodied Intelligence, and content creation.However, prior benchmarks, however, each emphasize different evaluation dimensions and lack a unified assessment of world-realism capability.To systematically evaluate World Models, we introduce the 4DWorldBench, which measures models across four key dimensions: Perceptual Quality, Condition-4D Alignment, Physical Realism, and 4D Consistency. The benchmark covers tasks such as Image-to-3D/4D, Video-to-4D, Text-to-3D/4D. Beyond these, we innovatively introduce adaptive conditioning across multiple modalities, which not only integrates but also extends traditional evaluation paradigms. To accommodate different modality-conditioned inputs, we map all modality conditions into a unified textual space during evaluation, and further integrate LLM-as-judge, MLLM-as-judge, and traditional network-based methods. This unified and adaptive design enables more comprehensive and consistent evaluation of alignment, physical Realism, and cross-modal coherence.Preliminary human studies further demonstrate that our adaptive tool selection achieves closer agreement with subjective human judgments.We hope this benchmark will serve as a foundation for objective comparisons and improvements, accelerating the transition from "visual generation" to "world generation." Our project can be found at https://yeppp27.github.io/4DWorldBench.github.io/.