WorldMemBench: A Comprehensive Benchmark for Memory of Video World Models

Computer Science > Computer Vision and Pattern Recognition arXiv:2610.19063 (cs) [Submitted on 13 Oct 2026] Title: WorldMemBench: A Comprehensive Benchmark for Memory of Video World Models Authors: Nan Zhou, Jianxiong Gao, Qiuyu Wang, Wen Wang, Jiahao Wang, Yufeng Yuan, Shuaiting Li, Hao Ouyang, Yujun Shen, Weiming Lu, Jun Xiao, Yueting Zhuang, Yongliang Shen.

计算机科学 > 计算机视觉与模式识别 arXiv:2610.19063 (cs) [提交于 2026 年 10 月 13 日] 标题：WorldMemBench：视频世界模型记忆的综合基准测试 作者：Nan Zhou, Jianxiong Gao, Qiuyu Wang, Wen Wang, Jiahao Wang, Yufeng Yuan, Shuaiting Li, Hao Ouyang, Yujun Shen, Weiming Lu, Jun Xiao, Yueting Zhuang, Yongliang Shen。

Abstract: As video models enable longer and more realistic interactive experiences, maintaining a world consistent with its history becomes essential. This requires both preserving previously observed content and ensuring coherent world evolution over time. World memory is a critical capability to evaluate alongside visual fidelity and video duration.

摘要：随着视频模型实现更长、更逼真的交互体验，保持一个与历史一致的世界变得至关重要。这既需要保留先前观察到的内容，也需要确保世界随时间推移的连贯演变。世界记忆是与视觉保真度和视频时长并列的一项关键评估能力。

We introduce WorldMemBench, a benchmark for systematically evaluating world memory in video world models. Across 260 test cases, it assesses static memory through the retention of human identity and appearance, object appearance and geometry, and environmental content and geometry, and dynamic memory through event progression and state persistence.

我们引入了 WorldMemBench，这是一个用于系统评估视频世界模型中世界记忆的基准测试。通过 260 个测试案例，它评估了静态记忆（通过对人物身份和外观、物体外观和几何形状以及环境内容和几何形状的保留）和动态记忆（通过事件进展和状态持久性）。

Using controlled camera departures and multi-revisit trajectories, the benchmark evaluates static content against fixed initial references and dynamic events against expected state transitions. It combines appearance and geometry metrics with event-specific assessments, and reports video quality and camera-following accuracy as complementary diagnostics.

该基准测试利用受控的摄像机偏离和多次重访轨迹，根据固定的初始参考评估静态内容，并根据预期的状态转换评估动态事件。它将外观和几何指标与特定事件的评估相结合，并报告视频质量和摄像机跟随准确性作为补充诊断。

Evaluation of nine representative models shows that strong static retention does not imply reliable dynamic state evolution, while repeated revisits reveal memory degradation beyond the first return. Human studies support the alignment between our automatic evaluations and human judgments. WorldMemBench provides a systematic testbed for diagnosing memory failures and guiding the development of persistent, temporally coherent video worlds.

对九个代表性模型的评估表明，强大的静态保留并不意味着可靠的动态状态演变，而重复重访揭示了在首次返回之后的记忆衰退。人类研究支持了我们的自动评估与人类判断之间的一致性。WorldMemBench 为诊断记忆故障和指导持久、时间连贯的视频世界的开发提供了一个系统的测试平台。