Ref4D-VideoBench：面向文本到视频生成模型的四维参考式评测
Ref4D-VideoBench: Four-Dimensional Reference-Based Evaluation of Text-to-Video Generative Models
为文生视频模型提供以参考视频为时空证据的细粒度评测协议，可直接用于诊断目标行为偏差、时序不一致与常识违背等实例级问题，替代纯文本提示下的无参考打分。
原文依据 we propose Ref4D-VideoBench, a reference-based, fine-grained, multi-dimensional benchmark for generated video evaluation
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论现有的大多数生成视频评测采用无参考范式。尽管近期的基准覆盖了多个维度，并与人类偏好呈现中等程度的相关性，但仅依赖文本提示削弱了现实世界的约束，难以对目标行为偏差、时序不一致和常识违背等实例级问题给出可问责且可解释的判断。在受控生成等具有明确预期的场景中，参考视频天然提供了丰富且无歧义的时空证据，从而能够实现更严格、更可信的评估。受此启发，我们提出 Ref4D-VideoBench，一个基于参考的、细粒度的、多维度的生成视频评测基准。Ref4D-VideoBench 包含 600 个高质量参考视频以及严格受证据约束的提示，并引入一套 12 项指标的结构化评测套件，覆盖四个关键维度：基础语义对齐、运动一致性、事件时序一致性和世界知识一致性。在八个文本到视频模型上的实验表明，与代表性的无参考框架相比，我们的方法在人类判断上取得了更强的一致性。我们的代码已发布于 https://github.com/TAILab-W/Ref4D-VideoBench。
英文原文摘要 1293 字符
Most existing evaluations of generated videos adopt a no-reference paradigm. Although recent benchmarks cover multiple dimensions and show moderate correlation with human preferences, relying solely on textual prompts weakens real-world constraints and makes it difficult to produce accountable and interpretable judgments on instance-level issues such as target behavior deviation, temporal inconsistency, and commonsense violations. In scenarios with explicit expectations, such as controlled generation, reference videos naturally provide rich, unambiguous spatio-temporal evidence, enabling stricter and more trustworthy assessment. Motivated by this, we propose Ref4D-VideoBench, a reference-based, fine-grained, multi-dimensional benchmark for generated video evaluation. Ref4D-VideoBench contains 600 high-quality reference videos with tightly evidence-bounded prompts, and introduces a 12-metric structured evaluation suite along four key dimensions: basic semantic alignment, motion consistency, event temporal consistency, and world knowledge consistency. Experiments on eight text-to-video models show that our method achieves stronger agreement with human judgments than representative no-reference frameworks. Our code is available at https://github.com/TAILab-W/Ref4D-VideoBench.