arXiv cs.AI· Xiaolin Liu, Yilun Zhu, Xiangyu Zhao, Xuehui Wang, Yan Li, Xin Li, Haoyu Cao, Xing Sun, Shaofeng Zhang, Xu Yang, Zhihang Zhong, Xue Yang·· 2 天前AI 评分30
Moment-Video：诊断视频 MLLM 对瞬时视觉事件的时间保真度
Moment-Video: Diagnosing Temporal Fidelity of Video MLLMs on Momentary Visual Events
AI 导读
研究者推出 Moment-Video 基准，用于评估视频 MLLM 对瞬时视觉事件的时间保真度，包含 1,000 个人工核验的视频问答对，覆盖 7 个领域、25 个子类及四类任务。在 34 个专有与开源 MLLM 的评测中，表现最佳的 Seed-2.0-Pro 总体准确率仅 39.6%，多数开源模型低于 25%。诊断显示，即便采样帧覆盖全部标注事件区间，模型仍存在大量错误，且视频越长性能越差。
正文 · 原文
Authors:Xiaolin Liu, Yilun Zhu, Xiangyu Zhao, Xuehui Wang, Yan Li, Xin Li, Haoyu Cao, Xing Sun, Shaofeng Zhang, Xu Yang, Zhihang Zhong, Xue Yang
Abstract:Video multimodal large language models (MLLMs) have made rapid progress in general and long-form video understanding, yet their ability to capture and use brief, answer-critical visual evidence remains underexplored. Momentary actions or state transitions may determine an answer while occupying only a small fraction of the video. Performance on broad video understanding tasks alone does not establish whether models can reliably identify and use this temporally localized evidence. We introduce Moment-Video, a benchmark for evaluating temporal fidelity through momentary visual event understanding. Questions are constructed around visually observable events or event sequences, requiring models to notice, count, describe, or reason about transient evidence. The benchmark contains 1,000 human-verified video-QA pairs across seven domains and 25 subcategories, covering four tasks: Temporal Occurrence, Temporal Counting, Action Description, and Temporal Reasoning. We evaluate 34 proprietary and open-source MLLMs. Under the main evaluation settings, the best-performing model, Seed-2.0-Pro, achieves 39.6% overall accuracy, while most open-source models remain below 25%. Diagnostic analyses show that denser frame sampling improves some models, but substantial errors persist even when sampled frames cover all annotated necessary event intervals in an audited subset. Performance also declines on longer videos, suggesting additional temporal-localization challenges. By exposing these limitations, Moment-Video provides a diagnostic benchmark to guide the evaluation and development of video MLLMs that more reliably understand brief, answer-critical visual events.
| Comments: | 31 pages, 12 figures, 12 tables |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2606.02522 [cs.CV] |
| (or arXiv:2606.02522v2 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2606.02522 arXiv-issued DOI via DataCite |
Submission history
From: Yilun Zhu [view email]
[v1]
Mon, 1 Jun 2026 17:32:20 UTC (8,465 KB)
[v2]
Wed, 14 Oct 2026 16:51:52 UTC (8,883 KB)
来源：arXiv cs.AI · arxiv.org