解决一句话对应多个视频片段的定位难题，提升多段落视频检索准确率。
Towards One-to-Many Temporal Grounding
- 设计新奖励函数，利用链式思考引导模型精准完整定位多个片段。
- 在56000样本数据集上实现43.65%的先进有效时间F1得分。
- 提出首个专门评估多段定位的基准与指标，适合视频理解研究者使用。
时序定位（TG）旨在将文本查询与对应的视频片段进行匹配。以往研究主要集中在单片段检索，但现实场景中常需为同一查询定位多个不连续的视频片段，我们称此为「一至多时序定位」（OMTG）。现有最先进的多模态大模型虽在一对一任务中表现优异，但在处理多段定位时因缺乏事件数量感知能力，常给出接近零的得分。为此，本文提出系统性解决方案：首先建立首个全面的OMTG基准，引入计数准确率（C-Acc）和有效时间F1（EtF1）作为评估指标；其次通过复杂构建流程创建高质量的56,000样本数据集；最后设计专用于OMTG的时序与字幕奖励函数，其中字幕奖励利用密集视频字幕的链式思考推理，显式引导策略优化以兼顾精确性与完整性。大量实验表明，所提模型在OMTG基准上取得43.65%的新高有效时间F1，优于Gemini 2.5 Pro和Seed-1.8分别达15.85%和15.61%。
原文摘要 · Abstract (English)
Temporal Grounding (TG) aims to localize video segments corresponding to a textual query. Prior research predominantly focuses on single-segment retrieval. Real-world scenarios, however, often require localizing multiple disjoint segments for a single query -- a setting we term One-to-Many Temporal Grounding (OMTG). Previous state-of-the-art MLLMs, optimized for one-to-one settings, struggle in this context, often yielding near-zero scores due to a lack of event cardinality perception. To bridge this gap, we present a systematic solution with three key contributions. First, we establish the first comprehensive OMTG benchmark, introducing Count Accuracy (C-Acc) and Effective Temporal F1 (EtF1) as evaluation metrics. Second, we curate a high-quality OMTG dataset comprising 56k samples through a sophisticated construction pipeline. Third, we develop novel temporal and caption reward functions specifically designed for OMTG. In particular, the caption reward leverages Chain-of-Thought reasoning over dense video captions to explicitly guide policy optimization toward both preciseness and completeness. Extensive experiments show our model achieves a new state-of-the-art EtF1 of 43.65\% on OMTG Bench, outperforming Gemini 2.5 Pro and Seed-1.8 by 15.85\% and 15.61\%, respectively. Project Page: https://insomniaaac.github.io/OMTG/
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。