引力驱动的语义对齐用于文本视频检索
Gravitation-Driven Semantic Alignment for Text Video Retrieval
把跨模态语义分布的对齐拆成「语义引力」（由不确定性导出的语义质量与语义距离）和「几何重叠」两个正交因子，并给出 O(D) 的闭式替代来避免难解的联合积分——做视频-文本检索或任何需要处理多对多语义歧义、区分难负样本的跨模态对齐任务时，这套概率分布匹配打分可以直接替换确定性嵌入的余弦相似度。
原文依据 decomposes the alignment of cross-modal semantic distributions into two orthogonal factors inspired by the Gravitational Force
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论「多对多」固有的语义歧义——即一个视频匹配多个文本、反之亦然——加剧了文本视频检索的难度。主流的确定性嵌入难以捕捉均值语义，而现有的概率方法由于施加了僵化的不确定性先验，或忽略了相似度与不确定性之间的交互，无法区分难负样本。为此，我们提出了一个受物理学启发的新框架（GraviAlign），它受万有引力启发，将跨模态语义分布的对齐分解为两个正交因子：（1）语义引力，通过由不确定性导出的「语义质量」与「语义距离」来度量分布中心之间的引力式对齐；（2）几何重叠，量化分布之间的交集。每个因子都具有独立的否决权，以拒绝那些存在错位或重叠不佳的匹配。此外，GraviAlign 为难以求解的联合积分提供了一种高效的（O(D)）、有理论依据的替代方案。在 DiDeMo、MSR-VTT 和 ActivityNet 上的大量实验证明了我们的有效性与优越性，扎实的消融研究也证实了两个新组件不可或缺。
英文原文摘要 1241 字符
The inherent semantic ambiguity of "many-to-many", where one video matches multiple texts and vice versa, aggravates the difficulty in text-video retrieval. The dominant deterministic embeddings only struggle to capture the mean semantics, while existing probabilistic methods fail to distinguish hard negatives for their imposing rigid uncertainty priors or ignoring the interaction between similarity and uncertainty. To this end, we propose a novel physics-inspired framework (GraviAlign) that decomposes the alignment of cross-modal semantic distributions into two orthogonal factors inspired by the Gravitational Force: (1) Semantic Attraction measuring gravitational alignment between distribution centers via uncertainty-derived "semantic mass" and "semantic distance"; (2) Geometric Overlap quantifying distribution intersection. Each factor has independent veto power to reject those matches with misalignment or poor overlap. Additionally, GraviAlign offers an efficient (O(D)), theoretically grounded alternative to intractable joint integrals. Extensive experiments on DiDeMo, MSR-VTT, and ActivityNet demonstrate our effectiveness and superiority, and solid ablation studies confirm the indispensability of two novel components.