发表机构
Southeast University（东南大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本研究通过2x2因子实验，发现分类法感知监督目标比几何选择更显著提升层级检索性能，且真实语义对齐至关重要。
AI 中文摘要
基础视觉模型提供了强大的通用表示，但高类别级检索准确率并不必然意味着嵌入尊重目标语义分类法。我们研究冻结的DINOv2特征上的严格显式分类法图像检索，并提出问题：当层级检索改进时，有多少变化与分类法感知监督的组织相关，又有多少与欧几里得-双曲几何选择相关？我们使用严格的跨类标准评估更高层级，排除更细粒度的匹配，并比较使用分类法距离回归或分类法感知的监督对比目标训练的欧几里得和双曲投影。一个计算匹配的2x2几何x损失因子设计使用相同的768-256-32投影器容量、优化调度、批次顺序和固定的100轮预算；损失轴表示回归到分类法-SupCon目标族对比。在CUB上，目标族对比在平均层级mAP（严格中/高层平均，排除类和叶）中，欧几里得空间为+0.0487，双曲空间为+0.0414，而几何对比为+0.0102和+0.0030。在NABird父类不相交检索中，相应的目标族对比为+0.0467和+0.0440，而几何对比为+0.0017和-0.0009。语义对齐控制显示，真实分类法显著优于保持结构的打乱层级，而NABird曲率/半径控制不支持更强的负曲率作为观察到的层级增益的解释。在两个分类法中，回归到分类法-SupCon对比总体上大于评估的几何对比；语义对齐也单独重要，而几何仍然依赖于层级。
英文摘要
Foundation vision models provide strong generic representations, yet high class-level retrieval accuracy does not necessarily imply that an embedding respects a target semantic taxonomy. We study strict explicit-taxonomy image retrieval on frozen DINOv2 features and ask: when hierarchical retrieval improves, how much of the change is associated with the organization of taxonomy-aware supervision, and how much with the Euclidean-hyperbolic geometry choice? We evaluate higher levels with strict cross-class criteria that exclude finer-grained matches, and compare Euclidean and hyperbolic projections trained with taxonomy-distance regression or a taxonomy-aware supervised contrastive objective. A compute-matched 2 x 2 Geometry x Loss factorial uses the same 768-256-32 projector capacity, optimization schedule, batch order, and fixed 100-epoch budget; the Loss axis denotes the Regression-to-Taxonomy-SupCon objective-family contrast. On CUB, the objective-family contrasts in mean hierarchy mAP (strict middle/high average, excluding Class/Leaf) are +0.0487 in Euclidean space and +0.0414 in hyperbolic space, compared with geometry contrasts of +0.0102 and +0.0030. On NABirds Parent-disjoint retrieval, the corresponding objective-family contrasts are +0.0467 and +0.0440, whereas geometry contrasts are +0.0017 and -0.0009. A semantic-alignment control shows that the true taxonomy substantially outperforms a structure-preserving shuffled hierarchy, while a NABirds curvature/radius control does not support stronger negative curvature as the explanation for the observed hierarchy gains. Across the two taxonomies, the Regression-to-Taxonomy-SupCon contrasts are larger in aggregate than the evaluated geometry contrasts; semantic alignment also matters separately, while geometry remains hierarchy-dependent.
Comments17 pages total: 9-page main paper (including references) + 8-page supplementary material; 3 figures and 2 main-paper tables
