Aperture：压缩令牌的合并一致性旋转状态
Aperture: Merge-Consistent Rotary States for Compressed Tokens
- The Chinese University of Hong Kong, Shenzhen（香港中文大学（深圳））
- Shenzhen Loop Area Institute（深圳河套学院）
机构由 AI 辅助整理，请以论文原文为准。
AI总结：
本研究提出Aperture方法，通过存储令牌加权支持的傅里叶矩在旋转频率下实现压缩令牌的合并一致性，证明最小实维度并验证其在视频问答中的效果。
AI中文摘要：
令牌压缩将多个位置的内容合并，但旋转位置嵌入通常为合并后的令牌分配一个坐标。我们探究哪些位置信息必须在后续合并中保留。Aperture在模型的旋转频率下存储令牌加权支持的傅里叶矩。我们证明，在足以满足所选期望旋转交互的连续状态中，这些矩具有最小的实维度。表示的质量使更新具有可加性；注意力归一化仍然是独立的读出选择。均匀间隔给出中心旋转乘以sinc增益。我们刻画了中心决定间隔宽度的情况，并构造了中心不决定间隔宽度的匹配示例。数值检查验证了加权支持实现。在训练的时间读取器中，压缩迁移随增益校准和特征放置而变化。在预先指定的原生视频问答比较中，存储支持达到65.63%的准确率，而部署的合并规则为67.12%。这些结果将压缩下的精确位置保留与下游收益区分开来。
英文摘要：
Token compression combines content from several positions, yet rotary position embeddings usually assign the merged token one coordinate. We ask what positional information must survive later merges. Aperture stores Fourier moments of the token's weighted support at the model's rotary frequencies. We prove that these moments have minimal real dimension among continuous states sufficient for the selected expected rotary interactions. Represented mass makes updates additive; attention normalisation remains a separate readout choice. Uniform intervals give a centre rotation times a sinc gain. We characterise when centres determine interval widths and construct matched examples where they do not. Numerical checks verify the weighted-support implementation. In trained temporal readers, compression transfer varies with gain calibration and feature placement. In a prespecified native video question-answering comparison, stored support reaches $65.63\%$ accuracy versus $67.12\%$ for the deployed merging rule. These results separate exact positional preservation under compression from downstream benefit.
