解读同步间隙:扩散变换器内部的隐藏机制
Interpreting the Synchronization Gap: The Hidden Mechanism Inside Diffusion Transformers
AI总结:
研究揭示扩散变换器中同步间隙的机理,发现其为内在架构属性,且在强耦合下消失,仅在最终层出现,低频结构先于高频细节承诺。
摘要:
最近的扩散过程理论模型,概念化为耦合奥恩斯坦-乌尔岑贝克系统,预测了交互时间尺度的层次结构,从而存在正向过程不同阶段的模式同步间隙。然而,由于这些预测依赖连续时间和可分析的分数函数,仍不清楚这种现象如何在实际部署的深度离散架构中表现。在本工作中,我们研究了同步间隙如何在预训练的扩散变换器(DiTs)中机制性实现。我们通过将两个生成轨迹嵌入到联合令牌序列中,并通过对称的交叉注意门调节变量耦合强度g,构建了显式的复制耦合架构。通过线性化分析注意差异,我们展示了复制交互的机理分解。我们通过跟踪预训练DiT-XL/2模型的承诺和每层内部模式能量,经验验证了我们的理论框架。我们的结果揭示:(1)同步间隙是DiTs的内在架构属性,即使在外部耦合关闭时仍存在;(2)根据我们的空间路由界限预测,强耦合下间隙完全消失;(3)间隙严格深度局部化,仅在Transformer的最终层中急剧出现;(4)全局低频结构始终在局部高频细节之前承诺。最终,我们的发现为扩散变换器如何解决生成模糊性提供了机理解释,将物种化转变隔离到网络的终端层。
英文摘要:
Recent theoretical models of diffusion processes, conceptualized as coupled Ornstein-Uhlenbeck systems, predict a hierarchy of interaction timescales, and consequently, the existence of a synchronization gap between modes that commit at different stages of the reverse process. However, because these predictions rely on continuous time and analytically tractable score functions, it remains unclear how this phenomenology manifests in the deep, discrete architectures deployed in practice. In this work, we investigate how the synchronization gap is mechanistically realized within pretrained Diffusion Transformers (DiTs). We construct an explicit architectural realization of replica coupling by embedding two generative trajectories into a joint token sequence, modulated by a symmetric cross attention gate with variable coupling strength g. Through a linearized analysis of the attention difference, we show that the replica interaction decomposes mechanistically. We empirically validate our theoretical framework on a pretrained DiT-XL/2 model by tracking commitment and per layer internal mode energies. Our results reveal that: (1) the synchronization gap is an intrinsic architectural property of DiTs that persists even when external coupling is turned off; (2) as predicted by our spatial routing bounds, the gap completely collapses under strong coupling; (3) the gap is strictly depth localized, emerging sharply only within the final layers of the Transformer; and (4) global, low frequency structures consistently commit before local, high frequency details. Ultimately, our findings provide a mechanistic interpretation of how Diffusion Transformers resolve generative ambiguity, isolating speciation transitions to the terminal layers of the network.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。