面向深度伪造定位的不一致性感知多模态薛定谔桥
Inconsistency-aware Multimodal Schrodinger Bridge for Deepfake Localization
把薛定谔桥当作无需显式加噪去噪的一致性度量与跨模态信息选择器，用一致性统计非对称分配桥步数，可迁移到任何需要抑制单侧噪声的多模态时序定位任务。
原文依据 a lightweight coarse bridge first proposes candidate intervals and estimates cross-modal consistency
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论音视频深度伪造定位要求输出区间级结果，以作为时序证据。尽管近期取得进展，在单侧或异步伪造情形下，对称融合会传播跨模态噪声，损害高精度定位。我们提出 IaMSB，一种不一致性感知的多模态薛定谔桥（SB），联合估计跨模态一致性并执行区间级定位。与扩散模型不同，SB 最小化路径分布差异，无需显式加噪或去噪即可给出一致性分数。借助薛定谔桥，IaMSB 将一致性估计、跨模态信息选择与桥步调度统一在一个框架中。具体而言，一个轻量级粗桥首先提出候选区间并估计跨模态一致性；这些统计量用于选择跨模态见证信号，并以非对称方式在各模态间分配桥步。随后，一个精化桥执行步数调优的融合，输出精化且时间对齐的区间。IaMSB 能够预判单侧与异步伪造，并利用带步数分配的瓶颈式跨模态交互抑制噪声传递、避免不必要的迭代。在多个基准上，IaMSB 稳定了严格 IoU 下的边界精度，将 AP@0.95 提升 3~10%，并带来更好的高精度定位，尤其在单侧伪造情形下。
英文原文摘要 1410 字符
Audio-visual deepfake localization demands interval-level outputs that serve as temporal evidence. Despite recent progress, symmetric fusion under single-sided or asynchronous forgeries propagates cross-modal noise, degrading high-precision localization. We present IaMSB, an inconsistency-aware multimodal Schrodinger Bridge (SB) that jointly estimates cross-modal consistency and performs interval-level localization. Unlike diffusion models, SB minimizes path-distribution discrepancy and yields consistency scores without explicit noise injection or denoising. With the Schrodinger Bridge (SB), IaMSB unifies consistency estimation, cross-modal information selection, and bridge-step scheduling in one framework. Specifically, a lightweight coarse bridge first proposes candidate intervals and estimates cross-modal consistency; these statistics select cross-modal witness signals and allocate bridge steps asymmetrically across modalities. A refinement bridge then performs step-tuned fusion and outputs refined, time-aligned intervals. IaMSB anticipates single-sided and asynchronous forgeries and, using bottlenecked cross-modal interaction with step allocation, suppresses noise transfer, avoids unnecessary iterations. Across benchmarks, IaMSB stabilizes strict-IoU boundary precision, raising AP@0.95 by 3~10%, and yields improved high-precision localization, particularly for single-sided forgeries.