PhaseWin 搜索框架实现高效的目标级解释
PhaseWin Search Framework Enable Efficient Object-Level Interpretation
把子模子集选择的归因从二次代价的贪心搜索换成近线性的分阶段粗到细窗口搜索，可直接套用到 Grounding DINO、Florence-2 这类目标级多模态模型的区域归因与显著性解释上，让高保真归因在真实推理预算内可用。
原文依据 PhaseWin achieves over 95% of greedy attribution faithfulness using only 20% of the computational budget
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论归因对于解释目标级基础模型至关重要。近期基于子模子集选择的方法已取得很高的保真度，但其效率限制阻碍了在真实场景中的实际部署。为此，我们提出 PhaseWin，一种新颖的相位窗口搜索算法，能够以接近线性的复杂度实现高保真的区域归因。PhaseWin 用分阶段的由粗到细搜索替代传统的二次代价贪心选择，结合自适应剪枝、窗口化细粒度选择与动态监督机制，在极大减少模型评估次数的同时紧密逼近贪心行为。在理论上，PhaseWin 在温和的单调子模假设下保留了接近贪心的近似保证。在实验上，PhaseWin 仅使用 20% 的计算预算即达到贪心归因保真度的 95% 以上，并在使用 Grounding DINO 与 Florence-2 的目标检测和视觉定位任务上持续优于其他归因基线。PhaseWin 为目标级多模态模型的可扩展、高保真归因确立了新的技术水平。代码见 https://github.com/Qihuai27/phasewin-search。
英文原文摘要 1236 字符
Attribution is essential for interpreting object-level foundation models. Recent methods based on submodular subset selection have achieved high faithfulness, but their efficiency limitations hinder practical deployment in real-world scenarios. To address this, we propose PhaseWin, a novel phase-window search algorithm that enables faithful region attribution with near-linear complexity. PhaseWin replaces traditional quadratic-cost greedy selection with a phased coarse-to-fine search, combining adaptive pruning, windowed fine-grained selection, and dynamic supervision mechanisms to closely approximate greedy behavior while dramatically reducing model evaluations. Theoretically, PhaseWin retains near-greedy approximation guarantees under mild monotone submodular assumptions. Empirically, PhaseWin achieves over 95% of greedy attribution faithfulness using only 20% of the computational budget, and consistently outperforms other attribution baselines across object detection and visual grounding tasks with Grounding DINO and Florence-2. PhaseWin establishes a new state of the art in scalable, high-faithfulness attribution for object-level multimodal models. Code is available at https://github.com/Qihuai27/phasewin-search.