提出在线原型检测方法，提升医疗视觉语言模型在部署时的异常输入识别能力。
PROTON: Prototype-Based Test-Time Online OOD Detection for Medical VLMs
- 用测试时高置信度预测构建在线原型库，结合嵌入空间距离与匹配得分
- 在多种分布外数据上提升23.9%以上AUROC，无需训练或提示工程
- 适合医疗图像分类部署场景，尤其适用于无标注数据的零样本应用
医疗视觉语言模型（VLMs）可实现零样本临床图像分类，但在部署阶段可靠检测分布外（OOD）输入仍是难题。现有静态评分方法对不同偏移类型表现不一：在FLAIR数据集上，最大概念匹配（MCM）对远端分布外（far-OOD）达76.4% AUROC，但对协变量偏移（如超广角眼底图像）仅42.4%，近乎随机。根源在于协变量偏移输入在softmax空间与分布内样本无法区分，却在VLM嵌入空间占据独特区域。为此，我们提出PROTON（基于原型的测试时在线OOD检测），一种轻量级后处理模块：通过高置信度测试预测动态维护原型库，并利用流级方差统计自适应融合原型距离与MCM得分，无需模型修改、训练数据或提示工程。在眼科基准FLAIR + FIVES上，PROTON使MCM在协变量偏移下提升+23.9 AUROC，语义偏移+8.8，远端分布外+8.1，是首个无需分层提示或标注数据即可全面提升的零样本方法。代码见https://github.com/GenMI-Lab/PROTON，项目页为https://genmi-lab.github.io/PROTON。
原文摘要 · Abstract (English)
Medical vision-language models (VLMs) enable zero-shot clinical image classification, yet reliably detecting out-of-distribution (OOD) inputs at deployment remains an open problem. No static scoring method works across all shift types: Maximum Concept Matching (MCM) on FLAIR achieves 76.4% AUROC for far-OOD but only 42.4% for covariate shifts such as ultra-wide-field fundus images, effectively random. We trace this to a structural mismatch: covariate-shifted inputs are indistinguishable from in-distribution samples in softmax space, yet occupy distinct regions in the VLM embedding space. To exploit this untapped signal, we propose PROTON (PROtotype-based Test-time ONline OOD detection), a lightweight post-hoc module that maintains an online prototype bank from high-confidence test predictions and adaptively fuses prototype distance with MCM scoring via stream-level variance statistics, requiring no model modification, training data, or prompt engineering. On the ophthalmology benchmark FLAIR + FIVES, PROTON improves MCM by +23.9 AUROC on covariate shift, +8.8 on semantic shift, and +8.1 on far-OOD, making it the only zero-shot method to improve all three without hierarchical prompts or labeled data. Code is available at https://github.com/GenMI-Lab/PROTON, and the project page is available at https://genmi-lab.github.io/PROTON.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。