何时解配对：医学视觉上下文学习中配对依赖性的调控
When to Unpair: Regulating Pairing Dependence in Medical Visual In-Context Learning
Cheng Wan · Chenjun Li · Qingyu Zhao
中文摘要
视觉上下文学习（visual in-context learning, ICL）天然契合标注稀缺的医学影像场景：它利用若干带标签的支持图像—标签对展示输入到输出的映射，标签整体指示所请求的任务。为诊断模型对单个配对的依赖程度，本文设计了一种测试时错配（test-time derangement）方法：在保持查询图像、支持图像集合及标签多重集不变的情况下，将每个支持标签重新分配给另一张支持图像。由此产生的配对差距（pairing gap，即打乱配对与匹配配对的性能之差）表明，四款已发布的模型在不同程度上均存在配对依赖。进一步分析一个此类配对训练的模型发现，即便使用真实未篡改的支持集，仍存在支持相关的伪影区域与病灶尺寸偏差，且对支持标签错位敏感。为调控此种依赖，本文提出晚期解配课程（late unpairing curriculum, LUC）：训练初期使用匹配配对，随后施加随机解配，将同一回合内每个支持标签替换为另一支持样本的标签。LUC 在两类骨干网络上几乎完全弥合配对差距，同时在所有评估任务类型上保持或提升匹配支持下的性能，优势还延伸至未见过任务与跨数据集回合，并减轻上述失效模式。在 BraTS 全肿瘤分割任务上，匹配支持 DSC 由 0.733 提升至 0.857，差距由 −0.184 缩至 −0.008。对一款已发布模型进行短时随机解配微调也可缩小差距。若将同等数量的解配训练轮次前置，即采用反向课程，则仍留下较大差距。这表明配对依赖由训练顺序塑造，而非仅由解配训练量决定。
关键要点
- 01提出测试时错配协议与配对差距指标，诊断出四款视觉上下文学习模型均存在不同程度的配对依赖
- 02分析发现匹配训练模型在真实支持集下出现伪影区域、病灶尺寸偏差及标签错位敏感等失效模式
- 03提出晚期解配课程 LUC，训练先匹配后随机解配，可在 BraTS 全肿瘤分割上将匹配支持 DSC 从 0.733 提升至 0.857，差距从 −0.184 缩至 −0.008
- 04LUC 优势可迁移至未见过任务与跨数据集回合，并缓解上述配对依赖型失效
- 05反向课程（同等解配量前置）留下大差距，说明配对依赖由训练顺序塑造而非解配训练量
解读
尚无解读。
原始英文摘要
arXiv:2610.10335v1 Announce Type: new Abstract: Visual in-context learning (ICL), well suited to label-scarce medical imaging, uses support image-label pairs to demonstrate input-output mappings, while the labels collectively indicate the requested task. We diagnose dependence on individual pairings with a test-time derangement that reassigns every support label to another support image while preserving the query, support images, and label multiset. The resulting pairing gap, defined as shuffled-minus-matched performance, shows that all four released models depend on the pairing, to widely varying degrees. Further analysis of a paired-trained model reveals support-associated spurious regions and lesion-size biases even with real, unaltered supports, alongside sensitivity to mis-registered support labels. To regulate this dependence, we introduce a late unpairing curriculum (LUC), which starts with matched training and then applies random unpairing, replacing each support label with that of another support in the same episode. LUC nearly closes the pairing gap on two backbones while maintaining or improving matched-support performance across all evaluated task types, with gains extending to held-out tasks and cross-dataset episodes. It also mitigates these failure modes. On BraTS whole-tumor segmentation, matched-support DSC rises from 0.733 to 0.857 while the gap shrinks from -0.184 to -0.008. In a released model, brief fine-tuning with random unpairing reduces the gap. A reversed curriculum that places the same number of unpairing epochs at the start of training leaves a large gap. This shows that pairing dependence is shaped by the order of training and not only by the amount of unpaired training.