只标每批一个样本，就能持续提升长期测试时自适应效果。
Effortless Active Labeling for Long-Term Test-Time Adaptation
- 每批选最接近源域与目标域边界的样本标注，提升单步学习效率。
- 用特征扰动快速识别关键样本，降低标注成本。
- 通过动态权重平衡标注与未标注样本影响，适合长期部署场景。
长期测试时自适应（TTA）因误差累积而困难重重。现有方法虽通过主动标注每批少量样本缓解问题，但标注负担随批次增加迅速上升。本文提出一种无需额外成本的主动标注策略：每批仅标注一个样本。首先，基于单步优化视角，选择最可能在一次迭代中有效提升模型性能的样本——即处于源域与目标域分布边界附近的样本；其次，提出利用特征扰动高效识别此类样本。进一步发现，标注样本与未标注样本产生的梯度幅度差异显著，因此引入两个动态权重来平衡二者对模型优化的影响。在ImageNet-C、-R、-K、-A及PACS等多个主流数据集上的大量实验表明，本方法在显著降低标注成本的同时，持续优于当前最优方法。
原文摘要 · Abstract (English)
Long-term test-time adaptation (TTA) is a challenging task due to error accumulation. Recent approaches tackle this issue by actively labeling a small proportion of samples in each batch, yet the annotation burden quickly grows as the batch number increases. In this paper, we investigate how to achieve effortless active labeling so that a maximum of one sample is selected for annotation in each batch. First, we annotate the most valuable sample in each batch based on the single-step optimization perspective in the TTA context. In this scenario, the samples that border between the source- and target-domain data distributions are considered the most feasible for the model to learn in one iteration. Then, we introduce an efficient strategy to identify these samples using feature perturbation. Second, we discover that the gradient magnitudes produced by the annotated and unannotated samples have significant variations. Therefore, we propose balancing their impact on model optimization using two dynamic weights. Extensive experiments on the popular ImageNet-C, -R, -K, -A and PACS databases demonstrate that our approach consistently outperforms state-of-the-art methods with significantly lower annotation costs.
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。