RDumb：一种质疑我们在持续测试时适应方面进展的简单方法
RDumb: A simple approach that questions our progress in continual test-time adaptation
AI总结：
本文提出 CCC 基准评估长期持续 TTA，并设计周期性重置模型的简单基线 RDumb，发现多数先进方法最终崩溃，而 RDumb 达到或超过最佳性能。
摘要：
测试时适应（TTA）允许在部署时将预训练模型更新到不断变化的数据分布。早期工作针对单个固定分布偏移测试这些算法，而近期工作提出并应用了在长时间尺度上进行持续适应的方法。为考察该领域所报告的进展，我们提出 Continually Changing Corruptions（CCC）基准，用于衡量 TTA 技术的渐近性能。我们发现，最终除一种方法外，所有最先进方法都会崩溃，并且表现比不适应模型更差，其中包括专门为抵抗性能崩溃而提出的模型。此外，我们引入一个简单基线“RDumb”，它定期将模型重置为预训练状态。在所有考察的基准中，RDumb 的表现优于或持平于此前提出的最先进方法。我们的结果表明，以往的 TTA 方法既不能有效规范适应以避免崩溃，也无法胜过一种简单的重置策略。
英文摘要：
Test-Time Adaptation (TTA) allows to update pre-trained models to changing data distributions at deployment time. While early work tested these algorithms for individual fixed distribution shifts, recent work proposed and applied methods for continual adaptation over long timescales. To examine the reported progress in the field, we propose the Continually Changing Corruptions (CCC) benchmark to measure asymptotic performance of TTA techniques. We find that eventually all but one state-of-the-art methods collapse and perform worse than a non-adapting model, including models specifically proposed to be robust to performance collapse. In addition, we introduce a simple baseline, "RDumb", that periodically resets the model to its pretrained state. RDumb performs better or on par with the previously proposed state-of-the-art in all considered benchmarks. Our results show that previous TTA approaches are neither effective at regularizing adaptation to avoid collapse nor able to outperform a simplistic resetting strategy.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
