English(EN) Low-Frequency Shortcuts in Texture-Driven Visual Learning

AI模型在纹理驱动的视觉学习中表现出低频捷径

作者 PulseAugur 编辑部 · [2 个来源] · 2026-06-02 11:11

研究人员发现，在纹理驱动的数据集上训练的视觉AI模型中存在一种新的捷径学习类型。这些模型倾向于依赖低频分量进行分类，即使关键信息在于更精细的细节中。通过修剪这些低频分量，模型在分布内数据的准确性提高了高达8%，并且对某些损坏的鲁棒性显著增强。 AI

影响识别出AI模型中的一种新漏洞，可能影响其在现实世界中纹理密集型应用中的可靠性。

排序理由该集群包含一篇详细介绍AI模型行为新发现的学术论文。

AI 生成摘要 · Google Gemini · 来自 2 个来源。我们如何撰写摘要 →

报道来源 [2]

arXiv cs.LG TIER_1 English(EN) · Utku \c{S}irin, Cathy Hou, David Alvarez-Melis, Stratos Idreos · 2026-06-03 04:00

Low-Frequency Shortcuts in Texture-Driven Visual Learning

arXiv:2606.03493v1 Announce Type: cross Abstract: Neural networks suffer from shortcut learning, where learned features generalize well to the training set but not to in-distribution (ID) or out-of-distribution (OOD) test sets. Existing studies are all based on a few standard ben…
arXiv cs.LG TIER_1 English(EN) · Stratos Idreos · 2026-06-02 11:11

纹理驱动视觉学习中的低频捷径

Neural networks suffer from shortcut learning, where learned features generalize well to the training set but not to in-distribution (ID) or out-of-distribution (OOD) test sets. Existing studies are all based on a few standard benchmarks, which are shape-driven. Numerous applicat…