arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2604.26409cs.CV
收藏

稀疏性作为关键:从潜在结构中解锁新的见解用于分布外检测

Sparsity as a Key: Unlocking New Insights from Latent Structures for Out-of-Distribution Detection

School of Integrated Technology, BK21 Graduate Program in Intelligent Semiconductor Technology(整合技术学院,智能半导体技术BK21研究生项目)

Ahyoung Oh, Wonseok Shin, Songkuk Kim

AI总结:

本文首次将稀疏自编码器应用于ViT的[CLS]令牌进行分布外检测,通过Top-k SAE解构密集特征,揭示了分布内数据的类激活模式,并提出基于核心能量剖面差异的评分函数,实现了在多个基准上的强性能。

摘要:

稀疏自编码器(SAEs)在解释大型语言模型(LLMs)方面表现出显著的成功,通过将密集表示分解为稀疏、语义组件。然而,它们在分析视觉转换器(ViTs)方面的潜力仍 largely under-explored。在本工作中,我们首次将SAEs应用于ViT [CLS]令牌进行分布外(OOD)检测,解决了现有方法依赖于纠缠特征表示的限制。我们提出了一种新的框架,利用Top-k SAE将密集的[CLS]特征解构为结构化的潜在空间。通过这种分析,我们揭示了分布内(ID)数据表现出一致的、类特定的激活模式,我们将其正式化为类激活轮廓(CAPs)。我们的研究发现了一个关键的结构不变量:尽管ID样本在CAPs中保持稳定的模式,OOD样本系统地破坏了这种结构。利用这一见解,我们引入了一个基于核心能量剖面差异的评分函数,以量化偏离理想激活剖面的程度。我们的方法在FPR95指标上实现了强性能,这对安全敏感应用至关重要,同时在多个基准上也实现了具有竞争力的AUROC。总体而言,我们的发现表明,通过SAEs揭示的稀疏、解构特征可以作为一种强大且可解释的工具,用于视觉模型中稳健的分布外检测。

英文摘要:

Sparse Autoencoders (SAEs) have demonstrated significant success in interpreting Large Language Models (LLMs) by decomposing dense representations into sparse, semantic components. However, their potential for analyzing Vision Transformers (ViTs) remains largely under-explored. In this work, we present the first application of SAEs to the ViT [CLS] token for out-of-distribution (OOD) detection, addressing the limitation of existing methods that rely on entangled feature representations. We propose a novel framework utilizing a Top-k SAE to disentangle the dense [CLS] features into a structured latent space. Through this analysis, we reveal that in-distribution (ID) data exhibits consistent, class-specific activation patterns, which we formalize as Class Activation Profiles (CAPs). Our study uncovers a key structural invariant: while ID samples preserve a stable pattern within CAPs, OOD samples systematically disrupt this structure. Leveraging this insight, we introduce a scoring function based on the divergence of core energy profiles to quantify the deviation from ideal activation profiles. Our method achieves strong results on the FPR95 metric, critical for safety-sensitive applications across multiple benchmarks, while also achieving competitive AUROC. Overall, our findings demonstrate that the sparse, disentangled features revealed by SAEs can serve as a powerful, interpretable tool for robust OOD detection in vision models.

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。

↑