arXiv 2602.063302026-02-06cs.CV

遏制幻觉：基于级联早期拒绝的信号与语义OOD检测解耦

Halt the Hallucination: Decoupling Signal and Semantic OOD Detection Based on Cascaded Early Rejection

School of Computer and Electronic Information, Nanjing Normal University, China（南京师范大学计算机与电子信息学院）；School of Artificial Intelligence and Information Technology, Nanjing University of Chinese Medicine, China（南京中医药大学人工智能与信息技术学院）

Ningkang Peng, Chuanjie Cheng, Jingyang Mao, Xiaoqian Peng, Feng Xing, Bo Zhang, Chao Tan, Zhichao Zheng, Peiheng Li, Yanhui Gu

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

CER框架通过级联早期拒绝机制，有效解耦信号与语义异常检测，显著提升OOD检测性能与效率。

摘要：

高效的和鲁棒的分布外（OOD）检测对于安全关键应用至关重要。然而，现有方法仍然对低级统计噪声执行全规模推理。这种计算不匹配不仅导致资源浪费，还会引发语义幻觉，即深度网络强行将物理异常解释为高置信度的语义特征。为此，我们提出了级联早期拒绝（CER）框架，通过粗到细的逻辑实现异常检测的分层过滤。CER包含两个核心模块：1）结构能量筛（SES），利用拉普拉斯算子在网络入口建立非参数性屏障，以高效拦截物理信号异常；2）语义感知的超球面能量（SHE）检测器，通过在中间层解耦特征幅度与方向以识别细粒度的语义偏差。实验结果表明，CER不仅将计算开销减少了32%，而且在CIFAR-100基准上实现了显著的性能提升：平均FPR95从33.58%大幅下降到22.84%，AUROC提高到93.97%。关键的是，在模拟传感器故障的真实世界场景中，CER表现出远超现有最先进方法的性能。作为一种通用插件，CER可以无缝集成到各种最先进模型中，以提供性能提升。

英文摘要：

Efficient and robust Out-of-Distribution (OOD) detection is paramount for safety-critical applications.However, existing methods still execute full-scale inference on low-level statistical noise. This computational mismatch not only incurs resource waste but also induces semantic hallucination, where deep networks forcefully interpret physical anomalies as high-confidence semantic features.To address this, we propose the Cascaded Early Rejection (CER) framework, which realizes hierarchical filtering for anomaly detection via a coarse-to-fine logic.CER comprises two core modules: 1)Structural Energy Sieve (SES), which establishes a non-parametric barrier at the network entry using the Laplacian operator to efficiently intercept physical signal anomalies; and 2) the Semantically-aware Hyperspherical Energy (SHE) detector, which decouples feature magnitude from direction in intermediate layers to identify fine-grained semantic deviations. Experimental results demonstrate that CER not only reduces computational overhead by 32% but also achieves a significant performance leap on the CIFAR-100 benchmark:the average FPR95 drastically decreases from 33.58% to 22.84%, and AUROC improves to 93.97%. Crucially, in real-world scenarios simulating sensor failures, CER exhibits performance far exceeding state-of-the-art methods. As a universal plugin, CER can be seamlessly integrated into various SOTA models to provide performance gains.

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。