arXiv 2605.103972026-05-11cs.CVcs.AI

AnomalyClaw：通过工具引导的反驳实现通用视觉异常检测代理

AnomalyClaw: A Universal Visual Anomaly Detection Agent via Tool-Grounded Refutation

Department of Computer Science and Engineering, Southern University of Science and Technology (SUSTech), Shenzhen, China（南方科技大学计算机科学与工程系，深圳，中国）；School of EEE, Nanyang Technological University (NTU), Singapore（南洋理工大学电子工程学院，新加坡）；CFAR, Agency for Science, Technology and Research (A*STAR), Singapore（科技研究局（A*STAR）的CFAR，新加坡）

Xi Jiang, Yinjie Zhao, Zesheng Yang, Feng Zheng

备注：: We release the agent, the benchmark, and the analysis artifacts at https://github.com/jam-cc/AnomalyClaw

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

本文提出AnomalyClaw，一种无需训练的视觉异常检测代理，通过多轮反驳过程提升异常判断。在CrossDomainVAD-12基准上，AnomalyClaw在多个模型上均取得显著提升，且引入自演化扩展提升模型性能。

摘要：

视觉异常检测（VAD）在许多实际领域至关重要，如工业检测、医学影像、基础设施监控和遥感。然而，不同领域中的具体异常定义、数据模态和标注标准使得单域训练的VAD模型难以迁移。视觉-语言模型（VLMs）在大规模跨域数据上预训练，能在任务指令下进行视觉感知，为跨域VAD提供有希望的解决方案。然而，单次推理的VLM判断不可靠，因为它们更依赖于先验知识而非正常样本参考或细粒度特征证据。因此，我们提出了AnomalyClaw，一种无需训练的VAD代理，将异常判断转化为多轮反驳过程。在每一轮中，代理提出候选异常并针对正常样本参考进行反驳，利用包含13种工具的库进行视觉验证、参考解析和冻结专家探测。在CrossDomainVAD-12基准（12个数据集）上，AnomalyClaw在单步直接推理上实现了稳定的宏AUROC提升，GPT-5.5上提升+6.23 pp，Seed2.0-lite上提升+7.93 pp，Qwen3.5-VL-27B上提升+3.52 pp。我们进一步引入了可选的口头化自演化扩展。它从内部分支分歧中构建在线规则书，无需Oracle标签。在Qwen3.5-VL-27B上，它实现了+2.09 pp的平均提升，与K=10 Oracle标签监督基线（+1.99 pp）相当。这些结果表明，代理反驳提升了VLMs的异常理解和推理能力，而不仅仅是聚合工具输出。

英文摘要：

Visual anomaly detection (VAD) is crucial in many real-world fields, such as industrial inspection, medical imaging, infrastructure monitoring, and remote sensing. However, the specific anomaly definitions, data modalities, and annotation standards across different domains make it difficult to transfer single-domain trained VAD models. Vision-language models (VLMs), pre-trained on large-scale cross-domain data, can perform visual perception under task instructions, offering a promising solution for cross-domain VAD. However, single-inference VLM judgments are unreliable, since they rely more on prior knowledge than on normal-sample references or fine-grained feature evidence. We therefore present AnomalyClaw, a training-free VAD agent that turns anomaly judgment into a multi-round refutation process. In each round, the agent proposes candidate anomalies and refutes each against normal-sample references, drawing on a 13-tool library for visual verification, reference parsing, and frozen expert probing. On the CrossDomainVAD-12 benchmark (12 datasets), AnomalyClaw achieves consistent macro-AUROC improvements over single-step direct inference with +6.23 pp on GPT-5.5, +7.93 pp on Seed2.0-lite, and +3.52 pp on Qwen3.5-VL-27B. We further introduce an optional verbalized self-evolution extension. It builds an online rulebook from internal-branch disagreement without oracle labels. On Qwen3.5-VL-27B, it delivers a +2.09 pp mean gain, comparable to a K = 10 oracle-label supervised baseline (+1.99 pp). These results show that agentic refutation improve anomaly understanding and reasoning of VLMs, rather than merely aggregating tool outputs.

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。