AnomalyR1: A GRPO-based End-to-end MLLM for Industrial Anomaly Detection
Preprint
Abstract
Industrial Anomaly Detection (IAD) poses a formidable challenge due to the scarcity of defective samples, making it imperative to deploy models capable of robust generalization to detect unseen anomalies effectively. Traditional approaches, often constrained by hand-crafted features or domain-specific expert models, struggle to address this limitation, underscoring the need for a paradigm shift. We introduce AnomalyR1, a pioneering framework that leverages VLM-R1, a Multimodal Large Language Model (MLLM) renowned for its exceptional generalization and interpretability, to revolutionize IAD. By integrating MLLM with Group Relative Policy Optimization (GRPO), enhanced by our novel Reasoned Outcome Alignment Metric (ROAM), AnomalyR1 achieves a fully end-to-end solution that autonomously processes inputs of image and domain knowledge, reasons through analysis, and generates precise anomaly localizations and masks. Based on the latest multimodal IAD benchmark, our compact 3-billion-parameter model outperforms existing methods, establishing state-of-the-art results. As MLLM capabilities continue to advance, this study is the first to deliver an end-to-end VLM-based IAD solution that demonstrates the transformative potential of ROAM-enhanced GRPO, positioning our framework as a forward-looking cornerstone for next-generation intelligent anomaly detection systems in industrial applications with limited defective data.Keywords
All Related Versions
This publication has 18 references indexed in Scilit:
- EIAD: Explainable Industrial Anomaly Detection Via Multi-Modal Large Language Models2025
- Dinomaly: The Less Is More Philosophy in Multi-Class Unsupervised Anomaly Detection2025
- Customizing Visual-Language Foundation Models for Multi-Modal Anomaly Detection and Reasoning2025
- LogiCode: An LLM-Driven Framework for Logical Anomaly DetectionIEEE Transactions on Automation Science and Engineering, 2024
- IM-IAD: Industrial Image Anomaly Detection Benchmark in ManufacturingIEEE Transactions on Cybernetics, 2024
- MAEDAY: MAE for few- and zero-shot AnomalY-DetectionComputer Vision and Image Understanding, 2024
- Deep Industrial Image Anomaly Detection: A SurveyMachine Intelligence Research, 2024
- PKU-GoodsAD: A Supermarket Goods Dataset for Unsupervised Anomaly Detection and SegmentationIEEE Robotics and Automation Letters, 2024
- Anomaly detection of defect using energy of point pattern features within random finite set frameworkEngineering Applications of Artificial Intelligence, 2023
- Beyond Dents and Scratches: Logical Constraints in Unsupervised Anomaly Detection and LocalizationInternational Journal of Computer Vision, 2022