LAD-Reasoner: Tiny Multimodal Models are Good Reasoners for Logical Anomaly Detection
This addresses the need for efficient and interpretable logical anomaly detection in industrial settings, though it is incremental as it builds on existing multimodal models and training methods.
The paper tackles logical anomaly detection by introducing a new task (RLAD) and a tiny multimodal model (LAD-Reasoner) that matches the performance of a much larger model (Qwen2.5-VL-72B) in accuracy and F1 score on the MVTec LOCO AD dataset, while producing more interpretable rationales.
Recent advances in industrial anomaly detection have highlighted the need for deeper logical anomaly analysis, where unexpected relationships among objects, counts, and spatial configurations must be identified and explained. Existing approaches often rely on large-scale external reasoning modules or elaborate pipeline designs, hindering practical deployment and interpretability. To address these limitations, we introduce a new task, Reasoning Logical Anomaly Detection (RLAD), which extends traditional anomaly detection by incorporating logical reasoning. We propose a new framework, LAD-Reasoner, a customized tiny multimodal language model built on Qwen2.5-VL 3B. Our approach leverages a two-stage training paradigm that first employs Supervised Fine-Tuning (SFT) for fine-grained visual understanding, followed by Group Relative Policy Optimization (GRPO) to refine logical anomaly detection and enforce coherent, human-readable reasoning. Crucially, reward signals are derived from both the detection accuracy and the structural quality of the outputs, obviating the need for building chain of thought (CoT) reasoning data. Experiments on the MVTec LOCO AD dataset show that LAD-Reasoner, though significantly smaller, matches the performance of Qwen2.5-VL-72B in accuracy and F1 score, and further excels in producing concise and interpretable rationales. This unified design reduces reliance on large models and complex pipelines, while offering transparent and interpretable insights into logical anomaly detection. Code and data will be released.