Reason-IAD: Knowledge-Guided Dynamic Latent Reasoning for Explainable Industrial Anomaly Detection

2026-02-10

AI Summary

Existing general-purpose multimodal large language models struggle to accurately identify category-specific fine-grained defects in industrial settings, resulting in insufficient detection accuracy and limited interpretability. To address this challenge, this work proposes a knowledge-guided dynamic latent-space reasoning framework that innovatively integrates retrieval-augmented category-specific textual knowledge, an entropy-driven optimizable implicit chain-of-thought token reasoning mechanism, and an information-theoretic dynamic visual patch injection strategy. This approach substantially enhances the model’s fine-grained understanding of industrial anomalies, outperforming state-of-the-art methods across multiple benchmarks while providing interpretable decision rationales.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Computer Vision: Multi-modal Vision
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Industrial anomaly detection demands precise reasoning over fine-grained defect patterns. However, existing multimodal large language models (MLLMs), pretrained on general-domain data, often struggle to capture category-specific anomalies, thereby limiting both detection accuracy and interpretability. To address these limitations, we propose Reason-IAD, a knowledge-guided dynamic latent reasoning framework for explainable industrial anomaly detection. Reason-IAD comprises two core components. First, a retrieval-augmented knowledge module incorporates category-specific textual descriptions into the model input, enabling context-aware reasoning over domain-specific defects. Second, an entropy-driven latent reasoning mechanism conducts iterative exploration within a compact latent space using optimizable latent think tokens, guided by an entropy-based reward that encourages confident and stable predictions. Furthermore, a dynamic visual injection strategy selectively incorporates the most informative image patches into the latent sequence, directing the reasoning process toward regions critical for anomaly detection. Extensive experimental results demonstrate that Reason-IAD consistently outperforms state-of-the-art methods. The code will be publicly available at https://github.com/chenpeng052/Reason-IAD.

Problem

• Industrial Anomaly Detection
• Multimodal Large Language Models
• Fine-grained Defect Patterns
• Explainability
• Category-specific Anomalies

Innovation

• knowledge-guided reasoning
• latent reasoning
• entropy-driven optimization
• retrieval-augmented module
• dynamic visual injection