SAGE: A Visual Language Model for Anomaly Detection via Fact Enhancement and Entropy-aware Alignment

2025-07-10

AI Summary

Current vision-language models exhibit poor generalization and weak interpretability in industrial anomaly detection, struggling with unseen anomaly categories and structured reasoning requirements. To address these limitations, we propose AD-PL—a dedicated framework for industrial anomaly reasoning. First, we construct a domain-specific dataset integrating industrial domain knowledge and expert preferences. Second, we design self-guided factual augmentation and entropy-aware direct preference optimization to enhance factual consistency and decision trustworthiness. Third, we introduce a multi-scale logical evaluation framework to quantitatively assess the semantic coherence and logical rigor of reasoning chains. Under zero-shot and one-shot settings, AD-PL achieves state-of-the-art performance on mainstream industrial benchmarks—including MVTec and VisA—while significantly improving model interpretability and cross-category generalization capability.

Technology Category

• Computer Vision: Visual Reasoning & Symbolic Representations
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning
• Machine Learning: Adversarial Learning & Robustness

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

While Vision-Language Models (VLMs) have shown promising progress in general multimodal tasks, they often struggle in industrial anomaly detection and reasoning, particularly in delivering interpretable explanations and generalizing to unseen categories. This limitation stems from the inherently domain-specific nature of anomaly detection, which hinders the applicability of existing VLMs in industrial scenarios that require precise, structured, and context-aware analysis. To address these challenges, we propose SAGE, a VLM-based framework that enhances anomaly reasoning through Self-Guided Fact Enhancement (SFE) and Entropy-aware Direct Preference Optimization (E-DPO). SFE integrates domain-specific knowledge into visual reasoning via fact extraction and fusion, while E-DPO aligns model outputs with expert preferences using entropy-aware optimization. Additionally, we introduce AD-PL, a preference-optimized dataset tailored for industrial anomaly reasoning, consisting of 28,415 question-answering instances with expert-ranked responses. To evaluate anomaly reasoning models, we develop Multiscale Logical Evaluation (MLE), a quantitative framework analyzing model logic and consistency. SAGE demonstrates superior performance on industrial anomaly datasets under zero-shot and one-shot settings. The code, model and dataset are available at https://github.com/amoreZgx1n/SAGE.

Problem

• Enhances anomaly detection with interpretable explanations
• Improves generalization to unseen industrial categories
• Aligns model outputs with expert preferences

Innovation

• Self-Guided Fact Enhancement for knowledge integration
• Entropy-aware Direct Preference Optimization for alignment
• Multiscale Logical Evaluation for consistency analysis