ZSG-IAD: A Multimodal Framework for Zero-Shot Grounded Industrial Anomaly Detection

TL;DR

ZSG-IAD is a multimodal framework that enables zero-shot industrial anomaly detection with grounded, interpretable results using vision-language features and structured reasoning.

Contribution

It introduces a novel multimodal, zero-shot anomaly detection framework with a language-guided grounding module and verifiable rewards for reliable, explainable industrial defect detection.

Findings

Strong zero-shot performance across multiple benchmarks

Produces transparent, physically grounded anomaly masks

Outperforms prior methods in interpretability and accuracy

Abstract

Deep learning-based industrial anomaly detectors often behave as black boxes, making it hard to justify decisions with physically meaningful defect evidence. We propose ZSG-IAD, a multimodal vision-language framework for zero-shot grounded industrial anomaly detection. Given RGB images, sensor images, and 3D point clouds, ZSG-IAD generates structured anomaly reports and pixel-level anomaly masks. ZSG-IAD introduces a language-guided two-hop grounding module: (1) anomaly-related sentences select evidence-like latent slots distilled from multimodal features, yielding coarse spatial support; (2) selected slots modulate feature maps via channel-spatial gating and a lightweight decoder to produce fine-grained masks. To improve reliability, we further apply Executable-Rule GRPO with verifiable rewards to promote structured outputs, anomaly-region consistency, and reasoning-conclusion coherence. Experiments across multiple industrial anomaly benchmarks show strong zero-shot performance and more transparent, physically grounded explanations than prior methods. We will release code and annotations to support future research on trustworthy industrial anomaly detection systems.