mlsanityData to 14 Oct 2026

Dual Anchors, Do It Better: Hierarchical Group Merging for Zero-Shot Anomaly Detection

arXiv 2608.11933 ↗First posted 12 Aug 2026cs.CVcited by 0

Not on the map. Why?

Abstract (arXiv)

Zero-shot anomaly detection (ZSAD) aims to identify anomalies in unseen domains, a setting that is particularly critical for industrial and medical applications where domain shifts are prevalent. However, most CLIP-based ZSAD methods anchor semantics solely on the text modality, making performance highly sensitive to prompt design and leading to weak visual grounding. To mitigate these limitations, we propose a Dual-Anchor framework that complements conventional text anchors with hierarchical image anchors constructed via a top-down grouping mechanism. This mechanism progressively aggregates local-to-global image features to form normal and abnormal group tokens, which serve as image anchors and act as gating signals in a Group-Gated Token Refiner to enhance the global representation. The refined image anchors are then fused with text prompts to construct dynamic state prompts. By jointly reinforcing visual and textual semantics, our framework stabilizes image-text alignment, reduces prompt dependency, and achieves strong generalization across 8 industrial and 6 medical benchmarks.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 22 of its 31 references that are in the map. First posted on arXiv 12 Aug 2026.

Builds on

22 of 31 references.

References · 22

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,744
  2. 2016Categorical Reparameterization with Gumbel-Softmaxcited by 2,739
  3. 2025DINOv3cited by 1,414
  4. 2016The Concrete Distribution: A Continuous Relaxation of Discrete Random Variablescited by 1,148
  5. 2022Token Merging: Your ViT But Fastercited by 1,021
  6. 2019Kvasir-SEG: A Segmented Polyp Datasetcited by 589
  7. 2022GroupViT: Semantic Segmentation Emerges from Text Supervisioncited by 531
  8. 2022SPot-the-Difference Self-Supervised Pre-training for Anomaly Detection and Segmentationcited by 400
  9. 2023WinCLIP: Zero-/Few-Shot Anomaly Classification and Segmentationcited by 315
  10. 2023AnomalyCLIP: Object-agnostic Prompt Learning for Zero-shot Anomaly Detectioncited by 271
  11. 2021VT-ADL: A Vision Transformer Network for Image Anomaly Detection and Localizationcited by 223
  12. 2016Skin Lesion Analysis toward Melanoma Detection: A Challenge at the International Symposium on Biomedical Imaging (ISBI) 2016, hosted by the International Skin Imaging Collaboration (ISIC)cited by 198
10 more
  1. 2023APRIL-GAN: A Zero-/Few-Shot Anomaly Classification and Segmentation Method for CVPR 2023 VAND Workshop Challenge Tracks 1&2: 1st Place on Zero-shot AD and 4th Place on Few-shot ADcited by 130
  2. 2021TokenLearner: What Can 8 Learned Tokens Do for Images and Videos?cited by 127
  3. 2024AdaCLIP: Adapting CLIP with Hybrid Learnable Prompts for Zero-Shot Anomaly Detectioncited by 127
  4. 2021Mixed supervision for surface-defect detection: from weakly to fully supervised learningcited by 101
  5. 2025AA-CLIP: Enhancing Zero-shot Anomaly Detection via Anomaly-Aware CLIPcited by 73
  6. 2023CLIP-AD: A Language-Guided Staged Dual-Path Model for Zero-shot Anomaly Detectioncited by 58
  7. 2024VCP-CLIP: A visual context prompting model for zero-shot anomaly segmentationcited by 50
  8. 2023HGFormer: Hierarchical Grouping Transformer for Domain Generalized Semantic Segmentationcited by 32
  9. 2025Bayesian Prompt Flow Learning for Zero-Shot Anomaly Detectioncited by 31
  10. 2023Bootstrap Fine-Grained Vision-Language Alignment for Unified Zero-Shot Anomaly Localizationcited by 10

Built on by

Not cited yet.