Research map / Vision-language and multimodal / CLIP and zero-shot
Exploring High-order-aware Prompt Learning for Zero-shot Anomaly Detection
Shun Wei, Jielin Jiang, Xiaolong Xu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add high-order-aware textual prompts (MoEPLer) (ablation) | AnomalyCLIP + HiPL · MVTec AD | image-level AUROC: 92.5 -> 93.5 | Improved |
| add high-order-aware visual prompts to both text and image encoders (ablation) | AdaCLIP + HiPL · MVTec AD | image-level AUROC: 87.5 -> 90.5 | Improved |
| add high-order-aware visual prompts to image encoder only versus text encoder only (ablation) | AdaCLIP + HiPL · MVTec AD | image-level AUROC: 88.5 vs 89.1 | Mixed |
| increase learnable word embedding length L from 8 to 12 (ablation) | PointAD + HiPL · MVTec 3D-AD | image-level AUROC: 80.0 -> 82.0 | Improved |
| increase word embedding length from 12 to 14 (ablation) | PointAD + HiPL · MVTec 3D-AD | image-level AUROC: 82.0 -> 79.1 | Hurt |
| increase number of experts K from 4 to 5 or 6 (ablation) | PointAD + HiPL · MVTec 3D-AD | image-level AUROC: 82.0 -> 80.7 | Hurt |
| use diverse expert architectures {Conv, Attn, MLP, Mamba} instead of all MLP (ablation) | AnomalyCLIP + HiPL · VisA | image-level AUROC: 82.6 -> 84.5 | Improved |
| dynamic neighbor generation instead of fixed neighbor number n=7 (ablation) | AnomalyCLIP + HiPL · MVTec AD | image-level AUROC: 92.3 -> 93.5 | Improved |
| plug HiPL into AnomalyCLIP | AnomalyCLIP · MVTec AD (2D) | image-level AUROC: 90.9 -> 93.5 | Improved |
| plug HiPL into PointAD | PointAD · MVTec 3D-AD | image-level AUROC: 77.3 -> 82.0 | Improved |
| plug HiPL into AdaCLIP on medical data | AdaCLIP · ColonDB | pixel-level AUROC: 85.4 -> 86.2 | Improved |
| plug HiPL into CoOp | CoOp · CIFAR10 | accuracy: 55.2 -> 56.0 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- WinCLIP — WinCLIP: Zero-/Few-Shot Anomaly Classification and Segmentation (CVPR 2023) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Bayesian Prompt Flow Learning for Zero-Shot Anomaly Detection (CVPR 2025 · similarity 0.90)
- PromptAD: Learning Prompts with only Normal Samples for Few-Shot Anomaly Detection (CVPR 2024 · similarity 0.89)
- Kernel-Aware Graph Prompt Learning for Few-Shot Anomaly Detection (AAAI 2025 · similarity 0.88)
- Towards Zero-Shot Anomaly Detection and Reasoning with Multimodal Large Language Models (CVPR 2025 · similarity 0.87)
- AnomalyCLIP: Object-agnostic Prompt Learning for Zero-shot Anomaly Detection (ICLR 2024 · similarity 0.87)
- AnomalyGPT: Detecting Industrial Anomalies Using Large Vision-Language Models (AAAI 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
