Research map / Vision-language and multimodal / CLIP and zero-shot
VarProtoAD: Variational Prototype-Conditioned Prompting for Zero-Shot Anomaly Detection
Mengyang Zhao, Zhuolin He, Haiyang Yu, Teng Fu, Ke Niu, Xiangyang Xue
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 K-means prototypes (ablation) | prompt-only ZSAD · VisA image-level | AUROC: 85.4 -> 89.1 | Improved |
| add variational vMF prototypes (ablation) | prompt-only ZSAD · VisA image-level | AUROC: 85.4 -> 90.0 | Improved |
| add variational vMF prototypes (ablation) | prompt-only ZSAD · VisA pixel-level | AUROC: 95.2 -> 96.3 | Improved |
| add variational vMF prototypes (ablation) | prompt-only ZSAD · ColonDB pixel-level | AUROC: 84.5 -> 88.3 | Improved |
| use vMF instead of K-means prototypes (ablation) | VarProtoAD · ColonDB pixel-level | AUROC: 87.2 -> 88.3 | Improved |
| add ungated cross-attention conditioning (ablation) | prompt-only ZSAD · VisA image-level | AUROC: 85.4 -> 88.2 | Improved |
| add scalar gate to cross-attention (ablation) | cross-attn conditioning · VisA image-level | AUROC: 88.2 -> 88.7 | Improved |
| use confidence-dependent gate instead of scalar gate (ablation) | cross-attn conditioning · VisA image-level | AUROC: 88.7 -> 90.0 | Improved |
| add implicit deviation prompts to explicit prompts (ablation) | VarProtoAD · VisA image-level | AUROC: 89.2 -> 90.0 | Improved |
| add explicit abnormal prompts to implicit prompts (ablation) | VarProtoAD · VisA image-level | AUROC: 89.1 -> 90.0 | Improved |
| explicit-only vs implicit-only abnormal prompts (ablation) | VarProtoAD · VisA image-level | AUROC: 89.2 vs 89.1 | No clear effect |
| use VarProtoAD instead of Bayes-PFL | ZSAD methods · 15 datasets overall image-level | AUROC: 94.7 -> 95.6 | Improved |
Closest papers by meaning
- Towards Zero-Shot Anomaly Detection and Reasoning with Multimodal Large Language Models (CVPR 2025 · similarity 0.90)
- Bayesian Prompt Flow Learning for Zero-Shot Anomaly Detection (CVPR 2025 · similarity 0.89)
- PromptAD: Learning Prompts with only Normal Samples for Few-Shot Anomaly Detection (CVPR 2024 · similarity 0.89)
- AnomalyCLIP: Object-agnostic Prompt Learning for Zero-shot Anomaly Detection (ICLR 2024 · similarity 0.88)
- Toward Generalist Anomaly Detection via In-context Residual Learning with Few-shot Sample Prompts (CVPR 2024 · similarity 0.88)
- Aligning and Prompting Anything for Zero-Shot Generalized Anomaly Detection (AAAI 2025 · similarity 0.87)
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.
