Research map / Label-efficient and robust learning / OOD and anomaly detection
Hunting Normality from Query Sample via Residual Learning for Generalist Anomaly Detection
Xiaolei Wang, Yuexin Wang, Tianhong Dai, Huihui Bai, Yao Zhao, Jimin Xiao
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 RFL alone (residual proxies directly hunt normality) to the NN baseline (ablation) | HNQ · MVTec AD | image AUROC: 93.2 -> 90.4 | Hurt |
| add RFL alone to the NN baseline (ablation) | HNQ · VisA | image AUROC: 82.3 -> 80.1 | Hurt |
| add NLS alone (initialized proxies) to the NN baseline (ablation) | HNQ · MVTec AD | image AUROC: 93.2 -> 93.8 | Improved |
| add NLS alone to the NN baseline (ablation) | HNQ · VisA | image AUROC: 82.3 -> 83.7 | Improved |
| combine RFL and NLS (residual guidance for normality learning) (ablation) | HNQ · MVTec AD | image AUROC: 93.2 -> 96.0 | Improved |
| combine RFL and NLS (ablation) | HNQ · VisA | image AUROC: 82.3 -> 87.3 | Improved |
| use residual R instead of mean support features as NLS value (ablation) | HNQ NLS · MVTec AD | image AUROC: 96.0 -> 92.7 | Hurt |
| use mean support features instead of R as RFL key (ablation) | HNQ RFL · MVTec AD | image AUROC: 96.0 -> 95.5 | Hurt |
| use mean support features instead of R as NLS key (remove residual guidance) (ablation) | HNQ NLS · MVTec AD | image AUROC: 96.0 -> 94.4 | Hurt |
| learn normality proxies guided by residuals instead of modeling residual distribution (ResAD), 4-shot | HNQ vs ResAD · BraTS (train on MVTec AD) | image AUROC: 74.9 -> 85.4 | Improved |
| replace ResAD residual distribution fitting with two attention modules | HNQ vs ResAD · MVTec AD | training time (hours): 20.6 -> 0.5 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- ResAD — ResAD: A Simple Framework for Class Generalizable Anomaly Detection (NeurIPS 2024) · better on 9 of 9 shared results
- PromptAD — PromptAD: Learning Prompts with only Normal Samples for Few-Shot Anomaly Detection (CVPR 2024) · better on 6 of 6 shared results
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
- Normal-Abnormal Guided Generalist Anomaly Detection (NeurIPS 2025 · similarity 0.93)
- Toward Generalist Anomaly Detection via In-context Residual Learning with Few-shot Sample Prompts (CVPR 2024 · similarity 0.89)
- ResAD: A Simple Framework for Class Generalizable Anomaly Detection (NeurIPS 2024 · similarity 0.88)
- One-to-Normal: Anomaly Personalization for Few-shot Anomaly Detection (NeurIPS 2024 · similarity 0.86)
- Open-Set Graph Anomaly Detection via Normal Structure Regularisation (ICLR 2025 · similarity 0.85)
- Beyond Normal References: Discriminative Few-Shot Anomaly Detection (NeurIPS 2026 · similarity 0.85)
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.
