Research map / Label-efficient and robust learning / OOD and anomaly detection
Unified multi-class anomaly detection
Trains one anomaly detector across classes using reconstruction, cross-modal constraints and class-generalizable features.
86 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2019–2026) · an approach in OOD and anomaly detection, keyword cluster “anomaly detection · normal · anomalies”
State History Tried Checklist Preprints Who Export
Where it stands
Not enough benchmark data Fewer than four papers report the same benchmark, so there is no record to follow.
18 papers in 2023–24 and 56 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic grew, from 8.5% to 21.6%.
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2019 · Deep Anomaly Detection with Outlier Exposure (ICLR 2019 · 385 citations)
- 2020 · Graph Embedded Pose Clustering for Anomaly Detection (CVPR 2020 · 204 citations)
- 2021 · CutPaste: Self-Supervised Learning for Anomaly Detection and Localization (CVPR 2021 · 1,006 citations)
- 2022 · A Unified Model for Multi-class Anomaly Detection (NeurIPS 2022 · 99 citations)
- 2023 · SimpleNet: A Simple Network for Image Anomaly Detection and Localization (CVPR 2023 · 619 citations)
- 2024 · RealNet: A Feature Selection Network with Realistic Synthetic Anomaly for Anomaly Detection (CVPR 2024 · 198 citations)
- 2025 · Dinomaly: The Less Is More Philosophy in Multi-Class Unsupervised Anomaly Detection (CVPR 2025 · 103 citations)
- 2026 · IAD-R1: Reinforcing Consistent Reasoning in Industrial Anomaly Detection (AAAI 2026 · 9 citations)
Earliest papers here
- Deep Anomaly Detection with Outlier Exposure (ICLR 2019 · 385 citations)
- Multi-View Anomaly Detection: Neighborhood in Locality Matters (AAAI 2019 · 39 citations)
- Transfer Anomaly Detection by Inferring Latent Domain Representations (NeurIPS 2019 · 34 citations)
What has been tried
45 papers here report 515 measured changes (a component added, removed or swapped, compared with a baseline): helped 75%, no clear effect 6%, hurt 16%, mixed 3%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Add AFS and RRS to feature reconstruction | MVTec-AD · Image AUROC | 94.46 / 95.67 -> 99.65 | Helped | RealNet: A Feature Selection Network with Realistic Synthetic Anomaly for Anomaly Detection (CVPR 2024 · 198 citations) |
| add Noisy Bottleneck | MVTec-AD · image-level AUROC | 98.41 -> 99.06 | Helped | Dinomaly: The Less Is More Philosophy in Multi-Class Unsupervised Anomaly Detection (CVPR 2025 · 103 citations) |
| Add patch-level and image-level residual learning to text prompt features | ELPV (2-shot) · AUROC | 0.733 -> 0.839 | Helped | Toward Generalist Anomaly Detection via In-context Residual Learning with Few-shot Sample Prompts (CVPR 2024 · 78 citations) |
| Add INP Extractor and INP-Guided Decoder | MVTec-AD · I-AUROC | 98.59 -> 99.53 | Helped | Exploring Intrinsic Normal Prototypes within a Single Image for Universal Anomaly Detection (CVPR 2025 · 50 citations) |
| Combine clustering with visual foundation models in C3 instead of using clustering alone | MVTec-AD · image-level AUC | 97.3 -> 97.8 | Helped | UniVAD: A Training-free Unified Model for Few-shot Visual Anomaly Detection (CVPR 2025 · 30 citations) |
| Add MEM | MVTec AD · I-AUROC | 98.42 -> 99.01 | Helped | UniNet: A Contrastive Learning-guided Unified Framework with Feature Selection for Anomaly Detection (CVPR 2025 · 29 citations) |
| Use hyperbolic embeddings instead of Euclidean embeddings (c=0) | VisA · AUPRO (%) | 93.4 -> 95.4 | Helped | Hyperbolic Anomaly Detection (CVPR 2024 · 28 citations) |
| Add feature and gradient regularization | MVTec LOCO · AUROC | +3.4 | Helped | Generating and Reweighting Dense Contrastive Patterns for Unsupervised Anomaly Detection (AAAI 2024 · 24 citations) |
| Add CNC to the base model | MVTec AD · I-AUROC | +1.3 | Helped | CNC: Cross-modal Normality Constraint for Unsupervised Multi-class Anomaly Detection (AAAI 2025 · 18 citations) |
| Replace normal-only reconstruction with maximizing reconstruction error on anomalous samples | MVTec AD · PRO | 94.4 -> 86.2 | Hurt | Boosting Fine-Grained Visual Anomaly Detection with Coarse-Knowledge-Aware Adversarial Learning (AAAI 2025 · 15 citations) |
| Add the semantic AD module to LTAD | MVTec step100 · detection AUROC | 87.05 -> 87.36 | Helped | Long-Tailed Anomaly Detection with Learnable Class Names (CVPR 2024 · 15 citations) |
| add PAPB contextual valid-action prediction | EgoPER (All) · EDA | +5.8% | Helped | Modeling Multiple Normal Action Representations for Error Detection in Procedural Tasks (CVPR 2025 · 7 citations) |
What reviewers will expect
Datasets and baselines named in the 21 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- MVTec-AD · 12 of 21 papers
- VisA · 11 of 21 papers
- BTAD · 5 of 21 papers
- Real-IAD · 4 of 21 papers
- MPDD · 4 of 21 papers
- ISIC2018 · 2 of 21 papers
- Chest X-ray · 2 of 21 papers
- MvTec 3D-AD · 2 of 21 papers
Baselines
- PatchCore · 14 of 21 papers
- PaDiM · 7 of 21 papers
- UniAD · 7 of 21 papers
- RD4AD · 6 of 21 papers
- SimpleNet · 5 of 21 papers
- WinCLIP · 5 of 21 papers
- DRA · 4 of 21 papers
- DeSTSeg · 4 of 21 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 74 papers; from OpenAlex and the papers' affiliations).
- Sun Yat-sen University · 3 papers
- Beijing University of Posts and Telecommunications · 3 papers
- Yonsei University · 3 papers
- Singapore Management University · 3 papers
- Shanghai Jiao Tong University · 3 papers
- Hebrew University of Jerusalem · 2 papers
- Peng Cheng Laboratory · 2 papers
- Mitsubishi Electric Research Laboratories · 2 papers
- Beijing Institute of Technology · 2 papers
- Tsinghua University · 2 papers
Use it
Download BibTeX of 33 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Anomaly detection with contaminated data (64 papers)
- Deep one-class and tabular anomaly detection (50 papers)
Elsewhere in the topic
- Provable and feature-based OOD detection (122 papers)
- Spurious-aware OOD detection (109 papers)
- Out-of-distribution generalization theory (71 papers)
- Uncertainty and misclassification detection (56 papers)
- Robust fine-tuning for OOD generalization (56 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).