Research map / Label-efficient and robust learning / OOD and anomaly detection
Anomaly detection with contaminated data
Handles noisy training data and distribution shifts using clustering, ensembles and outlier exposure.
64 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · 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.
16 papers in 2023–24 and 32 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic grew, from 7.5% to 12.4%.
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
- 2016 · A Framework for Outlier Description Using Constraint Programming (AAAI 2016 · 18 citations)
- 2019 · Learning Competitive and Discriminative Reconstructions for Anomaly Detection (AAAI 2019 · 21 citations)
- 2020 · Classification-Based Anomaly Detection for General Data (ICLR 2020 · 94 citations)
- 2022 · ADBench: Anomaly Detection Benchmark (NeurIPS 2022 · 252 citations)
- 2023 · Explicit Boundary Guided Semi-Push-Pull Contrastive Learning for Supervised Anomaly Detection (CVPR 2023 · 108 citations)
- 2024 · Anomaly Heterogeneity Learning for Open-set Supervised Anomaly Detection (CVPR 2024 · 37 citations)
- 2025 · Qsco: A Quantum Scoring Module for Open-Set Supervised Anomaly Detection (AAAI 2025 · 8 citations)
- 2026 · IDK-S: Incremental Distributional Kernel for Streaming Anomaly Detection (AAAI 2026 · 2 citations)
Earliest papers here
- A Framework for Outlier Description Using Constraint Programming (AAAI 2016 · 18 citations)
- Learning Competitive and Discriminative Reconstructions for Anomaly Detection (AAAI 2019 · 21 citations)
- Temporal Anomaly Detection: Calibrating the Surprise (AAAI 2019 · 8 citations)
What has been tried
37 papers here report 366 measured changes (a component added, removed or swapped, compared with a baseline): helped 63%, no clear effect 7%, hurt 21%, mixed 9%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Add HADG-generated heterogeneous distributions and CDL | AITEX, general setting, 10 examples · AUC | 0.892±0.007 -> 0.916±0.004 | Helped | Anomaly Heterogeneity Learning for Open-set Supervised Anomaly Detection (CVPR 2024 · 37 citations) |
| Increase Qsco circuit depth from ℓ = 2 to ℓ = 3 | General setting, 10 training anomalies across datasets · AUC | Mixed | Qsco: A Quantum Scoring Module for Open-Set Supervised Anomaly Detection (AAAI 2025 · 8 citations) | |
| use MEBin instead of fixed-threshold or Otsu binarization | MVTec AD · F1 | 0.499 -> 0.712 | Helped | AnomalyNCD: Towards Novel Anomaly Class Discovery in Industrial Scenarios (CVPR 2025 · 7 citations) |
| add pseudo-abnormal samples during training | Adult, MCAR, mr=0.2 · AUROC | 65.30 (4.36) -> 76.51 (2.12) | Helped | Unsupervised Anomaly Detection in The Presence of Missing Values (NeurIPS 2024 · 2 citations) |
| Add GBD | CIC-IDS · TP | 0.9422 -> 0.9979 (↑0.0557) | Helped | Dissect Black Box: Interpreting for Rule-Based Explanations in Unsupervised Anomaly Detection (NeurIPS 2024 · 2 citations) |
| Vary contamination ratio from 5% to 15% | Visual AD datasets · AUROC | No clear effect | An Evidence-Based Post-Hoc Adjustment Framework for Anomaly Detection Under Data Contamination (NeurIPS 2025 · 1 citation) | |
| use vector summation instead of the scalar anomaly score | 30 tabular datasets · Avg. Rank (AUC-ROC) | 3.7 -> 2.6 | Helped | Towards a Unified Framework of Clustering-based Anomaly Detection (ICML 2025 · 0 citations) |
| Add supervised cross-entropy loss to the contrastive objective | Scientific datasets · Embedding structure and class separation | Helped | AutoSciDACT: Automated Scientific Discovery through Contrastive Embedding and Hypothesis Testing (NeurIPS 2025 · 0 citations) | |
| Sample K = 6 detectors per time step instead of querying all detectors | Ozone · runtime (s) | 1746.19 -> 1002.05 | Helped | SEAD: Unsupervised Ensemble of Streaming Anomaly Detectors (ICML 2025) |
| Add auxiliary concept values to the prompts | Colored MNIST: Number · AUROC | 98.8 -> 98.5 | No clear effect | Language-Assisted Feature Transformation for Anomaly Detection (ICLR 2025) |
What reviewers will expect
Datasets and baselines named in the 19 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 · 4 of 19 papers
- CelebA · 3 of 19 papers
- ViSA · 2 of 19 papers
- CIFAR-10 · 2 of 19 papers
- MNIST · 2 of 19 papers
Baselines
- LOF · 4 of 19 papers
- kNN · 3 of 19 papers
- DeepSVDD · 3 of 19 papers
- DevNet · 2 of 19 papers
- OCSVM · 2 of 19 papers
- IForest · 2 of 19 papers
- DAGMM · 2 of 19 papers
- ECOD · 2 of 19 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 48 papers; from OpenAlex and the papers' affiliations).
- Tsinghua University · 4 papers
- Peng Cheng Laboratory · 3 papers
- Amazon · 2 papers
- The Chinese University of Hong Kong · 2 papers
- University of Wisconsin - Madison · 2 papers
- Peking University · 2 papers
Use it
Download BibTeX of 36 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
- Unified multi-class anomaly detection (86 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).