Research map / Label-efficient and robust learning / OOD and anomaly detection
ADBench: Anomaly Detection Benchmark
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use semi-supervised learning instead of fully supervised learning with 1% labeled anomalies | Anomaly detection on ADBench · 57 ADBench datasets, 1% labeled anomalies | median AUCROC: 60.84% -> 75.56% | Helped |
| Use semi-supervised learning instead of fully supervised learning with 5% labeled anomalies | Anomaly detection on ADBench · 57 ADBench datasets, 5% labeled anomalies | median AUCROC: 72.69% -> 80.95% | Helped |
| Increase labeled anomalies from 1% to 5% | Semi-supervised methods · 57 ADBench datasets | median AUCROC: 75.56% -> 80.95% | Helped |
| Use any one unsupervised algorithm rather than selecting among them | Unsupervised anomaly detection · 57 ADBench datasets | AUCROC | No clear effect |
| Choose LOF for local anomalies | Unsupervised anomaly detection · Local anomalies | AUCROC | Helped |
| Choose KNN for global anomalies | Unsupervised anomaly detection · Global anomalies | AUCROC | Helped |
| Use label-informed methods rather than the best type-matched unsupervised detector | Anomaly detection on synthetic anomaly types · Local, global, and dependency anomalies | AUCROC | Hurt |
| Use label-informed methods rather than OCSVM | Anomaly detection on synthetic anomalies · Clustered anomalies | AUCROC | Helped |
| Duplicate anomalies 6 times | Unsupervised methods · 57 ADBench datasets | median ΔAUCROC: -16.43% | Hurt |
| Duplicate anomalies 6 times | Semi-supervised methods · 57 ADBench datasets | median ΔAUCROC: -0.05% | No clear effect |
| Add irrelevant features | Supervised methods · 57 ADBench datasets, up to 50% irrelevant features | AUCROC degradation: ≤ 5% | No clear effect |
| Add irrelevant features | Unsupervised and most semi-supervised methods · 57 ADBench datasets, up to 50% irrelevant features | AUCROC degradation: up to 10% | Hurt |
Closest papers by meaning
- We Need to Rethink Benchmarking in Anomaly Detection (NeurIPS 2026 · similarity 0.90)
- The Elephant in the Room: Towards A Reliable Time-Series Anomaly Detection Benchmark (NeurIPS 2024 · similarity 0.86)
- An Evidence-Based Post-Hoc Adjustment Framework for Anomaly Detection Under Data Contamination (NeurIPS 2025 · similarity 0.86)
- ResAD: A Simple Framework for Class Generalizable Anomaly Detection (NeurIPS 2024 · similarity 0.85)
- Long-Tailed Anomaly Detection with Learnable Class Names (CVPR 2024 · similarity 0.84)
- ADPretrain: Advancing Industrial Anomaly Detection via Anomaly Representation Pretraining (NeurIPS 2025 · similarity 0.84)
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.