Research map / Time series / Anomaly detection

Multivariate anomaly detection

Detect anomalies across series using graph dynamics, causality and diffusion reconstruction.

37 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2018–2026) · an approach in Anomaly detection, keyword cluster “anomaly · detection · 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.

6 papers in 2023–24 and 27 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic held steady, from 42.9% to 45.8%.

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

2018 · Non-Parametric Outliers Detection in Multiple Time Series A Case Study: Power Grid Data Analysis (AAAI 2018 · 26 citations)
2019 · A Deep Neural Network for Unsupervised Anomaly Detection and Diagnosis in Multivariate Time Series Data (AAAI 2019 · 821 citations)
2021 · Graph Neural Network-Based Anomaly Detection in Multivariate Time Series (AAAI 2021 · 1,301 citations)
2022 · Deep Variational Graph Convolutional Recurrent Network for Multivariate Time Series Anomaly Detection (ICML 2022)
2023 · Detecting Multivariate Time Series Anomalies with Zero Known Label (AAAI 2023 · 62 citations)
2024 · Multi-Scale Dynamic Graph Learning for Time Series Anomaly Detection (Student Abstract) (AAAI 2024 · 3 citations)
2025 · Graph Mixture of Experts and Memory-augmented Routers for Multivariate Time Series Anomaly Detection (AAAI 2025 · 22 citations)
2026 · CANDI: Curated Test-Time Adaptation for Multivariate Time-Series Anomaly Detection Under Distribution Shift (AAAI 2026 · 2 citations)

Earliest papers here

Non-Parametric Outliers Detection in Multiple Time Series A Case Study: Power Grid Data Analysis (AAAI 2018 · 26 citations)
A Deep Neural Network for Unsupervised Anomaly Detection and Diagnosis in Multivariate Time Series Data (AAAI 2019 · 821 citations)
Graph Neural Network-Based Anomaly Detection in Multivariate Time Series (AAAI 2021 · 1,301 citations)

What has been tried

19 papers here report 196 measured changes (a component added, removed or swapped, compared with a baseline): helped 79%, no clear effect 3%, hurt 13%, mixed 5%. Papers mostly publish changes that helped, so "helped" is over-represented.

Ablations in recent papers

What was changed	Measured on	Result	Effect	Paper
Vary the number of experts from 1 to 4; 3 experts performs best	SWaT and WADI · AUROC	SWaT: 86.2±0.7 (1), 86.8±0.4 (2), 87.2±1.3 (3), 85.6±2.1 (4); WADI: 92.5±1.2 (1), 93.4±0.6 (2),	Mixed	Graph Mixture of Experts and Memory-augmented Routers for Multivariate Time Series Anomaly Detection (AAAI 2025 · 22 citations)
add causality graph sparsification	SWaT · ROC AUC	0.8559 -> 0.8690	Helped	GCAD: Anomaly Detection in Multivariate Time Series from the Perspective of Granger Causality (AAAI 2025 · 18 citations)
Replace learned channel correlation with CI, CD, or random masking	CICIDS, CalIt2, GECCO, MSL, SMD (Avg) · AUC-ROC	0.816 -> 0.755; 0.790; 0.782	Hurt	CATCH: Channel-Aware Multivariate Time Series Anomaly Detection via Frequency Patching (ICLR 2025 · 2 citations)
add reconstruction loss	PSM · F1	58.03 -> 65.85	Helped	Structured Temporal Causality for Interpretable Multivariate Time Series Anomaly Detection (NeurIPS 2025 · 1 citation)
add CPA-generated causality-preserving normal variations	SWaT · AUPRC	0.721 -> 0.764	Helped	Causality-Aware Contrastive Learning for Robust Multivariate Time-Series Anomaly Detection (ICML 2025 · 0 citations)
Replace reconstruction error with ChInf	SMD · F1	48.9 -> 52.1	Helped	Channel Matters: Estimating Channel Influence for Multivariate Time Series (NeurIPS 2025 · 0 citations)
Increase smoothing noise σ from 0.5 to 2.0	SMAP · F1-score	0.961 -> 0.913	Hurt	Fortifying Time Series: DTW-Certified Robust Anomaly Detection (NeurIPS 2025 · 0 citations)
increase time stride from 1 to 5	SMAP · AUC	0.73 -> 0.64	Hurt	TimeInf: Time Series Data Contribution via Influence Functions (ICLR 2025)

What reviewers will expect

Datasets and baselines named in the 16 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.

Datasets

SWaT · 8 of 16 papers
SMD · 7 of 16 papers
MSL · 7 of 16 papers
PSM · 7 of 16 papers
SMAP · 6 of 16 papers
WADI · 5 of 16 papers

Baselines

Anomaly Transformer · 8 of 16 papers
USAD · 6 of 16 papers
TimesNet · 5 of 16 papers
DCdetector · 4 of 16 papers
DAGMM · 3 of 16 papers
OmniAnomaly · 3 of 16 papers
CATCH · 3 of 16 papers
GDN · 2 of 16 papers

Recent preprints

Loading recent arXiv preprints…

Who publishes on it

Institutions on the most papers of this approach since 2023 (of 33 papers; from OpenAlex and the papers' affiliations).

Seoul National University · 4 papers
Zhejiang University · 2 papers
Xidian University · 2 papers

Use it

Download BibTeX of 27 key papers Open 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.

Same keyword cluster

Time series anomaly detection evaluation (47 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).