Overview

Tabular anomaly detection seeks unusual records in structured datasets, where anomalies may be rare, labels scarce, and features numeric, categorical, or mixed. Common approaches include isolation forests, neighborhood and density scores, reconstruction and energy models, and newer diffusion-based methods; some combine global and local evidence. Recent work also explores pretrained tabular models, in-context learning, synthetic anomalies, and adaptation across datasets to reduce the need for dataset-specific training. A central challenge is making detectors reliable under contaminated training data, distribution shifts, and different anomaly rates while choosing useful thresholds and producing understandable explanations.

Latest papers

CardsList

33 papers4 papers in the last four weeks, down 33% on the four weeks before. 0.0% of all new papers.

  1. TACTIC for Navigating the Unknown: Tabular Anomaly deteCTion via In-Context inference

    Mar 15, 2026Patryk Marszałek, Tomasz Kuśmierczyk, Marek ŚmiejaContextual Anomaly DetectionTabular Anomaly Detection

  2. Why Does Counterintuitive Likelihood Behavior Rarely Emerge in Tabular Anomaly Detection?

    Feb 10, 2026Donghwan Kim, Junghun Phee, Hyunsoo YoonAnomaly DetectionTabular Anomaly Detection

  3. Isolation-based Spherical Ensemble Representations for Tabular Anomaly Detection

    Oct 15, 2025Yang Cao, Sikun Yang, Hao Tian +5Tabular Anomaly DetectionIsolation Forest