Why Is the Counterintuitive Phenomenon of Likelihood Rare in Tabular Anomaly Detection with Deep Generative Models?

ICLR 2026 Conference Submission10784 Authors

Keywords: Deep Generative Models, Tabular Anomaly Detection, Likelihood Paradox of Deep Generative Model

Abstract: Deep generative models with tractable and analytically computable likelihoods, exemplified by normalizing flows, offer an effective basis for anomaly detection through likelihood-based scoring. We demonstrate that, unlike in the image domain where deep generative models frequently assign higher likelihoods to anomalous data, such counterintuitive behavior occurs far less often in tabular settings. We first introduce a domain-agnostic formulation that enables consistent detection and evaluation of the counterintuitive phenomenon, addressing the absence of precise definition. Through extensive experiments on 47 tabular datasets and 10 CV/NLP embedding datasets in ADBench, benchmarked against 12 baseline models, we demonstrate that the phenomenon, as defined, is consistently rare in general tabular data. We further investigate this phenomenon from both theoretical and empirical perspectives, focusing on the roles of data dimensionality and feature correlation difference. We find that likelihood-only detection with normalizing flows offers a practical and reliable approach for anomaly detection in tabular domains.

Primary Area: generative models

Submission Number: 10784