Towards One-for-All Anomaly Detection for Tabular Data

Tabular anomaly detection (TAD) aims to identify samples that deviate from the majority in tabular data and is critical in many real-world applications. However, existing methods follow a ``one model for one dataset (OFO)'' paradigm, which relies on dataset-specific training and thus incurs high computational cost and yields limited generalization to unseen domains. To address these limitations, we propose OFA-TAD, a generalist one-for-all (OFA) TAD framework that only requires one-time training on multiple source datasets and can generalize to unseen datasets from diverse domains on-the-fly. To realize one-for-all tabular anomaly detection, OFA-TAD extracts neighbor-distance patterns as transferable cues, and introduces multi-view neighbor-distance representations from multiple transformation-induced metric spaces to mitigate the transformation sensitivity of distance profiles. To adaptively combine multi-view distance evidence, a Mixture-of-Experts (MoE) scoring network is employed for view-specific anomaly scoring and entropy-regularized gated fusion, with a multi-strategy anomaly synthesis mechanism to support training under the one-class constraint. Extensive experiments on 34 datasets from 14 domains demonstrate that OFA-TAD achieves superior anomaly detection performance and strong cross-domain generalizability under the strict OFA setting. The source code is available at https://github.com/Shiy-Li/OFA-TAD.

Detecting unusual records in spreadsheets, like catching credit card fraud or spotting rare medical conditions, is a critical task. However, current AI tools usually require building and training a brand-new model from scratch for every single spreadsheet. This is highly time-consuming and makes it hard to quickly apply these safety tools to new, unseen problems. We wondered if we could train a single, universal model that works on any spreadsheet right out of the box. To achieve this, we developed OFA-TAD. Instead of focusing on the specific meaning of each column, our model simply looks at how isolated a data point is from its neighbors. By training this model once on a diverse collection of datasets, we taught it a universal rule for spotting anomalies. When tested on 34 different datasets, our single model successfully identified anomalies across entirely new domains without needing any retraining. This "train once, use anywhere" approach paves the way for faster, cheaper, and more accessible anomaly detection tools in the real world.