#anomaly detection

A new preprint argues that repeated rows in benchmark tables quietly inflate anomaly detection scores, and audits all 690 OddBench datasets to check how often it happens.