LLM as Detector: In-context learning for tabular anomaly detection

New arXiv paper proposes using LLMs with in-context learning for tabular anomaly detection, targeting violations of cross-feature dependencies. A second paper argues energy-based and reconstruction signals are complementary for this task.