LLMs and Flow Matching Advance Tabular Anomaly Detection · 2 papers

Two new research papers explore advanced techniques for anomaly detection in tabular data. The first paper, "LLM as Detector," proposes a framework that leverages Large Language Models (LLMs) for in-context learning to derive anomaly detection logic from structured normal-state knowledge, outperforming 15 state-of-the-art baselines across 24 datasets without requiring LLM fine-tuning. The second paper investigates unsupervised anomaly detection using flow matching, specifically comparing Time-Conditioned Contraction Matching (TCCM) with Forest-Flow under contaminated training data, and highlights the critical role of anomaly scoring functions like deviation and reconstruction scores for robust performance in financial contexts. AI

IMPACT These papers introduce novel methods for anomaly detection in tabular data, potentially improving accuracy and efficiency in financial and other data-intensive applications.

RANK_REASON Two academic papers published on arXiv detailing new methods for tabular anomaly detection.

Read on arXiv cs.LG →

• alphaXiv
• arXiv
• CatalyzeX
• DagsHub
• Decision score
• Deviation score
• Forest Flower
• Gotit.pub
• Hugging Face
• IArxiv
• Influence Flower
• LLM-Detector
• Reconstruction score
• ScienceCast
• TCCM
• Time-Conditioned Contraction Matching

• paper
• model release

AI-generated summary · Google Gemini · from 2 sources. How we write summaries →