Text-ADBench: Text Anomaly Detection Benchmark based on LLMs Embedding

2025-07-16

AI Summary

Text anomaly detection lacks standardized benchmarks, hindering rigorous methodological comparison and innovation. To address this, we introduce the first LLM-embedding-based benchmark for text anomaly detection, encompassing multiple models (GloVe, BERT, LLaMA, Mistral, OpenAI) and diverse domains. We systematically evaluate combinations of these embeddings with classical detectors—including KNN and Isolation Forest—under uniform experimental protocols. Key findings: embedding quality is the dominant factor governing detection performance; shallow detectors achieve competitive AUROC/AUPRC scores when paired with high-quality embeddings, matching deeper alternatives; and cross-model performance matrices exhibit low-rank structure, enabling efficient detector selection. All experiments employ comprehensive evaluation metrics (AUROC and AUPRC), and we publicly release the embedded datasets, source code, and evaluation framework. This benchmark establishes a reproducible, extensible foundation for advancing text anomaly detection research.

Technology Category

• Data Mining & Knowledge Management: Anomaly/Outlier Detection
• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Safety and Robustness

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling

Abstract

Text anomaly detection is a critical task in natural language processing (NLP), with applications spanning fraud detection, misinformation identification, spam detection and content moderation, etc. Despite significant advances in large language models (LLMs) and anomaly detection algorithms, the absence of standardized and comprehensive benchmarks for evaluating the existing anomaly detection methods on text data limits rigorous comparison and development of innovative approaches. This work performs a comprehensive empirical study and introduces a benchmark for text anomaly detection, leveraging embeddings from diverse pre-trained language models across a wide array of text datasets. Our work systematically evaluates the effectiveness of embedding-based text anomaly detection by incorporating (1) early language models (GloVe, BERT); (2) multiple LLMs (LLaMa-2, LLama-3, Mistral, OpenAI (small, ada, large)); (3) multi-domain text datasets (news, social media, scientific publications); (4) comprehensive evaluation metrics (AUROC, AUPRC). Our experiments reveal a critical empirical insight: embedding quality significantly governs anomaly detection efficacy, and deep learning-based approaches demonstrate no performance advantage over conventional shallow algorithms (e.g., KNN, Isolation Forest) when leveraging LLM-derived embeddings.In addition, we observe strongly low-rank characteristics in cross-model performance matrices, which enables an efficient strategy for rapid model evaluation (or embedding evaluation) and selection in practical applications. Furthermore, by open-sourcing our benchmark toolkit that includes all embeddings from different models and code at https://github.com/jicongfan/Text-Anomaly-Detection-Benchmark, this work provides a foundation for future research in robust and scalable text anomaly detection systems.

Problem

• Lack of standardized benchmarks for text anomaly detection evaluation
• Assessing embedding quality impact on anomaly detection performance
• Comparing shallow vs deep learning methods with LLM embeddings

Innovation

• Leverages diverse LLM embeddings for anomaly detection
• Systematically evaluates embedding-based detection methods
• Open-sources benchmark toolkit for future research