EMNLP Findings2025
Text Anomaly Detection with Simplified Isolation Kernel
Yang Cao, Sikun Yang, Yu-Jiu Yang, Lianyong Qi, Ming Liu
Abstract
Two-step approaches combining pre-trained large language model embeddings and anomaly detectors demonstrate strong performance in text anomaly detection by leveraging rich semantic representations. However, highdimensional dense embeddings extracted by large language models pose challenges due to substantial memory requirements and high computation time. To address this challenge, we introduce the Simplified Isolation Kernel (SIK), which maps high-dimensional dense embeddings to lower-dimensional sparse representations while preserving crucial anomaly characteristics. SIK has linear time complexity and significantly reduces space complexity through its innovative boundary-focused feature mapping. Experiments across 7 datasets demonstrate that SIK achieves better detection performance than 11 state-of-the-art (SOTA) anomaly detection algorithms while maintaining computational efficiency and low memory cost. All code and demonstrations are available at https://github.com/charles-cao/SIK.
Ask Lune
Know this paper in minutes, not an afternoon.
Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.
Lineage
- NLP-ADBench: NLP Anomaly Detection Benchmark cites LUNAR: Unifying Local Outlier Detection Methods via Graph Neural Networks
- A New Distributional Treatment for Time Series and An Anomaly Detection Investigation cites Isolation Distributional Kernel: A New Tool for Kernel based Anomaly Detection
Builds on5
- Language Models are Few-Shot LearnersTom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah et al.NeurIPS 2020 · 64,255 citations
- LUNAR: Unifying Local Outlier Detection Methods via Graph Neural NetworksAdam Goodge, Bryan Hooi, See-Kiong Ng, Wee Siong NgAAAI 2022 · 144 citations
- Isolation Distributional Kernel: A New Tool for Kernel based Anomaly DetectionKai Ming Ting, Bi-Cun Xu, Takashi Washio, Zhi-Hua ZhouKDD 2020 · 49 citations
- A New Distributional Treatment for Time Series and An Anomaly Detection InvestigationKai Ming Ting, Zongyou Liu, Hang Zhang, Ye ZhuVLDB 2022 · 15 citations
- NLP-ADBench: NLP Anomaly Detection BenchmarkYuangang Li, Jiaqi Li, Zhuo Xiao, Tiankai Yang et al.EMNLP Findings 2025 · 4 citations
Related papers
- LLMs Meet Isolation Kernel: Lightweight, Learning-free Binary Embeddings for Fast RetrievalZhibo Zhang, Yang Xu, Kai Ming Ting, Cam-Tu NguyenACL Findings 2026
- Can LLMs Find a Needle in a Haystack? A Look at Anomaly Detection Language ModelingLeslie Barrett, Vikram Sunil Bajaj, Robert J. KinganEMNLP Findings 2025 · 1 citation
- IDK-S: Incremental Distributional Kernel for Streaming Anomaly DetectionYang Xu, Yixiao Ma, Kaifeng Zhang, Zuliang Yang et al.AAAI 2026 · 1 citation
- Enhancing Two Steps Textual Anomaly Detection through Anisotropy MitigationPierre Fihey, Matthieu Labeau, Pavlo MozharovskyiACL 2026
- AnoLLM: Large Language Models for Tabular Anomaly DetectionChe-Ping Tsai, Ganyu Teng, Phillip Wallis, Wei DingICLR 2025
