Summary
Researchers Xiaona Zhou, Muntasir Wahed, Tianjiao Yu, Constantin Brif, and Ismini Lourentzou introduce VisAnomBench and VisAnomReasoner, addressing the poor performance of large language and multimodal models on anomaly pattern discovery in sequential data. Public anomaly detection benchmarks typically provide interval labels without natural-language explanations, making it hard to fine-tune vision-language models (VLMs) for grounded, interpretable decisions. VisAnomBench is a curated benchmark built from public time-series datasets, where high-quality anomaly explanations are generated by multiple large VLMs and filtered via fine-grained task-specific rewards. Fine-tuning on this benchmark yields VisAnomReasoner, a parameter-efficient VLM for time-series anomaly detection. On VisAnomBench it achieves more accurate anomaly localization, improving precision by at least 21.23 percentage points and F1 by 23.87 points, consistently outperforming all baselines. Additional experiments on the TSB-AD-U benchmark show strong cross-benchmark generalization, with precision gains of 9.57 points and F1 gains of 13.39 points. Paper: arXiv 2605.30344.
Paper Overview
Research Area: ML
Authors: Xiaona Zhou, Muntasir Wahed, Tianjiao Yu, Constantin Brif, Ismini Lourentzou
Published: 2026-05-28
arXiv: 2605.30344
Summary
Recent advances in vision-language models (VLMs) have achieved impressive performance on many tasks, yet prior studies report poor performance when large language or multimodal models are applied to anomaly pattern discovery in sequential data. Public anomaly detection benchmarks typically offer interval-level annotations but no natural-language explanations, making it difficult to fine-tune VLMs to produce grounded, interpretable decisions.
To fill this gap, the authors construct VisAnomBench, a curated benchmark built from public time-series datasets. High-quality anomaly explanations are generated using multiple large VLMs and filtered through fine-grained, task-specific rewards. By fine-tuning on this benchmark, they develop VisAnomReasoner, a parameter-efficient VLM for time-series anomaly detection.
Key Results
- VisAnomBench: VisAnomReasoner achieves more accurate anomaly localization, with precision improving by at least 21.23 percentage points and F1 by 23.87 percentage points, consistently outperforming all baselines.
- TSB-AD-U benchmark: Additional experiments demonstrate strong cross-benchmark generalization, with precision gains of 9.57 percentage points and F1 gains of 13.39 percentage points.
Links
- Paper: arXiv 2605.30344
*Auto-collected on 2026-06-01*
This page is an English static mirror generated for search and AI citation.
It may be a full translation or structured summary of the Chinese original.
Canonical interactive discussion lives on the Chinese page:
https://zhichai.net/topic/177980675