Tiny but Trusted: Efficient Vision-Language Reasoning for Time-Series Anomaly Detection

2026-05-28

AI Summary

This work addresses the lack of interpretability in existing vision-language models for temporal anomaly detection and the absence of natural language explanations in public datasets. To this end, the authors introduce VisAnomBench, the first benchmark for temporal anomaly detection annotated with high-quality natural language rationales, along with VisAnomReasoner, a parameter-efficient vision-language reasoning model. The proposed approach innovatively incorporates a task-oriented, fine-grained reward mechanism to select among multiple generated explanations and integrates temporal visual encoding with efficient fine-tuning strategies. Experimental results demonstrate that the model achieves significant improvements of 21.23% in accuracy and 23.87% in F1 score on VisAnomBench, and also outperforms current baselines on TSB-AD-U, confirming its effectiveness and strong generalization capability.

Abstract

Recent advances in Vision-Language Models (VLMs) have achieved impressive performance across many tasks, yet prior studies report unsatisfactory performance when applying large language or multimodal models to finding abnormal patterns in sequential data. Public anomaly detection benchmarks typically provide interval annotations but not natural-language rationales, making it difficult to fine-tune VLMs to produce grounded, interpretable decisions. To address this gap, we construct VisAnomBench, a curated benchmark built from public time-series datasets and augmented with high-quality anomaly explanations selected from multiple large VLMs using fine-grained, task-specific rewards. Through fine-tuning on this benchmark, we develop VisAnomReasoner, a parameter-efficient VLM for time-series anomaly detection. Experimental results on VisAnomBench show that VisAnomReasoner achieves more accurate anomaly localization and consistently outperforms all baselines, with improvements of at least 21.23 and 23.87 percentage points in precision and F1, respectively. Additional experiments on the TSB-AD-U benchmark demonstrate strong cross-benchmark generalization, with VisAnomReasoner improving precision and F1 by 9.57 and 13.39 percentage points, respectively.

Problem

• time-series anomaly detection
• Vision-Language Models
• interpretable decisions
• anomaly explanations
• benchmark annotation

Innovation

• Vision-Language Models
• Time-Series Anomaly Detection
• Parameter-Efficient Fine-Tuning
• Interpretable Reasoning
• Benchmark Construction