Time-RA: Towards Time Series Reasoning for Anomaly Diagnosis with LLM Feedback

TL;DR

Time-RA introduces a reasoning-based approach to time series anomaly detection, emphasizing interpretability and multimodal data integration, supported by a large-scale benchmark and open-source code.

Contribution

It reformulates TSAD into a generative, reasoning-oriented task and provides the first multimodal benchmark with extensive annotations and open-source resources.

Findings

Supervised fine-tuning improves diagnostic accuracy.

Visual representations enhance reasoning consistency.

Fine-tuned models outperform traditional baselines on unseen datasets.

Abstract

Time series anomaly detection (TSAD) has traditionally focused on binary classification and often lacks the fine-grained categorization and explanatory reasoning required for transparent decision-making. To address these limitations, we propose Time-series Reasoning for Anomaly (Time-RA), a novel task that reformulates TSAD from a discriminative into a generative, reasoning-intensive paradigm. To facilitate this, we introduce RATs40K, the first real-world large-scale multimodal benchmark with ~40,000 samples across 10 domains, integrating raw time series, textual context, and visual plots with structured reasoning annotations. Extensive benchmarking shows that while supervised fine-tuning and visual representations boost diagnostic accuracy and reasoning consistency, performance varies across complex scenarios. Notably, fine-tuned models demonstrate strong "plug-and-play" transferability, outperforming traditional baselines on unseen real-world datasets. Our work establishes a foundation for interpretable, multimodal time series analysis. All code (https://github.com/yyysjz1997/Time-RA) and the RATs40K dataset (https://huggingface.co/datasets/Time-RA/RATs40K) are fully open-sourced to facilitate future research.