TSEData-20K

Gold definitionUpdated Apr 2, 2026

Definition

TSEData-20K is a novel dataset specifically designed for advancing LLM-driven Time Series Anomaly Detection (TSAD). It focuses on enhancing models' reasoning abilities and multi-turn dialogue capabilities to better understand and explain anomalous behaviors in time series data.

At a glance

Executive summary

TSEData-20K is a new dataset designed to make AI models better at finding and explaining unusual patterns in time series data, especially through conversations. It helps overcome current AI limitations in reasoning and dialogue for anomaly detection, leading to more accurate and versatile systems.

TL;DR

A new dataset that teaches AI models to reason about and discuss unusual patterns in time series data.

Key points

Provides structured data for training LLMs in anomaly detection reasoning and multi-turn dialogue.
Solves the problem of inadequate reasoning, deficient multi-turn dialogue, and narrow generalization in LLM-driven Time Series Anomaly Detection.
Used by researchers and ML engineers developing advanced LLM-driven anomaly detection systems, particularly for time series data.
Improves upon existing LLM-driven AD methods by offering a dedicated dataset for complex reasoning and conversational capabilities.
Contributes to the research trend of enhancing explainability, reasoning, and interactive capabilities of LLMs in specialized domains like time series anomaly detection.

Use cases

Developing conversational AI agents for industrial fault diagnosis in sensor data, allowing engineers to query anomalies.
Training LLMs to explain complex financial market anomalies to analysts, providing detailed reasoning for unusual trends.
Benchmarking new LLM architectures for their ability to reason about intricate medical time series data, such as ECG or EEG signals.
Improving predictive maintenance systems by enabling LLMs to provide detailed, multi-turn anomaly reports and suggest mitigation strategies.
Creating interactive educational tools for teaching anomaly detection concepts using real-world time series data and LLM explanations.

Also known as

TSAD Dialogue Dataset, Time Series Anomaly Reasoning Dataset