LLMFew leverages large language models to tackle the challenge of few-shot multivariate time series classification in data-scarce industrial environments.

• Transforms complex multivariate time series data into formats LLMs can understand
• Enables effective classification with minimal training examples
• Demonstrates practical application in industrial and medical contexts
• Addresses a critical gap in time series analysis where labeled data is often limited

This research is particularly valuable for engineering applications where collecting large labeled datasets is costly or impractical, offering a path to implement advanced predictive maintenance and anomaly detection with fewer examples.

Large Language Models are Few-shot Multivariate Time Series Classifiers