This repository serves as a curated collection of papers, datasets, and models focused on the application of Large Language Models (LLMs) to time series analysis. It targets researchers and practitioners in machine learning, data science, and domain-specific fields like finance, healthcare, and IoT, providing a structured overview of this rapidly evolving research area.

How It Works

The project categorizes LLM applications in time series based on their integration points within a typical LLM pipeline: input (Prompting), tokenization (Quantization), embedding (Aligning), the LLM core (Vision as Bridge), and output (Tool Integration). This taxonomy helps users understand the different approaches to adapting LLMs for time series data, from treating it as raw text to developing specialized encoders or using visual representations as intermediaries.

Highlighted Details

• Comprehensive taxonomy categorizing LLM approaches for time series analysis.
• Extensive lists of relevant papers, datasets, and models with publication dates and affiliations.
• Includes a survey paper accepted to IJCAI 2024.
• Details on specific LLM architectures and their application domains (e.g., Finance, Health, IoT).

Maintenance & Community

The project is maintained by xiyuanzh and includes contributions from various academic institutions and companies. The primary resource is the survey paper, with ongoing updates likely tied to its continued development.

Licensing & Compatibility

The repository itself does not specify a license. The content is primarily a curated list of research papers and datasets, whose individual licenses would need to be consulted.

Limitations & Caveats

This is a curated list and does not provide direct code implementations or runnable models. The focus is on tracking research, not on providing a framework for direct use.