Time-Prompt: Unlocking LLMs for Time Series Forecasting
Time-Prompt is a framework proposed to activate large language models (LLMs) for time series forecasting. It constructs a unified prompt paradigm combining learnable soft prompts to guide LLM behavior and textualized hard prompts to enhance time series representations. A semantic space is designed to improve the LLM's understanding of the forecasting task. The approach addresses skepticism about LLMs' utility in time series forecasting, aiming to improve long-term forecasting performance.
Key facts
- Time-Prompt is a framework for activating LLMs in time series forecasting.
- It uses learnable soft prompts and textualized hard prompts.
- A semantic space is designed to enhance LLM understanding.
- The framework aims to improve long-term forecasting performance.
- The paper is from arXiv with ID 2506.17631.
- It addresses skepticism about LLMs in time series forecasting.
- The approach combines prompt engineering with time series representation.
- The framework is designed to unlock LLMs for time series tasks.
Entities
Institutions
- arXiv