LLLTime provides a method for zero-shot time series forecasting with large language models (LLMs). By transforming numerical data into text, the model predicts future series extensions, resembling natural sentence continuation.
LLMTime can outperform many popular time-series methods without any training on the target dataset.