TimeRAG: It's Time for Retrieval-Augmented Generation in Time-Series Forecasting

Mengxi Xiao, Zihao Jiang, Zhengyu Chen, Dong Li, Shuai Chen, Sophia Ananiadou, Jimin Huang, Min Peng, Qianqian Xie

Keywords: time series, large language model, retrieval augmented generation

Abstract: Time-series data are essential for forecasting tasks across various domains. While Large Language Models (LLMs) have excelled in many areas, they encounter significant challenges in time-series forecasting, particularly in extracting relevant information from extensive temporal datasets. Unlike textual data, time-series data lack explicit retrieval ground truths, complicating the retrieval process. To tackle these issues, we present TimeRAG, a novel retrieval-augmented approach tailored for time-series forecasting. Our method uniquely applies to continuous and complex temporal sequences, and it is trained using LLM feedback, effectively addressing the absence of ground truth and aligning the priorities of the retriever and the LLM. Experimental results demonstrate the effectiveness of TimeRAG, highlighting its ability to significantly enhance forecasting performance and showcasing the potential of LLMs in time-series prediction tasks.

Primary Area: learning on time series and dynamical systems

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 3551