Semantics-Enhanced Retrieval-Augmented Time Series Forecasting

2026-06-12

AI Summary

This work addresses the limitations of traditional time series similarity-based retrieval-augmented methods, which struggle to effectively capture predictive historical patterns in non-stationary scenarios. To overcome this challenge, the authors propose SERAF, a novel framework that introduces multimodal retrieval into time series forecasting for the first time. SERAF constructs dual perspectives by combining self-generated textual descriptions with raw numerical signals, enabling joint retrieval and fusion of numerical and semantic information. By moving beyond reliance on a single similarity metric, the approach achieves substantial performance gains over state-of-the-art methods across seven real-world datasets, demonstrating the efficacy of semantic augmentation in enhancing forecasting accuracy.

Abstract

Time series forecasting models often benefit from historical patterns. Inspired by Retrieval-Augmented Generation (RAG), recent research explored retrieving relevant historical time series segments to enhance forecasting. However, relying solely on time series similarity is often insufficient for retrieval under non-stationarity. To address this, we propose a multimodal approach: a \textbf{S}emantics-\textbf{E}nhanced \textbf{R}etrieval-\textbf{A}ugmented Time Series \textbf{F}orecasting framework, SERAF. Unlike mainstream approaches that depend only on time series similarity, SERAF conducts dual retrieval over the time series and their self-generated textual descriptions. It retrieves two complementary sets of historical patterns and corresponding futures, which are selectively and jointly used to guide future predictions. Experiments across seven real-world datasets demonstrate the effectiveness of SERAF in bridging numerical and semantic views of time series compared with state-of-the-art baselines.

Problem

• time series forecasting
• retrieval-augmented generation
• non-stationarity
• semantic retrieval
• multimodal retrieval

Innovation

• Retrieval-Augmented Forecasting
• Time Series Semantics
• Multimodal Retrieval
• Non-stationary Time Series
• Text-Enhanced Forecasting