SEER: Self-Evolving Event Reasoning and Retrieval for Time Series Forecasting

2026-10-02

AI Summary

This study addresses the limitations of conventional time series forecasting, which often neglects exogenous events, and existing retrieval-augmented methods that suffer from high noise and lack causal reasoning. To overcome these challenges, this work proposes a closed-loop forecasting framework that integrates large language models with retrieval-augmented generation (RAG). By leveraging reflective memory, the framework optimizes retrieval processes and constructs a self-evolving causal knowledge base. Furthermore, it pioneers the transformation of forecasting errors into a dual-decoupled feedback mechanism while strictly enforcing temporal boundaries to prevent data leakage. Experimental results demonstrate that the proposed approach consistently outperforms state-of-the-art foundation models and large language model baselines across six volatile benchmark datasets.

Abstract

Real-world time series are frequently driven by exogenous events and structural shifts, rendering conventional forecasting based solely on historical numerical observations insufficient. While language models can retrieve external news, standard retrieval-augmented approaches struggle with high noise, missing signals, and an inability to reason causally about event impacts. We propose SEER (Self-Evolving Event Reasoning and Retrieval), a closed-loop framework that dynamically optimizes event conditioning for time series forecasting. SEER translates prediction errors into two decoupled feedback mechanisms: (i) a reflective retrieval memory that refines subsequent search queries and filters spurious noise, and (ii) a persistent causal knowledge base that distills transferable domain dynamics. SEER enforces strict chronological boundaries across both event retrieval and reflection, preventing look-ahead bias and data leakage. Across six volatile time-series benchmarks, SEER consistently outperforms state-of-the-art time series foundation models and language model baselines.

Problem

• Time Series Forecasting
• Exogenous Events
• Retrieval-Augmented Generation
• Causal Reasoning
• Structural Shifts

Innovation

• Self-Evolving Event Reasoning
• Time Series Forecasting
• Reflective Retrieval Memory
• Causal Knowledge Base
• Closed-loop Framework