Bootstrapped Exploration with Causal Reasoning: A Training Paradigm for Adaptive Forecasting Agent

Time series forecasting is critical in domains such as finance, energy, and healthcare, yet real-world datasets often exhibit non-stationarity, noise, missing values, and distribution shifts, posing severe challenges for generalization. In practice, industry solutions typically rely on customized forecasting frameworks that combine imputation, decomposition, and specialized models. However, such frameworks are costly to engineer and maintain. Moreover, we observe that many frameworks suffer from the impacts of distribution shifts, which degrade their respective performance. It motivates a paradigm that transfers reliably across heterogeneous datasets while accumulating reusable strategy knowledge for large-scale, dynamic environments. Although large language model-based agents have recently shown strong reasoning and tool-use capabilities, existing approaches do not consistently adapt forecasting workflows across diverse time series. We identify two primary factors, including limited strategy-level supervision and the inherent complexity of mapping dataset-specific meta-features to effective forecasting strategies. To address these challenges, we propose BECRA, a novel agent training paradigm that learns forecasting intelligence through contrast-aware exploration and agent-level causal lesson extraction, without human-annotated supervision. BECRA distills symbolic strategy lessons that support in-context planning on unseen datasets, enabling zero-shot training adaptation.

Time series forecasting is important in areas such as finance, energy, and healthcare, but real world data is often noisy, incomplete, and constantly changing. Existing forecasting systems usually require experts to manually test many methods for each new dataset, which is costly and hard to scale. We propose BECRA, an AI agent training approach that learns from past forecasting attempts. It compares successful and unsuccessful strategies, extracts reusable lessons about when each strategy works, and uses these lessons to choose forecasting pipelines for new datasets without retraining. Our experiments show that BECRA improves forecasting performance, handles missing values and anomalies more robustly, and reduces repeated trial-and-error costs.