Cast-R1: Learning Tool-Augmented Sequential Decision Policies for Time Series Forecasting

2026-02-14

AI Summary

This work addresses the limitations of traditional time series forecasting methods, which struggle to autonomously gather evidence, reason about the future, or iteratively refine predictions in complex dynamic environments. The study reframes forecasting as a sequential decision-making problem and introduces a tool-augmented agent framework that leverages a memory mechanism for contextual evidence accumulation. By integrating modular tool invocation—such as statistical feature extraction and lightweight model assistance—with multi-step interactions, the agent enables reasoning-driven and self-reflective prediction refinement. The approach employs a two-stage learning strategy combining supervised fine-tuning and iterative reinforcement learning, augmented with curriculum learning. Extensive experiments on multiple real-world datasets demonstrate significant performance gains over state-of-the-art methods, validating the efficacy and potential of the agent-based paradigm for time series modeling.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Reasoning under Uncertainty: Sequential Decision Making
• Cognitive Modeling & Cognitive Systems: Agent Architectures

Application Category

• Search and Retrieval-Augmented AI: Agentic search
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Time series forecasting has long been dominated by model-centric approaches that formulate prediction as a single-pass mapping from historical observations to future values. Despite recent progress, such formulations often struggle in complex and evolving settings, largely because most forecasting models lack the ability to autonomously acquire informative evidence, reason about potential future changes, or revise predictions through iterative decision processes. In this work, we propose Cast-R1, a learned time series forecasting framework that reformulates forecasting as a sequential decision-making problem. Cast-R1 introduces a memory-based state management mechanism that maintains decision-relevant information across interaction steps, enabling the accumulation of contextual evidence to support long-horizon reasoning. Building on this formulation, forecasting is carried out through a tool-augmented agentic workflow, in which the agent autonomously interacts with a modular toolkit to extract statistical features, invoke lightweight forecasting models for decision support, perform reasoning-based prediction, and iteratively refine forecasts through self-reflection. To train Cast-R1, we adopt a two-stage learning strategy that combines supervised fine-tuning with multi-turn reinforcement learning, together with a curriculum learning scheme that progressively increases task difficulty to improve policy learning. Extensive experiments on multiple real-world time series datasets demonstrate the effectiveness of Cast-R1. We hope this work provides a practical step towards further exploration of agentic paradigms for time series modeling. Our code is available at https://github.com/Xiaoyu-Tao/Cast-R1-TS.

Problem

• time series forecasting
• sequential decision-making
• tool-augmented agents
• iterative reasoning
• long-horizon prediction

Innovation

• sequential decision-making
• tool-augmented agent
• memory-based state management
• iterative refinement
• reinforcement learning