Chronos-2: A foundation model for multivariate time series forecasting

📄 arXiv: 2610.41956v1 📥 PDF

作者: Abdul Fatir Ansari, Oleksandr Shchur, Jaris Küken, Andreas Auer, Boran Han, Pedro Mercado, Syama Sundar Rangapuram, Huibin Shen, Lorenzo Stella, Xiyuan Zhang, Mononito Goswami, Shubham Kapoor, Danielle C. Maddix, Pablo Guerron, Florian Saupe, Markus Reichstein, Tony Hu, Junming Yin, Nick Erickson, Prateek Mutalik Desai, Hao Wang, Huzefa Rangwala, George Karypis, Yuyang Wang, Michael Bohlke-Schneider

分类: cs.LG, cs.AI, stat.ML

发布日期: 2026-10-30

备注: 41 pages, 9 figures. Extended version of the Chronos-2 technical report (arXiv:2510.15821), adding scientific case studies (epidemiology, macroeconomics, carbon flux, hydrology, hierarchical forecasting), additional baselines, ablation studies, and calibration analysis. Code: https://github.com/amazon-science/chronos-forecasting

💡 一句话要点

提出Chronos-2以解决多变量时间序列预测中的交互建模问题

🎯 匹配领域: 支柱九：具身大模型 (Embodied Foundation Models)

关键词: 时间序列预测 基础模型 多变量建模 不确定性量化 科学应用

📋 核心要点

  1. 现有的时间序列预测模型通常需要针对每个特定领域的数据集单独开发，无法有效建模时间序列之间的交互关系。
  2. Chronos-2通过大规模预训练，能够在零-shot条件下建模多变量时间序列及其异质协变量之间的关系，提供更为准确的预测。
  3. 在多个基准测试中，Chronos-2的预测准确性平均提高了5.2个百分点，且在计算效率上显著优于现有的任务特定预测模型。

📝 摘要（中文）

准确的预测及不确定性量化在流行病学、经济学、能源系统和地球科学等多个科学应用中至关重要。然而，大多数预测模型仍需针对每个领域特定的数据集单独开发和训练。时间序列基础模型（TSFMs）旨在消除这一限制，通过大规模预训练使同一模型能够跨领域应用。Chronos-2作为一种零-shot预测的基础模型，能够建模多变量时间序列与异质协变量之间的交互关系，直接从历史上下文中推断时间序列之间的关系，并生成概率预测。Chronos-2在多个基准测试中表现出色，平均提高了5.2个百分点的技能评分，并在计算效率上也表现优异，能够在一秒内为100个时间序列生成预测，且在多种科学场景中超越了专门开发的方法。

🔬 方法详解

问题定义：本论文旨在解决多变量时间序列预测中，现有模型无法有效建模时间序列之间及外部驱动因素的交互关系的问题。

核心思路：Chronos-2通过大规模预训练，能够在不同领域间共享模型结构，直接从历史数据中推断时间序列之间的关系，而无需针对特定任务进行处理。

技术框架：Chronos-2的整体架构包括数据预处理、特征提取、关系建模和预测生成等主要模块，能够处理多种类型的输入数据。

关键创新：该模型的核心创新在于其能够同时考虑多变量时间序列及异质协变量的交互，突破了传统模型独立处理时间序列的局限性。

关键设计：在模型设计中，采用了适应性损失函数和高效的网络结构，使得模型在保持高准确度的同时，具备良好的计算效率。具体参数设置和网络结构细节在论文中进行了详细描述。

📊 实验亮点

Chronos-2在多个基准测试中表现优异，平均技能评分提高了5.2个百分点，胜率提升了10个百分点。同时，该模型在计算效率上也表现出色，能够在一秒内为100个时间序列生成预测，较最佳任务特定模型快50倍和1000倍。

🎯 应用场景

Chronos-2的研究成果在多个科学领域具有广泛的应用潜力，包括传染病预测、宏观经济建模、碳通量预测和径流模拟等。其高效的预测能力和准确性使其能够替代传统的领域特定模型，为科学研究和决策提供支持，未来可能推动跨领域的预测技术发展。

📄 摘要（原文）

Accurate forecasting with uncertainty quantification is critical in a variety of scientific applications ranging from epidemiology and economics to energy systems and earth sciences. Yet most forecasting models still need to be developed and trained separately for each domain-specific dataset. Time series foundation models (TSFMs) aim to remove this limitation and enable the application of the same model across domains through large-scale pretraining. However, most existing TSFMs treat each time series independently and therefore cannot model interactions among time series and external drivers. Here, we introduce Chronos-2, a foundation model for zero-shot forecasting that models the interactions between multivariate time series and heterogeneous covariates, which commonly arise in scientific applications. The model infers relationships among time series directly from historical context and generates probabilistic forecasts without task-specific treatment. Across various benchmarks, Chronos-2 achieves state-of-the-art accuracy on average by improving skill scores by 5.2 percentage points relative to the strongest baselines and increasing win rates by 10 percentage points. Chronos-2 is also computationally efficient, generating forecasts for 100 time series in under one second on a single mid-range GPU, 50x and 1000x faster end-to-end than the best task-specific point and probabilistic forecasters, respectively. In addition, we demonstrate its effectiveness in diverse scientific settings -- including infectious-disease forecasting, macroeconomic modeling, carbon-flux prediction, and runoff simulation -- where the model matches or surpasses specialized methods developed for each domain. These results show that a single pretrained model can capture structure shared across temporal systems.