Timer-M1: A Multivariate Time Series Foundation Model via Learning Primitives

2026-10-08

AI Summary

This study addresses the limited generalization of existing foundation models in complex scenarios by proposing a data synthesis pipeline and pretraining paradigm grounded in temporal and relational primitives. Methodologically, it introduces a gated two-dimensional Transformer block for efficient multivariate time series modeling, coupled with a dynamic cross-variable attention mechanism that enhances feature interactions to enable zero-shot forecasting. Experimental evaluations demonstrate that the proposed model achieves first place on both the FEV and TIME benchmarks and second place on GIFT-Eval. These results validate its state-of-the-art performance and strong generalization capabilities in multivariate time series forecasting tasks.

Abstract

We introduce Timer-M1, a pretrained multivariate time series foundation model that learns with primitives for zero-shot forecasting. Across domains, time series share elementary temporal and relational patterns, termed primitives, yet differ in how these primitives manifest and evolve across different contexts. Despite progress in zero-shot and task-general forecasting, existing foundation models may still struggle to generalize to complex real-world scenarios. To this end, we develop a primitive-based data synthesis and pretraining pipeline. The synthesis pipeline generates series with temporal primitives shared across domains and then assembles real and generated series into multivariate samples using relational primitives. Afterwards, samples are organized into episodes by assigning distinct channel roles as target variates, past-only covariates, and known-future covariates, ensuring that the model is optimized on predictable variates using available exogenous information. Technically, Timer-M1 further adapts gated two-dimensional Transformer blocks that dynamically allocate cross-variate attention across layers. Across three large-scale forecasting benchmarks, Timer-M1 ranks first on both FEV and TIME and second on GIFT-Eval among most recent time series foundation models. These results support effective primitive-based pretraining as a route to robust general forecasting technique across domains and task settings.

Problem

• multivariate time series
• foundation model
• zero-shot forecasting
• generalization

Innovation

• Multivariate Time Series Foundation Model
• Primitive-based Pretraining
• Zero-shot Forecasting
• Gated 2D Transformer
• Data Synthesis