QiYao-I: A Manifold Based Foundation Model for Irregular Multivariate Time Series Forecasting

2026-10-03

AI Summary

This study addresses the challenge of forecasting irregularly sampled multivariate time series, where existing models struggle with uneven temporal intervals and asynchronous cross-variable dependencies. To this end, we propose QiYao-I, a manifold-based foundation model that innovatively maps real-valued timestamps into a learnable manifold space. Specifically, it introduces a sampling-conditioned temporal manifold attention mechanism to precisely capture irregular temporal dynamics, alongside a frequency-aware dynamic cross-variable message passing strategy to effectively model asynchronous variable interactions. Integrated within zero-shot and few-shot learning frameworks, QiYao-I significantly outperforms state-of-the-art foundation models and end-to-end approaches across multiple real-world benchmarks, demonstrating superior generalization capability and predictive accuracy.

Abstract

Irregular multivariate time series forecasting is a challenging yet important problem in real-world applications, where observations are often irregularly sampled and asynchronously recorded across variables. Existing time series foundation models are mostly built on regularly sampled sequences, making them difficult to generalize to irregular time intervals and asynchronous cross-variable dependencies. To address these challenges, we propose QiYao-I, a manifold based foundation model for irregular multivariate time series forecasting. Specifically, we introduce a novel sampling-conditioned temporal manifold attention mechanism that maps real timestamps into a learnable temporal manifold feature space and injects temporal manifold biases into attention layers, enabling the model to capture both irregular time intervals and local sampling structures. Further, we propose a dynamic variable interaction mechanism with frequency awareness. It selectively performs cross-variable message passing under asynchronous observations. Extensive experiments on real-world irregular multivariate forecasting benchmarks demonstrate that QiYao-I achieves superior performance compared with both time series foundation models and end-to-end irregular forecasting models, showing strong generalization ability in zero-shot and few-shot settings.

Problem

• Irregular multivariate time series
• Time series forecasting
• Foundation model
• Asynchronous observations
• Irregular sampling

Innovation

• Manifold-based Foundation Model
• Irregular Multivariate Time Series
• Temporal Manifold Attention
• Dynamic Variable Interaction
• Frequency Awareness