QiYao-M: Multimodal Time Series Foundation Model with Role-Aware Modeling of Endogenous and Exogenous Modalities

2026-09-28

AI Summary

This study addresses the limitation of existing time series models that overlook the distinct predictive roles of endogenous and exogenous modalities by proposing a role-aware multimodal time series foundation model. Methodologically, the two modality types are modeled separately: an explicit evolution supervision mechanism is introduced for the endogenous modality, while a parameter-update-free retrieval-augmented strategy is designed for the exogenous modality. Additionally, surrogate training techniques are employed to achieve efficient optimization. Experimental results demonstrate that the proposed model exhibits superior forecasting performance across multiple benchmarks, both with and without exogenous modalities, thereby validating the effectiveness of role-decoupled modeling.

Abstract

Existing multimodal time series foundation models (TSFMs) typically model heterogeneous modalities through largely shared mechanisms, overlooking the distinct forecasting roles of endogenous and exogenous modalities. In this work, we propose QiYao-M, a role-aware multimodal TSFM that models the two types of modalities separately. For endogenous modalities, to capture how they evolve along with the underlying temporal dynamics, we introduce an Endo-Multimodal Predictor and Endo-Multimodal Supervision to explicitly learn their evolution from history to the future. For exogenous modalities, to generalize across domains and across various modality types and numbers under the scarcity of exo-multimodal pretraining data, we propose an Exo-Multimodal Retrieval Enhancer that enables rapid downstream adaptation without updating the TSFM parameters. We further introduce Endo-Modality Proxy Training to train this retrieval module without exogenous multimodal pretraining data. Extensive experiments across unimodal and multimodal benchmarks demonstrate strong forecasting performance in scenarios both with and without exogenous modalities.

Problem

• multimodal time series foundation model
• endogenous modalities
• exogenous modalities
• role-aware modeling
• forecasting

Innovation

• Multimodal Time Series Foundation Model
• Role-Aware Modeling
• Endogenous and Exogenous Modalities
• Retrieval Enhancer
• Proxy Training