Learning to Factorize and Adapt: A Versatile Approach Toward Universal Spatio-Temporal Foundation Models

2026-01-17

AI Summary

This work addresses key challenges in spatiotemporal foundation models, including poor cross-dataset generalization, high computational costs of joint pretraining, and spatial heterogeneity. To overcome these limitations, we propose FactoST-v2, a novel factorized disentanglement architecture that decouples universal temporal learning from domain-specific spatial adaptation. This design enables full weight transferability and arbitrary-length generalization. The temporal encoder is pretrained using random sequence masking, while a lightweight spatial adapter is constructed via meta-adaptive learning and prompt-based mechanisms. Extensive experiments demonstrate that FactoST-v2 achieves state-of-the-art accuracy across diverse domains, significantly outperforms existing foundation models in zero-shot and few-shot forecasting scenarios, and attains linearly superior inference efficiency—matching the performance of specialized expert models.

Technology Category

• Machine Learning: Deep Neural Architectures and Foundation Models
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Computer Vision: Other Foundations of Computer Vision

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization

Abstract

Spatio-Temporal (ST) Foundation Models (STFMs) promise cross-dataset generalization, yet joint ST pretraining is computationally expensive and grapples with the heterogeneity of domain-specific spatial patterns. Substantially extending our preliminary conference version, we present FactoST-v2, an enhanced factorized framework redesigned for full weight transfer and arbitrary-length generalization. FactoST-v2 decouples universal temporal learning from domain-specific spatial adaptation. The first stage pretrains a minimalist encoder-only backbone using randomized sequence masking to capture invariant temporal dynamics, enabling probabilistic quantile prediction across variable horizons. The second stage employs a streamlined adapter to rapidly inject spatial awareness via meta adaptive learning and prompting. Comprehensive evaluations across diverse domains demonstrate that FactoST-v2 achieves state-of-the-art accuracy with linear efficiency - significantly outperforming existing foundation models in zero-shot and few-shot scenarios while rivaling domain-specific expert baselines. This factorized paradigm offers a practical, scalable path toward truly universal STFMs. Code is available at https://github.com/CityMind-Lab/FactoST.

Problem

• Spatio-Temporal Foundation Models
• cross-dataset generalization
• domain heterogeneity
• computational efficiency
• universal modeling

Innovation

• factorized framework
• spatio-temporal foundation model
• temporal pretraining
• spatial adaptation
• meta adaptive learning