We propose a novel test-time computing paradigm of spatio-temporal forecasting.

We propose FactoST, a two-stage spatio-temporal foundation model that factorizes universal temporal pretraining from spatio-temporal adaptation, enabling efficient and generalizable forecasting across domains.

We propose RHYTHM, a new foundation model for human mobility prediction that combines temporal tokenization, hierarchical attention, and a frozen LLM backbone to model multi-scale spatio-temporal patterns efficiently.