Pythia: Toward Foundation World Models for Multimodal Time Series

Excerpt

arXiv:2610.05240v1 Announce Type: cross Abstract: Time-series foundation models offer a unified approach to forecasting across heterogeneous domains. Textual context and auxiliary observations provide complementary information about temporal dynamics, yet reusable multimodal predictive representations remain underexplored. We introduce Pythia, a foundation world model that learns context-conditioned latent dynamics across datasets through a joint-embedding predictive architecture. A stop-gradien