Collected sources and patterns will appear here. Add from search or the patterns library.
HistoricalTimeSeriesContext → ZeroShotForecast (via pre-trained transformer cross-domain representations)
Pre-trains a large transformer on a diverse collection of heterogeneous time series datasets from multiple domains (energy, finance, healthcare, traffic); at inference time, uses the pre-trained representations for zero-shot or few-shot forecasting on unseen time series without domain-specific retraining, treating historical context windows as in-context demonstrations.
Problem it solves
Classical time-series models require per-dataset fitting; a foundation model pre-trained on diverse series generalizes across domains via in-context learning.
Consumes
Emits
Establishes
The real projects this mechanism was found in. Attribution is the point — this is how the best teams actually do it.