NAISS
SUPR
NAISS Projects
SUPR
Foundation Models on Nationwide Registers for Pandemic Preparedness
Dnr:

NAISS 2026/4-340

Type:

NAISS Small

Principal Investigator:

Dominik Dietler

Affiliation:

Lunds universitet

Start Date:

2026-03-03

End Date:

2027-04-01

Primary Classification:

10201: Computer Sciences

Webpage:

Allocation

Abstract

Foundation models pretrained on population-scale event streams could provide early pandemic predictions before supervised methods become viable and potentially be used to model various aspects of the pandemic. We will train an autoregressive transformer on Swedish register data up to fixed cutoffs and evaluate out-of-time performance on COVID-19 outcomes against supervised baselines trained under the same constraints.