The details
Founding partners include Carnegie Mellon, Purdue, the Georgia Institute of Technology, Arizona State and George Washington University, along with the Astera Institute and the Open Research Community Accelerator. Hugging Face supplies the infrastructure. The problem they describe is that public research often sits in incompatible systems or legacy formats, with too little documentation to reuse. In the first challenge, teams use records through 2018 to forecast actual conditions from 2019 to 2026, including temperature, precipitation and the hottest day of the year. They also propose a climate adaptation idea. Final forecasting results are due in early 2027.
Why it matters
Public-access rules can require sharing without making results usable, so the challenges are a practical test of what reuse actually needs. Their findings could inform funding and policy. Limits remain: this is an announcement from the partners themselves, including the platform that would host the data, and the first results are months away. Forecasts must also beat simple baselines to show AI adds anything.