Hugging Face 10 29 announced Project ProtoStar: partnering with Carnegie Mellon, Purdue, Georgia Tech, Arizona State, and George Washington University (via the Open Forum for AI), Astera Institute, and the Open Research Community ORCA to promote publishing publicly funded research data, models, and code in AI-ready formats on open platforms, and to host a series of public challenges. Participating institutions join by signing a memorandum of understanding, with Hugging Face providing infrastructure.
Confidence: High (80) · Event time 2026-10-29
Rationale
- Although publicly funded research outputs are required to be open, they are often scattered across incompatible systems and legacy formats, lacking documentation and provenance, making them difficult for AI to discover and combine. Officially, Hugging Face hosts over 1.7 public datasets in core scientific domains, and the number of new scientific datasets released in the first nine months of 2026 was more than triple that of the same period last year.
- The first challenge, hosted by Carnegie Mellon University, opens on 11 2: using meteorological records from 210 U.S. cities up to 2018, predict annual average temperature, annual precipitation, hottest day, and cooling degree days for 2019 to 2026, comparing against simple baselines like historical means and linear trends to see if AI methods can achieve higher accuracy; also submit one urban climate adaptation policy recommendation. Final results will be published in early 2027 after full 2026 data is obtained.
Usage
- After 11 2, sign up for the first challenge on the Project ProtoStar page on Hugging Face, submit results as Hugging Face datasets, and view prediction scores on the public leaderboard; research institutions interested in joining can contact [email protected].
Risks
- Only an announcement from Hugging Face, with no media coverage; details on whether the challenge offers prizes or specific judging rules are not provided.
- The memorandum of understanding is non-binding, and it remains unclear which data each institution will actually open up.