原星挑战：气候风险情报
管理竞争登记、数据和提交材料
发表
2026年10月29日 (英语).
一项将大学、研究组织、工业界和开放科学界聚集在一起的倡议，使公共资助的研究更容易发现、再利用和借鉴。
今天，我们宣布了ProtoStar项目，这是一项通过让数据集，模型，代码更方便和可重复使用，特别是用于AI驱动的研究，来释放公共资助的科学研究潜力的倡议。 ProtoStar 带来会同研究机构，产业界，开放科学界支持以AI-即时，机器可操作的形式发布科学文物，并通过一系列公开挑战展示其价值。 由卡内基梅隆大学主办的首个挑战赛于11月2日开幕。 见下面如何参与。
该倡议的创始伙伴包括卡内基·梅隆大学、普杜大学、亚利桑那州立大学乔治亚理工学院和乔治·华盛顿大学，它们都作为大学联盟伙伴通过开放的AI论坛(OFAI)开展工作；阿斯特拉研究所；Hugging Face作为基础设施伙伴；开放研究社区加速器(ORCA)作为社区伙伴。
参与组织通过一份谅解备忘录，正式确定它们的参与，该备忘录支持以关于开放基础设施的AI准备格式发布适当的科学数据集、模型和代码。 有兴趣加入ProtoStar项目的组织可联系protostar@orcaopen.org。
美国联邦政府每年在研发方面投资约1900亿美元，而联邦政策越来越期望公开分享结果。 国家卫生研究所和国家科学基金会等机构加强了获得资助研究出版物的公众获取要求，并要求研究人员计划分享基本数据。 Hugging Face是NIH为共享研究数据而列出的十个通论寄存器之一。
但是，公开研究并不能自动使研究能够使用。 科学数据集、模型和代码往往分散在互不兼容的系统中，以静态格式或遗留格式储存，或发布，没有理解和再利用所需的文件、元数据和出处。 这些限制使得研究人员更难于在现有工作的基础上更上一层楼，AI系统也更难发现，解释和整合跨学科的科学资源。
ProtoStar项目旨在通过鼓励以AI-ready格式系统分享公共资助的科学文物，并在开放基础设施和实用出版指导的支持下，帮助缩小这一差距。 它借鉴了国家癌症研究所的AI数据准备挑战以及Hugging Face为AI编制公共部门数据的指南等努力。 目标是使研究组织更容易满足现有的公众获取的预期，同时提高它们发布的作品的科学价值。
对可再利用的科学资源的需求已经很明显。 Hugging Face拥有超过17,000个跨越核心科学领域的公共数据集，包括基因组学，蛋白质设计，化学，天气和气候，地球观测，和天文学。 如果将医学和数学包括在内，这个数字接近29 000个，还有相当数目的科学模型。
分享的速度也在加快。 研究人员在2026年的前9个月分享了Hugging Face上的新科学数据集比2025年同期多出三倍多，即使在排除了最大的个人上传器后，增长仍然相当可观。
图1. 按域(Q1 2023–Q3 2026)分列的Hugging Face季度新增公共科学数据集。 按创建日期计算数据集仍然公开。 资料来源：Hugging Face Hub数据，2026年10月。 探索数据。
当各组织以可重复使用的格式分享研究时，科学界可以扩大研究规模。 Hugging Face由NIH国家生物技术信息中心(NCBI)维护的18个公共寄存器集体收到超过1100万次下载，而NASAIMPACT的27个寄存器则收到约77万次下载，截至2026年10月26日。 文档也是使这些资源有用的一个重要部分：Hugging Face上90%的数据集下载来自完整有记录的数据集。
事实证明，公开的基因组数据对于进一步推动生物研究等领域的研究十分重要；Hugging Face的研究人员一直在利用公开可用的科学资源，包括最近来自HuggingFaceBio的发布，如Carbon-A和Carbon-X。 通过公开发布所产生的模型、数据集和工具，这些努力表明，现有的科学资源如何能够支持他人可以借鉴的新研究。
我们正通过参与者所采取的体制政策，通过ProtoStar项目进一步推动这种活动。 作为机构承诺的早期例子，CMU的开源方案办公室正在鼓励其研究人员分享合适的数据集，AI模型，以及Hugging Face的其他研究成果，同时认识到特定域存储器的持续重要性。
制作科学文物AI准备是重要的第一步。 我们还想表明，当这些资源可供再利用时，研究人员、开发者和更广泛的科学界能够取得什么样的成就。
ProtoStar的一个核心问题是：公共资助的科学数据集，模型，和代码，以AI-ready格式的发布，能否被随时发现，再利用，合并，并被扩展，以加速科学发现和公众利益？
为了探索这一点，ProtoStar将组织一系列公开挑战，每个挑战由一个参与的研究机构主办并围绕该机构公开拥有的发布的科学资源来构建。 这些挑战有两个相辅相成的目标：
- 证明科学再利用。 以公共资助的科学文物为基础，制作研究人员和开发者的具体实例，包括跨机构或学科综合资源的应用。
- 学点什么才能再利用查明技术障碍、文件空白、互操作性挑战以及能够为未来研究基础设施、供资决定和公共政策提供信息的成功做法。
由此产生的项目和经验教训将有助于建立一个实际证据基础，使公共资助的科学更容易获得和有用。
卡内基梅隆大学主办 2026年11月2日至13日
首发式"ProtoStar"挑战邀请参与者测试AI能否利用210个美国城市的历史天气数据来改进对城市气候条件的预测。 在卡内基梅隆大学研究的基础上，各团队将使用到2018年的气候记录来预测从2019年到2026年的实际状况，使其模型能够根据所观测到的天气而不是假设的未来情景来评价。
与会者将预测与城市规划有关的四项指标：年平均气温、年总降水量、一年中最热的一天和冷却度日，这是空调需求的衡量标准。 它们的预测将同简单的统计基线进行比较，包括历史平均值和线性趋势，以评估人工智能方法是否在准确性方面提供了有意义的改进。
除了预测之外，与会者还将根据从数据中汲取的教训提出一个气候适应政策的简短设想。 本部分鼓励团队将科学分析与实际的公众利益联系起来，如让城市做好极端高温的准备，提高能效，或管理不断变化的降水模式等。
团队将随这些产出提交Hugging Face数据集。 将通过公共领导板对照既定基线衡量预测错误来评估预测业绩，而政策设想则会得到单独的考虑。 一旦有完整的2026年天气记录，最终预报结果将在2027年初公布。
探讨CMU气候风险情报挑战
CMU挑战赛是本系列的第一场。 其他参与机构将依次承担各自的挑战，每个机构都强调不同的公共资助的科学资源和再利用的机会。 未来的挑战将是宣布通过Hugging Face上的Project ProtoStar页面。
无论是学术研究者、学生、独立开发者，还是在开放的科学基础设施上的公司，ProtoStar都提供将公共资助的研究投入工作的机会
我们还欢迎各大学、联邦研究机构、国家实验室、非营利组织和其他研究组织，它们希望使其科学资源更容易获得和再利用。 为了探讨参与或合作，请联系原星@orcaopen.org。
我们希望，ProtoStar将使公共资助的科学工作更容易接触到人民和能够在此基础上更上一层楼的工具，同时产生支持大规模开放科学所需的基础设施和做法的实际证据。
本文由 AI宝库自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。
Protostar Challenge: Climate Risk Intelligence
Manage competition registration, data, and submissions
Published
October 29, 2026
An initiative bringing together universities, research organizations, industry, and the open science community to make publicly funded research easier to discover, reuse, and build upon.
Today, we're announcing Project ProtoStar, an initiative to unlock the potential of publicly funded scientific research by making datasets, models, and code more accessible and reusable, particularly for AI-driven research. ProtoStar brings together research institutions, industry, and the open science community to support the release of scientific artifacts in AI-ready, machine-actionable formats, and to demonstrate their value through a series of open challenges. The first challenge, hosted by Carnegie Mellon University, opens on November 2. See How to Participate below.
The initiative's founding partners include Carnegie Mellon University, Purdue University, Georgia Institute of Technology, Arizona State University, and The George Washington University, all working through the Open Forum for AI (OFAI) as the university coalition partner; the Astera Institute; Hugging Face as the infrastructure partner; and the Open Research Community Accelerator (ORCA), through its Fostering Innovation Network, as the community partner.
Participating organizations formalize their involvement through a memorandum of understanding supporting the release of suitable scientific datasets, models, and code in AI-ready formats on open infrastructure. Organizations interested in joining Project ProtoStar can reach out to protostar@orcaopen.org.
The U.S. federal government invests approximately $190 billion annually in research and development, and federal policy increasingly expects the results to be shared publicly. Agencies such as the National Institutes of Health (NIH) and National Science Foundation (NSF) have strengthened public-access requirements for funded research publications and require researchers to plan for sharing the underlying data. Hugging Face is among the ten generalist repositories listed by NIH for sharing research data.
But making research publicly available does not automatically make it usable. Scientific datasets, models, and code are often scattered across incompatible systems, stored in static or legacy formats, or released without the documentation, metadata, and provenance needed to understand and reuse them. These limitations make it harder for researchers to build on existing work, and harder still for AI systems to discover, interpret, and combine scientific resources across disciplines.
Project ProtoStar aims to help close this gap by encouraging systematic sharing of publicly funded scientific artifacts in AI-ready formats, supported by open infrastructure and practical publishing guidance. It builds on efforts such as the National Cancer Institute's AI Data-Readiness Challenge and Hugging Face's guide to preparing public-sector data for AI. The goal is to make it easier for research organizations to meet existing public-access expectations while increasing the scientific value of the work they release.
The demand for reusable scientific resources is already evident. Hugging Face hosts more than 17,000 public datasets across core scientific domains, including genomics, protein design, chemistry, weather and climate, Earth observation, and astronomy. That figure approaches 29,000 when medicine and mathematics are included, alongside a comparable number of scientific models.
The pace of sharing is also accelerating. Researchers shared more than three times as many new science datasets on Hugging Face in the first nine months of 2026 as in the same period of 2025, with growth remaining substantial even after excluding the largest individual uploader.
Figure 1. New public science datasets on Hugging Face per quarter, by domain (Q1 2023–Q3 2026). Counts datasets still public, by creation date. Source: Hugging Face Hub data, October 2026. Explore the data.
When organizations share research in reusable formats, the scientific community can build on it at scale. The 18 public repositories maintained by NIH's National Center for Biotechnology Information (NCBI) on Hugging Face have collectively received more than 11 million downloads, while NASA IMPACT's 27 repositories have received approximately 770,000 downloads, as of October 26, 2026. Documentation is also an important part of making these resources useful: 90% of dataset downloads on Hugging Face are from fully documented datasets.
Publicly available genomic data has proved important in contributing to research that further advances fields such as biological research; researchers at Hugging Face have been building on openly available scientific resources, including recent releases from HuggingFaceBio such as Carbon-A and Carbon-X. By releasing the resulting models, datasets, and tools openly, these efforts demonstrate how existing scientific resources can support new research that others can build upon.
This is the kind of activity that we are further catalyzing via Project ProtoStar through institutional policies adopted by participating orgs. As an early example of institutional commitment, CMU's Open Source Programs Office is encouraging its researchers to share suitable datasets, AI models, and other research outputs on Hugging Face, while recognizing the continued importance of domain-specific repositories.
Making scientific artifacts AI-ready is an important first step. We also want to demonstrate what researchers, developers, and the broader scientific community can accomplish when those resources are made available for reuse.
A central question for ProtoStar is: Can publicly funded scientific datasets, models, and code, released in AI-ready formats, be readily discovered, reused, combined, and extended to accelerate scientific discovery and public benefit?
To explore this, ProtoStar will organize a series of open challenges, each hosted by a participating research institution and built around scientific resources that the institution has released openly. The challenges have two complementary goals:
- Demonstrate scientific reuse. Produce concrete examples of researchers and developers building on publicly funded scientific artifacts, including applications that combine resources across institutions or disciplines.
- Learn what makes reuse possible. Identify technical barriers, documentation gaps, interoperability challenges, and successful practices that can inform future research infrastructure, funding decisions, and public policy.
The resulting projects and lessons learned will help build a practical evidence base for making publicly funded science more accessible and useful.
Hosted by Carnegie Mellon University | November 2–13, 2026
The inaugural ProtoStar challenge invites participants to test whether AI can improve predictions of urban climate conditions using historical weather data from 210 U.S. cities. Building on research by Carnegie Mellon University, teams will use climate records through 2018 to predict actual conditions from 2019 to 2026, allowing their models to be evaluated against observed weather rather than hypothetical future scenarios.
Participants will forecast four indicators relevant to urban planning: average annual temperature, total annual precipitation, the hottest day of the year, and cooling degree days, a measure of demand for air conditioning. Their predictions will be compared against simple statistical baselines, including historical averages and linear trends, to assess whether AI approaches offer meaningful improvements in accuracy.
Beyond forecasting, participants will also propose a short climate adaptation policy idea informed by what they learned from the data. This component encourages teams to connect scientific analysis with practical public benefits, such as preparing cities for extreme heat, improving energy efficiency, or managing changing precipitation patterns.
Teams will submit a Hugging Face dataset with these outputs. Forecasting performance will be assessed through a public leaderboard measuring prediction errors against established baselines, while policy ideas will receive separate consideration. Final forecasting results will be published in early 2027, once the complete 2026 weather records are available.
Explore the CMU Climate Risk Intelligence Challenge
The CMU challenge is the first in this series. Other participating institutions will host their own challenges sequentially, each highlighting different publicly funded scientific resources and opportunities for reuse. Future challenges will be announced through the Project ProtoStar page on Hugging Face.
Whether you're an academic researcher, student, independent developer, or company building on open scientific infrastructure, ProtoStar offers opportunities to put publicly funded research to work and help us learn what is needed to make AI-ready open science more widely useful.
We also welcome universities, federal research agencies, national laboratories, nonprofits, and other research organizations interested in making their scientific resources more accessible and reusable. To explore participation or collaboration, reach out to protostar@orcaopen.org.
Our hope is that ProtoStar will make it easier for publicly funded scientific work to reach the people and tools that can build on it, while generating practical evidence about the infrastructure and practices needed to support open science at scale.
Auto-fetched and translated by AI Gem Box. All rights belong to the original author.