AI Foundation models in science (GenAI4EU) (RIA)

Topic updates

23 September 2025

15 September 2025

Reminder on proposal part B page limit and formatting conditions

1. Proposal part B page limit

The title, list of participants and sections 1, 2 and 3, together, should not be longer than page limit, indicated in a proposal part B template inside the Submission System. All tables, figures, references and any other element pertaining to these sections must be included as an integral part of these sections and are thus counted against this page limit.

The page limit will be applied automatically. At the end of a proposal part B template inside the Submission System you can see the structure of the actual proposal that you need to submit, please remove all instruction pages that are watermarked.

If you attempt to upload a proposal longer than the specified limit before the deadline, you will receive an automatic warning and will be advised to shorten and re-upload the proposal. After the deadline, excess pages (in over-long proposals/applications) will be automatically made invisible, and will not be taken into consideration by the experts. The proposal is a self-contained document. Experts will be instructed to ignore hyperlinks to information that is specifically designed to expand the proposal, thus circumventing the page limit.

Please, do not consider the page limit as a target! It is in your interest to keep your text as concise as possible, since experts rarely view unnecessarily long proposals in a positive light.

2.Proposal part B formatting conditions

The following formatting conditions apply (as listed in a proposal part B template inside the Submission System) and will be checked by the Agency during an admissibility check of submitted proposals.

The reference font for the body text of proposals is Times New Roman (Windows platforms), Times/Times New Roman (Apple platforms) or Nimbus Roman No. 9 L (Linux distributions).

The use of a different font for the body text is not advised and is subject to the cumulative conditions that the font is legible and that its use does not significantly shorten the representation of the proposal in number of pages compared to using the reference font (for example with a view to bypass the page limit).

The minimum font size allowed is 11 points. Standard character spacing and a minimum of single line spacing is to be used. This applies to the body text, including text in tables.

Text elements other than the body text, such as headers, foot/end notes, captions, formula's, may deviate, but must be legible.

The page size is A4, and all margins (top, bottom, left, right) should be at least 15 mm (not including any footers or headers).

Proposal part B template inside the Submission System document is tagged. Do not delete the tags; they are needed for our internal processing of information, mostly for statistical gathering. In that light, please do not move, delete, re-order, alter tags in any way, as they might create problems in our internal processing tools. Tags do not affect or influence the outcome of your application.

10 June 2025
Please note that due to a technical issue, during the first days of publication of this call, the topic page did not display the description of the corresponding destination. This problem is now solved. In addition to the information published in the topic page, you can always find a full description of destinations:

* destination 1: Achieving global leadership in climate-neutral, circular and digitised industrial and digital value chains;

* destination 2: Achieving technological leadership for Europe's open strategic autonomy in raw materials, chemicals and innovative materials;

* destination 4: Achieving open strategic autonomy in digital and emerging enabling technologies;

* destination 6: Digital and industrial technologies driving human-centric innovation,

that are relevant for the call in the Work Programme 2025 part for “Industry”. Please select from the work programme the destination relevant to your topic and take into account the description and expected impacts of that destination for the preparation of your proposal.

AI Foundation models in science (GenAI4EU) (RIA)

TOPIC ID: HORIZON-CL4-INDUSTRY-2025-01-DIGITAL-61

Type of grant: Call for proposals

General information

Programme:

Call: INDUSTRY (HORIZON-CL4-2025-01)

Type of action: HORIZON-RIA HORIZON Research and Innovation Actions

Type of MGA: HORIZON Lump Sum Grant [HORIZON-AG-LS]

Status: Forthcoming

Deadline model: single-stage

Planned Opening Date: 22 May 2025

Deadline dates: 23 September 2025 17:00 (Brussels time)

Topic description

Expected Outcome:

• Accelerate research and development in science, with focus on the domains of a) materials science, b) climate change science, c) environmental pollution science (including PFAS) and d) agricultural science ;
• Advance AI technology (not limited to Generative AI) tailored for scientific needs and potentially adaptable to other tasks in the area of application;
• Contribute to the development of foundation models in the areas of application, and pave the way for future funding of foundation models in a broader range of scientific disciplines;
• Advance solutions to societal or scientific challenges;
• Bridge existing knowledge gaps and induce interdisciplinarity by design across different fields necessary to advance the area of application; and
• Support open-source and open science, especially for research communities with limited access to modern AI tools.

Scope:

Foundation models in science are an evolving idea in the scientific community and go beyond the Generative AI trend[[Some examples in science include: Foundation model in materials science ([2401.00096] A foundation model for atomistic materials chemistry (arxiv.org), Helmholtz Foundation Models Initiative ( Helmholtz Foundation Model Initiative - Helmholtz Home), The Trillion Parameter Consortium (https://www.anl.gov/article/new-international-consortium-formed-to-create-trustworthy-and-reliable-generative-ai-models-for), NASA (NASA and IBM Openly Release Geospatial AI Foundation Model for NASA Earth Observation Data | Earthdata), the University of Michigan (Scientific Foundation Models (scifm.ai))]]. The purpose of this topic is to tap into their potential, and to advance the development of AI technology specifically tailored for the needs of science.

A foundation model^[1] can integrate information from various modalities of data. This model can then be adapted to a wide range of downstream, more specialized tasks. To build downstream applications, the foundation model is fine-tuned with additional training and task-specific examples. Therefore, a foundation model is itself incomplete but serves as the common basis from which many task-specific models can be built via adaptation.

In science, such foundation models could be trained on data from a specific scientific field and then be fine-tuned for a variety of tasks and used by a wider community in the field.

Proposals should address one of the following scientific domains:

• (A) Materials science: the development of new, innovative and advanced materials is essential for EU’s economic security and for achieving a competitive and sustainable industry (especially sectors such as energy, mobility, construction, health and electronics). Employing AI in the process of materials design, characteristics and discovery could significantly accelerate and scale potential innovative solutions.
• (B) Climate change science: advancing climate research is critical for achieving the EU's climate neutrality and resilience goals. AI foundation models can contribute to more accurate insights into climate dynamics, enhanced predictions of extreme weather events, regional impacts and the evolution of climate tipping points.
• (C) Environmental pollution sciences: advancing environmental sciences can support the detection and characterisation of pollution sources, as well as their pathways, distribution and impacts to the environment and human health. This is particularly relevant in the case of pollutants of concern, emerging and/or less known pollutants.
• (D) Agricultural sciences: advancing agricultural sciences research is critical to achieve a competitive, resilient and sustainable agricultural system. AI foundation models can contribute to enhance crop, livestock, soil and water management.

Proposals should focus on 1) developing foundation models (not limited to Generative AI) for science in the chosen domain; 2) showing a foundation model’s usefulness by adapting it to subtasks/scientific problems in the chosen domain; and 3) illustrating other possible areas of application.

The foundation models should provide researchers with access to essential AI-enabled capabilities for scientific discovery; employ the machine learning algorithms, models and architectures best suited for the chosen domain; be adaptable to different problems in the domain^[2]; and be based on a robust and reliable architecture, as any potential errors and problems would be propagated to the downstream applications.

The foundation models should be placed at the disposal of the scientific community as open models, including the source code and, where possible, training datasets and other associated assets needed for full reusability of the foundation models (unless justified otherwise). This will serve a wider scientific community, thus broadening access to such scientific infrastructure and facilitating the use and adaptation of the model to different problems. Proposers should provide a clear documentation on the use and limitations of the model, alongside case studies demonstrating the model's application to a variety of tasks/problems in the chosen domain.

Multidisciplinary research activities should involve both AI and domain scientists, and address some of the following:

• Conceptualisation and planning: the scope, objectives and expected outcomes of the foundation model;
• Suitable interfaces for domain experts without computer science background to contribute to and utilise the outcomes;
• Data identification, collection and management of (preferably diverse, multimodal) datasets through semantically annotation data schemas;
• Model development, validation, testing under relevant operational and environmental conditions (such as thermal gradients, fatigue, corrosion, etc.) and, as appropriate, model evaluation and benchmarking, for example DOME^[3];
• Integration of domain knowledge into the model (for example through machine readable representations like RDF (Resource Description Framework).

Proposals should:

• Prove access to high quality (multimodal) data needed for the development of the model. If in the process of developing the model, there is a need to create new data sets or adapt existing ones, they should follow the FAIR^[4] principles. Describe the data curation and quality control procedures that will be used to ensure the accuracy, completeness, and consistency of the training data.
• Contribute to efforts to reach common standards for data formats, metadata, taxonomies and ontologies.
• Demonstrate a strategy^[5] to access the computational resources needed for model training, evaluation/testing and inference.
• Propose a model architecture that is designed with transparency in mind
• Ideally, employ methodologies for integrating domain/interdisciplinary knowledge into the model and seek synergies with solutions that facilitate the managing and making sense of vast amounts of data (for example knowledge graphs).
• Identify at least four possible use cases and scientific challenges that can be addressed with the model and its adaptations.^[6]
• Identify and assess the potential risks of misuse of the foundation model.
• Propose a plan to make the model public, maintain and evolve it and promote it to the scientific community on a regular basis, in order to give visibility to the concept, discuss key findings and anticipate the technology evolution – possibly in synergy with other relevant projects.

Proposals should involve expertise in Social Sciences and Humanities (SSH), in the cases where legal and ethical experts should be involved to address data privacy, sharing agreements, and compliance with regulations.

Synergies with the selected projects from HORIZON-INFRA-2025-01-EOSC-06: Using Generative AI (GenAI4EU) for Scientific Research via EOSC are encouraged, where relevant. Proposals are encouraged to collaborate with established infrastructures such as the WeatherGenerator^[7] project.

International cooperation is encouraged, where the EU has reciprocal benefit, like the Trillion Parameter Consortium.^[8]

In this topic the integration of the gender dimension (sex and gender analysis) in research and innovation content is not a mandatory requirement.

[1] Foundation models is a term defined by the Center of Research on Foundation Models of Stanford University in: “On the Opportunities and Risks of Foundation Models”, https://arxiv.org/pdf/2108.07258.pdf

[2] An example in materials science, for inspiration only: *2401.00096.pdf (arxiv.org)

[3] https://dome-ml.org/

[4] Findable Accessible Interoperable Reusable data.

[5] In case the project plans to use the EuroHPC network, the EU-funded project EPICURE offers an application support service for EuroHPC: Epicure - European Commission (europa.eu)

[6] For materials science, examples include, but are not limited to: (for materials science) alternatives to hazardous materials like PFAS, materials that lower environmental footprint, materials for quantum technology, for higher capacity batteries, for more efficient photovoltaic devices, etc.; (for climate science) enhanced prediction of climate and weather extremes, early warning systems, forecasting of climate-driven migration, and monitoring of the global carbon budget, monitoring and measuring adaptation effectiveness; (for environmental pollution sciences) solutions for the detection and assessment of pollution, including pollutants of emerging concern; (for agricultural sciences) enhanced prediction of impact of plant pests, monitoring of animal health and welfare, monitoring of soil health or of water management in agriculture.

[7] https://weathergenerator.eu/

[8] Ref. Trillion Parameter Consortium (TPC) - Generative AI for Science and Engineering

Start submission

The submission system is planned to be opened on the date stated on the topic header.

Get support

Please read carefully all provisions below before the preparation of your application.

Online Manual is your guide on the procedures from proposal submission to managing your grant.

Horizon Europe Programme Guide contains the detailed guidance to the structure, budget and political priorities of Horizon Europe.

Funding & Tenders Portal FAQ – find the answers to most frequently asked questions on submission of proposals, evaluation and grant management.

Research Enquiry Service – ask questions about any aspect of European research in general and the EU Research Framework Programmes in particular.

National Contact Points (NCPs) – get guidance, practical information and assistance on participation in Horizon Europe. There are also NCPs in many non-EU and non-associated countries (‘third-countries’).

Enterprise Europe Network – contact your EEN national contact for advice to businesses with special focus on SMEs. The support includes guidance on the EU research funding.

IT Helpdesk – contact the Funding & Tenders Portal IT helpdesk for questions such as forgotten passwords, access rights and roles, technical aspects of submission of proposals, etc.

European IPR Helpdesk assists you on intellectual property issues.

CEN-CENELEC Research Helpdesk and ETSI Research Helpdesk – the European Standards Organisations advise you how to tackle standardisation in your project proposal.

The European Charter for Researchers and the Code of Conduct for their recruitment – consult the general principles and requirements specifying the roles, responsibilities and entitlements of researchers, employers and funders of researchers.

Partner Search help you find a partner organisation for your proposal.