Beyond generative AI: how to approach HORIZON-EIC-2026-PATHFINDERCHALLENGES-01-03

8th June 2026 at 12:26 pm

Artificial intelligence has made remarkable progress in recent years, particularly through large language models and generative AI. Yet despite their impressive capabilities, today’s systems still struggle with reasoning, abstraction and long-term planning. They can generate plausible answers, but often fail to explain their logic, adapt knowledge across contexts or reliably plan under uncertainty.

Inside the EIC Pathfinder Challenges 2026, the EIC Pathfinder Challenge HORIZON-EIC-2026-PATHFINDERCHALLENGES-01-03, also known as DeepRAP (Deep Reasoning, Abstraction & Planning), aims to address precisely these limitations. With four months left, now is the right time to start forming a competitive consortium for the 28 October 2026 deadline.

Why the EIC is looking beyond current AI

Current AI systems excel at recognising patterns from vast amounts of data. However, many remain fundamentally limited when confronted with situations that require causal understanding, common-sense reasoning, knowledge transfer across domains or adaptive decision-making.

The DeepRAP Challenge is based on the premise that future AI systems must move beyond statistical pattern matching and develop more advanced cognitive capabilities. The proposals should address one or more of the following cognitive capabilities:

• Deep Reasoning: understanding causal relationships, logic and complex decision processes
• Deep Abstraction: identifying higher-level concepts and transferring knowledge across contexts
• Deep Planning: setting goals, evaluating alternatives and adapting actions in dynamic environments

Importantly, proposals are not expected to tackle all three dimensions equally. However, applicants should clearly justify which capability or combination of capabilities they address and how these contribute to the development of next-generation cognitive AI.

Trustworthy cognitive AI: from principle to architecture

One of the most distinctive features of DeepRAP is its strong emphasis on trustworthiness as a central technical objective. The call expects projects to develop mechanisms that ensure:

• Explainability and transparency
• Fairness
• Risk evaluation
• Security and alignment with ethical and legal standards

This reflects broader European policy developments, including the AI Act and ongoing efforts to strengthen trustworthy AI ecosystems across Europe.

Applicants should therefore demonstrate not only how their systems perform, but also why users, regulators and society should trust their outputs. To this end, proposals should also propose new methods and metrics for evaluating and certifying reasoning and trustworthiness in AI.

Portfolio logic: where your proposal must fit

Like other Pathfinder Challenges, DeepRAP follows a portfolio-based approach. This means that proposals are evaluated not only on scientific excellence, but also on how they contribute to the overall diversity and coherence of the portfolio of funded projects. For example, two technically strong proposals addressing the same reasoning challenge with similar methodologies may compete for the same portfolio space. Conversely, a project that introduces a novel cognitive architecture, validation approach or application domain may strengthen portfolio diversity.

The final selection of projects will be made in a way ensuring maximum coverage across the following categories:

• Category 1 – Cognitive Function Capability: This category assesses which of the three core cognitive capabilities your proposal addresses: Deep Reasoning (DR), Deep Abstraction (DA), Deep Planning (DP), Multi-Capability Projects (MC).
• Category 2 – Technological Approach: Projects will be categorized based on their technological approach, such as Neuro-symbolic AI (NeSy), Advanced Deep Learning (ADL), Reinforcement Learning (RL), Cognitive Architectures (CA), Novel Interdisciplinary Frameworks (NIF), Formal Methods Integration (FMI), Multimodal Integration (MMI), Trustworthiness mechanisms Integration (TM), or others.
• Category 3 – Use Case and Application Domain: Projects will be mapped according to their primary real-world application focus, such as Scientific Discovery (SD), Decision Support Systems (DSS), Autonomous Systems (AS), Human-AI Collaboration (HAIC), Cybersecurity (CS), Industrial Applications (Ind), Social and Societal Applications (SSA).

Additionally, looks to the project’s capacity and strategy to engage in portfolio-wide collaboration and collective value creation, as required by the DeepRAP Challenge

• Category 4 – Synergy aspects: This category looks to the project’s capacity and strategy to engage in portfolio-wide collaboration and collective value creation, as required by the DeepRAP Challenge. Proposals will be categorised on specific dimensions of synergy, including: Benchmark Development (Ben), Interoperability (Int), Joint Pilots and Demonstrations (JPD).

The selected projects will also be assigned to lead and/or engage in portfolio-wide activities centred on the following priorities. The following figure summarises the portfolio building approach with eight expected selected projects:

Our Pathfinder track record

With years of hands-on experience in preparing successful proposals, we have become a trusted partner for ambitious researchers applying to the EIC Pathfinder programme. Our team at accelopment understands how to position high-risk research for competitive evaluation. Our track record extends back to the FET programme in Horizon 2020, the predecessor of EIC Pathfinder, where we contributed to pioneering projects such as CLASSY, FRINGE and Lumiblast. In Horizon Europe, we supported the preparation of successful EIC Pathfinder proposals, including CORENET, PEARL-DNA, POLINA, PIONEAR and BoneOscopy.

Whether you are preparing a first submission or considering a resubmission strategy, we provide tailored support across proposal positioning, impact development, consortium structuring and proposal writing. As in many of our successful collaborations, we can also participate as a project partner supporting communication, dissemination, exploitation and project management activities.

Our team is ready to help transform your breakthrough idea into a strong and compelling Pathfinder proposal. Contact us to discuss how our experience can strengthen your application.

If you want to be notified as soon as we publish any EIC-related news, you can subscribe to our blog posts and stay updated about the latest developments, trends, and results.

Dr. Johannes Ripperger
Research & Innovation Manager

Andreia Cruz
Research & Innovation Project Manager