The Helmholtz Association has a long-standing mission to address grand societal and scientific challenges in domains such as health, energy, climate, aerospace, transport, and materials science. As artificial intelligence (AI) continues to reshape scientific discovery, it is crucial to ensure that AI advancements are effectively harnessed to accelerate progress in these fields. One key mechanism for achieving this is through well-designed benchmarking datasets that define AI tasks in Helmholtz-relevant domains, motivating AI researchers to contribute solutions that directly impact scientific progress.
Challenge: The Need for Benchmarking in Reproducible and Trustworthy Science.
Helmholtz is uniquely positioned to address this gap. With vast, high-quality datasets spanning multiple scientific fields, it can pioneer cross-domain and multimodal benchmarks, enabling more comprehensive AI evaluations. By leveraging its data assets, Helmholtz can drive groundbreaking advancements and position itself as a leader in the emerging science of benchmarking - particularly in supporting foundation models designed for multimodal AI. To address this challenge, and to initiate a community of practice for benchmarking, and to support these first systematic benchmarking activities, the Helmholtz Association supports the hyperUNLOCK Initiative with this first call UNLOCK for benchmarking projects.
In this information event, we will present and explain the concept of the call, discuss the different objectives of the call and the selection criteria, and present the expected activities in 2026.
