1. The Closure Challenge:
We introduce a field-wide benchmark challenge for machine learning in Reynolds-averaged Navier-Stokes (RANS) turbulence modelling. Though open-source datasets exist for training data-driven turbulence closure models, the field has been notably lacking a standard benchmark metric and test dataset. The Closure Challenge is a curated collection of open-source datasets and evaluation code that remedies this problem. We provide a variety of high-fidelity training data in a standardized format, including mean velocity gradients. The test cases (periodic hills, square duct, and NASA wall-mounted hump) evaluate Reynolds number and geometry generalization, two key issues in the field. We present results from three early submissions to the challenge. This is an ongoing challenge, intended to continuously spur innovation in machine learning for turbulence modelling. Our goal is for this benchmark to become the standard evaluation for new machine learning frameworks in RANS.

The Closure Challenge is available at github.com/rmcconke/ml-turbulence-benchmark.
An accompanying paper has been posted on ArXiv, and you can refer to it when you cite the benchmark: 10.48550/arXiv.2603.28884

The challenge is a continuously ongoing event. You can join at any time.

Key Dates:
• October 2025: Dedicated website launch with complete details
• January 2026: Initial submissions due
• March 2026: Challenge progress updates at ML4Fluids Conference (4-6 March @ CWI Amsterdam)
• Post-March 2026: Dedicated online event for results (TBD)

Organizers: Tyler Buchanan, Paola Cinnella, Richard Dwight, and Ryley McConkey
To participate: Send an expression of interest to t.s.b.buchanan@tudelft.nl

2. The CYPHER Data Challenge: Machine-Learning-Enhanced Turbulent Combustion Closures for LES

The CYPHER data challenge focuses on benchmarking machine-learning approaches for modelling sub-filter closure terms in Large Eddy Simulation (LES) of turbulent lean premixed hydrogen flames. Using high-fidelity Direct Numerical Simulation (DNS) datasets, participants will develop and evaluate machine-learning models to predict unresolved quantities arising from turbulence–chemistry interaction, such as the sub-filter scalar flux of the filtered progress variable. The challenge aims to promote reproducible comparison of different model architectures while encouraging solutions that balance predictive accuracy and computational efficiency. Training and test datasets are derived from DNS of lean premixed hydrogen flames filtered at multiple resolutions. Model evaluation is performed automatically on the Codabench platform and considers both prediction accuracy and inference cost. The challenge is part of the CYPHER COST Action, which promotes collaboration between researchers and industry to accelerate the development of digital tools for renewable-fuel combustion technologies. Visit the dedicated challenge repository for datasets, problem specifications, and submission instructions.

Organizers: Pasquale Lapenna, Lorenzo Piu, Antonio Attili, Alessandro Parente, Paola Cinnella, Federica Ferraro, Anh Khoa Doan
Slides and links:apriori.readthedocs.io/en/latest/conferences/ML4Fluids2026.html
To participate (1st part):codabench.org/competitions/9173
To participate (2nd part): Send an expression of interest to pasquale.lapenna@uniroma1.it

3. The FLUIDSBENCH CFD emulator challenge

FluidsBench is a benchmark for Computational Fluid Dynamics (CFD) surrogates, designed to accelerate progress in the development of foundational AI models for fluids. Motivated by similar efforts in weather (WeatherBench 2) and early work on task specific efforts (CarBench), FluidsBench consists of an open-source evaluation framework, training and ground truth data available via external model hubs (e.g., HuggingFace), and a continuously updated website hosting the latest metrics and state-of-the-art leaderboards that will allow for testing of AI surrogate models. In-person and virtual workshops will be held (subject to acceptance) at popular fluids and ML events (e.g NeurIPS, ICML, ML4Fluids) to discuss the latest work and get community direction for this benchmarking effort. Please visit the website to signup to the mailing list to be notified when it'll be ready for submissions.

More details about the test cases, tasks, and challenge metrics will be published soon. STAY TUNED! Do not hesitate to reach out if you wish to contribute.

Organizers: Neil Ashton, Paola Cinnella, Richard Dwight, Astrid Walle, Ricardo Vineusa, Jean Kossaifi, Daniel Leibovici
To participate: visit fluidsbench.org to sign up to the mailing list and/or e-mail admin@fluidsbench.org