FluidsBench is an open benchmark for evaluating machine-learning models on realistic fluid dynamics problems. It brings together public datasets, shared evaluation tools and leaderboards to compare prediction accuracy and computational cost across models. This repository provides the tools and instructions to evaluate your model and submit results for review.
Evaluate your model with the published FluidsBench tools and submit its results for review.
- Choose a dataset: use the dataset index to find its quickstart, supported splits and submission availability.
- Follow the submission process: Submitting a result explains evaluation, validation and opening a pull request.
- Prepare your result package: follow the result format and model and compute reporting guide.
- Get help: ask in GitHub issues or the FluidsBench community.
- Preview the leaderboard: explore the metrics and CFD profile viewer. The preview contains no submitted results; publication follows review and approval.
Submit the result directory generated by your dataset's evaluator: metrics, profiles, evaluation evidence and model/compute records. Code, model weights and full prediction fields are optional. See the reproducibility requirements for details.
Use test data only for final evaluation. Maintainers review the package before publishing the result.
The former README sections now live in the submission workflow, package and version reference, and maintainer guide.
Repository code and leaderboard metadata use Apache 2.0. Submitters must have the right to publish their result material under its declared open licence. Upstream datasets and optional artifacts retain their own licences; allowed SPDX identifiers are enforced by the schema.
| Name | Affiliation |
|---|---|
| Neil Ashton | Mistral |
| Paola Cinnella | Sorbonne University |
| Astrid Walle | Pasteur Labs |
| Mohamed Elrefaie | MIT |
| Jean Kossaifi | NVIDIA |
| Ricardo Vinuesa | University of Michigan |
| Daniel Leibovici | NVIDIA |
| Faez Ahmed | MIT |
| Emanuele Tamburini | Sorbonne University |
| Tyler Buchanan | TU Delft |
Meet the scientific and industrial advisory board.
Visit the FluidsBench website for dataset information, benchmark news and community links.