AIXC Research / Open Source
Open Source

Open Science. Open Tools.

Benchmark datasets and evaluation tools for reproducible AI research. We believe reproducibility is the foundation of trustworthy science. If you cannot verify it, you should not trust it.

Datasets

Benchmark Datasets

Standardized evaluation datasets for reproducible comparison. Each dataset includes ground-truth labels, train/test splits, and baseline model performance.

AIXC-Mol
Molecular generation benchmark with binding affinity, synthesizability, and ADMET ground truth
142K molecules · 12 targets
AIXC-Fin
Financial causal factor benchmark with regime labels and interventional scenarios
8 markets · 15 years · 3 regimes
AIXC-Mat
Ceramic composition-property pairs with phase stability labels and thermodynamic constraints
47K compositions · 6 properties
AIXC-Causal
Cross-domain causal discovery with ground-truth DAGs, interventional data, and transfer pairs
200 DAGs · 4 domains · 50K samples
Research Tools
AIXC tools are built for researchers and engineers. We believe rigorous science accelerates discovery for everyone.
Contribute

How to Contribute

Five steps from first clone to merged pull request. We review all contributions within 48 hours and provide detailed feedback.

01
Fork & Clone
Fork the repo, clone locally, create a feature branch
02
Read the Docs
Review CONTRIBUTING.md and architecture documentation
03
Write Tests First
TDD approach: write failing tests, then implement
04
Submit PR
Open pull request with description, tests, and benchmarks
05
Review & Merge
48-hour review cycle with detailed feedback
Get Involved

Start Contributing

Join a growing community of researchers and engineers building the next generation of AI-driven scientific discovery tools.

GitHub Organization Read the Papers