# Contributing

Thanks for helping improve UFO-MNIST.

Useful contributions:

- loaders for additional ML frameworks
- reproducible benchmark scripts
- visualization notebooks
- documentation fixes
- source-manifest improvements

Benchmark submissions should include:

- model architecture
- preprocessing
- random seed
- number of epochs or iterations
- exact train/test split
- test accuracy and macro F1
- code needed to reproduce the result