Recent large model releases are reminding us that addressing fairness and bias in machine learning models is more important than ever. To help with this, we're releasing a new Hugging Face library: 🤗 Disaggregators!

Demo: https://huggingface.co/spaces/society-ethics/disaggregators

Library: https://github.com/huggingface/disaggregators

The demo shows examples of disaggregations with the LAION, MedMCQA (image attached), and The Stack datasets, along with handy code snippets.

This is in early development, so keep an eye out for more coming soon!