Open Intelligence

Open Intelligence brings together the lab’s contributions to open AI research, practical tools, and model licensing through collaboration with Matt White and colleagues across the Linux Foundation, PyTorch Foundation, University of Oxford, IBM, and the wider open-source community.

Model Openness Framework (MOF)

The Model Openness Framework evaluates the completeness and openness of machine-learning model releases across code, data, model artifacts, and documentation. Its three-tier classification supports transparency, reproducibility, and informed model selection. Xiao-Yang Liu is a co-author of the original framework paper with Matt White and collaborators.

Model Openness Tool (MOT)

The Model Openness Tool provides a practical reference implementation of the framework, helping the community document released components, assess their licenses, and classify model openness.

Model Openness Framework and Tool · MOT GitHub Repository

OpenMDW License

OpenMDW complements this work with a model-focused permissive license for released models and related code, data, and documentation. It clarifies the terms for using and sharing the materials included in a model distribution. Licensing and MOF classification address complementary questions: OpenMDW governs released materials, while MOF assesses the completeness and openness of the release.

OpenMDW Project · License Versions

Collaboration and Open Financial Intelligence

This collaboration provides a foundation for the lab’s Open Financial Intelligence projects, connecting openness and reproducibility with financial models, benchmarks, and agents. Matt White is now Founding Executive Director of the Open Intelligence Foundation and a Visiting Scholar at Columbia University. The Foundation advances open-source AI, open models, and open science.

Open Intelligence Foundation · Matt White’s Current Biography · MOF Specification

Research and Media

• Matt White, Ibrahim Haddad, Cailean Osborne, Xiao-Yang Yanglet Liu, Ahmed Abdelmonsef, Sachin Varghese, and Arnaud Le Hors. The model openness framework: Promoting completeness and openness for reproducibility, transparency, and usability in artificial intelligence. arXiv preprint arXiv:2403.13784 (2024).  Covered by the New York Times article "OpenWashing"