MusGO is composed of 13 categories: 8 essential and 5 desirable. Essential categories follow an openness-graded scale of three levels: closed (🔴), partial (🟠) and fully open (🟢). Instead, desirable categories are binary, indicating whether the element exists (⭐) or not. Framework criteria is detailed here.

Open source code • Training data • Model weights • Code documentation • Training procedure • Evaluation procedure • Research paper • Licensing

Model card • Datasheet • Package • User-oriented application • Supplementary material page

To help expand or refine the MusGO leaderboard, you can contribute by:

• Adding a new model to the leaderboard.
• Proposing updates or modifications to the evaluation of exisiting models.

To get started, head over to our How to contribute? page. It includes:

• Instructions for submitting models via pull request
• Guidelines for proposing changes or opening issues
• A checklist to support evidence-based assessments

Your feedback helps us maintain fair and up-to-date openness assessments. Thank you for contributing to the improvement of MusGO!

We’d love to hear your feedback, questions, or suggestions for improving the MusGO framework and the evaluation process. We encourage you to open up issues to:

• Flag a particular assessment in the current evaluations.
• Suggest improvements to the evaluation framework.
• Ask questions about how to interpret the categories.

You can also contact us directly at roser.batlle@upf.edu and laura.ibanez@upf.edu.

The MusGO framework is a living resource, developed through community collaboration, currently focused on assessing openness in music-generative AI. However, we are actively exploring complementary perspectives and refinements to further expand its scope and adaptability. We aim to better reflect the diverse ways in which music-generative systems can be understood, accessed, and used responsibly.

Updates will be shared once ready for community feedback.

The article MusGO: A Community-Driven Framework For Assessing Openness in Music-Generative AI was accepted at the 26th International Society for Music Information Retrieval (ISMIR) Conference. If our work is relevant to you, please cite it as follows:

@inproceedings{batlleroca2025musgo,
  author       = {Roser Batlle{-}Roca and Laura Ib{\'{a}}{\~{n}}ez{-}Mart{\'{\i}}nez and
                  Xavier Serra and Emilia G{\'{o}}mez and Mart{\'{\i}}n Rocamora},
  title        = {MusGO: {A} Community-Driven Framework for Assessing Openness in Music-Generative {AI}},
  booktitle    = {Proceedings of the 26th International Society for Music Information Retrieval Conference, {ISMIR} 2025, Daejeon, South Korea, September 21-25, 2025},
  pages        = {727--738},
  year         = {2025},
  url          = {https://doi.org/10.5281/zenodo.17706575},
  doi          = {10.5281/ZENODO.17706575},
  biburl       = {https://dblp.org/rec/conf/ismir/Batlle-RocaISGR25.bib},
}

This repository is an adapted version of Opening up ChatGPT: tracking openness of instruction-tuned LLMs. We are deeply grateful to the original creators, Andreas Liesenfeld, Alianda Lopez, and Mark Dingemanse, for their groundbreaking work on openness, transparency, and accountability in generative AI, which has inspired and shaped this project.

For more details, please refer to their papers:

• Liesenfeld, Andreas, Alianda Lopez, and Mark Dingemanse. 2023. “Opening up ChatGPT: Tracking Openness, Transparency, and Accountability in Instruction-Tuned Text Generators.” In CUI '23: Proceedings of the 5th International Conference on Conversational User Interfaces, July 19-21, Eindhoven. DOI: 10.1145/3571884.3604316.
• Andreas Liesenfeld and Mark Dingemanse. 2024. “Rethinking open source generative AI: open washing and the EU AI Act.” In The 2024 ACM Conference on Fairness, Accountability, and Transparency (FAccT '24). DOI: 10.1145/3630106.3659005.

This work has been supported by IA y Música: Cátedra en Inteligencia Artificial y Música (TSI-100929-2023-1), funded by the Secretaría de Estado de Digitalización e Inteligencia Artificial and the European Union-Next Generation EU, and IMPA: Multimodal AI for Audio Processing (PID2023-152250OB-I00), funded by the Ministry of Science, Innovation and Universities of the Spanish Government, the Agencia Estatal de Investigación (AEI) and cofinanced by the European Union.

We also thank our colleagues at the Music Technology Group at Universitat Pompeu Fabra for their thoughtful insights, constructive discussions and active engagement throughout the development of this work.