Charting and Navigating Hugging Face's Model Atlas
Eliahu Horwitz, Nitzan Kurer, Jonathan Kahana, Liel Amar, Yedid Hoshen
2025-03-14
Summary
This paper talks about creating a detailed map of AI models on Hugging Face to help people find and understand them, even though most models are poorly documented.
What's the problem?
There are millions of AI models available, but it’s hard to navigate them because many lack clear descriptions or connections, making it tough to see how they relate or improve over time.
What's the solution?
The researchers built a map using known model relationships and trained it to guess missing connections by spotting patterns like how models are copied, merged, or tweaked, then shared it as an interactive tool.
Why it matters?
This helps developers reuse existing models instead of building new ones from scratch, saving time and resources while making AI tools easier to understand and improve.
Abstract
As there are now millions of publicly available neural networks, searching and analyzing large model repositories becomes increasingly important. Navigating so many models requires an atlas, but as most models are poorly documented charting such an atlas is challenging. To explore the hidden potential of model repositories, we chart a preliminary atlas representing the documented fraction of Hugging Face. It provides stunning visualizations of the model landscape and evolution. We demonstrate several applications of this atlas including predicting model attributes (e.g., accuracy), and analyzing trends in computer vision models. However, as the current atlas remains incomplete, we propose a method for charting undocumented regions. Specifically, we identify high-confidence structural priors based on dominant real-world model training practices. Leveraging these priors, our approach enables accurate mapping of previously undocumented areas of the atlas. We publicly release our datasets, code, and interactive atlas.