Hugging Face has announced a collaboration with arXiv to make research papers more accessible and interactive. Starting today, Hugging Face Spaces is integrated with arXivLabs via a new Demo tab on arXiv paper pages, providing links to open-source demonstrations built by the community or the authors themselves.
Since its launch in October 2021, Hugging Face Spaces has hosted over 12,000 open-source machine learning demos. These demos, built with tools like Gradio and Streamlit, allow users to explore models and datasets without writing code. The integration means that on a paper's arXiv page, users can find and try demos that showcase the paper's contributions.
For example, the BERT paper's arXiv page now links to over 200 demos, ranging from simple model showcases to complex applications. This feature aims to broaden access to machine learning research, improve reproducibility, and allow diverse audiences to understand and critique models. Hugging Face is excited to see how the research community leverages this integration for better communication and interpretability.