Hugging Face Machine Learning Demos on arXiv

Hugging Face now embeds live Spaces demos directly on arXiv paper pages. Readers can run inference against a model without leaving the abstract.

For teams evaluating models for RAG or agentic pipelines, this cuts the paper-to-test loop from hours to minutes. Most developers still clone repos and configure local environments before deciding if a model fits — that's wasted time for initial filtering. Cloning before disqualifying is backwards.

ML engineers doing model selection for production use this to disqualify fast. Teams locked into closed APIs like OpenAI or Anthropic can skip it.