Community Evals: Because we're done trusting black-box leaderboards over the community

Current AI model evaluation is facing challenges, as high scores on benchmarks like MMLU do not always translate to real-world capability. To address this, Hugging Face is launching "Community Evals," a decentralized and transparent reporting system on its Hub. This feature allows dataset repositories to host leaderboards, models to store their own evaluation scores, and any user to submit results for a model via a pull request. The goal is to make the evaluation process more visible and reproducible by exposing what was evaluated, how, when, and by whom.