Papers with Code relaunches under Hugging Face with World Models, Agents, Robotics focus
Hugging Face has restored Papers with Code with benchmark tracking, metric data, and integration with Hugging Face Papers across three research domains.

Papers with Code is back online under Hugging Face management, bringing benchmark tracking and metric comparison tools to researchers focused on World Models, Agents, and Robotics.
Hugging Face acquired Papers with Code in 2023 but took the original site offline during migration. The restored version includes benchmark leaderboards for model comparison, metric datasets for each tracked domain, and direct integration with Hugging Face Papers for article discovery. Users can filter by research area and see which models lead on specific tasks.
The three focus areas reflect current priorities in open-source AI. World Models—systems that learn internal representations of environments—have seen renewed interest since the rise of video generation models. Agent research covers autonomous systems that plan and execute multi-step tasks. Robotics benchmarks track progress in embodied AI, where models control physical or simulated hardware.
Papers with Code originally launched as an independent project linking machine learning papers to their code implementations and benchmark results, becoming a standard reference for tracking state-of-the-art performance across computer vision, natural language processing, and reinforcement learning. The relaunch arrives as benchmark gaming—optimizing models specifically for leaderboard metrics rather than real-world performance—has become a recurring concern in AI research.