Research
Hugging Face paper studies scaling continued pre-training for embodied agents across virtual worlds
Researchers ran 120 continued pre-training configurations with models from 0.8B to 27B and 50 to 1,600 hours of data. They report the models are competitive with Minecraft specialists and frontier models in 3D worlds, but gaps remain in distant worlds.
Every outlet that ran it 1 article
Hugging Face PapersResearchLead
Scaling in Games: Continued Pre-Training For Embodied Agents In Diverse Virtual Worlds