RynnBrain: Open Embodied Foundation Models
🤖 AI Summary
This work addresses the absence of a unified, physically grounded multimodal foundation model for embodied intelligence, which hinders coherent perception, reasoning, and planning in real-world spatiotemporal dynamics. To this end, we propose the first unified embodied foundation model architecture that integrates four core capabilities: egocentric understanding, multi-scale spatiotemporal localization, physics-grounded reasoning, and physics-aware planning. The model employs a multi-scale Mixture-of-Experts (MoE) structure (2B/8B/30B-A3B) and task-customized post-training strategies, enabling strong performance across diverse downstream tasks—including navigation, vision-language-action (VLA) tasks, and complex spatial reasoning. Evaluated on 20 embodied benchmarks and 8 general visual understanding benchmarks, our model significantly outperforms existing approaches, demonstrating its effectiveness and adaptability as a general-purpose pretrained backbone for embodied AI.
Application Category
📝 Abstract
Despite rapid progress in multimodal foundation models, embodied intelligence community still lacks a unified, physically grounded foundation model that integrates perception, reasoning, and planning within real-world spatial-temporal dynamics. We introduce RynnBrain, an open-source spatiotemporal foundation model for embodied intelligence. RynnBrain strengthens four core capabilities in a unified framework: comprehensive egocentric understanding, diverse spatiotemporal localization, physically grounded reasoning, and physics-aware planning. The RynnBrain family comprises three foundation model scales (2B, 8B, and 30B-A3B MoE) and four post-trained variants tailored for downstream embodied tasks (i.e., RynnBrain-Nav, RynnBrain-Plan, and RynnBrain-VLA) or complex spatial reasoning tasks (i.e., RynnBrain-CoP). In terms of extensive evaluations on 20 embodied benchmarks and 8 general vision understanding benchmarks, our RynnBrain foundation models largely outperform existing embodied foundation models by a significant margin. The post-trained model suite further substantiates two key potentials of the RynnBrain foundation model: (i) enabling physically grounded reasoning and planning, and (ii) serving as a strong pretrained backbone that can be efficiently adapted to diverse embodied tasks.
Problem
Research questions and friction points this paper is trying to address.
embodied intelligence
foundation model
spatiotemporal dynamics
physically grounded reasoning
perception-reasoning-planning integration
Innovation
Methods, ideas, or system contributions that make the work stand out.
embodied intelligence
spatiotemporal foundation model
physically grounded reasoning
physics-aware planning
egocentric understanding
🔎 Similar Papers
No similar papers found.
