Embodied-R1.5: Evolving Physical Intelligence via Embodied Foundation Models

Embodied-R1.5 is a new Embodied Foundation Model (EFM) featuring 8 billion parameters and achieving state-of-the-art results on 16 out of 24 embodied visual language model benchmarks. It incorporates a Planner-Grounder-Corrector (PGC) framework for autonomous execution and self-correction in long-horizon tasks, supported by a data system of over 15 billion tokens and a multi-task balanced reinforcement learning approach. This model's ability to generalize across real-robot tasks and its availability as open-source resources, including weights and an evaluation framework, provide valuable tools for practitioners developing applications in physical intelligence and embodied AI.