【#Tech24H】The Zhiyuan Embodied AI Research Center has launched its second-generation integrated embodied system, GenieReasoner. To address the cross-modal alignment challenges between semantic reasoning and action control in Vision-Language-Action models, the system proposes a model architecture that supports unified discretized pre-training and utilizes flow matching techniques to alleviate the action precision limitations of traditional discrete tokenizers. The research center has concurrently open-sourced the Embodied Reasoning Intelligence Quotient benchmark, ERIQ, designed to decouple and quantitatively evaluate the reasoning capabilities of the embodied brain in VLAs. Experimental results show that GenieReasoner achieves strong performance in reasoning tasks and demonstrates generalization capabilities in cross-system real-world robot experiments.