无限镜元
基础模型与学习范式
无限镜元构建“智能体＋环境”双向演化的世界模型：智能体行动改变环境，环境反馈又更新智能体能力。该闭环针对具身训练中交互数据稀缺与虚实衔接困难的问题。
MirroS models both the agent and its environment: actions change the world, and the resulting feedback updates the agent. This dual loop addresses scarce interaction data and the gap between simulated and real physical-AI training.
