View: Robotic Models to Be Trained by LLMs via Real2Sim RL

YouJiacheng · x · 2026-08-22

A viewpoint suggests that many robotic models will be trained via LLM-driven real2sim (real-to-simulation) reinforcement learning, with the design process itself also directed by LLMs. This highlights the potential central role of LLMs in data generation and architectural design for embodied intelligence.

Original post →

More from Embodied

Embodied channel →