World Models as an Intermediary between Agents and the Real World

Sherry Yang

Large language model (LLM) agents trained using reinforcement learning has achieved superhuman performance in low-cost environments like games, mathematics, and coding. However, th…