Sheng-Hao Yuan

Towards Automated Reinforcement Learning: Applications and Prospects of Large Language Models as Cognitive Components in the Full RL Lifecycle

: Deep reinforcement learning (DRL) has made significant progress in recent years. However, it still faces multiple bottlenecks in real-world applications. Agents suffer from low sample efficiency, and designing effective reward functions remains very difficult. Furthermore, traditional DRL lacks general cognitive abil...