Pierluca D'Oro and Martin Klissarov
TalkRL: The Reinforcement Learning Podcast
00:00
Challenges of Training an RL Agent for NetHack
The chapter discusses the difficulties of training a reinforcement learning agent to play NetHack, a text-based game. The speakers explain the challenges of interpreting NetHack for a language model and describe their approach of training a reward model and separately training the RL agent to bridge the gap between the game and the language model.
Transcript
Play full episode
