Arash Ahmadian on Rethinking RLHF
TalkRL: The Reinforcement Learning Podcast
00:00
Exploring Reward Structures and Optimization Techniques in Reinforcement Learning
The chapter delves into discussions around reward structures in reinforcement learning, comparing different optimization techniques like banded formulation and vanilla policy gradient. It explores the impact of these discussions on preference training at a specific company and highlights the importance of curating reward signals for better optimization in RLHF.
Transcript
Play full episode
