Johnyhar

If I focus on Y & Z until the result is passable, do I then do the same for X? Or do I try hard with X (time which I would enjoy), even though it won't be appreciated? At that point why not work on a personal project which would give me satisfaction, instead of working.

Wouldn't RL training, with the goal of aligning the LLM with the reward function R(x, y), result in the outputs of the trained LLM maximizing said reward function? How different are the rewards of the N outputs in BoN sampling, to justify its cost.