Future RL training will incorporate current-generation AI systems to provide reward signals for next-generation AI

This is our current best-guess model of future AI development, and so the role of debate is to ensure that the LLM judgements used during training provide as accurate a reward signal as possible.

https://www.alignmentforum.org/posts/BB8o7b8A4Aykeksvw/debate-training-reduces-reward-hacking-in-rlaif