Comment by Sol-

Thanks for the links, very interesting.

Wonder how a "self-play" equivalent would look like for LLMs, since they have no easy criterion to evaluate how well they are doing like in Go (as mentioned in the videos).