Caleb Biddulph comments on Why not train reasoning models with RLHF?

[Caleb Biddulph]

Yeah, but there are probably other interesting takeaways