On: LLMs can't self-correct in reasoning tasks, DeepMind study finds

[tpush]

I mean, there's plenty of other meds whose exact precise functioning we don't know, just what results and side effects it has.