News for hackers. Stuff that matters.

A comment on LLMs can't self-correct in reasoning tasks, DeepMind study finds. Parent comment

Re:LLMs can't self-correct in reasoning tasks, DeepMind stud... by robertlagrant on Monday October 09, 2023 @04:44PM (#37825260)
> Generally, people in HN

These silly generalisations about HN are pointless. HN has a pretty diverse set of worldviews, approaches, and levels of expertise across a range of topics. There is zero point in lumping people into a single, or even a majority, behaviour.

[ Parent | on HN ]
Re:LLMs can't self-correct in reasoning tasks, DeepMind stud... by ryanklee on Monday October 09, 2023 @04:50PM (#37825333)
Yeah, disagree. I read most threads on these topics, and although I can't characterize people that don't post, I can certainly attest to the generality I'm pointing out. I haven't seen any discussions that defy it. They may exist, but they are atypical.

The idea that generalities can't be true or useful is pretty insidious, since it prevents reasoning about groups.

[ Parent | on HN ]
Re:LLMs can't self-correct in reasoning tasks, DeepMind stud... by robertlagrant on Monday October 09, 2023 @06:08PM (#37826081)
> The idea that generalities can't be true or useful is pretty insidious

This idea was not mentioned, whether or not you attest that it was.

[ Parent | on HN ]