3 ms·
> Hallucinations are an inherent property of how LLMs work. While technically true the hallucination rates on modern models is low and other checks can ensure
by hahahaa 2mo ago
> Hallucinations are an inherent property of how LLMs work.
While technically true the hallucination rates on modern models is low and other checks can ensure that by the time a human sees it it is most likely solid.
For research there is more danger as there is less feedback loop other than other LLM scrutinising the first. For research I get it to come to a conclusion but provide me with links so I can judge. More like advanced search.
- righthand 2mo ago> it it is most likely solid. Most likely? That’s not reassuring at all. So you’re saying the other checks can result in hallucinations?
- hahahaa 2mo agoMost likely solid is definitely better than a lot of human made PRs I have seen. But making a decision the human reviewer disagrees with or misunderstanding a spec but maybe not asking a follow up isn't hallucination in my book. Hallucination is just making stuff up without checking.
- kamranjon 2mo ago“While technically true the hallucination rates on modern models is low…” Isn’t this entirely context dependent? Where did you get the information that modern models have low hallucination rates? I’d love to see the benchmark if there is one, it seems like it would be useful to track.
- munksbeer 2mo agoPut it this way, the agent may well be hallucinating behind the scenes when it is coding for me, but I never see that, because it obviously verifies by compiling and testing the solution and won't show me something that is broken.
- hahahaa 2mo agoAnecdotal. Older models felt they were on drugs.