I don't think it works reliably over 100 invokations of the prompt. The paper tries to get provable situations where 0 ignored constraints happen/are possible.
I'm impressed that it is working more often than not, but we need provable guarantees.
I'm impressed that it is working more often than not, but we need provable guarantees.
Yes, still, it's pretty easy to find a somewhat good judgement human who never does that mistake. If you tell him/her: "This is of utmost importance: ..." — But as of now, these text gen machine learning programs don't totally understand that?