News for hackers. Stuff that matters.

A comment on Detecting when LLMs are uncertain. Parent comment

Re:Detecting when LLMs are uncertain by eddd-ddde on Friday October 25, 2024 @07:26PM (#41950968)
Couldn't that just, never get an answer?

Or maybe just says "i don't know" with full certainty.

[ Parent | on HN ]
Re:Detecting when LLMs are uncertain by zbentley on Friday October 25, 2024 @07:31PM (#41951020)
That would be extremely useful in some domains.
[ Parent | on HN ]
Re:Detecting when LLMs are uncertain by mumblemumble on Friday October 25, 2024 @07:49PM (#41951186)
Perhaps only if you can also be very certain that the output is correct whenever the logprobs don't trigger the filter.

If that's not the case then it might just trigger bad risk compensation behavior in the model's human operators.

[ Parent | on HN ]