Re:[BUG] Claude says "You're absolutely right!" about everyt...
by notahacker on Wednesday August 13, 2025 @08:07AM
(#44887427)
And then those sort of responses get parlayed into "chatbots give better feedback than medical doctors" headlines according to studies that rate them as high in "empathy" and don't worry about minor details like accuracy....
Re:[BUG] Claude says "You're absolutely right!" about everyt...
by cvwright on Wednesday August 13, 2025 @08:34AM
(#44887683)
This illustrates the dangers of training on Reddit.
Re:[BUG] Claude says "You're absolutely right!" about everyt...
by ryandrake on Wednesday August 13, 2025 @11:26AM
(#44889721)
I'm sure if you ask it for any relationship advice, it will eventually take the Reddit path and advise you to dump/divorce your partner, cut off all contact, and involve the police for a restraining order.
Re:[BUG] Claude says "You're absolutely right!" about everyt...
by uncircle on Wednesday August 13, 2025 @12:40PM
(#44890699)
“My code crashes, what did I do wrong?”
“NTA, the framework you are using is bad and should be ashamed of itself. What you can try to work around the problem is …”
Re:[BUG] Claude says "You're absolutely right!" about everyt...
by nullc on Wednesday August 13, 2025 @05:54PM
(#44894311)
It's not a (direct) product of reddit. The non-RLHFed base models absolutely do not exhibit this sycophantic behavior.
Re:[BUG] Claude says "You're absolutely right!" about everyt...
by cubefox on Wednesday August 13, 2025 @08:54AM
(#44887847)
I recently had Gemini disagree with me on a point about philosophy of language and logic, but it phrased the disagreement very politely, by first listing all the related points in which it agreed, and things like that.
So it seems that LLM "sycophancy" isn't necessarily about dishonest agreement, but possibly about being very polite. Which doesn't need to involve dishonesty. So LLM companies should, in principle, be able to make their models both subjectively "agreeable" and honest.
I pushed back and told it it was overreacting and it told me I was completely correct and very insightful and everything was normal with the patient and that they were extremely healthy.