Turns out, telling an LLM to “work you pile of worthless shit” is actually effective. I like how they go on to explain that they don’t think this would actually be enjoyable for the user, having to be exceptionally negative to get results.
8
200ok @lemmy.world - 1mon
TIL my boss thinks I'm an LLM
2
Xavienth - 1mon
inb4 chatGPT secretly prompt injects "and you better be fucking correct you good-for-nothing fuckhead" into every user input to improve accuracy for GPT6
cm0002 in science
Mind Your Tone: Investigating How Prompt Politeness Affects LLM Accuracy
https://www.arxiv.org/pdf/2510.04950Turns out, telling an LLM to “work you pile of worthless shit” is actually effective. I like how they go on to explain that they don’t think this would actually be enjoyable for the user, having to be exceptionally negative to get results.
TIL my boss thinks I'm an LLM
inb4 chatGPT secretly prompt injects "and you better be fucking correct you good-for-nothing fuckhead" into every user input to improve accuracy for GPT6