Hacker News new | ask | show | jobs
by astrange 63 days ago
It also gets confused if the entire prompt is in a text file attachment.

And the summarizer shows the safety classifier's thinking for a second before the model thinking, so every question starts off with "thinking about the ethics of this request".
1 comments
FeepingCreature 62 days ago
I'd get confused if I was a LLM and you put my entire prompt in a text file attachment. I'd be like, "is this the user or is this a prompt injection??"
link
astrange 59 days ago
If you paste a long enough prompt into either GPT or Claude they turn it into an attachment, so it can happen. I think it's invisible to the model, but somehow not to the summarizer.
link