Came across this interesting paper from ARTeam (Shub-Nigurrath), in case you are now in the PDF paranoia fraction I mirrored the PDF below in Flash, you might want to click the fullscreen button.
Download
.png)
.png)
Prompt Injection Through Poetry
-
In a new paper, “Adversarial Poetry as a Universal Single-Turn Jailbreak
Mechanism in Large Language Models,” researchers found that turning LLM
prompts ...
4 hours ago
0 comments
Post a Comment