- Published
- Source
- Schneier on Security
New paper: “Self-Jailbreaking: Language Models Can Reason Themselves Out of Safety Alignment After Benign Reasoning Training.” Abstract: We discover a novel and surprising phenomenon of unintentional misalignment in reasoning language models (RLMs), which we call self-jailbreaking. Specifically, after benign reasoning training on math or code domains, RLMs will use multiple strategies to circumvent their own…
MyCyberWay career service
Not sure where to start?
Book a free 20-minute fit call.
Tell us where you are and where you want to be. We'll tell you honestly which programme fits – including if the answer is none of them yet.
- Twenty minutes, free
- A straight recommendation
- A person replies
Headlines and short summaries only, except where a public body publishes under an open licence. Every story links to its publisher, who owns it. MyCyberWay does not edit or endorse third-party reporting.
