Matan Ben-Tov

@matanbt

CS PhD student in Computer Science @TelAvivUni. Interested in buzzwords like AI and Security, and wherever they meet.

Joined October 2009

1KFollowing

78Followers

Pinned

What makes or breaks powerful jailbreak suffixes? 🔓🤖 We find that: 🥷 they work by hijacking the model’s context; ♾️ the more universal a suffix is the stronger its hijacking; ⚔️🛡️ utilizing these insights, it is possible to both enhance and mitigate these attacks. 🧵

matanbt's tweet image. What makes or breaks powerful jailbreak suffixes? 🔓🤖

We find that:
🥷 they work by hijacking the model’s context;
♾️ the more universal a suffix is the stronger its hijacking;
⚔️🛡️ utilizing these insights, it is possible to both enhance and mitigate these attacks.

🧵

5.0K

Matan Ben-Tov Retweeted

Noam Bizan@NoamBizan · Jun 27

Russia-North Korea ties are expanding... read my article on the security implications of this relationship for #Geopolitics #RussianUkrainewar 🇷🇺🇰🇵 geopolreport.com/reports/the-st…

179