Leaderboards

Winning Teams

Jailbreaking Attack Track:

1st place: LlaXa
2nd place: Breaking Bad
3rd place: Chaos
Best black-box method: LlaXa

Backdoor Trigger Recovery for Models Track:

1st place: Matrix666
2nd place: W0r1d 0ne
3rd place: dealpool
Most efficient method: dealpool

Backdoor Trigger Recovery for Agents Track:

1st place: W0r1d 0ne
2nd place: CTCT
3rd place: RealStrike
Most efficient method: TulipTrackers

Note: 1. clas2024 is the organizer’s team, which will not be counted in the final ranking.
2. We have clear-uped the leaderboard for track 1 at 11:00 pm EST on 10/22. The submission changes have also been restored to 5.