Yixin Tan
Titles as the sources gave them. In 1 entry, with 0 checked quotations.
In the Base
Every entry they appear in.
Diagnose Before You Refine: Multi-Dimensional Judge Feedback for Query-Efficient Jailbreak Attacks
A preprint proposes Jev-Guided Strategy Bandit, a jailbreak method using multi-dimensional judge feedback, and reports an average HarmBench attack success rate of 89.8% with only 10 queries on open-source settings.