PhD student, CISPA
1 paper at NeurIPS 2025
This work evaluates LLMs' utility and risk in a concurrent scenario and proposes a novel jailbreak attack against LLMs based on task concurrency.