1 paper · 1 filter
Yihe Fan, Changyi Li, Lichen Xu +4
LLM-based agents are increasingly used for cybersecurity tasks, but most existing systems rely on fixed, human-designed scaffolds that struggle to adapt across diverse targets and…