Zhaoyu Hu
Titles as the sources gave them. In 1 entry, with 0 checked quotations.
In the Base
Every entry they appear in.
ProtoArena: Can AI Agents Discover and Exploit Real-World Cross-Component Protocol Vulnerabilities?
According to the description, a preprint introduces ProtoArena, a benchmark of 80 environments in which AI agents try to exploit vulnerabilities that depend on how network protocol components interact. Of five models tested, the authors report Claude Opus 4.8 scored highest: 48.8% with guidance and 12.5% in the black-box setting.