Review
Jul 2026
AIR-BENCH Live: An Evolving Safety Benchmark for Foundation Models
A multi-agent, persona-driven prompt generation algorithm generates realistic, multilingual prompts with minimal human review, leaving room for improvement with modern jail breaking techniques in AIR-BENCH Live, a self-evolving successor to AIR-BENCH 2024.
Rohan Naphade, Minzhou Pan, Bo Li
· 0 citations