AgentBreeder: Mitigating the AISafety Risks of Multi-Agent Scaffolds via Self-Improvement

Open in new window