Control-Data Flow Separation: Stable Prompt Optimization in Multi-Agent LLMs

Researchers propose a control-data flow separation approach for stable prompt optimization in multi-agent large language model (LLM) systems, addressing the entanglement of task-relevant content and execution-critical protocols such as message routing, output formatting, and termination signals. The method aims to improve optimization stability by decoupling these roles.

Prompt optimization can improve multi-agent LLM systems, but the prompts being optimized often serve two entangled roles: generating task-relevant content and specifying execution-critical protocols, such as message routing, output formatting, and termination signals, on which the underlying code re