Influencing LLM Multi-Agent Dialogue via Policy-Parameterized Prompts

Researchers propose a framework using policy-parameterized prompts to influence multi-agent LLM dialogue behavior without training. The approach treats prompts as actions and dynamically constructs them through five components to control conversation flow based on metrics like responsiveness and stance shift.

• →New framework treats prompts as parameterized actions to control LLM multi-agent behavior without requiring training.
• →System dynamically constructs prompts through five components based on current agent state.
• →Dialogue effectiveness measured across five indicators: responsiveness, rebuttal, evidence usage, non-repetition, and stance shift.
• →Experiments demonstrate that prompt parameterization can successfully influence dialogue dynamics in discussion scenarios.
• →Framework offers lightweight policy mechanism for advancing multi-agent systems and social simulation research.