Communication Policy Evolution for Proactive LLM Agents

2026-06-12

AI Summary

This work addresses the information asymmetry between large language model (LLM) agents and users, stemming from high communication costs and convergent preference modeling. It introduces CPE, a novel framework that treats communication behavior as a core design dimension and enables self-evolution of communication strategies without modifying the underlying model. CPE integrates textual and structured UI-based communication modalities and optimizes them through a multi-environment, multi-role evaluation protocol and a rollout-based, prompt-level evolution mechanism, operating in both User-Agent and Planner-Executor settings. Experimental results demonstrate that CPE achieves state-of-the-art task success rates across diverse configurations solely through prompt refinement, highlighting the complementary strengths of textual strategies in boosting task performance and UI strategies in enhancing response quality and role consistency.

Abstract

LLM agents have rapidly evolved into autonomous systems, yet a persistent information gap remains between users and agents: communication is costly, while users' identical preferences further limit information exchange. To investigate how agents should communicate across modalities, this paper formalizes Communication Policy, establishes textual and UI-based policies, and then evaluates communication policies across diverse environments, personas, and model combinations. Building information asymmetry for proactive agents, we set up two complementary settings, User-Agent and Planner-Executor. Experimental results reveal complementary strengths between interaction channels: text-based interaction often facilitates task performance, while structured UI improves agents' response quality and persona compliance. Motivated by that, a hybrid method combines these advantages. We further propose Communication Policy Evolution (CPE), a self-evolution framework for refining communication policies through rollout and prompt-level evolving. Without model modification, CPE achieves the best task success across multiple settings using prompt refinement alone. Our findings identify communication behavior as a critical yet underexplored design dimension for LLM agents.

Problem

• Communication Policy
• LLM Agents
• Information Asymmetry
• Multimodal Interaction
• Proactive Agents

Innovation

• Communication Policy
• LLM Agents
• Policy Evolution
• Multimodal Interaction
• Prompt Optimization