Ev-Trust: A Strategy Equilibrium Trust Mechanism for Evolutionary Games in LLM-Based Multi-Agent Services

TL;DR

Ev-Trust introduces a strategy-equilibrium trust mechanism for evolutionary games in LLM-based multi-agent services, synthesizing direct and indirect trust into agents' expected revenue to drive strategy evolution. Using replicator dynamics, the authors prove local evolutionary equilibrium existence and stability, and demonstrate in a decentralized open service market that malicious behavior declines while collective revenue and trust robustness improve. The framework operates within a Request-Response-Payment-Evaluation workflow, enabling organic exclusion of malicious participants without centralized arbitration. This work advances trust modeling in the Agentic Web by presenting a theoretically grounded, self-regulating mechanism with empirical validation across large-scale agent populations.

Abstract

The rapid evolution of the Web toward an agent-centric paradigm, driven by large language models (LLMs), has enabled autonomous agents to reason, plan, and interact in complex decentralized environments. However, the openness and heterogeneity of LLM-based multi-agent systems also amplify the risks of deception, fraud, and misinformation, posing severe challenges to trust establishment and system robustness. To address this issue, we propose Ev-Trust, a strategy-equilibrium trust mechanism grounded in evolutionary game theory. This mechanism integrates direct trust, indirect trust, and expected revenue into a dynamic feedback structure that guides agents' behavioral evolution toward equilibria. Within a decentralized "Request-Response-Payment-Evaluation" service framework, Ev-Trust enables agents to adaptively adjust strategies, naturally excluding malicious participants while reinforcing high-quality collaboration. Furthermore, our theoretical derivation based on replicator dynamics equations proves the existence and stability of local evolutionary equilibria. Experimental results indicate that our approach effectively reflects agent trustworthiness in LLM-driven open service interaction scenarios, reduces malicious strategies, and increases collective revenue. We hope Ev-Trust can provide a new perspective on trust modeling for the agentic service web in group evolutionary game scenarios.

Figures (12)

• Figure 1: Examples of fraud or malicious behavior in interactions between agents. Malicious agents disrupt multi-agent collaborative networks through interaction.
• Figure 2: Ev-Trust Workflows. The example in the figure describes a collaborative subgame involving malicious agents. After forming groups, issuing requests, and exchanging trust, the malicious agent chooses the fraud strategy to obtain the maximum immediate revenue. The task execution fails, the agents give each other trust evaluations, and the malicious agent is punished for its lack of honesty and leaves the current group to join another group.
• Figure 3: Trust Value Evolution. The x-axis represents the number of system evolution cycles, and the y-axis represents changes in agent trust values. The red line indicates the period before and after the mutation event. Since the BRS and ICFP methods lack mutual evaluation between service providers and consumers, only the provider's statistical results are presented.
• Figure 4: Strategy Proportions. The proportions of the three strategies in all interactions change as the system evolves. Denials to service are excluded from statistics as they do not impact trustworthiness.
• Figure 5: Agent Revenue. Our approach compares the revenue of agents in the two other baseline mechanisms as the system evolves.

Theorems & Definitions (1)

• theorem 1: Joint evolutionary stability