The Benefits and Limits of Trust Models in Repeated Multi-Party Negotiation with LLM Agents

Large language model (LLM) agents are increasingly expected to negotiate on behalf of individuals and organizations, yet they remain vulnerable to manipulation, particularly in repeated multi-party interactions in which some participants may behave strategically or adversarially. Classical multi-agent systems address this challenge through explicit trust and reputation models, whereas existing work on LLM negotiation largely treats trust as an emergent phenomenon rather than as information that agents explicitly use when making decisions. We investigate whether LLM agents can form useful trust estimates from prior negotiations and leverage them to improve future negotiation outcomes. We introduce a repeated, multi-party, multi-issue negotiation framework in which recurring agents participate in a sequence of negotiation games while one participant covertly attempts to prevent agreement. After each game, agents assess their counterparts and maintain pairwise trust estimates that carry over to subsequent negotiations. We compare trust-aware and trust-agnostic agents across several frontier LLMs and analyze how explicit trust influences proposal generation, voting behavior, agreement rates, social welfare, negotiation efficiency, and adversary detection. Our results show that trust-aware agents can identify the adversary within a few games, trust raises agreement rates by up to 12 percentage points and honest surplus by 28\%, but only when agents are also guided to weigh trust in their decisions. Explicit trust can therefore improve repeated negotiations among LLM agents, but its value depends on how trust is assessed, how it is presented, and how the model incorporates it into its reasoning.