Talk is Cheap, Communication is Hard: Dynamic Grounding Failures and Repair in Multi-Agent Negotiation

Grounding is the collaborative process of establishing mutual belief sufficient for the current communicative purpose. While static grounding maps language to a shared, externally observable context, dynamic grounding is a joint activity where meaning is negotiated through interaction. Current multi-agent LLM benchmarks focus on static, one-shot tasks, overlooking the ability to repair grounding breakdowns across turns. We introduce an iterated, multi-turn negotiation game in which two agents allocate shared resources toward private projects with verifiable jointly optimal outcomes. An oracle baseline confirms that individual agents can identify Pareto-optimal allocations in isolation, yet interacting dyads consistently fail to reach them across open- and closed-source models. Our investigation reveals four failure modes: (1) coordination degradation in shifting dyads lacking shared interaction history; (2) a reliance on perfunctory fairness---equal resource splits---over reward-maximizing coordination; (3) an anchoring effect where initial proposals are treated as axiomatic rather than negotiable; and (4) failures in referential binding, where agents lose track of commitments across turns. These results characterize dynamic grounding as a critical and understudied axis of multi-agent coordination, and our framework enables systematic study of interventions including information-sharing prompts and theory-of-mind elicitation.