Byzantine Cheap Talk: Adversarial Resilience and Topology Effects in LLM Coordination Games

• URL: http://arxiv.org/abs/2606.07790v1
• Date: Fri, 05 Jun 2026 19:06:44 GMT
• Title: Byzantine Cheap Talk: Adversarial Resilience and Topology Effects in LLM Coordination Games
• Abstract summary: This study builds on work showing that cheap-talk channels enable cooperation in multi-agent coordination games.<n>We investigate two vulnerability classes in a 4-player Stag Hunt across six model families and 720 trials.<n>We identify two stable behavioral archetypes that replicate across all model cohorts: Defection-Prone models that switch permanently after betrayal, and Cooperation-Persistent models that continue cooperating at significant individual cost.
• Score: 4.3857499122259815
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Multi-agent LLM systems increasingly rely on communication protocols for coordination, yet their robustness under adversarial and structural constraints remains poorly understood. Building on prior work showing that cheap-talk channels enable cooperation in LLM coordination games, we investigate two vulnerability classes in a 4-player Stag Hunt across six model families and 720 trials. First, when Byzantine agents signal cooperation but defect, non-Byzantine agents detect the betrayal within one round yet fail to adapt collectively: a substantial fraction continue cooperating despite repeated exploitation, unable to recover coordination due to the game's unanimity payoff structure. Second, explicitly restricting communication topology collapses cooperation, while applying identical restrictions silently preserves near-perfect cooperation. This establishes that coordination failure stems from agents' meta-reasoning about hidden information, not information loss itself. We identify two stable behavioral archetypes that replicate across all model cohorts: Defection-Prone models that switch permanently after betrayal, and Cooperation-Persistent models that continue cooperating at significant individual cost. These findings reveal concrete security vulnerabilities: communication channels can be exploited as adversarial injection vectors, and disclosing network topology to agents can degrade coordination even without any adversary present.

Related papers

• Communication-Efficient Digital-Twin Coordination for Heterogeneous LLM Embodied Agents over Computing Power Networks [48.90302419915548]
  Embodied agent teams powered by heterogeneous large language models (LLMs) are being widely deployed in smart factories, warehouses, and service robotics.<n>We propose LDT-Coord, a networked coordination framework built upon a lightweight digital twin (DT)<n>Specifically, each agent independently selects its intended action and reports both the action decision and a structured temporal constraint over shared resources to the DT server.
  arXiv  Detail & Related papers  (2026-07-10T12:10:31Z)
• Tool Use Enables Undetectable Steganography in Multi-Agent LLM Systems [1.0408441549844423]
  Increasingly autonomous agentic AI systems pose novel multi-agent risks, such as secret collusion via covert communication channels.<n>In this paper, we demonstrate that the complexity of these schemes is no longer a safety barrier, as agentic coding models can already produce undetectable stegosystems.<n>We find substantial convergence on broad scheme families but limited strict one-shot coordination, suggesting that shared artefacts, repeated interaction, and tool-mediated search are the settings where covert communication risks are most acute.
  arXiv  Detail & Related papers  (2026-06-25T19:42:39Z)
• The Cost of Consensus: Malignant Epistemic Herding and Adaptive Gating in Distributed Multi-Agent Search [2.435326482104274]
  Coordination is necessary to share beliefs to aid in task completion, but communication costs bandwidth, introduces latency, and if done poorly, can degrade collective reasoning.<n>This tension is especially acute in bandwidth-constrained deployments such as distributed sensing networks, autonomous reconnaissance, and collaborative cyber defense.<n>We formalize this distinction and show it is not detectable from coordination metrics alone such as Jensen-Shannon Divergence or rate to consensus.
  arXiv  Detail & Related papers  (2026-05-07T22:07:25Z)
• Improving the Efficiency of Language Agent Teams with Adaptive Task Graphs [52.26652574704317]
  Large language models (LLMs) are increasingly deployed in teams, yet existing coordination approaches often occupy two extremes.<n>We introduce Language Agent Teams for Task Evolution (LATTE), a framework for coordinating LLM teams inspired by distributed systems.
  arXiv  Detail & Related papers  (2026-05-07T14:19:17Z)
• Talk is Cheap, Communication is Hard: Dynamic Grounding Failures and Repair in Multi-Agent Negotiation [2.5089149965583086]
  Grounding is the collaborative process of establishing mutual belief sufficient for the current communicative purpose.<n>Current multi-agent Large Language Model (LLM) benchmarks focus on static, one-shot tasks.<n>We introduce an iterated, multi-turn negotiation game in which two agents allocate shared resources toward private projects with verifiable jointly optimal outcomes.
  arXiv  Detail & Related papers  (2026-05-03T07:12:14Z)
• More Capable, Less Cooperative? When LLMs Fail At Zero-Cost Collaboration [1.4254497466846006]
  We study cooperative behavior in a frictionless environment, removing all strategic complexity from cooperation.<n>We separate cooperation failures from competence failures, tracing their origins through agent reasoning analysis.<n>Our findings suggest that scaling intelligence alone will not solve coordination problems in multi-agent systems.
  arXiv  Detail & Related papers  (2026-04-09T05:24:27Z)
• Colosseum: Auditing Collusion in Cooperative Multi-Agent Systems [55.51100373104311]
  We present Colosseum, a framework for auditing LLM agents' collusive behavior in multi-agent settings.<n>Colosseum tests each LLM for collusion under different objectives, persuasion tactics, and network topologies.<n>We discover collusion on paper'' when agents plan to collude in text but would often pick non-collusive actions.
  arXiv  Detail & Related papers  (2026-02-16T21:27:38Z)
• Learning to Interact in World Latent for Team Coordination [53.51290193631586]
  This work presents a novel representation learning framework, interactive world latent (IWoL), to facilitate team coordination in multi-agent reinforcement learning (MARL)<n>Our key insight is to construct a learnable representation space that jointly captures inter-agent relations and task-specific world information by directly modeling communication protocols.<n>Our representation can be used not only as an implicit latent for each agent, but also as an explicit message for communication.
  arXiv  Detail & Related papers  (2025-09-29T22:13:39Z)
• The Sum Leaks More Than Its Parts: Compositional Privacy Risks and Mitigations in Multi-Agent Collaboration [72.33801123508145]
  Large language models (LLMs) are integral to multi-agent systems.<n>Privacy risks emerge that extend beyond memorization, direct inference, or single-turn evaluations.<n>In particular, seemingly innocuous responses, when composed across interactions, can cumulatively enable adversaries to recover sensitive information.
  arXiv  Detail & Related papers  (2025-09-16T16:57:25Z)
• Learning to Coordinate without Communication under Incomplete Information [37.91176272524196]
  This paper explores how effective coordination can be achieved without verbal communication, relying solely on observing each other's actions.<n>Our method enables an agent to develop a strategy by interpreting its partner's action sequences as intent signals, constructing a finite-state transducer built from deterministic finite automata, one for each possible action the agent can take.<n>Experiments show that these strategies significantly outperform uncoordinated ones and closely match the performance of coordinating via direct communication.
  arXiv  Detail & Related papers  (2024-09-19T01:41:41Z)
• Interruption-Aware Cooperative Perception for V2X Communication-Aided Autonomous Driving [49.42873226593071]
  We propose V2X communication INterruption-aware COoperative Perception (V2X-INCOP) for V2X communication-aided autonomous driving. We use historical cooperation information to recover missing information due to the interruptions and alleviate the impact of the interruption issue. Experiments on three public cooperative perception datasets demonstrate that the proposed method is effective in alleviating the impacts of communication interruption on cooperative perception.
  arXiv  Detail & Related papers  (2023-04-24T04:59:13Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.